Stereochemical Course of the Reaction Catalyzed by the Cyclic GMP Phosphodiesterase from Retinal Rod Outer Segments*

The stereochemical course of hydrolysis catalyzed by the cyclic GMP phosphodiesterase from bovine retinal rod outer segments was determined. The S, diastereomer of guanosine 3',5'-cyclic monophosphorothioate was hydrolyzed by cyclic GMP phosphodiesterase in Hz"0 to give ['60,'80]guanosine 6'-monophosphoro- thioate. This isotopomer was reacted with diphenyl phosphorochloridate to form the two diastereomers of P1-(5'-guanosyl) P2-(diphenyl) 1-thiodiphosphate. The "P NMR spectrum of this mixture of diastereomers was identical to that obtained from ['60,'80]guan-osine 5'-monophosphorothioate resulting from the hy- drolysis of the R, diastereomer of guanosine S'-p-ni-trophenyl phosphorothioate by snake venom phos- phodiesterase. This finding indicates that the "0 is bridging in the R, diastereomer of the P1-(5'-guan-osyl) P2-(diphenyl) 1-thiodiphosphate

The cyclic GMP phosphodiesterase of retinal rod outer segments is a key enzyme in visual excitation (reviewed in Stryer, 1986;Liebman et al., 1987). Light triggers a cascade leading to the activation of this enzyme. The subsequent hydrolysis of cGMP causes cGMP-gated cation channels in the plasma membrane to close (Fesenko et al., 1985;Zimmerman et al., 1985). Retinal rod cyclic GMP phosphodiesterase has great catalytic power. The ratio of the turnover number of the enzyme to its Michaelis constant, kOat/Km, is 6 X lo7 " 1 s-l , close to the diffusion-controlled limit. We report here the elucidation of the stereochemical course of the retinal rod cyclic GMP phosphodiesterase reaction. This information is essential for unraveling the catalytic mechanism. If H20 directly adds to bound cGMP in one step, the reaction is expected to proceed with inversion of configuration at the phosphorus atom. If however, a nucleophilic group on the enzyme first adds to cGMP to form a covalent intermediate, and H20 then attacks in-line to displace the enzyme, the overall reaction should proceed with retention of configuration. Several methods for analyzing the stereochemical course of * This research was supported by Grant EY-02005 from the National Eye Institute and Collaborative Research Grant 0011/86 from NATO. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. enzyme-catalyzed phosphoryl and nucleotidyl transfer reactions have been devised and applied (reviewed in Knowles, 1980;Frey, 1982;Lowe, 1983;Gerlt et al., 1983;. Determination of the stereochemical course of a reaction requires substrates in which phosphorus is a chiral center. For the cyclic GMP phosphodiesterase reaction, this could be accomplished by employing cGMP in which the phosphorus is made chiral by the presence of two different isotopes of oxygen at the two nonbridging positions. Alternatively, the phosphorothioate analog cGMPS,' in which one of the nonbridging oxygens is replaced by sulfur, could be used. Both methods have been successfully applied in elucidating the stereochemical course of CAMP phosphodiesterases (Burgers et al., 1979b;Coderre et al., 1981;Jarvest et al., 1982). Diastereomers of cGMPS are known to be slowly hydrolyzed by retinal rod cyclic GMP phosphodiesterase. The turnover numbers for the S, and R, isomers are 3.7 and 0.2 s-', respectively, compared with 4000 s-l for cGMP (Zimmerman et al., 1985). We hydrolyzed the S, diastereomer of cGMPS in H2lS0 and determined the configuration of the product of the cyclic GMP phosphodiesterase reaction by a new method. The isotopomer formed by hydrolysis was treated with diphenyl phosphorochloridate to form two diastereomers, and the reaction mixture was analyzed by 31P NMR spectroscopy. This new approach is simple to apply and should be generally applicable in determining the stereochemical course of phosphoryl and nucleotidyl transfer reactions in which phosphorothioate analogs can serve as substrates.

