Association with the SRC Family Tyrosyl Kinase LYN Triggers a Conformational Change in the Catalytic Region of Human cAMP-specific Phosphodiesterase HSPDE4A4B

The cAMP-specific phosphodiesterase (PDE) HSPDE 4A4B(pde46) selectively bound SH3 domains of SRC family tyrosyl kinases. Such an interaction profoundly changed the inhibition of PDE4 activity caused by the PDE4-selective inhibitor rolipram and mimicked the enhanced rolipram inhibition seen for particulate, compared with cytosolic pde46 expressed in COS7 cells. Particulate pde46 co-localized with LYN kinase in COS7 cells. The unique N-terminal and LR2 regions of pde46 contained the sites for SH3 binding. Altered rolipram inhibition was triggered by SH3 domain interaction with the LR2 region. Purified LYN SH3 and human PDE4A LR2 could be co-immunoprecipitated, indicating a direct interaction. Protein kinase A-phosphorylated pde46 remained able to bind LYN SH3. pde46 was found to be associated with SRC kinase in the cytosol of COS1 cells, leading to aberrant kinetics of rolipram inhibition. It is suggested that pde46 may be associated with SRC family tyrosyl kinases in intact cells and that the ensuing SH3 domain interaction with the LR2 region of pde46 alters the conformation of the PDE catalytic unit, as detected by altered rolipram inhibition. Interaction between pde46 and SRC family tyrosyl kinases highlights a potentially novel regulatory system and point of signaling system cross-talk.

The cAMP-specific phosphodiesterase (PDE) HSPDE 4A4B(pde46) selectively bound SH3 domains of SRC family tyrosyl kinases. Such an interaction profoundly changed the inhibition of PDE4 activity caused by the PDE4-selective inhibitor rolipram and mimicked the enhanced rolipram inhibition seen for particulate, compared with cytosolic pde46 expressed in COS7 cells. Particulate pde46 co-localized with LYN kinase in COS7 cells. The unique N-terminal and LR2 regions of pde46 contained the sites for SH3 binding. Altered rolipram inhibition was triggered by SH3 domain interaction with the LR2 region. Purified LYN SH3 and human PDE4A LR2 could be co-immunoprecipitated, indicating a direct interaction. Protein kinase A-phosphorylated pde46 remained able to bind LYN SH3. pde46 was found to be associated with SRC kinase in the cytosol of COS1 cells, leading to aberrant kinetics of rolipram inhibition. It is suggested that pde46 may be associated with SRC family tyrosyl kinases in intact cells and that the ensuing SH3 domain interaction with the LR2 region of pde46 alters the conformation of the PDE catalytic unit, as detected by altered rolipram inhibition. Interaction between pde46 and SRC family tyrosyl kinases highlights a potentially novel regulatory system and point of signaling system cross-talk.
Although it has long been appreciated that cAMP plays a pivotal role in controlling a wide range of cellular processes, the complexity of the signaling system responsible for the generation, detection, and degradation of this second messenger has only recently become apparent. Thus nine forms of adenylyl cyclase able to generate cAMP, and around 30 forms of cyclic nucleotide phosphodiesterase (PDE) 1 isoenzymes able to de-grade cAMP, have been identified together with multiple forms of protein kinase A (PKA) and a large family of PKA-anchoring proteins (1)(2)(3)(4)(5)(6)(7). A number of these species have been shown to exhibit specific intracellular distributions. For example, the localization of particular adenylyl cyclase isoenzymes to distinct regions of the plasma membrane in polar cells (8), the targeting of PDE4 isoenzymes via their N-terminal alternatively spliced regions (3, 9 -14), and the binding of PKA isoenzymes to anchor proteins have all been noted (1,4). These arrangements demonstrate a propensity to organize the components of the cAMP signal transduction system within defined regions of the three-dimensional matrix of the cell interior. This, undoubtedly, provides the molecular basis of the compartmentalization of cAMP signaling that has been noted in a number of different cell types (1,2,4). That specific adenylyl cyclase and PDE isoenzymes can be regulated through the action of a variety of intracellular signaling systems suggests that cAMP signaling may be regulated in distinct fashions in different cells (2).
The cAMP-specific phosphodiesterase, PDE4 enzyme family is encoded by four genes which each produce a series of isoenzymes through alternative mRNA splicing (3,6,7). Selective inhibitors of these enzymes are currently being developed as anti-inflammatory agents that appear likely to be of potential therapeutic use in a variety of disease states including asthma, rheumatoid arthritis, and AIDS (15)(16)(17)(18). Each PDE4 isoenzyme has a unique N-terminal region which, in the case of the so-called "long" isoenzyme, are connected to the catalytic domain by two regions that provide unique signatures of this enzyme family. These are the Upstream Conserved Regions, UCR1 and UCR2 (19). One feature that distinguishes isoenzymes of each PDE4 class is the nature of their two Linker Regions (3) namely LR1, which connects UCR1 to UCR2, and LR2, which connects UCR2 to the catalytic domain (3).
There is a growing realization that proteins involved in intracellular signaling systems can be recruited and organized within distinct intracellular domains (4). SH3 domains, which are found in various families of proteins including adapter and cytoskeletal proteins and signal transducing proteins such as SRC family tyrosyl kinases, can serve such a function (20,21).
They are distinct, self-folding, globular units of ϳ60 residues that confer protein-protein interaction by binding proline-rich regions on acceptor proteins. Interactions involving SH3 domains allow the assembly of functionally active complexes that serve to control conformational switches in a number of key regulatory pathways (20).
We have shown (12) that the N-terminal region of the long human PDE4A isoenzyme pde46 appears to be responsible, at least in part, for its intracellular targeting in COS7 cells. Such membrane association was accompanied (12) by a profound increase in the sensitivity of the particulate form of this enzyme to inhibition by the archetypal PDE4-selective inhibitor rolipram. Additionally, this interaction resulted in a striking change in the kinetics of enzyme inhibition by rolipram, implying that particulate association had triggered a conformational change in this enzyme. Here we identify a distinct proline-and arginine-rich stretch of sequence located within the LR2 region of pde46 which confers SH3 domain binding and triggers a conformational change in the catalytic unit of this enzyme . PCR using primers ESH1 and ESH4. After purification on an agarose gel the mutated 5Ј regions were then inserted into to pSVsport-h6. 1 using XhoI (in h6.1 sequence) and BamHI (in the pSVsport multiple cloning site and in primer ESH1 sequence) restriction sites, replacing the unmodified region. The sequences of the resultant constructs were then confirmed.
Generation of the ⌬b-pde46 Mutant-The deletion ⌬b (Fig. 1), encompassing amino acids 313-320 inclusive, of HSPDE4A4B (pde46) (19) was generated by site-directed mutagenesis using the QuickChange Mutagenesis system (Stratagene Ltd., Cambridge, UK) according to the manufacturer's instructions. This employed the plasmid pSV.SPORT-h46 (12) together with the complementary oligonucleotide primers GSd-b1, 5Ј-CGAGAAAAACAGCAACCGCCCCCG-3Ј (sense) and GSdb2, 5Ј-CGGGGGCGGTTGCTGTTTTTCTCG-3Ј (antisense) in a PCR reaction. Upon completion, the reaction mix was treated with the restriction enzyme DpnI, and digested samples were transformed into competent Escherichia coli strain XL1-Blue. This generated the plasmid pSV.h-PDE46-⌬b. Confirmation of the mutation was obtained by sequencing of miniprep DNA.
Generation of a GST Fusion Protein of a VSV Epitope-tagged Form of the Human LR2 Region-PCR was used to generate the LR2 region (3) (Fig. 1) of pde46 with a C-terminal VSV epitope tag using the synthetic oligonucleotide primers GS-h46-LR2 5Ј-GCGGGATCCATGCCATCAC-CCACG-3Ј (sense) and GS-h46-LR2 5Ј-TGCTCTAGATTACTTTCCCA-GCCTGTTCATCTCTATATCGGTGTACTGTAAGTGTGGTAC-3Ј (antisense) and pde46 as template DNA. The PCR fragment generated was treated with BamHI and XbaI, and the digested fragment was purified before ligation into the BamHI/XbaI sites of the plasmid pcDNA3, yielding the plasmid pcDNA-LR2-VSV. This was used as a template for PCR with the following oligonucleotide primers: h46 pGEX 5Ј, 5Ј-CGCGGA-TCCCATCACCCACG-3Ј, and h46 pGEX 3Ј, 5Ј-CGGCTCGAGTTACTT-TCCCAGCC-3Ј. The resultant PCR fragment was digested with BamHI and XhoI; the fragment was gel-purified and ligated into BamHI, XhoI cut pGEX-5X-1 (Amersham Pharmacia Biotech) to form an in-frame fusion with the GST gene. The resultant plasmid was designated pGEX-LR2-VSV. Sequences of all constructs were confirmed.
