Amino Acid Residues Conferring Ligand Binding Properties of Prostaglandin I and Prostaglandin D Receptors

Using chimeras of the mouse prostaglandin (PG) I receptor (mIP) and the mouse PGD receptor (mDP), we previously revealed that the cyclopentane ring recognition by these receptors is specified by a region from the first to third transmembrane domain of each receptor; recognition by this region of mIP is broad, accommodating the D, E, and I types of cyclopentane rings, whereas that of mDP binds the D type of PGs alone (Kobayashi, T., Kiriyama, M., Hirata, T., Hirata, M., Ushikubi, F., and Narumiya, S. (1997) J. Biol. Chem. 272, 15154–15160). In the present study, we performed a more detailed chimera analysis, and narrowed the domain for the ring recognition to a region from the first transmembrane domain to the first extracellular loop. One chimera with the replacement of the second transmembrane domain and the first extracellular loop of mDP with that of mIP bound only iloprost. The amino acid substitutions in this chimera suggest that Ser50 in the first transmembrane domain of mIP confers the broad ligand recognition of mIP and that Lys75 and Leu83 in the second transmembrane domain of mDP confer the high affinity to PGD2 and the strict specificity of ligand binding of mDP, respectively.

Prostanoids are a group of cyclooxygenase metabolites of C-20 unsaturated fatty acids and consist of the prostaglandins (PGs) 1 and the thromboxanes (1). PGs share a prostanoic acid as a common structure and have two structural features to discriminate each other. First, they have functional groups attached at the 9-and 11-positions of a cyclopentane ring, which classifies them into four types: D, E, F, and I. Second, they are classified into three series, 1, 2, and 3, by the number of double bonds in the side chains (Fig. 1). Thromboxane A 2 has an oxane ring instead of the cyclopentane ring. These prostan-oids act on cell surface receptors specific to each type. We and others have cloned eight types and subtypes of the prostanoid receptors (2)(3)(4)(5)(6)(7)(8)(9)(10)(11). They are the PGD receptor (DP), the EP 1 , EP 2 , EP 3 , and EP 4 subtypes of the PGE receptor, the PGF receptor (FP), the PGI receptor (IP), and the thromboxane A 2 receptor (TP). These studies revealed that the prostanoid receptors belong to the G protein-coupled rhodopsin type receptor superfamily. Expression of each cloned receptor confirmed that the binding affinities of these receptors to prostanoid molecules are primarily determined by the cyclopentane ring structures of ligands. For example, DP shows the highest affinities to the D type of PGs (PGD 1 , PGD 2 ). An exception is IP, which can accommodate the cyclopentane rings of the I and E types of PGs. This receptor binds PGE 1 with as high affinity as PGI 2 . However, the binding affinity to another E type of PGs, PGE 2 , is much lower. PGI has a unique configuration of the ␣-side chain due to the presence of an additional ring attached to the cyclopentane ring. It is believed that PGE 1 without a double bond in the ␣-side chain can mimic this configuration of PGI molecule and bind to IP, but this is not achieved by PGE 2 (12). These results suggest that the ligand binding specificity of mIP is determined by two ways; one is the less strict specificity for the cyclopentane ring structures, and the other is the recognition of the configuration of the side chains.