MATERIALS AND METHODS
Cyclic GMP phosphodiesterase of retinal rod outer segments was isolated from frozen bovine retinas and purified on a hydroxyapatite column (Wensel and Stryer, 1986). Purified cyclic GMP phosphodiesterase was activated by treatment with 0.1 mg/ml trypsin for 5 min at room temperature (Hurley and Stryer, 1982). Tryptic digestion was stopped by the addition of soybean trypsin inhibitor (1 mg/ml). Cyclic GMP phosphodiesterase activity was assayed by measuring the change in pH resulting from proton release (Liebman and Evanczuk, 1982). Trypsin-activated cyclic GMP phosphodiesterase (2.2 mg/ml) was stored in 50% glycerol containing 150 mM sodium phosphate (pH 7.2), 0.5 mM dithiothreitol. The specific activity of the enzyme was 1200 pmol/min/mg (1200 units/mg). Snake venom phosphodiesterase (from Crotalus durissus, 2 mg/ml, 1.5 units/mg) was from Boehringer Mannheim. H2I80 (98.37% "0,0.75% "0, and 0.88% l6O) was purchased from Ventron, Karlsruhe, Federal Republic of Germany.
Reverse phase HPLC on ODS Hypersil was performed with a Waters dual pump 6000 A system in combination with a Waters model 680 gradient controller and a model 440 UV detector. For The abbreviations used are: GMPS, guanosine 5'-monophosphorothioate; S,and R,-cGMPS, the S, and R, isomers of guanosine 3',5'-cyclic monophosphorothioate; TEAB, triethylammonium bicarbonate; HPLC, high performance liquid chromatography. elution, a linear gradient ofeither 0-50% (system A) or 0-15% (system B) acetonitrile in 100 mM triethylammonium bicarbonate (TEAB) over 15 min was used. 31P NMR spectra were recorded with a Bruker WP2OoSY spectrometer operating at 81.01 MHz with quadrature detection and 'H broad-band decoupling in 5-mm tubes. For measurements of spectra under anhydrous conditions, 507-TR tubes from Wilmad Glass Co. (Buena, NJ) were used. Chemical shifts refer to 85% phosphoric acid as external standard and are negative when upfield from this reference.
Tri-n-octylamine was dried by storage over solid KOH. Dioxane was dried by refluxing over sodium in the presence of 4-benzoylbiphenyl until a persistent green color had developed. It was then distilled, and aliquots were removed under an argon atmosphere. Deuterated dioxane was dried by storage over sodium wire.
Synthesis of Diastereomers of Guanosine 5"p-Nitrophenyl Phosphorothioate-N-Isobutyrylguanosine 5'-bis(p-nitrophenyl)phosphorothioate (Eckstein and Kutzke, 1986) (500 mg, 720 pmol) was suspended in 10 ml of dioxane, and 8 ml of 25% aqueous ammonia were added. The reaction vessel was tightly sealed with a septum, and the mixture was kept at 50 "C overnight. The reaction solution was evaporated, and the residue was dissolved in HZ0 and extracted with ether. The aqueous phase was evaporated to remove traces of ether and then applied to a DEAE-Sephadex A-25 column (2 X 30 cm) which was eluted with a linear gradient of 1.5 liters each of Hz0 and 400 mM TEAB. The product was eluted at about 260 mM TEAB. Fractions containing product were combined, evaporated to dryness, and the residue was co-evaporated with methanol to remove traces of buffer. The yield of the mixture of diastereomers was 165 pmol. The two diastereomers had retention times in HPLC (system A) of 8.97 and 9.17 min and chemical shifts in the 31P NMR spectrum (DzO) of 52.61 and 52.19 ppm.