Preparation and Generation of SH3 Domain Fusion Proteins-GST fusion proteins of the various SH3 domains employed in this study were generated as described previously by us (24) with the exception of those of c-Abl (kind gift from Dr. D. Baltimore, MIT, Cambridge, MA) and c-Crk (kind gift from Dr. S. Fischer, INSERM, Paris, France). The procedures used to grow transformed E. coli and then to induce, isolate, and purify various GST fusion proteins were done as described in detail by us before (24).
Pull-down Assays with GST Fusion Proteins-This was performed using a modification of a procedure described previously by us (24). Volumes of slurry containing 400 g of fusion protein immobilized on glutathione-agarose were pelleted, and the supernatants were discarded. Within each assay, volumes taken were equalized with washed beads. The pellets were resuspended in 100 g of crude cytosol of COS7 cells transiently transfected to express the PDE4A form, diluted to a final volume of 200 l in KHEM buffer (50 mM KCl, 50 mM HEPES-KOH, pH 7.2, l0 mM EGTA, 1.92 mM MgCl 2 ) containing 1 mM dithiothreitol and protease inhibitor mixture. The immobilized fusion protein and cytosol were incubated together for 10-min "end-over-end" at 4°C. The beads were then collected by centrifugation for 5 s at high speed in a benchtop centrifuge, and the supernatant was retained as the unbound fraction. The beads were held on ice and washed three times with 400 l of KHEM containing 1 mM dithiothreitol and protease inhibitor mixture over a 15-min period. These washes were pooled along with the unbound fraction and aliquots taken for PDE assay and Western blotting. Bound PDE was eluted from the beads by incubating three times in 100 l of elution buffer (10 mM glutathione, 50 mM Tris-HCl, pH 8.0), for 10 min at 4°C. The eluted fractions were pooled and aliquots taken for PDE assay and Western blotting.
Binding of PDE4 Species to SH3 Fusion Proteins Prior to Kinetic Analyses-COS7 cells were transfected with either native or the indicated mutant forms of pde46 and h6.1. They were lysed in KHEM buffer, and the high speed supernatant fraction was divided into 100-l aliquots, each of which contained typically 180 -200 g of protein.
In these experiments each 100-l aliquot of the COS7 cell high speed supernatant fraction was incubated with 200 g of the indicated SH3 domain-GST fusion protein that had been purified from bacterial extracts and immobilized on glutathione-agarose. The immobilized fusion protein and indicated PDE form were incubated end-over-end at 4°C for 10 min. The beads were washed 3 times in KHEM buffer over a 15-min period and eluted by 3ϫ 100-l washes in ice-cold elution buffer (see above). The eluted fractions were then pooled and taken for imme-diate analysis of PDE activity. In some instances an additional 200 g of SH3 domain GST fusion protein was added to the eluted fraction. Over the time of the PDE assay Ͼ95% of each of the PDE forms remained bound to the SH3 (LYN/FYN/SRC) fusion proteins. No binding occurred to GST alone, and GST addition to PDE assays did not elicit any change in rolipram inhibition of the PDEs analyzed herein.
Transfection and Subcellular Fractionation of COS7 Cells-Cells were grown and transfected as described before by us in some detail (12). Disruption of COS7 cells was done as described previously by us in some detail (9,12). This procedure routinely yielded (14, 25) a P1 pellet (1,000 ϫ g av for 5 min) and P2 pellet (60 min at 100,000 ϫ g av ) as well as a high speed supernatant (sn). The homogenization procedure was complete in that there was no detectable latent lactate dehydrogenase activity present in either of the pellet fractions, indicating the absence of cytosolic proteins and Ͼ98% disruption. The high speed supernatant, P1 and P2 fractions, contained 34 Ϯ 3, 25 Ϯ 2, and 41 Ϯ 2% of the total protein, respectively (means Ϯ S.D.; n ϭ 8 experiments).
SDS-PAGE and Western Blotting-Samples of cellular fractions (2-50 mg of protein) were boiled for 5 min after being resuspended in Laemmli (26) buffer and were separated on 10% acrylamide gels. Gels were routinely run at 40 mA/gel for 4 -5 h with cooling. Transfected PDE was detected by Western blotting with human PDE4A-specific antisera (12) following transfer to nitrocellulose. Immunoreactive bands were identified by using anti-rabbit peroxidase-linked IgG and the Amersham Pharmacia Biotech ECL detection system as described before by us (12).
Immunoprecipitation of Fusion Proteins-The GST fusion protein GST-LR2-VSV (25 g) was diluted to 500 l in immunoprecipitation buffer (50 mM Tris-HCl, pH 8.0, 100 mM NaCl, 1 mM EDTA, 0.02% Triton X-100) containing an equal amount of GST or LYN-SH3 GST fusion protein. Fusion protein mixtures were incubated for 1 h at 4°C with anti-VSV antiserum (1:50 v/v) or mouse IgG (1:50 v/v) after which protein A-agarose beads (1:50 v/v) were added. After incubation for an additional hour, immune complexes were pelleted by centrifugation at 14,000 ϫ g av and washed three times in immunoprecipitation buffer. Following washing, immunoprecipitates were subjected to SDS-PAGE and Western blotting with anti-GST antiserum (1:5,000 v/v). Labeled bands were identified by using anti-rabbit peroxidase-linked IgG (1: 10,000 v/v) and visualized with the Amersham Pharmacia Biotech ECL protocol.
Co-immunoprecipitation of pde46 with LYN-COS7 and COS1 cells were transfected with 10 g of plasmid pcDNA3 containing a cDNA of full-length pde46 by the DEAE-Dextran transfection method. Seventy two hours post-transfection cells were harvested in RIPA buffer (55 mM Tris-HCl, pH 7.4, 132 mM NaCl, 22 mM sodium fluoride, 11 mM sodium pyrophosphate, 1.1 mM EDTA, 5.5 mM EGTA) and then lysed with 8 strokes of a 26 1 ⁄2-gauge needle attached to a disposable syringe. Unbroken cells were removed by centrifugation at 1,000 ϫ g av for 5 min. The resulting supernatant was centrifuged at 100,000 ϫ g av for 45 min at 4°C to yield a high speed supernatant fraction (S2) and a high speed pellet fraction (P2). 500 g of S2 was mixed with 30 l of pre-equilibrated protein A-agarose and incubated for 30 min at 4°C. Beads were removed by centrifugation at 2,000 ϫ g av for 5 min, and the cleared lysates were incubated with 4 g/ml anti-LYN antisera in the presence of protein A-agarose beads for 3 h at 4°C. Beads were collected by centrifugation (2,000 ϫ g av for 5 min) and washed three times with lysis buffer. 0.2 mg (protein) of the high speed (P2) pellet fraction was treated with 500 l of lysis buffer containing 0.5% Nonidet P-40. They were then incubated for 30 min at 4°C and then centrifuged at 100,000 ϫ g av for 45 min at 4°C. The resulting supernatant was mixed with 30 l of pre-equilibrated protein A-agarose and incubated for 30 min at 4°C. Beads were removed by centrifugation at 2,000 ϫ g av for 5 min, and the cleared supernatant was incubated with 4 g/ml anti-LYN antiserum in the presence of protein A-agarose beads for 3 h at 4°C. Beads were collected by centrifugation (2,000 ϫg av for 5 min) and washed three times with lysis buffer. Co-immunoprecipitation of pde46 with LYN was analyzed by immunoblotting with anti-human PDE4A antiserum (12) and a LYN mAb.
