Inorganic pyrophosphatase from bovine retinal rod outer segments.

Rod outer segments from bovine retina contain a higher level of intracellular inorganic pyrophosphatase (EC 3.6.1.1) activity than has been found in any other mammalian tissue; the specific activity in extracts of soluble outer segment proteins is more than 6-fold higher than in extracts from bovine liver and more than 24-fold higher than in skeletal muscle extracts. This high activity may be necessary to keep inorganic pyrophosphate concentrations low in the face of the high rates of pyrophosphate production that accompany the cGMP flux driving phototransduction. We have begun to explore the role of inorganic pyrophosphatase in photoreceptor cGMP metabolism by 1) studying the kinetic properties of this enzyme and its interactions with divalent metal ions and anionic inhibitors, 2) purifying it and studying its size and subunit composition, and 3) examining the effects of pyrophosphate on rod outer segment guanylyl cyclase. Km for magnesium pyrophosphate was 0.9-1.5 microM, and the purified enzyme hydrolyzed > 885 mumol of PPi min-1 mg-1. The enzyme appears to be a homodimer of 36-kilodalton subunits when analyzed by gel electrophoresis and density gradient centrifugation, implying that kcat = 10(3) s-1, and kcat/Km = 0.7-1 x 10(9) M-1 s-1. The enzyme was inhibited by Ca2+ at submicromolar levels: 28% inhibition was observed at 138 nM [Ca2+], and 53% inhibition at 700 nM [Ca2+]. Imidodiphosphate acted as a competitive inhibitor, with Ki = 1.2 microM, and fluoride inhibited half-maximally approximately 20 microM. Inhibition studies on rod outer segment guanylyl cyclase confirmed previous reports that pyrophosphate inhibits guanylyl cyclase, suggesting an essential role for inorganic pyrophosphatase in maintaining cGMP metabolism.

Rod outer segments from bovine retina contain a higher level of intracellular inorganic pyrophosphatase (EC 3.6.1.1) activity than has been found in any other mammalian tissue; the specific activity in extracts of soluble outer segment proteins is more than &fold higher than in extracts from bovine liver and more than 24-fold higher than in skeletal muscle extracts. This high activity may be necessary to keep inorganic pyrophosphate concentrations low in the face of the high rates of pyrophosphate production that accompany the cGMP flux driving phototransduction. We have begun to explore the role of inorganic pyrophosphatase in photoreceptor cGMP metabolism by 1) studying the kinetic properties of this enzyme and its interactions with divalent metal ions and anionic inhibitors, 2) purifying it and studying its size and subunit composition, and 3) examining the effects of pyrophosphate on rod outer segment guanylyl cyclase. K,,, for magnesium pyrophosphate was 0.9-1.6 PM, and the purified enzyme hydrolyzed >886 rmol of PPI min" mg". The enzyme appears to be a homodimer of 36-kilodalton subunits when analyzed by gel electrophoresis and density gradient centrifugation, implying that koat = los s-', and &,/X,,, = 0.7-1 X 10' M" s-'.
The enzyme was inhibited by Ca2+ at submicromolar levels: 28% inhibition was observed at 138 nM [Ca"'], and 53% inhibition at 700 nM [Ca"']. Imidodiphosphate acted as a competitive inhibitor, with Ki = 1.2 PM, and fluoride inhibited half-maximally -20 PM. Inhibition studies on rod outer segment guanylyl cyclase confirmed previous reports that pyrophosphate inhibits guanylyl cyclase, suggesting an essential role for inorganic pyrophosphatase in maintaining cGMP metabolism.
Inorganic pyrophosphate (PPi) is produced in living cells by numerous metabolic pathways (Stetten, 1960), including synthesis of polymers such as DNA, RNA, protein, and polysaccharides, as well as synthesis of small molecules such as cGMP and CAMP. PPi is also produced by oxidative phosphorylation and glycolysis. Consequently, the rate of PPi production varies with the metabolic activity in cells (Klemme, 1976) and has to be counterbalanced by efficient catabolism. In all cells studied, there is an inorganic pyro-* This work was supported by National Institutes of Health Grant EY07981. Some of this work has been presented in preliminary form Wensel, 1990,1991). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$To whom correspondence should be addressed. Tel.: 713-798-6994; Fax: 713-796-9438. phosphatase (PPase,' EC 3.6.1.1) activity that catalyzes rapid conversion of PPi into Pi (reviewed by Cooperman (1982) and Cooperman et al. (1992)). It has been suggested that PPi may regulate many enzyme activities without actually participating in the reaction (Khandelwal and Kamani, 1980;Wheeler and Lowenstein, 1980). In spite of its importance in metabolism and the fact that PPase activity varies widely among mammalian tissues (Shatton et aL, 1981), almost nothing is known about how this enzymatic activity is regulated in mammals. It has been suggested that inhibition of PPase by Ca2+ may represent a mechanism for hormonal control of PPase activity (Davidson and Halestrap, 1987;Hachimori et al., 1983), but this kind of regulation has not been clearly demonstrated. While PPase from rat liver mitochondria has the requisite Ca2+ sensitivity (Ki = 67 nM for CaPPi) (Davidson and Halestrap, 1989), cytoplasmic PPase from porcine brain has been reported to require non-physiological levels of Ca2+ for inhibition (Hachimori et aL, 1983).