The prostanoid receptors can be functionally grouped into three categories; the relaxant receptors, the contractile receptors, and the inhibitory receptor (13). The relaxant receptors, consisting of IP, DP, EP 2 , and EP 4 , mediate increases in cAMP and induce smooth muscle relaxation. Sequence homology among these functionally related receptors is higher than that between the three separate groups (11,14). According to the phylogenetic tree constructed upon these homologies, the EP 2 and EP 4 subtypes evolved from a primitive EP, and that IP, DP, and EP 2 evolved together after the evolution of EP 4 subtype. The amino acid sequences of mDP, mEP 2 , and mIP show 58% identity each other in the transmembrane domains (Fig. 2), whereas those of mIP and mEP 4 show 44% identity. Taking advantage of the high homology between IP and DP, we previously constructed various chimeric receptors from mIP and mDP, and examined the regions conferring the ligand binding properties of these receptors (15). When the region extending from the sixth transmembrane domain to the carboxyl terminus of mIP was replaced with the corresponding region of mDP, this IP N-V /DP VI-C receptor acquired the ability to bind PGD 2 and PGE 2 without decreasing the affinities of mIP to iloprost and PGE 1 . These binding characteristics did not change when the fourth and fifth transmembrane domains of mIP was further replaced with the corresponding regions of mDP. However, when the COOH-terminal end of the second transmembrane domain (Phe 102 ) to the second intracellular loop of mIP was further replaced with those of mDP, this IP N-II(Val101) /DP II(Leu83)-C receptor (formerly designated as IP N-II /DP III-C (Ref. 15)), markedly decreased the affinities to PGE 1 , PGE 2 , and iloprost, and bound the D type of PGs alone, as did IP N-I /DP II-C and mDP. These results suggest that the domains recognizing the ring structure and the side chain configuration are located in different regions. The sixth to seventh transmembrane domain of mIP confers the specificity of mIP to discriminate a structural difference in the ␣-side chain between PGE 1 and PGE 2 , and hence determines the selectivity between PGE 1 and PGE 2 . The binding domain for the cyclopentane ring localizes in a region containing the first to third transmembrane domain. This region of mIP shows the broad binding properties for the cyclopentane ring and accommodates the D, E, and I types, whereas that of mDP is strict and recognizes the D type alone, and that the COOH-terminal end of the second transmembrane domain (Phe 102 ) to the second intracellular loop of mDP confers the strict specificity of ligand binding of mDP.
In this study, to identify the amino acid residues conferring these ligand binding properties of mIP and mDP, we have first constructed a series of more detailed chimeric receptors from mIP and mDP in the region from the first to third transmem-brane domain, and then introduced point mutation(s) in these regions. Mutant receptors were expressed in COS-7 cells, and their ligand binding properties were examined. We report here the identification of an amino acid residue in the first transmembrane domain of mIP conferring the broad cyclopentane ring recognition of mIP, and two amino acid residues in the second transmembrane domain of mDP conferring the high affinity to PGD 2 and the strict specificity of ligand binding of mDP.

EXPERIMENTAL PROCEDURES
Materials-PGD 2 , PGE 1 , PGE 2 , and PGF 2␣ were generous gifts from Ono Pharmaceuticals Co. Ltd. (Osaka, Japan). (5,6,8,9,12,14, (16). The BalI-EcoRV fragment of CP302, a cDNA of mIP (8), and the Asp718-BamHI fragment of PGc9, a cDNA of mDP (9), were subcloned into the EcoRV sites and the Asp718 and BamHI sites of pCMX, respectively. Four chimeric receptors and 10 point mutants (Fig. 3A) were then constructed by a PCR-based strategy using primer pairs and templates described in Table I. Five ng of each template was used for amplification by PCR in a reaction mixture containing 10 mM Tris-HCl (pH 8.0), 50 mM KCl, 2 mM MgCl 2 , 1 mM 2-mercaptoethanol, 10% dimethyl sulfoxide, 0.2 mM dNTPs, 2.5 units of Ex Taq polymerase (Takara Co. Ltd., Kyoto, Japan), and 20 pmol of each primer in a total volume of 20 l. After a denaturation step at 94°C for 3 min, 20 cycles of an amplification step (94°C for 1 min, 50°C for 1 min, 72°C for 1 min) were carried out and followed by a final elongation step of 3 min at 72°C. PCR products were electrophoresed, excised, purified using a DNA fragment purification kit (Toyobo Co. Ltd., Osaka, Japan), inserted into pCMX, and sequenced by the dideoxy chain termination method. These mutant receptors have no insertion or deletion in the amino acid sequences.
