Identification of Inhibitor Specificity Determinants in a Mammalian Phosphodiesterase*

Mammalian phosphodiesterase types 3 and 4 (PDE3 and PDE4) hydrolyze cAMP and are essential for the regulation of this intracellular second messenger in many cell types. Whereas these enzymes share structural and biochemical similarities, each can be distinguished by its sensitivity to isozyme-specific inhibitors. By using a series of chimeric enzymes, we have localized the region of PDE4 that confers sensitivity to selective inhibitors. This inhibitor specificity domain lies within a short sequence at the carboxyl terminus of the catalytic domain of the protein, consistent with the competitive nature of inhibition by these compounds. Surprisingly, the identified region also includes some of the most highly conserved residues among PDE isoforms. A yeast-based expression system was used for the isolation and characterization of mutations within this area that confer resistance to the PDE4-specific inhibitor rolipram. Analysis of these mutants indicated that both conserved and unique residues are required for isoform-specific inhibitor sensitivity. In some cases, combined point mutations contribute synergistically to the reduction of sensitivity (suppression of IC50). We also report that several mutations display differential sensitivity changes with respect to distinct structural classes of inhibitors.

Mammalian phosphodiesterase types 3 and 4 (PDE3 and PDE4) hydrolyze cAMP and are essential for the regulation of this intracellular second messenger in many cell types. Whereas these enzymes share structural and biochemical similarities, each can be distinguished by its sensitivity to isozyme-specific inhibitors. By using a series of chimeric enzymes, we have localized the region of PDE4 that confers sensitivity to selective inhibitors. This inhibitor specificity domain lies within a short sequence at the carboxyl terminus of the catalytic domain of the protein, consistent with the competitive nature of inhibition by these compounds. Surprisingly, the identified region also includes some of the most highly conserved residues among PDE isoforms. A yeast-based expression system was used for the isolation and characterization of mutations within this area that confer resistance to the PDE4-specific inhibitor rolipram. Analysis of these mutants indicated that both conserved and unique residues are required for isoformspecific inhibitor sensitivity. In some cases, combined point mutations contribute synergistically to the reduction of sensitivity (suppression of IC 50 ). We also report that several mutations display differential sensitivity changes with respect to distinct structural classes of inhibitors.
cAMP is a ubiquitous intracellular second messenger that activates a family of cAMP-dependent protein kinases. The physiological interpretation of changes in cAMP concentration is cell type-specific, presumably reflecting differences in the expression of cAMP-dependent protein kinase isoforms and their substrates. Likewise, there are families of enzymes that regulate cAMP synthesis (adenylyl cyclases) and cAMP degradation (phosphodiesterases or PDEs 1 ) (1). Mammalian PDEs represent a particularly large grouping of isozymes that have been categorized into families based on sequence similarities (2)(3)(4)(5). These groupings also correlate with distinct catalytic and pharmacological characteristics. Most PDE families are represented by multiple genes, and many of these genes encode multiple proteins that arise from alternate splicing and downstream initiation (1, 6 -9). In addition, post-translational modifications of PDE proteins can modify their biochemical properties (10 -12).
PDEs within a family share extensive sequence identity, whereas PDEs from different families display lower degrees of relatedness (25-40% identity) that are confined primarily to their catalytic domains (3,4,13). The sequence conservation shared among all PDEs is likely to represent residues that are directly involved in cyclic nucleotide hydrolysis as well as those that play a role in determining structural features required for catalytic function. Indeed, deletion analysis has demonstrated that the catalytic domains of PDEs from different families contain the area of highest conservation (14 -20). In addition, mutagenesis of strictly conserved residues within this approximately 270 -290 amino acid region often leads to reduction or loss of enzyme activity (14,21,22), confirming that there is a direct association between conserved sequence and conserved function. A corollary to this model is that some of the sequence variations among PDEs should underlie the distinct biochemical properties of each enzyme.
Perhaps the best means for distinguishing among PDE isoforms has been through the unique pharmacological profiles they display. A variety of PDE-inhibiting drugs has been developed for use in the study of these enzymes as well as the treatment of several different pathologies (23)(24)(25)(26). These compounds appear to behave primarily as active site competitors and should, therefore, interact with the most highly conserved PDE region. Many of these inhibitors, however, show remarkable selectivity for one PDE isoform over another. Indeed, such specificity is believed to be critical for their therapeutic value.
Enzymes encoded by four PDE4 genes (PDE4A, PDE4B, PDE4C, and PDE4D) have been targets of particular interest in the search for therapeutically useful inhibitors (27,28). Sensitivity to rolipram has served as a defining feature for these enzymes, although recently, more potent and promising inhibitors have been described (29 -31). In general, truncated PDE4 constructs that have enzymatic activity also retain most of their isoform-specific drug sensitivity (15), consistent with the model that inhibitor specificity determinants are intimately incorporated within the catalytic domain. Using random mutagenesis and a positive selection carried out in a yeast model system (15), several residues were previously identified as being required for rolipram sensitivity of a mammalian PDE4B enzyme. The altered codons were localized within the known catalytic domain of the enzyme. Unexpectedly, however, the three mutated residues were not specific to PDE4 (rolipramsensitive) enzymes but were found also in PDE3s. The PDE3 genes (PDE3A and PDE3B) encode cAMP-hydrolyzing enzymes that are resistant to rolipram. PDE3s are sensitive, however, to inhibition by trequinsin and other drugs that do not affect PDE4 enzymes. Therefore, the residues previously identified by drug resistance selection, although clearly important and perhaps necessary in PDE4, are apparently not sufficient for rolipram binding when in the context of the related PDE3 enzyme.