In rod outer segments (ROS) of the vertebrate retina, the second messenger cGMP plays a pivotal role in phototransduction (reviewed by Stryer (1991) and Pugh and Lamb (1990)). Light initiates an enzymatic cascade that leads to hydrolysis of cGMP and, in turn, to closure of cGMP-gated cation channels in the plasma membrane. A secondary effect of this closure is a reduction in intracellular [Ca'+] (Yau and Nakatani, 1985;McNaughton et al., 1986;Ratto et aL, 1988) that appears to be essential for normal kinetics of recovery from the light response (Lamb et al., 1986) and for the phenomenon of light adaptation (Matthews et al., 1988, Nakatani andYau, 1988). Lowering Caz+ markedly enhances the activity of the ROS guanylyl cyclase (Pepe et aL, 1986;Koch and Stryer, 1988;Dizhoor et aL, 1991). Studies of *'O incorporation into guanine nucleotides in intact retinas have demonstrated that cGMP is produced in the dark at a rate of 29 PM s-', and in the light the rate increases to 130 PM s" (Ames et al., 1986).
These high levels of cGMP production must also generate high levels of PPi, a by-product of the cyclization. This ion can inhibit cellular enzymes (Khandelwal and Kamani, 1980) including a cyclic nucleotide phosphodiesterase (Cheung, 1966), guanylyl cyclase (Hakki and Sitaramayya, 1990;Hayashi and Yamazaki, 1991) and adenylyl cyclase (Johnson and Shoshani, 1990), and its export from cells can lead to pathological conditions (Pritzker, 1986), so it seems likely that the PPase that catalyzes its hydrolysis is either maintained at high levels of activity at all times, or is regulated in synchrony with guanylyl cyclase. A high level of PPase activity has in fact been reported to be present in bovine ROS (Hakki and Sitaramayya, 1990). Because very little is known about retinal PPase or about PPase regulation in animals in general, we The abbreviations used are: PPase, inorganic pyrophosphatase; MOPS, 4-morpholinepropanesulfonic acid; DTT, dithiothreitol; HPLC, high pressure liquid chromatography; ROS, rod outer segments.

24634
Photoreceptor Inorganic Pyrophosphatase 24635 decided to purify and study the enzyme(s) from bovine ROS. Because Ca2+ plays an important role i n the recovery phase of the rod cell's light response when PPi production is expected to be maximal, and because Ca2+ is a potent inhibitor of yeast PPase (Ridlington and Butler, 1972), we have tested the sensitivity of ROS PPase to Ca2+ in the physiologically relevant concentration range, and have studied the effects of PPi on ROS guanylyl cyclase. We have also examined the effects on both PPase and guanylyl cyclase of imidodiphosphate, a nonhydrolyzable PPi analogue that has been reported to reduce a current component attributed to activation of guanylyl cyclase by low Ca2+ in gecko rods (Detwiler and Rispoli, 1989). For reverse phase HPLC, buffer E was water with 0.1% (v/v) trifluoroacetic acid, and buffer F was isopropano1:acetonitrile:water (3:2:2) with 0.08% trifluoroacetic acid.

Reagents
Rod Outer Segments and Protein Preparation-For preparative purposes, ROS were prepared from frozen dark dissected bovine retinas by the sucrose flotation method as described (Fung and Stryer, 1980). For analysis of PPase localization, purer ROS were prepared by the sucrose gradient method of Papermaster and Dryer (1974). The ROS band was collected, washed with ROS buffer, resuspended in buffer containing 24% (w/w) sucrose, and loaded onto a second identical sucrose gradient. Following centrifugation the gradient was divided into 10 fractions, each of which was analyzed for PPase activity and for 500-nm rhodopsin absorbance (in 1.5% N,N-dimethyldodecylamino-N-oxide). The ROS preparations were carried out in dim red light.