The Chimeric IP N-Ex1 /DP III-C Receptor-Fragments N-1 and C-1 were amplified by PCR with primer pairs and the template shown in Table I (Fig. 3B). Fragment N-1 was digested with Asp718 and Psp1406I, and fragment C-1 was digested with Psp1406I and BamHI. Both digested fragments were ligated into the Asp718 and BamHI sites of pCMX-mIP.
The Chimeric IP N-I /DP II-C Receptor-Fragments N-2 and C-2 were amplified by PCR with primer pairs and the template shown in Table I ( Fig. 3B). Fragment N-2 was digested with Asp718 and SpeI, and fragment C-2 was digested with SpeI and BamHI. Both digested fragments were ligated into the Asp718 and BamHI sites of pCMX-mIP.
The Chimeric IP N /DP I-C Receptor-Fragments N-3 and C-3 were amplified by PCR with primer pairs and the template shown in Table I ( Fig. 3B). Fragment N-3 was digested with Asp718 and SnaBI, and fragment C-3 was digested with SnaBI and BamHI. Both digested fragments were ligated into the Asp718 and BamHI sites of pCMX-mIP.
The Chimeric DP N-I /IP II-Ex1 /DP III-C Receptor-Fragments N-4 and C-4 were amplified by PCR with primer pairs and the template shown in Table I (Fig. 3B). Fragment N-4 was digested with Asp718 and Eco47III, and fragment C-4 was digested with Eco47III and BamHI. Both digested fragments were ligated into the Asp718 and BamHI sites of pCMX-mIP.
Four mutants were constructed by site-directed mutagenesis of this chimeric DP N-I /IP II-Ex1 /DP III-C receptor.  The Chimeric CR R107Q Receptor-Fragments N-5 and C-5 were amplified by PCR with primer pairs and the template shown in Table I ( Fig. 3B). Fragment N-5 was digested with Asp718 and BsiWI, and fragment C-5 was digested with BsiWI and BspEI. Both digested fragments were ligated into the Asp718 and BspEI sites of pCMX-mDP.
The Chimeric CR T94K Receptor-Fragment N-6 was amplified by PCR with primer pairs and the template shown in Table I (Fig. 3B). Fragment N-6 was digested with PstI, and fragment C-6 was obtained by digestion with PstI and BspEI of pCMX-IP N-Ex1 /DP III-C . Both digested fragments were ligated into the PstI and BspEI sites of pCMX-mDP.
The Chimeric CR A19P Receptor-Fragments N-7 and C-7 were amplified by PCR with primer pairs and the template shown in Table I ( Fig. 3B). Fragment N-7 was digested with Asp718 and NarI, and fragment C-7 was digested with NarI and BspEI. Both digested fragments were ligated into the Asp718 and BspEI sites of pCMX-mDP.
The Chimeric CR G22S Receptor-Fragments N-8 and C-8 were amplified by PCR with primer pairs and the template shown in Table I ( Fig. 3B). Fragment N-8 was digested with Asp718 and Eco47III, and fragment C-8 was digested with Eco47III and BspEI. Both digested fragments were ligated into the Asp718 and BspEI sites of pCMX-mDP.
Five mutants were then constructed by site-directed mutagenesis of CR T94K receptor.
The Chimeric CR T94K/F96L Receptor-Fragment N-9 was amplified by PCR with primer pairs and the template shown in Table I (Fig. 3B). Fragment N-9 was digested with PstI, and fragment C-9 was obtained by digestion with PstI and BspEI of pCMX-IP N-Ex1 /DP III-C . Both digested fragments were ligated into the PstI and BspEI sites of pCMX-mDP.
The Chimeric CR T94K/A100M Receptor-Fragments N-10 and C-10 were amplified by PCR with primer pairs and the template shown in Table I (Fig. 3B). Fragment N-10 was digested with Asp718 and BsiWI, and fragment C-10 was digested with BsiWI and BspEI. Both digested fragments were ligated into the Asp718 and BspEI sites of pCMX-mDP.