Because catalytic activity and pharmacological sensitivity appear to be inextricably intertwined, we adopted a chimera approach to systematically localize the inhibitor specificity domain within the much larger catalytic region. Sequences from PDE3 and PDE4 were used to create chimeric enzymes that were analyzed for their kinetic properties and drug sensitivity. Extensive mutagenesis within the inhibitor specificity-determining region was then used to identify specific residues involved in drug sensitivity. Mutants selected for resistance to rolipram also showed resistance to other PDE4 inhibitors, but in several cases the magnitude of the effects varied greatly suggesting that the mutants themselves can act as probes that may sense structural features important for inhibitor function.
Plasmids, Cloning, and PCR-BglII sites were introduced at positions 775 and 1168 of PDE4B (GenBank TM accession number J04563) and positions 2465 and 2847 of PDE3A (GenBank TM accession number M91667, original clone generously provided by Vincent Manganiello, National Institutes of Health) using PCR mutagenesis. The primers used are shown below. Restriction sites are underlined and the positions on the PDE4B and PDE3A sequence of the 3Ј base of each primer are indicated in parentheses. ACGI⌬4 (16) was used as a template for PDE3, and PDE4⌬4 (this work) was used as a template for PDE4. The upstream ADH promoter primer CTGCACAATATTTCAAGCTATACC and downstream primers TTCCAGATCTGTGAAGACAGCATC(759) and GCTCAGATCTGCACAATGTACCAT(1153) containing BglII sites were used to amplify the amino end of PDE4, whereas downstream primers CTCCAGATCTGGGATATTCCCAGACAGA(2448) and ATTG-AGATCTGCCAACTTTATACACATT(2819) were used to amplify the amino end of PDE3. The downstream ADH terminator primer GACA-ACCTTGATTGGAGACTTGAC and upstream primers TCACAGATCT-GGAAATCCTGGCTGC(794) and GTGCAGATCTGAGCAACCCTACC-AA(1187) containing BglII sites were used to amplify the carboxyl end of PDE4, and upstream primers TCCCAGATCTGGAGTTGATGGCGC-TGT(2488) and TGGCAGATCTCAATGGTCCAGCTAAAT(2869) containing BglII sites were used to amplify the carboxyl end of PDE3. The amino fragments were digested with SalI and BglII and the carboxyl fragments with BglII and NotI. These PCR fragments were then initially ligated into pBluescript (Stratagene). Fragments were then digested out of pBluescript and ligated into an epitope tag-modified form of the yeast expression vector pADNS (32) using the restriction enzymes SalI and NotI. To recreate wild type PDE4 and PDE3 containing BglII sites, the amino and carboxyl fragment pairs of PDE4 and PDE3 were ligated together. To create chimeras CH1, CH2, and CH3, the amino fragments of PDE4 were ligated to the corresponding carboxyl fragments of PDE3. The complementary ligations were done to create chimeras CH5 and CH6. CH4 was made using an internal PDE4 fragment bordered by PDE3 sequences.
Chimeras CH7-CH10 were made by replacing portions of the BglII-NotI fragment of CH6 with corresponding sequences of PDE3. These fragments were generated by PCR. They were then ligated back to CH6 digested with BglII and NotI. For CH7-CH9, the carboxyl-terminal PDE3 fragments were generated by using the upstream primers GATCGGTACCCAGAAGAAGAGGAGGAA(3132) containing a KpnI site, GATCCTGCAGGACTAATGCCTGGAAA(3074) containing a PstI site, GATCAAGCTTCATCTCTCACATTGTGG(3028) containing a Hin-dIII site, and the downstream primer GATCGCGGCCGCTTATCCAT-TATTTTCACAGGTTTCC containing a NotI site. This produced 87-, 145-, and 191-bp fragments, respectively. The 87-bp PCR fragment was digested with KpnI and NotI and, together with a BglII-KpnI PDE4 fragment, ligated into the BglII-NotI cut vector creating CH7. Using the upstream primer GTGCAGATCTGAGCAACCCTACCAA(1187) containing a BglII site and the downstream primers GATCCTGCAGAT-GCCCAGGTCTCCCACAATG(1352) containing a PstI site, and GAT-CAAGCTTACCTGGGACTTTTCCACAGA(1306) containing a HindIII site, 219-and 172-bp PDE4 fragments were generated, respectively. The 219-bp PDE4 fragment was digested with BglII and PstI, and the 145-bp PDE3 fragment was digested with PstI and NotI. These fragments were ligated together into BglII-NotI-digested CH6 to create chimera CH8. The 172-bp PDE4 fragment was then digested with BglII and HindIII, and the 191-bp PDE3 fragment was digested with HindIII and NotI. These fragments were then ligated into BglII-NotI-digested CH6 to create chimera CH9. Finally, to create chimera CH10 the upstream primer TGGCAGATCTCAATGGTCCAGCTAAAT(2869) containing a BglII site and the downstream primer GATCGAATTCATT-GACAATACCATCTGTCC containing an EcoRI site were used to generate a 77-bp PDE3 fragment. This was digested with BglII and EcoRI. Using upstream primer GATCGAATTCTTCCAACAGGGAGACAAAGA containing an EcoRI site and a downstream ADH terminator primer, a 271-bp PDE4 fragment was generated and digested with EcoRI and NotI. This EcoRI-NotI fragment was ligated with the BglII-EcoRI PDE3 fragment into BglII-NotI-digested CH6.
In Vivo Heat Shock Assay-To check each PDE construct for activity, heat shock assays were done in the following way. Chimeras were transformed into PP5 yeast cells and plated onto synthetic complete minus leucine (SC-Leu) medium. A single colony from each chimera transformant was chosen and patched onto SC-Leu medium and grown for 3 days to establish starvation conditions. Patches were then replica plated onto two YPD plates. The first plate, used to monitor replica transfer and growth, was equilibrated at room temperature. The second plate, used to measure heat shock resistance, was preheated to 55 o C for 1 h. The preheated plate was heat-shocked at 55 o C for 10 min. Plates were then allowed to cool to room temperature and incubated at 30 o C for 2 days.