Soluble ROS proteins including PPase were extracted by homogenizing ROS using a Potter-Elvehjem homogenizer in ROS buffer and separating the soluble proteins from the membranes by centrifugation (30 min, 40,000 rpm, Beckman type 45 Ti rotor). The supernatant from this procedure is referred to as "ROS extract," and was used for many of the PPase assays. Protein concentrations were determined by the Coomassie Blue binding method (Bradford, 1976).
Other Bovine Tissues-Samples of bovine liver, brain, and skeletal muscle were obtained at a local slaughterhouse within a few minutes after slaughter, and frozen in liquid nitrogen, closely mimicking the process by which the frozen retinas are prepared, except for the slightly longer time lag involved in the dissection of the retinas. Tissues were homogenized in ROS buffer using a glass-Teflon homogenizer, and soluble proteins separated from cellular debris in a TL-100 ultracentrifuge at 40,000 rpm for 20 min. The supernatant solutions of soluble proteins were used for PPase assays.
PPase Assay-PPase activity was determined by measuring phosphate (Pi) generated from substrate PPi. For most routine purposes, PPase assays were performed by using a molybdate-blue based colorimetric assay (Bartlett, 1959). The standard assay was performed in a 100-pl total volume containing PPase assay buffer and sufficient enzyme to yield total PPI hydrolysis of less than 10% during a 10min incubation at 37 "C. The reaction was terminated by adding 40 pl of 10 N H2SO4. For K,,, and Ki determinations or assays in the presence of Pi, [32P]PPi was used as substrate with 1 mM MgCL (for K,,, and Ki) or 3 mM MgCIz, and PPi and Pi were separated either by polyethyleneimine cellulose thin layer chromatography (1 M KHzPO4) or by isobutanol extraction of their molybdate complexes as described by Springs et al. (1981). One unit of PPase activity corresponds to 1 pmol of PPi hydrolysis per min.
Inhibition of ROS PPase by fluoride was measured using the standard assay conditions and 0-10 mM NaF. Because imidodiphosphate interferes with the colorimetric assay, ["PIPPi was used for measuring inhibition of PPase by imidodiphosphate. Both imidodiphosphate and PPi concentrations were varied, using 1 mM M e , 0.2 mM EGTA, 50 mM Tris-HC1 (pH 7.5).
Guunylyl Cyclnse Assays-Guanylyl cyclase activity was determined using the method of Koch and Stryer (1988) with the following modifications. ROS membranes stripped with low salt buffer to remove cGMP phosphodiesterase and recoverin were used as the source of enzyme, 2 mM Mnz+ or 5 mM M e was substituted for 12 mM M e , and isobutylmethylxanthine and EGTA were not included, as they were found under these conditions not to increase cGMP production significantly. To examine the effect of PPi on guanylyl cyclase, PPi was added to give a final concentration ranging from 0 to 12 mM in the case of M$+ or 0 to 6 mM in the cast of Mn2+. Additional M e or Mn2+ was added to some samples as indicated to ensure an excess of divalent metal ions over PPi.
Determination of Free Ca2+-The concentration of free Ca2+ was determined using the indicator dye Fura-2 by the method of Grynkiewicz et al. (1985) using either an Aminco-Bowman spectrofluorimeter modified as described elsewhere (Ramdas et al., 1991) or an SLM 8000 spectrofluorimeter.
Purification of PPase-ROS extract from 300 retinas was diluted 2-fold with water and loaded onto a hydroxylapatite column (2.4 X 7 cm), which had been equilibrated with buffer A. The column was washed with 50 ml of buffer A and then with a linear 400-ml gradient from buffer A to buffer B, and the eluent was collected in 5-ml fractions. Fractions 35-52, which contained the peak of PPase activity, were pooled and concentrated to 5 ml with an ultrafiltration apparatus using Amicon YM-30 membranes. The sample was desalted by dilution to 50 ml with buffer C and repeated reconcentration (three times, final volume 3 ml) in the same apparatus, and then injected onto a weak anion exchange HPLC column (Waters Protein Pak DEAE 5 PW, 7.5 X 75 mm) and eluted with a 60-min linear gradient (1 ml/min) from buffer C to 10% buffer D, 90% buffer C, both at pH 8.0. The fractions containing the main peak of PPase activity were pooled, concentrated, and desalted as described above. The sample was then reinjected onto the same column and eluted using conditions identical to those for the first injection, except that the gradient was prolonged to 80 min. This concentration/desalting/injection procedure was repeated a third time, using an 80-min gradient from buffer C to 10% buffer D, both at pH 8.5. The PPase-containing fractions from the final HPLC step were analyzed by 12% SDS-polyacrylamide gel electrophoresis to exmine the purity of the preparation. Reverse phase HPLC was carried out using a Vydac Protein C4 column and a linear gradient from 30% buffer F, 70% buffer E to 80% buffer F. Fractions were dried under vacuum, redissolved in PPase assay buffer, and assayed for PPase activity.