The Chimeric CR T94K/F102L Receptor-Fragment C-11 was amplified by PCR with primer pairs and the template shown in Table I (Fig. 3B). Fragment C-11 was digested with PstI and BspEI, and fragment N-11 was obtained by digestion with PstI of pCMX-CR T94K . Both digested fragments were ligated into the PstI and BspEI sites of pCMX-mDP.
The Chimeric CR T94K/V103A Receptor-Fragment C-12 was amplified by PCR with primer pairs and the template shown in Table I (Fig. 3B). Fragment C-12 was digested with PstI and BspEI, and fragment N-12 was obtained by digestion with PstI from pCMX-CR T94K . Both digested fragments were ligated into the PstI and BspEI sites of pCMX-mDP.
The Chimeric CR T94K/S109Q Receptor-Fragments N-13 and C-13 were amplified by PCR with primer pairs and the template shown in Table I (Fig. 3B). Fragment N-13 was digested with Asp718 and BsiWI, and fragment C-13 was digested with BsiWI and BspEI. Both digested fragments were ligated into the Asp718 and BspEI sites of pCMX-mDP.
The Chimeric CR L25M/L30V Receptor-Fragments N-14 and C-14 were amplified by PCR with primer pairs and the template shown in Table I (Fig. 3B). Fragment N-14 was digested with Asp718 and Eco47III, and fragment C-14 was digested with Eco47III and BspEI. Both digested fragments were ligated into the Asp718 and BspEI sites of pCMX-mDP.
Ligand Binding Studies-Each receptor was transiently expressed in COS-7 cells cultured in 15-cm dishes by transfecting with 20 g of plasmid DNA by the lipofection method (17). After culture for 60 h, the cells were harvested, washed once, and suspended in a buffer containing 20 mM Hepes-NaOH (pH 7.4) containing 5 mM MgCl 2 , 140 mM NaCl, and 5 mM KCl. Binding assays were performed at 4°C for 1 h essentially as described previously (9). In competition experiments, the cells were incubated with 20 nM [ 3 H]PGD 2 or 20 nM [ 3 H]iloprost in the presence of various concentrations of PGD 2 , PGE 1 , PGE 2 , PGF 2 ␣, or iloprost. The incubation was terminated by the addition of 2 ml of ice-cold 10 mM Tris-HCl (pH 7.4) (the washing buffer), and the mixture was rapidly filtered through GF/C filters (Whatman International Ltd., Maidstone, United Kingdom). The filter was then washed with 5 ml of the washing buffer three times. The radioactivity on the filter was measured in 5 ml of Clear-Sol scintillation mixture (Nakalai Tesque, Kyoto, Japan). Nonspecific binding was determined in the presence of 500-fold excess of unlabeled ligands in the incubation mixture. K i values were calculated from IC 50 values of radioligand binding as described previously (18).

RESULTS
We previously revealed that the binding domain for the cyclopentane ring of mIP extending from the first to third transmembrane domains has the broad recognition, accommodating the D, E, and I types of cyclopentane rings, whereas that of mDP has the strict recognition, accommodating the D type alone. We also found that further replacement of the region containing the COOH-terminal end of the second transmembrane domain (Phe 102 ) to the second intracellular loop of mIP with that of mDP resulted in loss of the broad recognition of mIP and gain of the strict specificity of mDP. This IP N-II(Val101) / DP II(Leu83)-C receptor, formerly designated as IP N-II /DP III-C (15), markedly decreased the affinities to PGE 1 , PGE 2 , and iloprost, and bound PGD 2 alone. These results suggest that this region of mDP has amino acid residue(s) determining the strict specificity of ligand binding of mDP.