LYN SH3 Binding to Protein Kinase A-phosphorylated pde46 -The cytosolic fraction of pde46 from transfected COS7 cells, containing 8 pmol/min/l enzyme, and the equivalent amount of cytosol (50 g) from mock-transfected cells were each made up to 200 l in PKA assay buffer (0.2 mM ATP, 0.1 mM [ 32 P]ATP, 10 mM MgCl 2 , 30 mM ␤-mercaptoethanol, 10% glycerol, 100 mM Tris-HCl, pH 7.5) and incubated in the presence of 2 units of PKA for 30 min at room temperature. The sample was then divided into two equal aliquots, one of which was incubated in the presence of a LYN SH3-GST (100 g) fusion protein, and the other was incubated in the presence of GST (100 g) alone. After a 30-min incubation on ice, 50 l of glutathione beads were added to each aliquot and incubated end-over-end for 30 min at 4°C. The beads were then washed 3 times in 0.5 ml of phosphate-buffered saline containing protease inhibitors to remove any unbound proteins before being washed 3 times (5 s centrifugation at 10,000 ϫ g av ) in 100 l of elution buffer (5 mM glutathione, 50 mM Tris-HCl, pH 8) to release the GST, the LYN SH3-GST fusion protein, and any proteins that may have bound to them. The eluates were then incubated for 30 min on ice with an antiserum specific for the unique C terminus of human PDE4A (12). Antibody-antigen complexes were then precipitated using protein A beads and washed 3 times in 0.5 ml of phosphate-buffered saline. The resultant pellets were boiled in Laemmli sample buffer (26) and applied to the lanes of an 8% polyacrylamide gel. After running the gel, the proteins were transferred to nitro-cellulose, and radioactive bands were detected using x-ray film and PhosphorImager plates.
Binding of h6.1 to P1 Pellet Fraction and Competition with GST-LR2-VSV-Using procedures described by us before (27,28) h6.1 was synthesized in vitro using the coupled, single tube STP3 T7 transcription/ translation system (Novagen) according to the manufacturer's instructions. Briefly, 1 g/reaction of pSVsport-h6.1 DNA template was added to STP3 Transcription Mix (10 l total reaction volume) and incubated at 30°C for 15 min. Following the transcription step, 30 l of STP3 Translation Mix and 40 Ci of [ 35 S]methionine was added to the reaction tube. The reaction volume was made to 50 l with nucleasefree water, and then the tube was incubated at 30°C for 60 min. This was used in binding experiments as described before by us (27,28). Here, 5 l of in vitro synthesized h6.1 was incubated for 10 min at 4°C with 10 g of P1 pellet fraction, and the indicated amount of either GST-LR2-VSV fusion protein or GST (control) in a total volume of 500 l of binding buffer (55 mM Tris-HCl, pH 7.4, 132 mM NaCl, 22 mM sodium fluoride, 11 mM sodium pyrophosphate, 1.1 mM EDTA, 5.5 mM EGTA) containing protease inhibitor mixture. Pellets were then collected by centrifugation (2,000 ϫ g av , 5 min, 4°C), washed 3 times with binding buffer, subjected to SDS-PAGE, and Western-blotted. Radioactive bands were identified by PhosphorImager analysis.
Immunofluorescence Analyses-48 h after transfection cells were transferred onto coverslips (18 ϫ 18 mm) at about 40% confluency. They were grown for a further 24 h and then fixed for 30 min in 4% paraformaldehyde in Tris-buffered saline (TBS). Cells were permeabilized with 3 changes of 0.2% Triton in TBS for 15 min, and, following four 5-min blocking incubations with 20% goat serum and 4% bovine serum albumin, were labeled for 2 h with polyclonal antibodies raised against specific peptide sequences of the C-terminal region of PDE 4A4B (12) or RNPDE4A5 (10,11). Alternatively a monoclonal anti-PDE4A antibody, also raised against the C-terminal region of PDE4A4B, was used. Labeling was detected using a tetramethyl rhodamine isothiocyanateconjugated goat anti-rabbit IgG or Alexa 594-conjugated goat antimouse IgG for 1 h. Co-staining of cells was achieved using a monoclonal mouse anti-LYN antibody at a dilution of 1:100 and a polyclonal antibody raised to a specific peptide sequence of the G s ␣ subunit of GTPbinding protein. Localization of proteins was visualized using FITCconjugated goat anti-mouse or anti-rabbit IgG. All incubations were performed at room temperature. Cells were visualized using a laserscanning confocal microscope using an Axiovert 100 microscope with a X63/1.4NA plan apochromat lens, as described previously (12).
PDE Assay-cAMP PDE activity was assessed at 30°C by a modification of the two-step procedure of Thompson and Appleman (29) as described previously by us (30). Initial rates were taken from linear time courses of activity. Mock transfections, with vector only, as described before (12,13), did not alter the endogenous COS cell PDE activity. As a routine we subtracted the residual COS cell PDE activities done in parallel experiments from those activities found in PDEtransfected cells. Protein was measured by the method of Bradford (31) using bovine serum albumin as a standard. Dose-effect inhibitor analyses were done with 1 M cAMP as a substrate. Racemic rolipram was dissolved in 100% Me 2 SO as a 10 mM stock and diluted in 20 mM Tris-HCl, pH 7.4, 10 mM MgCl 2 buffer for dilution in the assay. Residual Me 2 SO did not affect PDE activity over the ranges used here. Inhibitor studies were analyzed using KaleidaGraph (Synergy Software, Reading, PA). To define K m values, data from PDE assays were done over a range of cAMP substrate concentrations and then analyzed by fitting to the hyperbolic form of the Michaelis-Menten equation using an iterative least squares procedure (Ultrafit; with Marquardt algorithm, robust fit, experimental errors supplied; Biosoft, Cambridge, UK) (12,25).

RESULTS AND DISCUSSION
Through alternative mRNA splicing each of the four PDE4 genes generates a number of isoenzymes that have unique N-terminal regions (3,6,7,19,32). Such splicing, however, occurs at two distinct splice junctions that generate so-called "short" and "long" PDE4 isoenzymes. The long forms are characterized by the presence of two regions that provide a unique signature for members of the PDE4 multigene family (19), namely the so-called UCR1 and UCR2 regions (Fig. 1). These are located between the unique N-terminal region of each isoenzyme and their catalytic unit. The region that connects UCR1 and UCR2 has been called (3) linker region 1 (LR1) and that region which connects UCR2 to the catalytic region has been called linker region 2 (LR2). These linker regions vary dramatically in sequence between the different PDE4 families, with the most profound differences seen for the LR2 region (3). In contrast to the long PDE4 isoenzymes, however, the short isoenzymes lack a UCR1 region.