performed as described (Baehr et al., 1979). Crude PPase (ROS Size of Native PPase-Sucrose density centrifugation analysis was extracts, 500 pg of total protein), was applied to a linear gradient of 5-15% sucrose (w/v) in a total volume of 3 ml of ROS buffer, along with molecular weight marker proteins: hemoglobin, fluorescein isothiocyanate-labeled carbonic anhydrase, and ROS cGMP phosphodiesterase (purified as described by Wensel and Stryer (1990)). After centrifugation at 400,000 X g for 3 h (Beckman TL-100 rotor), the gradient was collected in thirteen 2 2 0 4 fractions which were assayed for PPase activity, as well as for trypsin-activated phosphodiesterase (Liebman and Evanczuk, 1982;Wensel and Stryer, 1986), for 409-nm hemoglobin absorbance, and for carbonic anhydrase fluorescence.
Electrophoresis-SDS-polyacrylamide gel electrophoresis was carried out essentially according to the method of Laemmli (1970) using 12% acrylamide. Protein bands were visualized with Coomassie Blue stain.

RESULTS
PPase Actiuity in ROS-The PPase activity detected in ROS varied from preparation to preparation, as might be expected for a soluble enzyme within a membrane organelle Photoreceptor Inorganic Pyrophosphatase that is fairly leaky because of preparation from frozen retinas. The highest specific activity of PPase that we have observed in ROS-derived samples is 1.0 unit/mg rhodopsin in ROS suspensions, or 2.4 units/mg in extracts, but values as low as 0.7 unit/mg have also been observed in some extracts. Because PPase activity is ubiquitous, we were concerned initially that the activity observed could be due to contamination from other cells in the homogenate, even though the high level of activity suggested that contamination was not likely to be the major source of PPase. As a test of the origin of the activity, we prepared somewhat purer ROS using small scale sucrose gradient centrifugation according to the method of Papermaster and Dreyer (1974) and then centrifuged the purified ROS on a second sucrose gradient and analyzed the activity in each fraction. The results (Fig. 1) indicated that the enzyme copurified with rhodopsin, and not with soluble proteins nor with more rapidly sedimenting membranes. Although the ROS tend to be rather leaky under our assay conditions, we observed an increase (about 2-fold) in apparent specific activity when the membranes were sonicated, as expected for an enzyme trapped within membranes.
Because the PPase activity in ROS is soluble (approximately 20% of the activity remains on the membranes after washing in either isotonic or low salt buffers; data not shown) the activity copurifying with rhodopsin must be trapped within sealed membrane compartments. PPase activity is present in all the supernatants obtained at those stages of our standard ROS preparation at which sucrose is diluted to allow the ROS to sediment. This activity is presumably derived from other cells in the retina as well as from leaky ROS. Importantly, the specific activity of PPase was always higher in the final ROS pellet (typically 0.7-2.4 units/mg soluble protein) than in the retinal homogenate from which the outer segments were isolated (typically 0.2-0.8 unit/mg). Therefore, unless some non-ROS vesicles with extremely high PPase activity happen to copurify with the ROS, it is fairly certain that the activity observed is in fact derived from ROS. We cannot, of course, rule out the possibility that some fraction of the total activity represents contamination from inner segments or from other cell types and organelles; for example the small amount of membrane-bound activity could represent some contamination from mitochondria, which have been reported to contain a membrane bound PPase (Baykov et al., 1989).
Extracts of soluble proteins from other bovine tissues ex-

dient.
Rod outer segments prepared from bovine retinas by sucrose density gradient centrifugation were washed and applied to a second (identical) sucrose gradient. After centrifugation, fractions were collected and assayed for rhodopsin content (absorbance at 500 nm: open circles) and for pyrophosphatase activity, before (triangles) and after (filled circles) sonication. A value of 1 corresponds to 0.014 unit of activity. amined for comparison had significantly lower activity under our assay conditions. They were compared to a typical preparation of ROS extract (2.4 units/mg) assayed in parallel under identical conditions. PPase activity in brain extracts was the highest (0.83 unit/mg), consistent with the high levels found in other mammalian brains (e.g. 0.85 unit/mg in porcine brain extracts (Hachimori et al., 1983) and 0.23 unit/mg in rat brain extracts (Shatton et al., 1981)), and with the higher levels of PPase mRNA present in retina as compared to brain (Yang and Wensel, 1992). The activity in liver extracts (0.36 unit/mg) was somewhat lower than in brain, but, as found in rodents (e.g. -0.8 unit/mg in rat or mouse liver extracts (Shatton et al., 1981) and 0.3 unit/mg in mouse liver extracts (Irie et al., 1970)), much higher than the activity we found in extracts from bovine skeletal muscle (50.1 unit/mg). Thus, the level of PPase activity expressed in different bovine tissues can vary by more than 24-fold.