To define the regions conferring the ligand binding properties of mIP and mDP to discriminate the cyclopentane ring structure in more detail, the mIP, mDP, and six chimeric receptors including IP N-III /DP IV-C , IP N-Ex1 /DP III-C , IP N-II(Val101) / DP II(Leu83)-C , IP N-I /DP II-C , IP N /DP I-C , and DP N-I /IP II-Ex1 /DP III-C (Fig. 4) were expressed in COS-7 cells for binding studies. The cells were incubated with 20 nM amounts of either [ 3 H]iloprost or [ 3 H]PGD 2 , and the binding properties were analyzed by competition with PGD 2 , PGE 1 , PGE 2 , PGF 2␣ , and iloprost. We used in the present study intact cell suspensions for the binding assay, because more reproducible results were obtained in cell suspensions than in cell lysates used in the previous study (15). Representative analyses are shown in Fig. 4, and the results of several analyses are summarized in Table II. As shown in Fig. 4A, mIP showed a selective binding to iloprost and PGE 1 with the K i values of 62 Ϯ 4 and 417 Ϯ 59 nM, respectively (Table II), consistent with previous reports on the cloned mouse IP receptor (8) and on native IP receptor in various cells (19,20). Also consistent with previous reports on the cloned mouse and human DP receptors (9, 21) and on native human DP receptor (22), only PGD 2 effectively displaced [ 3 H]PGD 2 binding to mDP with a K i value of 11 Ϯ 2 nM (Fig. 4B, Table II). We then examined the binding properties of the chimeric receptors. As described previously (15), IP N-III /DP IV-C bound PGD 2 , PGE 1 , PGE 2 , and iloprost with the K i values of 555 Ϯ 48, 100 Ϯ 15, 155 Ϯ 9, and 51 Ϯ 7 nM, respectively (Fig. 4A, upper right, and Table II). Similar ligand binding properties were shown by IP N-Ex1 /DP III-C , which has a further substitution of the third transmembrane domain. They bound PGD 2 , PGE 1 , PGE 2 , and iloprost with the K i values of 275 Ϯ 88, 55 Ϯ 6, 55 Ϯ 6, and 21 Ϯ 3 nM, respectively (Fig. 4A, lower left, and Table II). These results indicate that the domain determining the binding specificity for the cyclopentane ring of mIP localizes in a region containing the first transmembrane domain to the first extracellular loop, and the third transmembrane domain is exchangeable. In contrast, the binding of PGE 1 , PGE 2 , and iloprost was almost abolished when the COOH-terminal end of the second transmembrane domain (Phe 102 ) to the first extracellular loop (Cys 122 ) of mIP was then further replaced (Fig. 4B, upper left, and Table II). This IP N-II(Val101) /DP II(Leu83)-C receptor bound PGD 2 alone. This binding specificity was consistent with that obtained by our previous study using the lysates of cells expressing this chimeric receptor (15), although the affinity was much lower than that found in the previous analysis; we found a difficulty in reproducing the previous high affinity of [ 3 H]PGD 2 binding to this chimera in cell lysates. The above results suggest that some amino acid residue(s) from Leu 83 to Cys 104 of mDP are responsible for the exclusion of prostanoid molecules other than the D type from binding to the receptor. Similar strict binding specificity was exhibited by IP N-I /DP II-C and IP N /DP I-C , which have further substitution of the second and the first transmembrane domains (Fig. 4B, Table II). Notably, these IP N-I /DP II-C and IP N /DP I-C receptors showed approximately 4and 160-fold increases in the binding affinity to PGD 2 compared with that of IP N-II(Val101) /DP II(Leu83)-C , respectively. IP N / DP I-C bound PGD 2 with almost the same K i value of 10 Ϯ 2 nM as that of mDP (Table II). These results indicate that the strict specificity of ligand binding of mDP is conferred at least in part by some amino acid residue(s) in the COOH-terminal end of the second transmembrane domain and the first extracellular loop of mDP and that some amino acid residue(s) in the first and the second transmembrane domains of mDP confers the high affinity to PGD 2 . We then replaced the second transmembrane domain and the first extracellular loop of mDP with that of mIP. This DP N-I /IP II-Ex1 /DP III-C receptor bound only iloprost with a K i value of 389 Ϯ 76 nM, and the binding of PGD 2 , PGE 1 , or PGE 2 was negligible (Fig. 4A, lower right, and Table II).