Interactions between SH3 domains and putative target proteins have been investigated in a number of laboratories by using specific SH3 domains expressed as in-frame fusion proteins with GST (20,21). We have used such an approach to demonstrate (24) that the long rat PDE4A isoenzyme, rpde6, can interact with the SH3 domains from a variety of proteins, showing an apparent preference for interaction with the SH3 domains of SRC family tyrosyl kinases. This interaction was mediated exclusively by the extreme N-terminal alternatively spliced region of 112 residues, as both a different long rat PDE4 isoenzyme (rpde39) and an engineered rat PDE4A species (Met 26 -RD1), which lacks all N-terminal sequence up from UCR2, failed to bind to the SH3 domains of SRC family tyrosyl kinases. Within the N-terminal alternatively spliced region of rpde6 are three motifs of the form that might be predicted to interact with SH3 domain-containing proteins, including those of SRC family tyrosyl kinases. These consist of a "core" PXXP motif together with a closely associated arginine residue, yielding a motif of the general form either PXXPXXR or RXXPXXP. Three similar motifs are found ( Fig. 1) in the N-terminal region of pde46, the human homologue of rpde6. We show here (Table  I and Fig. 2) that pde46 was able to bind to GST fusion proteins expressing the SH3 domains of the tyrosyl kinases LYN and SRC. There is an apparent specificity in this interaction. Relative to the binding of pde46 to SRC SH3 (set at unity) both LYN SH3 (4.9 Ϯ 0.5 times as effective) and FYN SH3 (4.4 Ϯ 0.3 times) bound more effectively, whereas ABL SH3 bound similarly (0.7 Ϯ 0.2 times), and CRK SH3 (0.05 Ϯ 0.03 times) and LCK SH3 (0.04 Ϯ 0.02 times) showed little if any interaction (data are means Ϯ S.D. for n ϭ 3 separate experiments; analyses done by Western blotting of pull-down assays as in Fig. 2; 100 g of COS cell lysate used plus 2ϫ 200 g of fusion proteins as described in detail under "Experimental Procedures". Previously, we have shown (24) that an N-terminal truncated rodent PDE4A form called Met 26 -RD1, was unable to bind to the SH3 domain of SRC. In marked contrast to this, however, we noted in this study that a cognate human PDE4A N-terminal truncate, exemplified by the form h6.1 (Fig. 1), was still able to bind to the SH3 domains of LYN and SRC (Table I and Fig. 2). The PDE4A homologues, pde46 (human) and rpde6 (rat), are extremely similar proteins, particularly regarding their alternatively spliced N-terminal regions and core catalytic unit. However, there are certain differences in their primary structure (3,19) which appear to be highly localized. The most obvious differences relate to insertions in pde46 found both toward its C terminus and within its LR2 region (3). Indeed, the sequence of the exon 8-encoded LR2 region appears to be hypervariable not only between the various PDE4 families but also between cognate PDE4A isoenzymes from different mammals (3,32). This latter point is clearly evident upon comparison of the sequence of the LR2 region from human and rat PDE4A species ( Fig. 1 and Ref. 32). Inspection of this region shows that the human, but not the rat, LR2 region ( Fig. 1) is characterized by the presence of an RXXPXXP motif contained within a highly proline-and arginine-rich stretch of sequence. As such, the human PDE4A LR2 region is of a form that might be expected to be able to interact with SH3 domains. This difference between the rat and human LR2 regions (51) might  (19) and the N-terminal truncated species h6.1 (22). Identified features are the unique N-terminal alternatively spliced region of pde46 and which contains three putative SH3 binding motifs starting at Pro 3 , Pro 37 , and Pro 61 ; the upstream conserved regions UCR1 and UCR2 which are unique to PDE4 family members and start at Ser 140 and Gln 228 , respectively; the linker region LR1, which connects UCR1 and UCR2 and starts at Phe 195 ; linker region LR2 which starts at Met 305 and connects UCR2 to the putative catalytic unit which starts at Met 332 ; two other putative SH3 binding regions, at Pro 684 and Pro 819 , found toward the C-terminal region of the protein which start at Glu 701 . h6.1 has 9 non-native amino acids at its N terminus. The form of the ⌬a and ⌬b mutants, which have deletions within the LR2 region, is shown together with a comparison of the sequences of the human and rat LR2 regions so as to indicate the absence of the PRPRP region in the rat LR2 and the differences in the human PPPPP region with that seen. A putative SH3 binding motif of the form RXXPXXP can be identified in the human, but not the rat, LR2 region as R 315 PRPSQP.
then offer an explanation as to why the human truncate, h6.1 (Fig. 2), but not the rat truncate Met 26 -D1 (24), was able to bind to SH3 domains of SRC and LYN. In order to evaluate this, we made two h6.1 mutants. In ⌬a-h6.1 the entire proline-and arginine-rich region found within human PDE4A-LR2 was deleted ( Fig. 1). In the ⌬b-deletion mutant a stretch of sequence found uniquely within the human, but not the rat, PDE4A LR2 region was removed (Fig. 1). This region contained the "PRPRP" repeat sequence whose deletion served also to destroy the RXXPXXP motif found within human LR2. Expressed transiently in COS7 cells, both of these deletion mutants were found as active species exhibiting K m values for cAMP of 2.7 Ϯ 0.4 and 2.3 Ϯ 0.5 M for the ⌬a-h6.1 and ⌬b-h6.1 mutants, respectively (mean Ϯ S.D.; n ϭ 3). Their activities were inhibited by the PDE4-selective inhibitor rolipram in a dose-dependent fashion, with IC 50 values that were similarly both to each other and to h6.1 (Table II). However, neither of these deletion mutants was able to bind to the SH3 domains of either LYN (Fig. 2) or SRC (data not shown), and each was found exclusively in the high speed supernatant (cytosol) fraction of these cells (Table III). In addition to the putative SH3 binding motifs found in the extreme N-terminal region of pde46 and within the LR2 region, two other possible SH3 binding motifs are evident toward the C terminus of the protein (Fig. 1). However, we believe it unlikely that these are capable of interacting with the SH3 domains of LYN, FYN, and SRC because similar motifs are evident in the rodent PDE4A isoenzyme, rpde39, and the N-terminal PDE4A truncate, Met 26 -RD1, both of which failed to interact with SH3 domains (24). Consistent with this, we show here (Fig. 2) that both the human ⌬a-h6.1 and ⌬b-h6.1 mutants, each of which exhibits the C-terminal PXXPXXR motifs (Fig. 1), failed to bind to SH3 domains. We thus propose that the sites of interaction of SRC family tyrosyl kinase SH3 domains with PDE4A species are limited to the N-terminal alternatively spliced regions found in common in the human and rat homologues, pde46 and rpde6, together with an additional site found uniquely in the LR2 region of the human PDE4A isoenzyme, pde46.
In order to define whether the LR2 region of human PDE4A would indeed interact with an SRC family tyrosyl kinase SH3 domain, we generated a VSV epitope-tagged version of LR2 as a GST fusion protein. This species migrated as a single, 33-kDa species on SDS-PAGE ( Fig. 3A; track h) when detected using a polyclonal antibody specific for GST (12). The GST fusion protein of LYN SH3 was similarly detected and migrated as a single, 36-kDa species on SDS-PAGE (Fig. 3A, track i). Using a mAb specific for VSV we were able to immunoprecipitate the VSV epitope-tagged LR2-GST fusion protein (Fig. 3A, track c) but not the GST fusion protein of LYN SH3 (Fig. 3A, track e). However, after mixing these two chimeras together, we were then able to co-immunoprecipitate them both, indicated by the doublet evident on SDS-PAGE (Fig. 3A, track g), using the VSV-specific mAb but not by using a nonspecific mAb (Fig. 3A,  track f). Indeed, a nonspecific mAb was ineffective in causing immunoprecipitation of either chimera added alone or in combination (Fig. 3, tracks b, d, and f). These experiments demonstrate that the proline-and arginine-rich LR2 region that will be found within all human PDE4A isoenzymes (32) can indeed interact with the LYN SH3 domain. Indeed, as purified LYN SH3-GST and LR2-VSV-GST proteins were used in this study, this indicates that SH3 domain interaction with the human PDE4A LR2 region was direct and did not involve any intermediary protein or any post-translational modification.
We have previously maintained that h6.1 (12,22), like the rodent N-terminal truncate Met 26 -RD1 (9, 10), was an entirely cytosolic species. However, in those studies we only analyzed the high speed (P2) pellet and the supernatant fraction. Given the ability of h6.1 to interact with SH3 domains, we have re-investigated the intracellular localization of these two species in this study. This analysis (Table III) served to confirm that h6.1 was located in the cytosol and not in the P2 fraction. However, we were also able to show ( Table III) that ϳ23% of the total h6.1 immunoreactivity was associated with the P1 particulate fraction. In marked contrast to this the rat truncate, Met 26 -RD1 was exclusively located within the high speed supernatant fraction (Table III). This difference in association of h6.1 and Met 26 -RD1 with the P1 particulate fraction may explain the previously reported differences in intracellular distribution of these two species inferred from immunofluorescence analyses done using laser scanning confocal microscopy (10,12). In these studies the immunofluorescence of Met 26 -RD1 was evenly spread throughout the cell, indicative of a solely cytosolic distribution (9), whereas h6.1 showed a nonuniform distribution through the cell interior, with an increased labeling associated with the perinuclear cytoskeleton (12). This would be consistent with the association of h6.1 with a nuclear/ cytoskeletal P1 subcellular fraction in addition to the high speed supernatant/cytosol fraction. In order to determine if association of h6.1 with the P1 pellet fraction was attributable to its proline-and arginine-rich LR2 region, we analyzed (Table  III) the intracellular distribution of the LR2 region deletion mutants, ⌬a-h6.1 and ⌬b-h6.1 (Fig. 1). Such mutants were, like Met 26 -RD1, found exclusively in the high speed supernatant (cytosol) fraction (Table III). This suggests that the association of h6.1 found with the P1 particulate fraction involves an interaction between its LR2 region and an SH3 domain-containing protein. Consistent with this we were able to demonstrate (Fig. 3B) that the VSV-LR2-GST fusion protein was able to prevent in vitro synthesized h6.1 from binding to a P1 pellet isolated from untransfected COS7 cells. As a control we were able to demonstrate that GST was ineffective in this regard (Fig. 3B). This provides further support to the notion that h6.1 associates with the P1 pellet fraction through an interaction involving its LR2 domain.