Hydrolysis of Guurwsine 5'-p-Nitrophenyl Phosphorothioate by Snake Venom Phosphodiesterase-A mixture of diastereomers of guanosine 5'-p-nitrophenyl phosphorothioate (413 pmol) was dissolved in 100 pl of H2"0, and the solution was evaporated to dryness. The residue was dissolved in 2.5 ml of H2'0 containing 100 mM Tris-HCI (pH 8.9) and 25 mM MgCl,. Snake venom phosphodiesterase (75 pl of 3 units/ml) was added, and the solution was incubated at 37 "C for 1 h. HPLC analysis (system A) indicated that the diastereomer with the shorter retention time had been hydrolyzed. The solution was applied to a DEAE-Sephadex column (2 X 30 cm) which was eluted with a linear gradient of 1 liter each of Hz0 and 400 mM TEAB. The undigested S,-isomer was eluted at 300 mM TEAB. The product, I'60,'80]GMPS, was eluted at 330 mM TEAB. Fractions containing ['60,'80]GMPS were combined and evaporated to dryness. The residue was co-evaporated with methanol to remove traces of buffer. The yield of product was 100 pmol. The residue was dissolved in water and the material purified by preparative HPLC (system A). The yield was 35 pmol, and the retention time was 4.3 min.
Phosphorothwate Analogs of cGMP-S,-and R,-cGMPS were prepared as a mixture of diastereomers and separated as described (Eckstein and Kutzke, 1986). Their configurations were assigned by comparing their 31P NMR shifts (54.99 and 56.69 ppm, respectively) with those of the corresponding diastereomers of cUMPS (54.46 and 56.04 ppm, respectively). The diastereomer of cUMPS with the low field resonance was shown by x-ray crystallographic analysis to have the Rp configuration (Hinrichs et al., 1987). Hence, the diastereomer of cUMPS with the high field resonance was the S, configuration. As the chemical shifts of the diastereostereomers of cUMPS and cGMPS are very similar, the configuration of the diastereomer of cGMPS with the high field resonance can also be assigned as S,.
Enzymatic Hydrolysis of S,-cGMPS-S,-cGMPS (50 pmol) was evaporated in 100 ~1 H,"0 and the residue dissolved in 5 ml of H2'0 containing 80 mM Tris-HC1 (pH 8.0) and 4 mM MgClZ. To this solution was added 165 pg (75 pl, 200 units) of retinal rod cyclic GMP phosphodiesterase. After overnight incubation at 30 "C, analysis by HPLC (system B) indicated that the reaction had gone to completion. The solution was applied to a QAE-Sephadex A-25 column (3 X 25 cm) which was eluted with a linear gradient of 1 liter each of 100 and 400 mM TEAB. Fractions of approximately 20 ml were collected. The product was eluted at approximately 280 mM TEAB and a trace of unreacted substrate at 240 mM TEAB. Fractions containing product were combined and evaporated to dryness in uacuo. The residue was reevaporated twice with methanol to remove traces of buffer. The yield of the triethylammonium salt of ['60,''O]GMPS was 32 *mol (63%). The product was homogeneous by HPLC analysis (system B). The 31P NMR spectrum indicated that less than 2% of [160,160] GMPS had been formed in this reaction.
Synthesis of PI-(.5'-Guanosyl) P2-fDipheny.yli 1-Thiodiphosphate-The triethylammonium salt of ['60,'60]GMPS (10 pmol) was dissolved in water and passed over a Merck I ion exchange column (5 X 25 cm) in the pyridinium form. The column was washed with 750 ml of water. The eluate was evaporated to dryness in uacuo and the residue dissolved in approximately 50 ml of methanol by gentle heating with a fan. After complete solution had occurred, 20 pmol (8.84 ml) of dry tri-n-octylamine was added and the solution evaporated in uactm to dryness again. The residue was evaporated twice with approximately 2 ml of anhydrous dioxane. During all these operations, the rotary evaporator was connected to a balloon filled with argon to reestablish atmospheric pressure. All subsequent additions were made with a syringe through a septum with care taken to maintain a dry argon atmosphere in the reaction vessel. The residue was dissolved by adding 400 pl of dry dioxane and 150 pl of dry dsdioxane. To this solution was added 25 pmol (10.93 p l ) of dry tri-noctylamine and 12 pmol (2.47 pl) of diphenyl phosphorochloridate. After reaction for 1 h, 10 pl of a 50 mM solution of 8-hydroxyquinoline in dry dioxane was added to complex trace metals. This solution was transferred with a syringe to an NMR tube that was filled with an argon atmosphere and closed with a septum.