These results indicate first that the region of mIP from the second transmembrane domain to the first extracellular loop contains amino acid residue(s) conferring the iloprost binding, and second that amino acid residue(s) in the first transmembrane domain of mIP contribute to its broad binding properties.
To identify the amino acid residue(s) determining the broad binding properties of mIP, four amino acid residues in the first transmembrane domain of mIP, Pro 47 , Ser 50 , Met 53 , and Val 58 , which were identical in mIP and mEP 2 but differed in mDP, were identified (Fig. 5). The corresponding mDP amino acid residues in the first transmembrane domain of DP N-I /IP II-Ex1 / DP III-C (Ala 19 , Gly 22 , Leu 25 , and Leu 30 ) were changed to these mIP amino acids by site-directed mutagenesis (Fig. 3A). The resultant CR A19P , CR G22S , and CR L25M/L30V receptors were expressed, and their ligand binding properties were analyzed using [ 3 H]iloprost as a radioligand (Fig. 6). CR A19P , in which Pro 47 (mIP) was substituted for Ala 19 (mDP) of DP N-I /IP II-Ex1 / DP III-C , showed no difference in the binding properties from DP N-I /IP II-Ex1 /DP III-C (Fig. 6, A and B , Table III). However, when Ser 50 (mIP) was substituted for Gly 22 (mDP) of DP N-I / IP II-Ex1 /DP III-C , the resultant CR G22S receptor recovered the ability to bind PGE 1 , PGE 2 , and PGD 2 with the K i values of 65 Ϯ 12, 75 Ϯ 18, and 503 Ϯ 73 nM, respectively. This receptor bound iloprost with an affinity almost comparable to that of  To identify the amino acid residue(s) conferring the high affinity to PGD 2 , two amino acid residues in the second transmembrane domain of mIP, Thr 94 and Arg 107 , which were identical in mIP and mEP 2 but differed in mDP, were identified (Fig. 5). These amino acid residues of DP N-I /IP II-Ex1 /DP III-C was individually changed to the corresponding mDP amino acid residues, Lys 75 and Gln 88 (Fig. 3A). Ligand binding of the resultant CR T94K and CR R107Q receptors was analyzed using [ 3 H]iloprost as a radioligand (Fig. 6). CR R107Q , in which Gln 88 (mDP) was substituted for Arg 107 (mIP) of DP N-I /IP II-Ex1 / DP III-C , showed no difference in the binding properties from DP N-I /IP II-Ex1 /DP III-C (Fig. 6D, Table III). On the other hand, CR T94K acquired the ability to bind PGD 2 with the K i value of 23 Ϯ 7 nM, an affinity comparable to that of mDP, 11 Ϯ 2 nM (Figs. 4B and 6E, Tables II and III). This mutant receptor retained its iloprost binding, indicating that the exclusion of iloprost is exerted by other residue(s).
To identify the amino acid residue(s) to exclude iloprost from binding to mDP, five mutant receptors were constructed from CR T94K . Because mEP 2 , like mDP, does not bind iloprost, four amino acids that are identical in mDP and mEP 2 but differ in mIP, and one amino acid that differs between mDP, mEP 2 , and mIP were identified in the second transmembrane domain and  Table I. Numbers in parentheses indicate nucleotide numbers of the 5Ј-and 3Ј-termini of each fragment corrected for the residue numbers in the original mIP or mDP receptor cDNA (7,8). Restriction sites used for construction are indicated. the first extracellular loop (Fig. 5). Four of five amino acids were in the second transmembrane domain, and one was in the first extracellular loop. These mIP amino acid residues of CR T94K were individually changed to the corresponding mDP residues by site-directed mutagenesis (Fig. 3A). Ligand binding was analyzed using [ 3 H]PGD 2 as a radioligand (Fig. 7). CR T94K bound PGD 2 and iloprost with the K i values of 7 Ϯ 1 and 93 Ϯ 11 nM (Fig. 7A, Table IV). Individual substitution of Leu 77 (mDP), Met 81 (mDP), Ala 84 (mDP), and Gln 90 (mDP) for Phe 96 (mIP), Ala 100 (mIP), Val 103 (mIP), and Ser 109 (mIP) showed an affinity to PGD 2 almost comparable to that of CR T94K . As for iloprost binding, one mutant receptor, CR T94K/F96L showed about the same affinity as CR T94K itself (Fig. 7B, Table IV). On the other hand, each of the CR T94K/A100M , CR T94K/V103A , and CR T94K/S109Q receptors showed the lower affinity to iloprost than CR T94K , albeit about the 1.4 -2-fold difference ( Fig. 7 (C,  E, and F), Table IV). A more marked reduction in the binding affinity to iloprost was observed when Leu 83 (mDP) was substituted for Phe 102 (mIP). The resultant CR T94K/F102L receptor showed about a 5-fold decrease in the binding affinity to iloprost in comparison to CR T94K without an appreciable decrease in the binding of PGD 2 (Fig. 7D, Table IV).