To check for drug sensitivity, an in vivo liquid heat shock assay was done. PP5 yeast cells transformed with wild type, chimera, or mutant PDE constructs were seeded into 10 ml of SC-Leu medium and grown overnight in a 30 o C shaker to a density of 3 ϫ 10 7 cells or greater. Cultured yeast cells were then spun down and the pellets resuspended in 10 ml of YPD. These cultures were incubated in a 30 o C shaker for 3 days to establish starvation conditions. 0.5-ml aliquots of each culture were taken, and the appropriate PDE inhibitors were added. For rolipram treatment, a 100 mM stock of rolipram (Biomol) solution in 100% ethanol was used to give a final concentration of 1 mM rolipram in culture. For trequinsin, a 0.13 M trequinsin (Biomol) solution in water was used to give a final concentration of 325 M in culture. These cultures were incubated overnight in a 30 o C shaker. Drug treatments were repeated the following day and cultures incubated for an additional 4 h in a 30 o C shaker. Cultures were then heat-shocked at 52 o C for 25 min in a shaking water bath to allow for uniform heat treatment and plated at the appropriate dilution. Surviving colonies were then counted after 2 days of incubation at 30 o C, and the following scale was used to assess survival level: Ϫ, consistently fewer than 100 colonies (usually below 50); ϩ, greater than 10 2 colonies; 2ϩ, greater than 10 3 colonies; 3ϩ, greater than 10 4 colonies; 4ϩ, greater than 10 5 colonies; 5ϩ, greater than 10 6 colonies.
In Vitro PDE Assays, Inhibition Studies, and Western Analysis-To measure the kinetic properties and the IC 50 of chimeric and mutant proteins, in vitro PDE assays and inhibition studies were performed. Reactions were done in a 20-l volume containing 0.1 M Tris, 10 mM MgCl 2 , 100 g of 5Ј-nucleotidase from Crotalus adamanteus (Sigma), and 0.25-5 g of total protein. Yeast protein extracts were prepared as described previously (33). For K m determination, cAMP concentrations over the range of 0.01 to 300 M were used. The final range for each mutant was chosen after an initial analysis in the 0.01 to 10 M range. The percent of [2, H]cAMP (NEN Life Science Products) varied from 0.01 to 15%. K m and V max values were calculated from a Lineweaver-Burk plot created using Microsoft Excel. For inhibition studies, the cAMP concentrations used were less than the K m so that IC 50 values approached K i values. RP73401, zardaverine, and CP-293,121 were generously provided by Dr. Ray Owens (Celltech Therapeutics, Berkshire, UK), Dr. Christian Schudt (Byk-Guiden Pharmazeutika, Konstanz, Germany), and Dr. John Cheng (Pfizer), respectively. Reactions were incubated in a 30 o C water bath for 10 min. These were loaded on ion-exchange columns (Bio-Rad) and washed with 50% ethanol. EcoLite scintillation fluid (ICN) was added to the wash/eluate, and radioactivity was measured by scintillation spectrometry. IC 50 values were determined using SigmaPlot.
For Western blots, 50 g of total protein was separated on 0.1% SDS, 8% polyacrylamide gels and transferred onto Hybond ECL nitrocellulose membrane (Amersham Pharmacia Biotech). The membranes were probed with a monoclonal antibody (12CA5) directed against the hemagglutinin epitope (1:3000 dilution). Proteins were visualized using ECL (Amersham Pharmacia Biotech).
Selection of Drug-resistant PDE Mutants-A library of randomly mutagenized DNA fragments encoding the 109-residue PDE4 inhibitor specificity domain (ISD, Fig. 1) was made using a modification of an established PCR protocol (34). An upstream primer, GTGCAGATCT-GAGCAACCCTACCAA, containing a BglII site and a downstream ADH terminator sequence primer were used to generate the library. Four PCR reactions were carried out, each of which contained a limiting amount of one of the four dNTPs. Reactions were done in a 50-l volume containing 1 g of each primer, 200 M dNTP including dITP (where the limiting dNTP is at 30 M), 2.5 mM MgCl 2 , 10 mM Tris-HCl, pH 8, 50 mM KCl, and 10 ng of CH6 DNA for template. PCR was done using standard reaction conditions. Fragments were digested with BglII and NotI and subcloned into pADPD⌬2 cut with BglII-NotI. Ligations were transformed into competent XL1-Blue cells (Stratagene). 6 g of DNA from each library was then transformed into PP5 yeast cells and subjected to in vivo liquid heat shock in the presence of rolipram. In order to enhance the number of rolipram-resistant mutants over background, we first pooled the colonies from this initial selection. A new "enriched" library was prepared by extracting plasmid DNA from the selected yeast colonies and transforming this material into bacterial cells followed by a large scale DNA preparation. This DNA was then transformed into PP5 cells, and an in vivo heat shock treatment in the presence of rolipram was done. Individual clones were then isolated from the second round of selection. To confirm that rolipram resistance was caused by a mutant PDE construct, DNA was extracted, purified, and retransformed into PP5 yeast cell and subjected to another round of heat shock treatment in the presence of rolipram. The mutated region of each construct was then subcloned into an otherwise wild type PDE4 and retested before being sequenced to determine the site of mutations.