RESULTS
The S, diastereomer of cGMPS was hydrolyzed by retinal rod cyclic GMP phosphodiesterase in H2"O. Unreacted S,-cGMPS and ['60,'80]GMPS, the product of the phosphodiesterase reaction, were readily separated by reverse phase HPLC (system B). Their retention times were 9.2 and 4.64 min, respectively. HPLC analysis indicated that the substrate was completely converted to product after overnight incubation as described under "Materials and Methods." Unexpectedly, S,-cGMPS and GMPS could not be separated by the standard ion-exchange chromatography system employing DEAE-Sephadex. Complete separation was achieved, however, by chromatography on QAE-Sephadex. The 31P NMR spectrum of the ['60,'80]GMPS isolated from the reaction mixture showed one strong signal located at 43.9 ppm; a minor signal (less than 2%) was present 0.034 ppm downfield from the major one. Addition of unlabeled GMPS demonstrated that the small downfield peak was ['60,'60]GMPS and therefore that the major peak was ['60,'sO]GMPS. The magnitude of this "0-induced chemical shift is consistent with previously reported values (Lutz et al., 1978;Cohn, 1982;Lowe, 1983). This result shows that the reaction proceeded with the expected incorporation of l 8 0 and that essentially no label had been lost either in the reaction or in the isolation.
The aim of this study was to determine whether hydrolysis of S,-cGMPS (compound 1 in Fig. 1) in H2180 yielded the S , (2) or R, (3) isotopomer of [160,'80]GMPS. The experimental strategy for distinguishing between these isotopomers was to react the ['60,1s0]GMPS product with diphenyl phosphorochloridate to give the two diastereomers of P1-(5'-guanosyl) PP-(diphenyl) 1-thiodiphosphate (Fig. 1) and record their NMR spectra. These spectra were then interpreted by comparing them with spectra of products derived from the action of snake venom phosphodiesterase, an enzyme whose stereochemical course is known (Burgers et al., 1979a). The phosphorothioate region of the 31P NMR spectrum of the mixture of diastereoisomers obtained from ['60,'80]GMPS, that was produced by the hydrolytic action of retinal rod cyclic GMP phosphodiesterase, is shown in Fig. 2 A . The four lines in this spectrum arise from a pair of diastereomers (either 4 and 5, or 6 and 7). Each compound gives rise to two lines because of coupling between its two phosphorus atoms. Compound 4 can be distinguished from 5 , and 6 can be distinguished from 7 by measuring the "0-induced chemical shift. "0 in a nonbridging position (compounds 4 and 7) is expected t o produce a larger shift than that given by "0 in the bridging position (5 and 6) because the "0-induced shift depends on bond order (Lowe et al., 1979). To determine this shift, a mixture of ['60,1sO]GMPS derived from hydrolysis by retinal rod cyclic GMP phosphodiesterase and ['60,'60]GMPS was reacted with diphenyl phosphorochloridate. The phosphorothioate region of the NMR spectrum of the mixture of reaction products is shown in Fig. 2B. The pair of lines centered a t 47.96 ppm ( J P~,~ = 31.5 Hz) shows an "0 shift of 0.023 ppm. The other pair, centered a t 47.76 ppm = 28.3 Hz) shows a shift of 0.037 ppm. Hence, the pair centered at 47.96 ppm comes from 5 or 6, whereas the pair centered at 47.76 ppm comes from 4 or 7.