Although an interaction involving the proline-and argininerich sequences in LR2 may account for the association of h6.1 with the P1 fraction, it cannot fully account for the association of full-length pde46 with the P1 fraction. For the ⌬b-pde46 Procedures," to determine the association of full-length (pde46) and N-terminally truncated (h6.1) human PDE4A forms with the SH3 domains of LYN and SRC expressed as GST fusion proteins. The association of PDE4A species was detected either by assaying PDE activity (1 M cAMP as substrate) or by immunoblotting with a PDE4A-specific antiserum raised against a C-terminal peptide. Binding is expressed as a percentage of the total activity/immunoreactivity present in the incubation mixture prior to pull down of the GST fusion proteins on glutathione-agarose. Native GST was used as a control. pde46 immunoreactivity was quantified as described in detail before by us (12)    mutant, whose LR2 domain has been modified to ablate binding to SH3 domains, was still found associated, albeit at a reduced level, with the P1 fraction (Table III). Similarly, the rodent PDE4A isoenzyme, rpde6, which lacks the proline-and arginine-rich insert seen in the human LR2 region, was also found associated with the P1 fraction (Table III). These data suggest that the N-terminal portion of these long PDE4A isoenzymes is also involved in the targeting of these isoenzymes to components found within the P1 fraction. Immunoblotting COS7 cells (Fig. 4A) with a mAb specific for the SRC family tyrosyl kinase, LYN, identified two immunoreactive components of 56 Ϯ 1 and 53 Ϯ 2 kDa. These reflect the sizes of the two known splice variants of LYN (33). Subcellular fractionation studies showed that this kinase was located within both the P1 and P2 particulate fractions of COS7 cells and was not found expressed in the cytosol (Fig. 4A). We have previously shown (12) that particulate pde46 could not be solubilized using either high salt concentrations or the non-ionic detergent Triton X-100, indicating that it is likely to be associated with the cytoskeleton. Similarly, under identical conditions of either high [NaCl] or with Triton X-100 or a combination of both, we were unable to release LYN from particulate fractions (Fig. 4B). This suggests that pde46 and LYN may both be similarly associated with the particulate fractions. However, using an RIPA/Nonidet P-40 extraction method, which other investigators (see e.g. Ref. 34) have used to extract SRC family tyrosyl kinases from cells and determine associated proteins through immunoprecipitation, then we were able to achieve the solubilization of both pde46 and LYN. This allowed us to demonstrate that solubilized pde46 and LYN could be co-immunoprecipitated using an anti-LYN antiserum (Fig. 5A), whereas a nonspecific antiserum was ineffective (Fig. 5A). The pde46 that was co-immunoprecipitated with LYN was catalyt-ically active. As noted for pde46 bound to LYN SH3 (see below), the activity of LYN-bound pde46 was similar (96 Ϯ 12%; mean Ϯ S.D.; n ϭ 3) to that of pde46 found in the cytosol of transfected COS7 cells (100%; 1 M cAMP as substrate). The amount of pde46 that could be co-immunoprecipitated with LYN was variable, being in the range of 32-54% (n ϭ 3). This, undoubtedly, reflects the fact that complexes are likely to be disrupted by varying extents during the lysis, extraction, and immunoprecipitation procedures. Nevertheless, this procedure clearly identified an interaction occurring between these two proteins. In contrast to this, however, we were unable to coimmunoprecipitate LYN and pde46 from the high speed supernatant (cytosol fraction) of COS7 cells (Fig. 5A, left-hand  panel).
In order to address further the possibility that pde46 and LYN interact in intact COS7 cells, we utilized a confocal scanning microscopy approach. Immunofluorescence analyses of COS7 cells was done (Fig. 5B) using an antiserum specific for pde46 (rhodamine; red) and mAb specific for LYN (fluorescein; green). We noted (Fig. 5B) that pde46 immunofluorescence was evident both at discrete cortical regions of the cell periphery and also in a filamentous network surrounding the nucleus that reflects the major cytoskeletal arrangement in these cells (12). The distribution of LYN immunofluorescence (Fig. 5B) was remarkably similar to that of pde46 and, indeed, overlay of the images yielded an essentially uniform "yellow" image, suggesting that both LYN and pde46 were highly co-localized in COS7 cells. In contrast to analyses with pde46, it is clear that the distribution of the short rat PDE4A isoenzyme, RD1, in these cells was entirely different from that of LYN and could be clearly distinguished (Fig. 5B). In addition the plasma membrane-located guanine nucleotide regulatory protein, G s ␣, was shown to have a punctate location at the cell plasma membrane Panel E shows Phosphor-Imager data indicating that LYN SH3 is able to bind labeled pde46 after phosphorylation by PKA and [ 32 P]ATP. The identity of this labeled band was confirmed using immunoblotting with PDE4A-specific antisera. Panel F shows that LYN SH3-GST (4 g of protein) was able to pull down all of the immunoreactive pde46 in a cytosol/high speed supernatant fraction from forskolin-treated COS7 cells (100 g of protein).
which was readily differentiated from that of pde46 (Fig. 5B). Such data provide independent evidence in support of an interaction between LYN and pde46 occurring in the particulate fraction of COS7 cells. Our confocal analyses might also imply that in intact COS7 cells the major fraction of pde46 is essentially particulate/LYN-associated. It is thus possible that cellular disruption may lead to the release of a fraction of pde46, but not LYN, which is subsequently found in the high speed supernatant (cytosol) fraction.
Within the sequence of pde46 are consensus motifs for protein kinase A to elicit phosphorylation. We have thus attempted to ascertain whether pde46 would still be able to interact with LYN SH3 under conditions of elevated cAMP where protein kinase A might be expected to be activated and pde46 could possibly become phosphorylated. We show here (Fig. 2D) that whereas pde46 can indeed be phosphorylated by PKA in vitro, this phosphorylated species is still capable of binding to LYN SH3 (Fig. 2E). We also noted that pde46 obtained from COS7 cells treated with the adenylyl cyclase activator, forskolin, and the nonspecific PDE inhibitor, isobutylmethylxanthine, in order to increase intracellular cAMP levels, did not alter the ability of pde46 to complex LYN SH3 (Fig. 2F). Under the conditions of such an experiment it is not known to what extent pde46 might have become phosphorylated by PKA; nevertheless, the entire soluble pde46 component could be bound to LYN SH3 (Fig. 2F). Such data indicate that under conditions of elevated cAMP levels pde46 can still associate with the LYN SH3 domain.
PDE4-selective inhibitors serve as potent anti-inflammatory agents of potential therapeutic use (15)(16)(17)(18). Of these, rolipram has served as the paradigm. However, an enigma associated with this particular compound is its apparent ability to discriminate between two conformational states of PDE4 isoenzymes that exhibit different affinities for inhibition by rolipram  50 values given as means Ϯ S.D. for n ϭ 3 separate experiments in each instance. Values are expressed in M rolipram and were obtained from dose-effect studies performed as described under "Experimental Procedures." Assays were done with 1 M cAMP as substrate. No differences in the K m values using cAMP as substrate were noted for the various species bound to SH3 fusion proteins compared with those values found for the corresponding enzyme expressed in the various subcellular fractions (Ͻ5% change; n ϭ 3). Similarly, the various mutants showed similar K m cAMP values compared with the parent species. The form of the Dixon plot is indicated as linear (L) or nonlinear (NL), namely parabolic with, given in parentheses (n), the degree of nonlinearity (cooperativity) as assessed by fitting Dixon plot data to the equation y ϭ mx n ϩ C, where n is equivalent to the Hill coefficient and is equal to unity for an enzyme obeying simple Michaelian kinetics of inhibition. Data are given for enzymes expressed in COS-7 unless stated otherwise. PDE activities (nmol/min/mg protein) of enzymes are given in parentheses for the source of enzyme assayed. In experiments utilizing complexes of PDE enzymes and SH3 domains then, 200 g of protein COS-7 cell lysate was used together with 2ϫ 200 g of the various GST fusion proteins as described under "Experimental Procedures."