Site-directed Mutagenesis-To determine the contribution to rolipram resistance of each mutation comprising the double mutants, individual mutations were introduced into PDE4⌬2. In addition, mutations were created on residues that diverge between PDE4 and PDE3 to independently probe the role of these unique sites to rolipram resistance. Divergent residues in PDE4 were replaced by corresponding PDE3 residues. Mutagenesis was done on PDE4⌬2 in a pSKBluescript (Stratagene) construct using a Bio-Rad Muta-Gene Phagemid In Vitro Mutagenesis kit. Primers used to introduce these mutations contained unique restriction sites to facilitate the identification of mutants as follows: M409K, TCCCTGTTGGAATTCCTCCTTGATGCGATCAG-

Chimeric PDEs Reveal an Inhibitor Specificity Domain-To
localize the PDE region that encodes the determinants of inhibitor specificity, we created a series of chimeric PDEs composed of sequences from a representative human type 3 enzyme (HSPDE3A) and a representative rat type 4 enzyme (RNPDE4B1). The chimera approach was necessary in order to examine the drug sensitivity properties of regions that are too small to retain catalytic activity by themselves. We employed the PDE3⌬5 deletion mutant (16) and the PDE4⌬4 deletion mutant (see "Experimental Procedures") as the parent constructs. In each case, these enzymes display both catalytic activity and isozyme-specific drug sensitivity. As shown in Fig.  1, a single BglII restriction endonuclease site was first introduced into one of two locations within the catalytic regions of parent PDE3 and PDE4 constructs. For PDE4, the nucleotide changes were silent with respect to the encoded protein. For PDE3, each new restriction site produced one codon change resulting in the introduction of a PDE4 residue. All of the created BglII-bounded fragments were first reassembled to demonstrate that these nucleotide changes did not affect the production of active PDE3 or PDE4 enzymes ( Fig. 2A and data not shown). The individual PDE3 and PDE4 fragments were then recombined to create a variety of chimeric proteins (Fig.  1). To test the activity of these chimeras, each construct was transformed into yeast cells devoid of endogenous PDEs and FIG. 1. Chimera constructs derived from PDE3 and PDE4 sequences. PDE4⌬4, shown in black, is an aminoand carboxyl-terminal truncation of RNPDE4B1 that includes residues 140 -499. PDE3⌬5, shown in gray, is an aminoterminal truncation of HSPDE3A that includes residues 511-1141 (16). The arrows indicate the introduced BglII restriction sites used as junctions for chimera formation (except for CH4, each construct has either one or the other BglII site). The hatched area of each construct indicates the most highly conserved region among PDEs of different families. Note that PDE3 has a slightly larger hatched region to reflect an additional 44 amino acids found in this enzyme but in no other known PDE type. This "insert" region lies to the left of both chimera junction points. The inhibitor specificity domain (ISD) is indicated. Heat shock resistance (ϩ/Ϫ) indicates the ability of each construct, when transformed into PP5 cells, to rescue the lethality resulting from heat shock. then checked for their ability to rescue heat shock sensitivity (15,33). Only chimeras CH1 and CH6 had PDE activity, although Western analysis showed that all chimeras were expressed and of approximately the expected molecular weight ( Fig. 2A). The inactivity of the other constructs may indicate that those junctions have disrupted the structural integrity of the proteins, implying a structural sensitivity for the catalytic site of these and perhaps other PDEs.
In the case of CH1, the PDE4-derived fragment alone contained the full catalytic domain. This construct was informative because it indicated that the carboxyl-terminal PDE3derived fragment was not inhibitory in a chimera context and was therefore not likely to be responsible for the inactivity of chimeras CH2, CH3, or CH4. The CH6 chimera was composed of essential catalytic sequences from both PDE3 and PDE4. Although CH6 contains primarily PDE3 sequences (439 amino acids) with a relatively small contribution from PDE4 (109 amino acids), this chimera showed sensitivity to rolipram (PDE4-specific inhibitor) and resistance to trequinsin (PDE3specific inhibitor) both in vivo (Table I) and in vitro (Table II). These data localize the PDE4 inhibitor specificity domain to within the 109 amino acids from the carboxyl terminus of the catalytic region.
To define further the inhibitor specificity domain, four additional chimeras that replaced PDE4 sequences with PDE3 se-quences were created (Fig. 1). Of these, only two (CH7 and CH8) showed catalytic activity as judged by their ability to confer heat shock resistance, although Western analysis indicated that all were expressed and of approximately the predicted molecular weight (Fig. 2B). Biochemical analysis of the catalytically active chimeric proteins (Table III) showed that the CH6 chimera has an elevated K m and reduced V max compared with the PDE4⌬4 enzyme, which is itself slightly altered from full-length wild type PDE4 (15). The CH7 and CH8 constructs, which harbor greater PDE3 content, showed a further increase in K m .