DISCUSSION
Clarification of the ligand binding domain of receptors is important in understanding how each receptor responds selectively to its agonist, and ultimately contributes to development of more selective agonists and antagonists. The cyclopentane ring structures of each PG molecules are the primary determinant of the ligand binding affinity and specificity of each PG receptor. In this study, to identify the domains and amino acid residue(s) of the prostanoid receptor recognizing the ring structure, not the side chain configuration, we constructed a series of mutant receptors from mIP and mDP based on their high homology of the amino acid sequences. We started with the conclusion of our previous chimeric receptor study (15) that the binding domain for the cyclopentane ring localizes in a region containing the first to third transmembrane domain, and first conducted a more detailed chimera analysis. Using the IP N-Ex1 / DP III-C receptor, we first found that the ring recognition by IP and DP is done by a region spanning from the first transmembrane domain to the first extracellular loop of each receptor, and the third transmembrane domain is exchangeable. This finding is to correct our previous proposal that the third transmembrane domain plays a critical role in the ring recognition (15). The previous proposal was based on our incorrect interpretation of the data obtained with the IP N-II(Val101) / DP II(Leu83)-C receptor. This chimera was designated as IP N-II / DP III-C in the previous study (15), and the exclusion of PGs other than PGD 2 by this receptor was interpreted solely due to the replacement of the third transmembrane domain of mDP. The above finding that the third transmembrane domain is exchangeable in the prostanoid receptors is rather surprising, because this domain plays an important role in recognition of ligands in the receptors for various biogenic amines with an Asp in this domain working as a counterion for the amine group of the ligands (23). Whether this finding is also applicable to other prostanoid receptors such as EP, FP, and TP can be tested by replacing the above region of one receptor with the corresponding region of the other receptor. The next finding obtained by the present chimera analysis was that the replacement of a region containing the second transmembrane domain and the first extracellular loop of DP with that of IP conferred the iloprost binding without binding of the D and E types of PGs. IP and DP are believed to have been evolved with EP 2 from the primitive PGE receptor (14). The EP 2 receptor does not bind iloprost, whereas IP binds both iloprost and PGE. These findings taken together indicate that, during evolution, IP acquired the iloprost binding without losing the PGE binding, mainly by mutations in the region from the second transmembrane domain to the first extracellular loop. Similarly, the acquisition of PGD 2 binding by DP appears to have been made also by mutations in this region with, in this case, a concomitant loss of PGE binding, because IP N-I /DP II-C showed the selective binding to PGD 2 . Exclusion of PGE binding in DP appears to be exerted also by mutation(s) in the first transmembrane domain, for DP N-I /IP II-Ex1 /DP III-C did not show the PGE binding. Indeed, a mutation of Gly in the first transmembrane domain of this receptor to the corresponding amino acid, Ser, in IP and EP 2 recovered the PGE binding in the receptor.