Complementary evidence comes from the phosphate region of the NMR spectrum ( Fig. 3). Compounds 5 and 6 are expected to show an upfield "0-induced shift because "0 is directly bonded to the phosphorus atom of the phosphate group, whereas 4 and 7 should show no such shift. In fact, the pair of lines centered a t -24.04 ppm shows an "0 shift of 0.023 ppm, whereas the pair centered at -23.97 ppm shows no shift (compare Fig. 3, A and B). Hence, the -24.04 ppm pair comes from 5 or 6 and the -23.97 pair from 4 or 7.
The final step in the analysis was to determine whether the reaction of ['60,'s0]GMPS with diphenyl phosphorochloridate produced 4 and 5 or 6 and 7. A known isotopomer of ['60,'80]GMPS was needed to answer this question. R,- ['60,'sO]GMPS was produced by hydrolyzing R,-guanosine 5"pnitrophenyl phosphorothioate (compound 8 in Fig. 4) in H2lSO using snake venom phosphodiesterase, an enzyme known to form products with retention of configuration (Burgers et al., 1979a;1979~). Reaction of R,-[160,'80]GMPS produced in this way with diphenyl phosphorochloridate gave compounds 6 and 7 (Fig. 4). As depicted in Fig. 5, the 31P NMR spectra of 6 and 7, together with their l60,l6O counterparts, correspond to those shown in Figs. 2B and 3B. Hence, the products of the retinal rod cyclic GMP phosphodiesterase series of reactions are also 6 and 7. If 4 and 5 had been produced, the lsO-induced shift of the 47.96-ppm doublet in  I  I  I  I  I  I  I~I  I  I  I  I FIG. 3. 31P NMR spectrum of the phosphate region of P1-(5'-guanosyl) P2-(diphenyl) 1-thiodiphosphate derived from the retinal phosphodiesterase reaction product. The reaction mixtures for spectra A and B were identical with those of Fig. 2, A and E. Parameters were the same as for Fig. 2, except that the offset was -660 Hz, and the number of transients for spectrum A was 2000.
the phosphorothioate region (Fig. 2B) would have been larger than that of the 47.76 ppm doublet; in the phosphate region (Fig. 3 B ) , the -23.97-ppm doublet rather than the -24.04ppm doublet would have shown the 180-induced shift. The finding that the products derived from the retinal rod cyclic GMP phosphodiesterase reaction were 6 and 7 indicates that Rp-['60,'80]GMPS was formed from Sp-cGMPS. Hence, the reaction proceeded with inversion of configuration at the phosphorus atom.  4. Scheme for using snake venom phosphodiesterase to generate compounds of known stereochemistry. Hydrolysis by this enzyme is known to proceed with retention of configuration (Burgers et al., 1979a.

DISCUSSION
An important question concerning the mode of action of retinal rod phosphodiesterase is whether it catalyzes cyclic GMP hydrolysis directly in one step or does so in two steps through a covalent enzyme intermediate. The most direct experimental approach to answering this question is to elucidate the stereochemical course of the reaction. A chiral phosphorus is required. The phosphorus can be made chiral either by employing two different isotopes of oxygen at the two nonbridging positions in the cyclic phosphate or by substituting sulfur for one of the oxygens. We have adopted the second strategy, using the phosphorothioate analog cGMPS, because the synthesis of the substrate as well as the analysis of the product of the cyclic GMP phosphodiesterase reaction are simpler. The stereochemical course of hydrolysis of cyclic GMP, the physiologic substrate, is almost certainly the same as that of the phosphorothioate analog. Studies of 10 enzymes Gerlt et al., 1983), including the nucleoside 3',5'-cyclic phosphodiesterase from bovine heart (Burgers et al., 1979b;Coderre et al., 1981) and yeast (Jarvest et al., 1982), have shown identical stereochemical courses for phosphate esters (as determined by the oxygen isotope method) as for their phosphorothioate analogs.