The contribution of the 109-amino acid region to isozyme inhibitor specificity was explored using a range of PDE inhibitors. The CH6 chimera showed only a minor decrease in sensitivity to rolipram (Table II) compared with the PDE4⌬4 enzyme. Sensitivity to RP73401, a highly potent PDE4 inhibitor (29), showed a more substantial decrease. This result suggested that some determinants that govern RP73401 inhibition reside upstream of the 109-amino acid PDE4 sequence found in CH6. Inhibition by zardaverine, a PDE3/PDE4 dual-specificity drug (35), was unaffected compared with the PDE4⌬4 enzyme. Similarly, there was no significant change in response to the broad spectrum PDE inhibitor IBMX. The reduced PDE4 content chimeras (CH7 and CH8) generally showed noticeable decreases in sensitivity to all the PDE4 inhibitors and to IBMX. None of the chimeras, however, showed any degree of sensitivity to the PDE3-specific inhibitors trequinsin or cGMP despite the fact that these enzymes have catalytic domains composed primarily of PDE3 sequences. These results support the designation of the 109-amino acid PDE4 region found in chimera CH6 as the principal inhibitor specificity domain. They also suggest that some PDE3 inhibitor specificity determinants may reside in the corresponding region of that enzyme (replaced by PDE4 sequences in CH6). The smaller 70-amino acid region of PDE4 present in CH8 did still confer a much greater specificity for PDE4 inhibitors (rolipram, RP73401) than PDE3  Fig. 1) were analyzed by Western blot (50 g per lane) using antibody to the hemagglutinin epitope fused to the amino terminus of each PDE. The 4b1, 4b2, 4b3, 4b12, 3b1, and 3b2 constructs are PDE4-only and PDE3-only enzymes recreated from PCR product fragments that result in the introduction of BglII restriction sites. These constructs were used to confirm that these same fragments, used for chimera constructions, were unaltered. B, similar Western analysis for PDE4⌬4, PDE3⌬5, and the refined chimeras derived from CH6. A background hemagglutinin antibody-reactive band of approximately 50 kDa is also present.   inhibitors (trequinsin, cGMP) relative to the parental constructs, although the degree of selectivity was significantly diminished compared with CH6. Genetic Selection of Drug-resistant Mutants Identifies Determinants of Inhibitor Specificity-To identify the residues responsible for PDE4 drug sensitivity, we subjected the DNA encoding the 109 codon inhibitor specificity domain to random mutagenesis. The mutagenized fragments were then cloned into a PDE4⌬2 (15) construct. This enzyme, although still truncated relative to wild type PDE4, has additional aminoterminal sequences not found in PDE4⌬4. The library of mutant PDE4 constructs was then transfected into phosphodiesterase-deficient yeast cells and subjected to rolipram treatment followed by heat shock. This procedure leads to the selection of drug-resistant, but catalytically active, mutants. Isolated mutants were retested to confirm the rolipram resistance phenotype and then sequenced. Interestingly, all mutations were confined to the 70-amino acid region within the inhibitor specificity domain that is present in CH8 (Table IV and Fig. 3). This result is consistent with the suggestion from the biochemical analysis of the chimeras that the carboxyl-terminal residues of the 109-amino acid domain may be only indirectly involved in isozyme specificity. Two separate mutants (A2 and G1) both showed mutations of the same codon (isoleucine 408), indicating a relatively high level of mutagenesis.
As indicated in Table IV, only two out of six mutants contained single missense mutations. The other mutants each had two altered codons. This result raised the possibility that some mutations may have been silent and were fortuitously isolated due to the high level of mutagenesis. To test this possibility, each mutation was reintroduced individually into a PDE4⌬2 construct. In the absence of drug, all single and double mutants conferred heat shock resistance indicating that catalytic activity was still intact. Heat shock of rolipram-treated cells was used to assess the contribution of each mutant to drug resistance. Indeed, in the case of mutant G1 resistance to rolipram clearly segregates with the I408T mutation indicating that the K439R mutation was fortuitous and uninvolved. Note that a separate mutation of Ile-408 was also isolated as a single mutation (A2) that similarly showed rolipram resistance. The mutations that make up mutants A1, C1, and G2 each showed drug resistance effects on their own, however, suggesting a more complex relationship. For C1, each mutation was itself able to confer rolipram resistance with the combined double mutant showing a somewhat stronger resistance. The individual mutations found in the G2 mutant were also shown to each contribute on their own to rolipram resistance. In this case, the difference in resistance between the double mutant and each of the single mutants was even larger than for the C1 mutant. Interestingly, only one of the mutations (M409K) that comprise the A1 mutant was, on its own, able to produce some rolipram resistance. The other mutation (E425G) gave no significant increase in survival compared with the wild type (rolipramsensitive) PDE4 construct. The M409K alone, however, is not able to account for the high level of rolipram resistance seen in the double mutant. This suggested that the E425G mutation, although unable to give drug resistance on its own, is able to facilitate or enhance the drug resistance of the M409K mutation.
The enzymatic properties of the mutants were also examined. Mutations of Ile-408 had the most dramatic effect, leading to a large increase in K m values (Table V). This was true for both the A2 mutation (I408N) and the G1 double mutation as well as the I408T mutation that makes up part of G1. The similarity of both K m and V max values for the G1 and the I408T mutants supported the conclusion, derived from the heat shock survival analysis, that the K439R mutation found in G1 was indeed fortuitous and did not contribute to changes in enzyme behavior. The C1 mutant also showed a significant increase in K m . When isolated, the Y401C and F412S mutations that make up the C1 compound mutant each had elevated K m values, but their effects appear to be greater than additive. The A1 double mutant enzyme had a K m value that was essentially unchanged from PDE4⌬2. However, the individual mutation that appeared to be most responsible for A1 drug resistance (M409K) had a noticeably elevated K m value. These data reinforce the observed synergistic effect seen for the A1 mutations with regard to drug resistance. This suggests that, with respect to biochemical properties, the A1 mutations work together to produce an enzyme with near-native K m for cAMP but substantial rolipram resistance. Finally, the T1 mutant is also relatively unaffected in K m but has a notably lower V max (60-fold below PDE4⌬2) which may account for its somewhat reduced heat shock survival, even in the absence of drug (Table IV).