PDE4A form
Source of enzyme assayed IC 50 (15,16). The molecular basis underlying these two proposed conformational states, at least as regards PDE4A isoenzymes, is unknown. Of particular interest then was our striking observation (12) that, expressed transiently in COS7 cells, the particulate form of pde46 was found to be profoundly more sensitive to inhibition by rolipram than the cytosolic form of pde46. Not only this, but the kinetic mechanism of rolipram inhibition of particulate-associated pde46 appeared to be altered in a fashion that implied that the enzyme in this fraction had undergone a conformational change (12). Here we evaluated the ability of rolipram to inhibit pde46 when complexed with the SH3 domains of both LYN ( Fig. 6A and Table II) and FYN (Table II) provided as GST fusion proteins. These data clearly demonstrated that SH3-bound pde46 exhibited a profound increase in its sensitivity to inhibition by rolipram. Indeed, the IC 50 values for inhibition of LYN SH3-and FYN SH3-associated pde46 were ϳ17and ϳ3-fold lower, respectively, than those observed for cytosolic pde46 expressed in COS7 cells (Table II). Such differences were of a similar magnitude (ϳ13-fold) to that reported previously by us (12) for particulate-associated pde46 (IC 50 ϭ 0.2 M) compared with cytosolic pde46 (IC 50 ϭ 2.6 M) expressed in transiently transfected COS7 cells. We determined, as a control, that the addition of purified GST to cytosolic pde46 did not affect the sensitivity of this enzyme to inhibition by rolipram (Table II). In a further control we attempted to determine whether complex formation per se would alter the sensitivity of pde46 to rolipram by analyzing cytosolic pde46 which had been immunoprecipitated with an antiserum specific to PDE4A. This immobilized form of pde46 did not show any change in the sensitivity and kinetics of inhibition by rolipram (Table II). Thus particulate pde46 did not show an altered sensitivity to rolipram inhibition merely on the basis that it was immobilized; rather, the specific interaction of an appropriate SH3 domain with its LR2 region was required. Dixon plots of the reciprocal of reaction velocity versus rolipram concentration have been shown to be linear for the cytosolic form of pde46 but parabolic for the particulate enzyme expressed in COS7 cells (12). Such a profound change is indicative of an altered kinetic status of the particulate form of pde46. This has been interpreted (12) as being due to a conformation-induced change in the particulate enzyme that altered the kinetics of inhibition from simple competitive to partial competitive in nature. Although two alternative possibilities have been suggested (3), namely that the particulate fraction of COS7 cells might contain pde46 as either a mixture of two conformational states with different affinities for rolipram or one exhibiting negatively cooperative kinetics of inhibition, in both instances SH3 domain interaction would have to achieve a  5. Interaction of pde46 and LYN in COS7 cells. Panel A, the P2 pellet fraction (500 g of protein) from both COS7 and COS1 cells that had been transiently transfected with a plasmid encoding pde46 was solubilized using RIPA buffer and then subjected to immunoprecipitation using either an anti-LYN polyclonal antiserum or a control antiserum. The immunoprecipitate was then subjected to SDS-PAGE and immunoblotting. In the upper panel immunoblots were developed using an antiserum specific for PDE4A and in the lower panel a mAb specific for LYN. Cytosol (500 g of protein) from pde46-transfected COS1 and COS7 cells (as indicated) was subjected to immunoprecipitation using polyclonal antisera specific for either LYN or SRC as well as a nonspecific antisera (control). Blots were then developed using an antibody specific for PDE4A. Panel B, COS7 cells were analyzed using a laser scanning confocal microscope. The upper set of pictures relates to cells transfected to express pde46. A series of 0.25-m optical sections were analyzed for the co-localization studies. Here we show an optical section taken through the center of the cell that has been probed with a PDE4A-specific antiserum (detected with goat anti-rabbit IgG labeled with FITC; rhodamine red), and also with a specific LYN mAb (detected with a goat anti-mouse IgG labeled with FITC; fluorescein green). The pde46 "image" (pde46, red) and LYN image (LYN, green) are given together with a combined (superimposed) image (pde46 ϩ LYN combined) plus a combined image of another cell (pde46 ϩ LYN combined (2)). The pattern of immunofluorescence of both LYN and pde46 in all optical sections (ϳ30 at 0.25 nM) taken through these cells was very similar and in each instance yielded a uniform yellow image indicative of a high degree of co-localization seen in three different transfection studies. Analyses were also done to compare the distribution of pde46 with G s ␣ and LYN with the rat short form PDE4A, RD1. In the pde46 analysis with G s ␣ a PDE4A mAb was employed, and anti-mouse IgG was labeled with ALEXA594 (red) and used for its detection, whereas G s ␣ was probed with a specific rabbit polyclonal antibody with detection using an anti-rabbit IgG labeled with FITC (green). In the RD1 analysis with LYN, a PDE4A-specific antiserum was used to probe RD1 with detection using goat anti-rabbit IgG labeled with FITC (red) and LYN was probed using a specific mAb with detection using a goat anti-mouse IgG labeled with FITC (green). Analyses were repeated using different combinations of fluorescent labeled antisera with similar results .   FIG. 6. Rolipram inhibition of LYN SH3-bound pde46. Panel A shows the dose-dependent inhibition, by rolipram, of soluble cytosolic pde46 (q) and of LYN SH3-bound pde46 (E); panel B shows the Dixon plot transform of the reciprocal of the PDE activity against [rolipram] for cytosolic pde46 (q) and LYN-SH3 bound pde46 (E); panel C shows the dose-dependent inhibition, by rolipram, of ⌬b-pde46 found in the soluble cytosolic fraction (q), the P2 fraction (f), and when LYN SH3 bound (E); panel D shows the Dixon plot transform for particulate-bound ⌬b-pde46 (f) and LYN SH3-bound ⌬b-pde46 (E). PDE4 enzymes were all expressed in COS7 cells. Where appropriate, in these experiments 200 g of protein of COS7 cell lysate was used together with 2ϫ 200 g of the various GST fusion proteins as described underr "Experimental Procedures." Assays were done at 1 M cAMP substrate using preparations with activities shown in Table II. Data are mean Ϯ S.D. for n ϭ 3 separate experiments. conformational change in pde46 which led to altered kinetics of rolipram inhibition. In a similar fashion to that seen for particulate-associated pde46, we see here that Dixon replots (Fig.  6B) for rolipram inhibition of LYN SH3-bound pde46 were both parabolic with similar Hill coefficients that were strikingly less than unity (Table II). Similarly, analyzing the N-terminal truncate h6.1 bound to the SH3 domains of both LYN and FYN, we also observed a profound increase in sensitivity to inhibition by rolipram compared with the soluble enzyme ( Fig. 7A and Table  II) as well as parabolic Dixon plots compared with the linear ones seen for the cytosolic form of h6.1 expressed in COS7 cells ( Fig. 7B and Table II). Furthermore, we show here that, as with pde46, the particulate (P1) associated form of h6.1 also exhibited an enhanced sensitivity to inhibition by rolipram and parabolic Dixon replots, compared with its soluble (cytosolic) form (Table II and Fig. 7, panels A and B).