When employing phosphorothioates to determine the stereochemical course, the reaction is performed in H2180 to obtain an isotopomer of GMPS. The configuration of the isotopomer is then determined by forming a pair of diastereomers. The differentiation between the two diastereomers is based on the observations that l80 induces an upfield chemical shift in the 31P NMR spectrum (Lutz et al., 1978;Cohn, 1982;Lowe, 1983) and that this shift increases with the bond order of the P-0 bond (Lowe et al., 1979). For adenosine or 2'deoxyadenosine nucleotides, the configuration of the isotopomers can be elucidated by stereospecific phosphorylation  with adenylate and pyruvate kinase. The kinase reaction yields the S,-isomer of the corresponding nucleoside 5'-0-(1thiotriphosphate). However, this method could not be used here because adenylate kinase does not phosphorylate GMP. A more generally applicable method has been described by  in which the isotopomer of the nucleoside 5'-phosphorothioate is methylated by diazomethane or dimethyl sulfate to form the nucleoside phosphorothioate S-methyl and 0-methyl triester. The products can again be distinguished by 31P NMR spectroscopy to define the stereochemical course of the reaction. The stereochemical course of the reactions catalyzed by mung bean nuclease (Hamblin et aL, 1987) and mucosa phosphodiesterase  have been determined in this way using nucleoside phosphorothioates as substrates. We were concerned that methylation of the guanine base of GMP might complicate the interpretation of the NMR spectra. Consequently, we searched for an alternative method that would be generally applicable.
It occurred to us that reaction of the isotopomer of guanosine 5'-monophosphorothioate with diphenyl phosphorochloridate should result in formation of a pair of diastereomers of P1-(5'-guanosyl) P2-(dipheny1)l-thiodiphosphate as long as the oxygens rather than the sulfur atom of the phosphorothioate group were acting as the nucleophiles. This was shown to be the case because activation of nucleoside 5 'phosphorothioates with diphenyl phosphorochloridate and subsequent reaction with pyrophosphate results in the formation of the nucleoside 5'-O-(l-thiotriphosphates) (Eckstein and Gindl, 1967;Eckstein and Goody, 1976;Sheu et al., 1979;Richard and Frey, 1983). Experience with 31P NMR spectra of phosphorothioate analogues of nucleotides, in particular those of nucleoside 5'-O-(l-thiotriphosphates) (Chen and Benkovic, 1983) and nucleoside 3',5'-cyclic phosphorothioates (Eckstein and Kutzke, 1986), shows that the chemical shifts are virtually independent of the bases of the nucleoside. Consequently, the relationship established here between chemical shift and configuration of diastereomers of P1-(5'guanosyl) P2-(diphenyl) 1-thiotriphosphate is likely to be applicable to other nucleotides.
We have found that hydrolysis of a phosphorothioate analog of cyclic GMP proceeds with inversion of configuration at the phosphorus atom. This result is most simply interpreted in terms of an in-line SN2 mechanism (Knowles, 1980), in which the attacking H20 occupies one apex of a trigonal bipyramid and the 3'-hydroxyl-leaving group occupies the other apex. A two-step reaction involving an in-line attack by a nucleophilic group on the enzyme followed by an in-line attack by water would lead to retention of configuration, which is not observed. It is interesting to note that most phosphoryl and nucleotidyl transferase reactions (-90% of more than 50 studied) proceed with inversion of configuration at the phosphorus atom . In fact, the stereochemical course of hydrolysis by retinal rod cyclic GMP phosphodiesterase is the same as observed previously for cyclic AMP phosphodiesterases from bovine heart (Burgers et a!., 1979b) and yeast (Jarvest et al., 1982). Another common feature is that all three enzymes hydrolyze the S, diastereomer of the phosphorothioate analog of their cyclic nucleotide substrate more rapidly than the R, diastereomer. The amino acid sequences of regions of the three phosphodiesterases are very similar (Charbonneau et al., 1986;Chen et al., 1986;Ovchinnikov et aE., 1987). It seems likely that the three enzymes evolved from a common ancestor. Their mechanistic similarity may be an expression of their evolutionary kinship. Acknowledgments-We wish to thank Dr. J. Ott for many helpful suggestions, U. Kutzke for expert technical assistance, and B. Seeger for his expertise in recording the NMR spectra.