As noted above, the mutations that contribute significantly to rolipram resistance are all found within the PDE4-derived 70 amino acids that are common to all of the active chimeras. In fact, all of these mutations cluster within a 49-residue stretch which shows significant conservation with PDE3 (Fig. 3). Half of the drug resistance mutations were found in residues that are shared by both PDE4 and PDE3. This includes Ile-408, the site where two independent mutations each led to dramatic  Phosphodiesterase-deficient yeast (PP5) were transformed with vector only (pADNS), a parental PDE4 construct (PDE4⌬2), the isolated rolipram-resistant mutants (A1, A2, C1, G1, G2, or T1) or individual mutations re-introduced into a PDE4⌬2 background. Cells were grown in the absence or presence of rolipram, as indicated, and tested for heat shock (HS) survival as described under "Experimental Procedures." The level of survival was judged by colony formation and is indicated in each column (Ϫ, Ͻ50; ϩ, Ͼ10 2 ; 2ϩ, Ͼ10 3 ; 3ϩ, Ͼ10 4 ; 4ϩ, Ͼ10 5 ; 5ϩ, Ͼ10 6 ).

Enzymes
ϪRolipram/HS ϩRolipram/HS Trp-4043Arg 2ϩ 2ϩ effects on both enzyme function (Table V) and rolipram sensitivity (Table IV and Table VI). It should be noted that previously reported rolipram resistance mutations, two of which also reside in this region, are in residues shared with PDE3 (15). In addition, the newly characterized mutations include four residues that are not conserved between PDE4 and PDE3. This result is consistent with a model for isozyme-specific drug interactions that exploit both conserved and unique residues. PDE Mutants Show Differential Resistance to Inhibitors-We next performed assays to quantify the degree of rolipram resistance conferred by the isolated mutations, as well as key individual mutations that were part of the double mutants. Mutations of residue Ile-408 (mutants A2 and G1), which is conserved between PDE4 and PDE3, led to remarkably high IC 50 values for rolipram (Table VI). Recall that these mutations also confer the largest increase in K m values (Table V) indicating that the enzymes, while still active, have been altered in their response to both substrate and inhibitor. By contrast, the double mutants that show K m values unchanged or only slightly elevated from PDE4⌬2 (mutants A1 and G2) do have significantly increased IC 50 values for rolipram (21-fold increase for mutant A1 and 5-fold for G2). These mutations appear to preferentially affect the inhibitor response while leaving substrate interactions relatively unaffected. The T1 mutant (W404R), which is also unaltered in substrate K m , showed the smallest increase in rolipram IC 50 (less than 2-fold). The ability to identify mutants with such low level drug resistance reflects the degree of sensitivity of the in vivo heat shock selection technique.
Several other compounds known to inhibit PDE4 isozymes were also tested. In general, mutations that reduced sensitivity to rolipram also reduced sensitivity to the highly potent PDE4 inhibitors RP73401 and CP-293,121 (36). However, mutants A1 and G2 showed considerably less resistance to RP73401 (3-fold and no change, respectively) than to rolipram (21-and 5-fold, respectively) or CP-293,121 (16-and 4.8-fold, respectively). When the dual-specificity inhibitor zardaverine was used, these mutants behaved distinctly: mutant A1 showed strong zardaverine resistance (43-fold change) whereas mutant G2 was unaffected in its response to the same drug. In addition, the Y401C and F412S mutations that make up the C1 mutant also showed significant disparities with regard to their effects on each drug. The Y401C mutant was far more refractory to CP-293,121 (50-fold change) than to the other potent inhibitor, RP73401 (2.8-fold change). It had an intermediate effect on rolipram and zardaverine (8-and 16-fold, respectively). In contrast, the F412S mutant was least affected for CP-293,121 (4.7-fold). These data support the model that, at least in the case of PDE3 and PDE4, chemically distinct inhibitors work through both common and unique determinants on the enzyme.
In this context, the mutants themselves seem able to distinguish among various drugs and, by extension, their active chemical groups.
Interestingly, the mutants were generally less affected in their response to IBMX, the broad spectrum PDE inhibitor. The exceptions to this were the mutations of Ile-408 (mutant A2 and to a lesser extent mutant G1), as well as the F412S mutant, which resulted in strong blocks to the IBMX response. Recall that the Ile-408 mutations also had the greatest effects on K m as well as resistance to the isozyme-specific inhibitors.
Analysis of Directed Residue Swaps-We pursued an independent, site-directed mutagenesis approach to the identification of inhibitor specificity determinants. Mutations were confined to the 70-amino acid PDE4 region where the selected mutations reside. In addition, only residues that diverge between PDE4 and PDE3 were chosen for alteration. Specifically, codons were swapped from the PDE3 sequence into the equivalent position in PDE4⌬2, essentially creating microchimeras. This approach was adopted in order to minimize the likelihood of creating severely disrupted enzymes since all mutations are substitutions from a fully active and related PDE. The codon swap strategy also incorporated the use of multiple changes (within a cluster of non-conserved residues) which, it was reasoned, might reveal cooperative or even synergistic effects that would otherwise be overlooked.
A series of swap mutations was created and tested first for PDE activity and then screened for rolipram resistance using the heat shock survival assay (Fig. 4). All mutants showed PDE activity and, in the absence of drug, conferred survival at levels near those seen for wild type PDE4 (data not shown). Several of TABLE V K m and V max values for mutant PDEs All mutations are in the PDE4⌬2 background. Results were calculated from analysis of samples prepared in a yeast expression system (see "Experimental Procedures"). Enzyme velocities were normalized for total protein in the yeast extract. Four concentrations of cAMP were used in each determination, and all data points were done in triplicate.  Vertical lines indicate amino acid identities. The ⌬ symbols denote residues altered in the selected rolipram-resistant mutants (excluding Lys-439 which neither confers nor enhances drug resistance, see text). The * symbols indicate the positions of two previously described drug resistance mutations (15). The short gray underlines show the PDE4 residues that when changed to those found in PDE3 at the same positions result in rolipram resistance (see Fig. 4). The heavy black lines indicate the sequences that were switched in CH7 and CH8. the mutants showed weak but reproducible resistance to rolipram. It should be noted that this level of drug resistance would probably not have been sufficient to allow isolation through the heat shock selection protocol where resistant mutants initially represent only a minor proportion of the library being tested. In two cases (SM12 and SM13), single amino acid changes were found to give resistance equivalent to that conferred by larger group swaps (SM3 and SM10, respectively). Three mutations (four codon changes) that each gave rise to low level drug resistance were then combined and reanalyzed. The combination mutant (SM9/12/13) showed high level rolipram resistance (approximately 10 4 colonies following heat shock). This result provides another example of synergistic effects in drug resistance and clearly demonstrates that residues unique to PDE4 are indeed key determinants in inhibitor specificity. The residues identified by this analysis include Met-409 which was mutated to a lysine in one of the mutants selected from the random library (mutant A1) and shown to confer drug resistance on its own as well as to have a synergistic effect.