We would like then to suggest that it is SH3 domain binding to the LR2 region that is responsible for triggering a conformational change in the catalytic unit of the particulate bound forms of pde46 and h6.1 and that is detected by a profound alteration in the kinetics of rolipram inhibition of these enzymes. As indicated by the studies described above, it would seem that the particulate, but not soluble (cytosolic), form of pde46 is associated with LYN in COS7 cells. We thus propose that it is the interaction between the LR2 region of pde46 and h6.1 with the LYN SH3 domain that underpins the altered kinetics of rolipram inhibition seen for the particulate forms of both these enzymes (Table II). In order to try and gauge this we analyzed (Fig. 6, panels C and D) the kinetics of rolipram inhibition of ⌬b-pde46. This mutant, although expected to interact with LYN SH3 through its N-terminal region, has a disrupted LR2 region (Fig. 1) which, on the basis of studies done using ⌬b-h6.1 (Fig. 2), will be unable to interact with the SH3 domain of LYN. Analyses were done comparing the soluble, cytosolic form of ⌬b-pde46 with both this form bound to LYN SH3 and the ⌬b-pde46 form expressed in the particulate fraction of transfected COS7 cells (Table II and Fig. 6 panels C and D). As distinct from native pde46, the ⌬b-pde46 mutant exhibited dose-effect curves for rolipram inhibition that were similar in form for both the cytosolic, the particulate, and the LYN SH3-bound states (Fig. 6, panels C and D). Additionally, Dixon plots of the reciprocal of PDE activity against rolipram concentration were also linear for both the LYN SH3-bound and the particulate-bound forms of ⌬b-pde46 (Fig. 6, D), as well as those for ⌬b-h6.1 (Table II). Interestingly, the IC 50 values derived for inhibition of ⌬b-pde46 by rolipram in these various states mirrored that exhibited by the cytosolic form of pde46 (Table II), indicating that this deletion "traps" the enzyme in the low affinity state for rolipram inhibition. Additional evidence that it is essential to have an LR2 region able to interact with SH3 domains to elicit a heightened sensitivity to inhibition by rolipram can also be demonstrated by analysis of rpde6, the rodent homologue of pde46. We have shown previously (11) that particulate rpde6 did not exhibit any increase in sensitivity to rolipram inhibition compared with its cytosol form, and we confirm this observation here (Table II). In addition, we show here that association of cytosolic rpde6 with LYN SH3 did not engender any change in sensitivity to inhibition by rolipram (Table II) and, furthermore, Dixon analyses of rolipram inhibition of particulate-associated rpde6 were linear and indicative of kinetics of simple competitive inhibition by rolipram (Table II and data not shown), unlike those for particulate pde46 (12). Such data are all consistent with the importance of the proline-and arginine-rich sequences found in the human PDE4A LR2 region for the SH3-and particulate associationmediated change in rolipram kinetics of the human PDE4A enzyme, pde46.
While rolipram serves as the paradigm for a PDE4-selective inhibitor that can discriminate between the two proposed conformational states of PDE4 isoenzymes, it has also been suggested that certain other selective PDE4 inhibitors are not able to be similarly discriminatory (15,16). Thus the PDE4-selective inhibitor SB207499 (Fig. 8A) has been demonstrated (35) to inhibit similarly the two conformational "states" that PDE4 enzymes(s) adopt and that can be discriminated by rolipram. On this basis, if SH3 interaction with the LR2 region of pde46 offered a means of "switching" PDE4 conformation, one might predict that such an interaction would not affect inhibition of pde46 by SB207499. Indeed this seems to be the case (Fig. 8B) as the IC 50 values for SB207499 of the cytosolic, particulate (P2), and LYN SH3-bound forms of pde46 were very similar at 0.12 Ϯ 0.02, 0.10 Ϯ 0.02, and 0.15 Ϯ 0.04 M, respectively (n ϭ 3 separate experiments; means Ϯ S.D.). RP73401 (Fig. 8A) has been shown (36) to be unable to discriminate between the two conformational states of PDE4 detected by altered rolipram  Table II. inhibition. We observed here that RP73401 inhibited the cytosolic, P2 particulate and LYN SH3-associated forms of pde46 similarly (Fig. 8C) with IC 50 values of 1.0 Ϯ 0.2, 1.5 Ϯ 0.3, and 0.9 Ϯ 0.4 nM, respectively (n ϭ 3 separate experiments). Thus the particulate and LYN SH3-bound forms of pde46 appear to reflect a distinct conformational state of the enzyme that is detectable by increased sensitivity to inhibition by rolipram. As pde46 is expressed in many cell types (6), it is possible that such a conformer may provide at least a fraction of the high affinity rolipram-inhibited fraction reported by many investigators in crude cell systems (3,16,(37)(38)(39).
It is thus likely that pde46 expressed in different cell types may exhibit markedly different sensitivities to inhibition by rolipram, dependent upon whether its LR2 region is interacting with an SH3 domain-containing protein or not. Indeed, the magnitude of any alteration in rolipram IC 50 value may reflect the nature of the interacting SH3 domain containing protein, as pde46 association with LYN SH3 and FYN SH3 yielded bound enzymes that exhibited rather different IC 50 values for inhibition by rolipram (Table II). The putative interaction between human PDE4A isoenzymes and SH3-containing proteins might not only be expected to occur in cell particulate fractions but also with cytosolic SH3 containing proteins in a cell typespecific fashion. This may provide an explanation for observa-tions indicating that pde46 and h-pde1, a species equivalent to h6.1, expressed in different systems from the COS7 cells used in this study, yielded parabolic Dixon plots for rolipram inhibition (36,40,41). Intriguingly, both of these groups showed that apparently normal kinetics of rolipram inhibition ensued when grossly N-terminal truncated species, which lacked the LR2 region, were used (36,41). Certainly, if soluble pde46, expressed in COS1 cells (41), and soluble h6.1 (h-pde1), expressed in Saccharomyces cerevisiae (36,40), were able to interact with a cytosolic, SH3 domain-containing protein then such truncation, which served to remove the LR2 domain, would be expected to convert the complex kinetics of rolipram inhibition to a simple Michaelian state. As the COS1 cell system was experimentally accessible to us, we took this as a model system to evaluate our proposal. We were able to confirm the observations of Owens et al. (41) that, unlike expression in the cytosol of COS7 cells (Table II), pde46 expressed in the cytosol of COS1 cells yielded parabolic Dixon plots for rolipram inhibition (Table II). Additionally, we also expressed h6.1 in COS1 cells and found (Fig. 9), again in dramatic contrast to expression in COS7 cells, nonlinear Dixon plots for rolipram inhibition ( Fig. 9 and Table II). In both these instances the cytosolic forms of these enzymes expressed in COS1 cells were considerably more sensitive to inhibition by rolipram than FIG. 8. Inhibition of pde46 by the PDE4-selective inhibitors RP73401 and SB207499. Panel A, structure of rolipram, RP73401, and SB207499; panel B, SB207499 dose inhibition data for the inhibition of pde46 activity from transiently transfected COS7 cells using the high speed supernatant/cytosol (s/n) fraction (q), the particulate (P2) pellet fraction (f), and the cytosol fraction bound to LYN SH3 (E); panel C, RP73401 dose inhibition data for the inhibition of pde46 activity from transiently transfected COS7 cells using the high speed supernatant/ cytosol (s/n) fraction (q), the particulate (P2) pellet fraction (f), and the cytosol fraction bound to LYN SH3 (E). Where appropriate, in these experiments 200 g of protein of COS7 cell lysate was used together with 2ϫ 200 g of the various GST fusion proteins as described under "Experimental Procedures." The dose-effect data represent the average of three separate experiments using different cell transfections. Assays were done at 1 M cAMP substrate using preparations with activities shown in Table II. when they were expressed in COS7 cells (Table II). However, we were also able to demonstrate that ⌬b-h6.1, the mutant whose crippled LR2 region was unable to interact with SH3 domains (Fig. 2), exhibited normal Michaelian kinetics of rolipram inhibition when expressed in COS1 cells ( Fig. 9 and Table  II). This mutant was also similarly sensitive to inhibition by rolipram when it was expressed in both COS1 and COS7 cells (Table II). Furthermore, when rpde6, the rat homologue of pde46, was expressed in COS1 cells, the cytosolic form of this enzyme exhibited normal Michaelian kinetics of inhibition by rolipram (Table II) as did the cytosolic rat N-terminal PDE4A truncate, Met 26 -RD1 (Fig. 9). Such observations indicate the importance of an intact human PDE4A LR2 domain in allowing cytosolic PDE4A species expressed in COS1 cells, rather than COS7 cells, to exhibit non-Michaelian kinetics of rolipram inhibition. We suggest that in COS1 cells, as distinct from COS7 cells, cytosolic pde46 may be interacting through its LR2 domain with an SH3-containing protein, and it is this interaction that is responsible for switching its conformation and thus altering its kinetics of inhibition by rolipram. To investigate this possibility, we examined whether pde46 could be co-immunoprecipitated with an SRC family tyrosyl kinase from the cytosol of COS1 cells. Although we were unable to co-immunoprecipitate pde46 from the cytosol of either COS1 or COS7 cells with LYN (Fig. 5A), we discovered that pde46 could be immunoprecipitated with SRC itself from the high speed supernatant fraction of COS1 cells but not from COS7 cells (Fig. 5A). The pde46 enzyme that was co-immunoprecipitated with SRC kinase from the high speed supernatant fraction of COS1 cells exhibited a similar (102 Ϯ 18%; mean Ϯ S.D.; n ϭ 3) activity to that of pde46 (100%; 1 M cAMP substrate) expressed in the high speed supernatant fraction of COS7 cells, where it was not associated with SRC kinase (Fig. 5A). This is consistent with the lack of change in the PDE activity of pde46 we observed upon its association with SRC SH3 (see below). The amount of pde46 that could be immunoprecipitated with SRC kinase from the high speed supernatant fraction of COS1 cells varied in a range of 20 -38% (range; n ϭ 3) of the total pde46 in the fraction. This may indicate that either not all of the pde46 is bound to SRC kinase in the cytosol of COS1 cells or that complexes have been disrupted during the processes of cell lysis, fractionation, and immunoprecipitation. Nevertheless, such data clearly indicate that such an interaction does occur in the high speed supernatant analyzed. This suggests that PDE assays done on this fraction will possibly monitor two populations of pde46, one which is bound to SRC kinase and one which is not. That population bound to SRC kinase will exhibit an increased sensitivity to inhibition by rolipram. Thus the presence of two enzyme populations of pde46, found in the high speed supernatant fraction of COS1 cells, with very different sensitivities to inhibition by rolipram will in itself serve to exacerbate the nonlinear, parabolic nature of Dixon replots (3). Thus the association of cytosolic pde46 with SRC kinase in COS1 cells, but not COS7 cells, may provide the molecular basis of the differences in kinetics of rolipram inhibition of this human PDE4A enzyme expressed in these two different cell backgrounds.