In search for amino acid residues responsible for the identified effects of replacing the above mentioned domains between IP and DP, we engineered a series of mutant receptors in which amino acid residues derived from one receptor were mutated to the corresponding amino acids of the other receptor in DP N-I / IP II-Ex1 /DP III-C . By this procedure, we identified Ser 50 of IP responsible for conferring the PGE as well as PGD binding back in the chimeric receptor as mentioned above, and Lys 75 in the second transmembrane domain of DP required for the high affinity PGD 2 binding. Interestingly, FP has a His residue at the corresponding position. Rehwald et al. (24) recently mutated this His to various amino acids, and found the decrease in binding of PGF 2␣ to the FP receptor. Based on the pH effects on   (5) 10 Ϯ 2 (7) 11 Ϯ 2 (13) Ͼ2000 PGF 2␣ Ͼ2000 Ͼ2000 Ͼ2000 Ͼ2000 Ͼ2000 Ͼ2000 Ͼ2000 Ͼ2000 B max (fmol/1 ϫ 10 6 cells) 452 Ϯ 85 (10) 500 Ϯ 9 (3) 117 Ϯ 28 (6) 693 Ϯ 300 (5) 93 Ϯ 13 (5) 619 Ϯ 183 (7) 145 Ϯ 20 (13)  the PG binding to the mutant receptors, they indicated that His at this position may work in the PGF 2␣ binding as a hydrogen bond donor. Our findings that Lys at this position of DP works to recognize the ring structure of the D type of PGs indicate that amino acid at this position may form a hydrogen bond with a functional group attached to the ring and not with a carboxyl group as suggested by these authors. The above identifications are consistent with the findings on the chimeric receptors described above. Also consistent is the identification of amino acid residues preventing iloprost binding in the DP receptor. The chimera study using IP N-II(Val101) /DP II(Leu83)-C suggested that they localize in the COOH-terminal end of the second transmembrane domain and the first extracellular loop. The following point mutation study identified four amino acid residues, Met 81 , Leu 83 , Ala 84 , and Gln 90 of DP, attenuating the binding of iloprost to DP, three of which localize in the above defined region. Although each of these mutations decreases the iloprost binding to these mutants by only 1.5-5-fold, in combination, as  seen in DP, they could be sufficient to exclude iloprost as observed in native DP. Interestingly, three of them are present in the COOH-terminal end of the second transmembrane domain, and only one is in the extracellular loop, suggesting that the iloprost exclusion is carried out mainly by the transmembrane domains and not as a consequence of filter action of the extracellular loops as recently shown for the EP 3 receptor (25). Thus, we identified several amino acids in the ligand binding domains of IP and DP conferring the binding properties of each receptor. However, it should be mentioned that this mutation analysis was carried out based on the DP N-I /IP II-Ex1 /DP III-C chimera receptor. The converse analysis should be performed on IP N-I /DP II-C to identify amino acid residues of DP in the first transmembrane domain to increase the affinity for PGD 2 binding, and, similarly, IP residues in the second transmembrane domain to enable iloprost binding. These analyses combined together will tell us exactly which amino acid residues in IP and DP determine the binding selectivity of each receptor, and such knowledge may help to define the ligand binding domains of other prostanoid receptors such as EP 1 , EP 2 , EP 3 , EP 4 , FP,  and TP. PGs have two structural features, a cyclopentane ring and the side chains, and the receptors are supposed to recognize both of these structures and stabilize the ligand binding. Already much evidence has accumulated to suggest that the Arg residue conserved in the seventh transmembrane domain makes a hydrogen bond with the carboxyl group of prostanoid molecules (reviewed in Ref. 13). We previously used the IP/DP chimeras and suggested that the recognition of the side chain of iloprost and PGE 1 takes place in the sixth and/or seventh transmembrane domains (15). Recently, Kedzie et al. (26) introduced the Leu 304 3 Tyr substitution in the seventh transmembrane domain of the EP 2 receptor and found that this mutation caused an approximately 100-fold increase in the affinity to iloprost. These results taken together with our previous and present studies suggest that the ligand binding pocket of the prostanoid receptors are formed mainly by the first, second and seventh transmembrane domains, of which the former two are involved in the recognition of the ring structure and the latter in that of the side chains.