To test the possibility that amino acid swaps from PDE3 might transfer a degree of sensitivity to a PDE3-specific inhibitor, we screened the mutants for heat shock survival after trequinsin treatment. No trequinsin sensitivity was observed for any mutant enzyme including the highly rolipram-resistant SM9/12/13 (Fig. 4). This result indicated that the PDE4 residues identified as among those essential for rolipram sensitivity are not, when replaced with their PDE3 counterparts, sufficient for a response to trequinsin. DISCUSSION The similarity in biochemical properties among enzymes from distinct phosphodiesterase families is believed to reflect the sequence identity shared among them. Indeed, the highest level of sequence conservation is localized within their catalytic domains. In addition, site-directed mutagenesis of highly conserved PDE residues has been an important tool in the identification of amino acids required for catalysis (14,21,37). Conversely, the same structure/function paradigm would predict that the distinct pharmacological sensitivity profiles of the PDE isozymes derive from the contribution of residues unique to each enzyme. We used chimeric PDEs made between enzymes with shared substrate biochemistry but different pharmacological properties to localize the sequence determinants of inhibitor specificity in PDE4B to a relatively short (109 amino acids) region at the carboxyl terminus of the catalytic domain (Fig. 5). Interestingly, several lines of evidence have placed critical substrate-interacting residues within this same 109residue sequence (14,21,38,39).
The chimeric enzyme showed clear PDE4-type behavior such as inhibition by rolipram, although there was a slight decrease in sensitivity to this drug. The more substantial reduction in RP73401 sensitivity (approximately 30-fold) suggested that for this highly potent inhibitor some determinants lie upstream of the proposed 109-amino acid inhibitor specificity region. Strikingly, the 439 PDE3-derived residues, which make up most of the catalytic domain along with some upstream sequences, neither blocked sensitivity to PDE4 inhibitors nor conferred sensitivity to PDE3 inhibitors (trequinsin or cGMP). Whatever PDE3 specificity determinants may reside within this region are therefore not sufficient to confer sensitivity to these compounds, and our results suggest that at least some such determinants are located in the PDE3 region corresponding to those sequences replaced with the PDE4 inhibitor specificity domain. The fact that the same chimera was completely sensitive to zardaverine, the dual specificity PDE3/PDE4 inhibitor, is consistent with this enzyme being a true chimera that retains properties common to both parent enzymes. Further work, expanding this analysis to include chimeras made from other PDE isoforms, will be required to ascertain whether the inhibitor specificity domains are in general localized to the same relative position within each catalytic region.
Attempts to localize further the PDE4 inhibitor specificity region using constructs with decreased PDE4 content resulted in enzymes (CH7 and CH8) that, while still resistant to PDE3 inhibitors, lost a significant degree of sensitivity to rolipram, zardaverine, and the broad spectrum PDE inhibitor, IBMX. These chimeras do not appear to represent seamless fusions of related enzymes but are likely to be somewhat disrupted in some aspect of their conformation. Indeed, two other chimeras with junctions in this domain, as well as chimeras with junctions further upstream, were found to be completely nonfunctional. These data suggest that PDEs in general may show a high level of structural sensitivity.
The mutations isolated using an in vivo drug resistance selection of enzymes carrying randomly generated mutations in the inhibitor specificity domain all reside within the first 70 residues. This observation is consistent with the results for CH7 and CH8, described above, which also suggested that the carboxyl-terminal third of this domain may not include critical specificity determinants for the drugs examined here. It should be noted, however, that this short carboxyl-terminal stretch of  (39). Whereas our in vivo mutant selection protocol was most likely not exhaustive (only one of the identified residues was hit twice), the localization of the eight mutations within the first 70 amino acids did reinforce the proposed importance of this shorter region in sensitivity to pharmacological compounds as well as in enzymatic function. It should be noted that two out of three previous mutations that confer rolipram resistance to PDE4 also reside in this 70 amino acid sequence (see Ref. 15; Fig. 3). Interestingly, of the 10 drug resistance mutations described here and in the previous study, 8 are found within a single short exon of rat PDE4B (40). In addition, 6 of the 10 mutations are found clustered within a 30-residue sequence implicated in substrate specificity (38). This region encompasses a hydrophilic stretch that has a high probability of exposure to the solvent milieu, consistent with the possibility of direct involvement in both substrate and inhibitor binding. The true cartography of these interactions, however, will only be ascertained through direct structural analysis.