Catalytic activity toward cAMP did not appear to alter subsequent to the interaction of pde46, expressed in the cytosol of COS7 cells, with LYN SH3. Thus the free enzyme exhibited a K m value of 2.6 Ϯ 0.6 M, whereas that complexed to LYN SH3 exhibited a K m value of 2.9 Ϯ 0.5 M (mean Ϯ S.D.; n ϭ 3 separate experiments). In addition there was little change in the V max value of the LYN SH3-bound enzyme, which was 97 Ϯ 18% (n ϭ 3) that of the enzyme found in the high speed supernatant (100%). Similar data were found for rpde6, with K m values of 2.4 Ϯ 0.7 and 3.2 Ϯ 0.4 (mean Ϯ S.D.; n ϭ 3 separate experiments) for the free and LYN SH3-bound enzymes, respectively. Additionally the V max value of the LYN SH3-complexed enzyme was 96 Ϯ 8% (mean Ϯ S.D.; n ϭ 3 separate experiments) that of the free enzyme, indicating that no change in activity occurred. CONCLUSION We would like to suggest that pde46 and, indeed, presumably all human PDE4A isoenzymes will be able to undergo a conformational change in their catalytic unit that is controlled by a "switch" located in their exon 8-encoded (32) LR2 region. This can be operated by interaction with the SH3 domains of certain proteins, with a propensity for the involvement of those of SRC family tyrosyl kinases. Such an interaction may be of relevance in that tyrosyl protein kinases, such as LYN, are of crucial importance in controlling the function of a variety of cell types, including T-cells (42) and monocytes (43), where PDE4-selective inhibitors can potently inhibit cell function and exert antiinflammatory effects (15). The switch region in human PDE4A enzymes is located within LR2. It appears to be critically dependent upon an insert that is found in human but not in rat PDE4A forms (32). This contains the five residue repeat PRPRP (residues 314 -318) that is required to create a putative SH3 binding motif of the form, RXXPXXP. There is a precedent for conformational changes in proteins that are subject to interaction with SH3 domain containing proteins in that the binding of FYN SH3 to the HIV-1 Nef protein causes a change in the conformation of the proline-rich helical region in Nef (44). PDE4 enzymes are multidomain proteins, and it is possible that an SH3 domain-induced change in the structure of the LR2 region could trigger an alteration in the conformation of the catalytic unit that is detected by altered inhibition by the competitive inhibitor, rolipram. Certainly there is a precedent for this in that Conti and co-workers (45,46) have shown that protein kinase A-mediated phosphorylation within the UCR1 region of HSPDE4D3 can activate this isoenzyme. Such a modification also appears to increase the sensitivity of PDE4D3 to  Table II. inhibition by rolipram through a conformational change which is distinct from that which leads to enzyme activation (47). Such data indicate that conformational changes in PDE4 Nterminal regions are able to alter the conformation and functioning of the catalytic unit. The suggestion has been made (16,39) that PDE4 enzymes may be able to adopt distinct conformations that are reflected by different susceptibilities to inhibition by rolipram. The PKA-mediated phosphorylation of HSPDE4D3 may present one possible route for this to be achieved (47,48). Here we suggest that an additional route may be provided by human PDE4A isoenzymes when they interact through their LR2 region with appropriate SH3 domain-containing proteins. That this interaction causes such a profound change in the sensitivity of these enzymes to inhibition by rolipram might be expected to have consequences for the development of PDE4 inhibitors. For example, the same PDE4A isoenzyme may exhibit very different sensitivities to inhibition by certain PDE4-selective inhibitors in different cells, subcellular fractions, and changes in the conditions of cell activation. The identification of such a molecular switch also offers the possibility of developing agents able to interact with the human PDE4A LR2 domain which either mimic or ablate the action triggered by SH3 binding and thus may either enhance or depress the ability of PDE4A enzymes to be inhibited by compounds akin to rolipram.
There is considerable similarity between the sequences of the catalytic regions of various PDE family members (3,6,7). Colicelli and co-workers (49,50) have devised an ingenious means of identifying residues within PDE4 which are essential for selective inhibition by rolipram. These all map within a discrete region of ϳ70 residues which encompasses part of the C-terminal end of the catalytic unit and part of the C-terminal region itself (Fig. 1). This "inhibitor selectivity" region is encoded within a single exon (exon 15) that is unique to the PDE4 family (32). It is thus tempting to suggest that if this inhibitor selectivity region defines specificity for rolipram inhibition, then the binding of SH3 domains to the LR2 region of human PDE4A isoenzymes may trigger conformational changes within the inhibitor selectivity region. Intriguingly, it has been noted (51) that certain mutations in the inhibitor selectivity region, while exerting severe effects on inhibition of the enzyme by rolipram, have little effect on inhibition by RP73401. This indicates that there are binding interactions within the inhibitor selectivity region that are specific for each of these PDE4selective inhibitors. Indeed, structural differences between rolipram and RP73401 (Fig. 8A) would indicate that this might be expected. The conformational change elicited by the interaction of an SRC family tyrosyl kinase SH3 domain with the LR2 region of human PDE4A may then extend to specific regions at the C-terminal end of the catalytic unit that are probed by rolipram and not by RP73401. Thus subtle and localized changes ensue through this interaction. The precise nature of such changes will require delineation subsequent to threedimensional structural analyses of a human PDE4A which includes the LR2 domain.
We suggest that the particulate form of pde46 expressed in COS7 cells interacts with the tyrosyl kinase LYN. The basis of such a contention lies upon our observations that (i) these species can be co-immunoprecipitated, (ii) they were shown to co-localize in immunofluorescence studies, (iii) they are both found in the P1 and P2 pellet fractions of COS7 cells where they are similarly resistant to release by Triton X-100 and high [NaCl], and (iv) that the functional change in rolipram inhibition seen for the particulate form of pde46, compared with its cytosolic form, can be mimicked upon the binding of the cytosolic form to a LYN SH3. We localize the site of interaction of LYN SH3 which leads to altered rolipram inhibition to the LR2 region of pde46 using both a deletion strategy, comparison with rat homologues, and by demonstrating the interaction of a chimeric LR2 region with a chimeric LYN SH3 region. Although a diverse range of proteins exhibit SH3 domains, the number of proteins that have been identified as being able to bind them is still relatively small. This study identifies a possible contender and points to the SH3 domain interaction triggering a conformational change in the multi-domain PDE4A enzyme. Notwithstanding this, the putative physiological role of an interaction between the LR2 region of pde46 and an SH3-containing protein remains to be ascertained.