Four of the selected drug-resistant mutants described here were shown to carry two mutations. In two cases (C1 and G2), each mutation made a contribution to the drug resistance as judged by in vivo heat shock assay. In another case (A1), one of the mutations had no effect on its own but synergistically enhanced resistance by the other mutation. The IC 50 analysis for these isolates also showed that individual mutations could not account for the properties of the double mutants. For instance, the C1 compound mutant showed drug resistance that was generally far greater than the additive effects of its isolated single mutants (except for IBMX where the double mutant had an intermediate level of resistance). These data demonstrate that double mutations can be valuable tools in the identification of residues involved in drug resistance, and perhaps other properties. They also suggest that aggressive mutagenesis may be particularly useful in uncovering some mutant phenotypes when a strong selective pressure can be employed. The same phenomenon was observed in the codon swap experiments. Whereas some individual swap mutations gave modest resistance, the combined mutations gave far greater than an additive result.
Inhibitor specificity determinants identified by the random mutagenesis studies included residues that are conserved between PDE4 and PDE3 as well as residues that are found in PDE4 but not PDE3. The involvement of conserved amino acids in sensitivity to isozyme-specific drugs may in part reflect a role in the conformation of the active site that affects the binding of both substrate and competitive inhibitor compounds. Indeed, two independent mutations at position 408 (leucine, isoleucine, or valine in all known PDEs) gave the greatest drug resistance and also showed the most dramatic increase in K m , suggesting an important role for this conserved hydrophobic residue in the structural integrity of the active site. This interpretation does not rule out the possibility that binding of some drugs may also include direct contact with residues that are highly conserved among PDE isoforms. For specificity, however, there must also be involvement of determinants that appear in combination only in a given PDE isoform or that are completely isozyme-specific (Fig. 5). PDE4 positions 409 (Met) and 430 (Cys) are each different in PDE3 and most other PDE isozymes, although they are found in PDE1 which is not FIG.4. PDE3-PDE4 amino acid swap mutations. The sub-region of the PDE4 inhibitor-specificity domain encompassing all the identified drug-resistance point mutations is shown at top. Below are the positions and identities of the PDE3 residues that were swapped into PDE4⌬2 (mutant names are given at left). The results from liquid culture heat shock survival (drug-resistance) assays following treatment with rolipram (ROLI) or trequinsin (TREQ) are given at right. The levels of survival for rolipram-treated cells following heat shock fell into three categories: as follows: Ϫ, consistently fewer than 100 colonies (similar to the PDE4⌬2 control); ϩ, approximately 200 colonies; 3ϩ, equal to or greater than 10 4 colonies. For trequinsin-treated cells, survival was rated at 5ϩ indicating equal to or greater than 10 6 colonies.

FIG. 5. Model for inhibitor-specificity determinant localization.
A chimeric PDE composed of primarily PDE3 sequences (speckled) with a region of PDE4-derived sequences (white) is shown. Substrate (cAMP) and inhibitors (e.g. rolipram and RP73401) are shown at right and at the active site with lines indicating potential enzyme interaction points. AϪ, residues common to PDE3 and PDE4 that are important for substrate and inhibitor binding; BϪ, residues common to PDE3 and PDE4 that are important for binding of each inhibitor (but not substrate); C and DϪ, residues specific to PDE4 and localized within the inhibitor specificity domain that are required for PDE4-specific inhibitor action and which may interact with chemical features that distinguish one inhibitor from another. rolipram-sensitive and which diverges markedly from PDE4 at other positions within the inhibitor specificity domain. Our drug resistance analysis has also identified several residues only found in PDE4. PDE4 positions 401 (Tyr) and 425 (Glu) are distinct from all known PDEs. Interestingly, in the selected mutants Y401C and E425G, these residues were altered to amino acids found in another PDE type (PDE8 and PDE7, respectively). This suggests that at these positions, perhaps only a limited number of amino acids are compatible with enzyme activity.
In the case of Glu-425, this is also one of the few residues in the inhibitor specificity domain that shows degeneracy within the four-member gene family of PDE4 enzymes. PDE4C, which has been reported to show reduced sensitivity to rolipram compared with the -4A, -4B, and -4D enzymes (41,42), has an aspartic acid residue at this position. It should also be noted that the Drosophila melanogaster Dunce protein, which shows a remarkable degree of structural and functional conservation with the mammalian PDE4s but is resistant to rolipram (43), differs from PDE4B at this and two other sites (Ile-408 and Val-437) identified by our random mutagenesis/selection and residue swapping.
There have been several reports demonstrating that PDE4 enzymes exhibit both low affinity (M) and high affinity (nM) binding to rolipram (18,44,45). It has been proposed that these differences represent a single binding site that can exist in two distinct conformations, the regulation and physiological significance of which is under investigation but not yet understood (46,47). In most PDE4 preparations from either mammalian or yeast cells, only a portion of the enzyme exhibits high affinity binding properties (18,45), and the truncated forms of the enzymes used in this study show little or no such behavior (data not shown), consistent with other studies (18,22,48). It is likely, however, that the mutations characterized here will have a similar impact on high affinity rolipram binding since both high and low affinity interactions are believed to work through the same binding site. It should also be noted that the action of some newer and more potent drugs such as RP73401 and CP-293,121, which bind and inhibit PDE4s in the nanomolar range, are not significantly affected by the conformational shift that controls rolipram affinity (18,47, and this work).
We also observed that several of the selected rolipram-resistant mutants showed differential effects toward distinct classes of inhibitors. This behavior is consistent with an inhibitorenzyme interaction model that invokes both shared and unique contact points for each of the multiple compounds that show competitive inhibition of a given PDE. Such drug-specific interaction profiles may indeed be reflected in the distinct and wide ranging IC 50 values seen for isozyme-specific inhibitors (e.g. rolipram versus RP73401 or CP-293,121 versus RP73401). Mutations in a residue that mediates interaction with a particular chemical group would be expected to have effects restricted to inhibitors containing that moiety (Fig. 5). This model, and the data presented here, also suggest that PDE mutants themselves might serve as useful probes in the characterization of new compounds with potential therapeutic value.