Dependence of Activity on Pyrophosphate and Divalent Cat-
ions-The [PPJ dependence of the initial hydrolytic velocity is shown in Fig. 2. In this experiment the calculated K, was 1.47 p~, and in a number of experiments values between 0.9 and 1.5 p~ were consistently obtained, using either ROS extracts or highly purified PPase. This value is significantly lower than the K, values reported for other PPases which range from 5 p~ (Yoshida et al., 1982) to 700 pM (Chen et al., 1973), and it is also much lower than the value of 26 pM previously reported for ROS PPase (Hakki and Sitaramayya, 1990). The discrepancy in K , for ROS PPase may be due to the reported use of concentrations of PPi between 50 p~ and 1 mM for the K , determinations in the previous study. When When the effects of cations were tested, using 3 mM PPi as substrate and 3 mM added cation, it was found that M F confers maximal activity (100%) on ROS PPase. Some activity is also found in the presence of the same concentration of Co2+ (14.2%), A13+ (9.9%), Ni2+ (8.3%), Zn2+ (7.3%), Cu2+ (7.0%), or Mn2+ (6.6%). Similar results have been reported for other PPases such as those from yeast (Butler and Sperow, 1977;Janson et al., 1979) and pig cartilage (Felix and Fleisch, 1975). Since MgZ+ has maximum potential to activate the enzyme and is the probable cofactor in vivo, we further investigated the MgZ+ concentration dependence of this activation, as shown in Fig. 2 (inset). At 3 mM PPi with varying MgClz concentrations, PPase had maximal activity at about 3 mM M$+ and declined in activity at MP'+/PPi ratios greater than 1. This result is consistent with the finding for other PPases that both MgPPi and Mg2PPi complexes are substrates, with a slightly lower maximal velocity for the dimagnesium species (Moe and Butler, 1972a;Felix and Fleisch, 1975;Pynes and Younathan, 1967).
Inhibition by Ca2+-Submillimolar levels of EGTA enhanced the activity of PPase in ROS approximately 8-fold (data not shown). This effect may be partly due to removal of Ca2+, and partly due to removal of trace heavy metal ions which also bind EGTA. Inhibition of the EGTA-activated enzyme by addition of Ca2+ confirmed that Ca2+ does inhibit ROS PPase. Fig. 3 shows the [Ca"] dependence of this inhibtion, determined by using varying ratios of Ca2+ to EGTA and measuring free [Ca"] with the indicator dye fura-2. The result shows that under our standard assay conditions 138 nM [Ca"] leads to 28% inhibition, and greater than 50% inhibition is observed at 700 nM free [Ca"]. The concentration of CaPPi under the latter condition is estimated to be -70% that of free Ca", while the concentration of MgPPi is slightly less than 2 mM. Inhibition by Ca2+ seems to be an intrinsic property of the enzyme, as Ca2+ effects were not significantly different with highly purified ROS PPase as compared to crude extracts of soluble proteins. When Ca2+ sensitivity of PPase was examined in crude extracts of soluble proteins from other bovine tissues, it was found to be similar, although at every Ca2+ concentration examined, inhibition of ROS PPase was slightly greater (data not shown).
Anionic Inhibitors of PPuse-Imidodiphosphate, a nonhydrolyzable PPi analogue, was found to be a potent inhibitor of ROS PPase. Fig. 4A shows a Dixon (1953) plot of the inverse velocity as a function of imidodiphosphate concentration at three different substrate concentrations. The results indicate that inhibition is competitive with Ki = 1.2 p~. Fluoride, an anion commonly used to stimulate G-proteins (including transducin) and adenylyl cyclase, is an inhibitor of many PPases (Felix and Fleisch, 1975). We have found that it inhibits ROS PPase with half-maximal inhibition at 20 PM (Fig. 4B), a much lower concentration of F-than is normally used for G-protein stimulation.
Inhibition of Guanylyl Cyclase by Pyrophosphate-PPi was an effective inhibitor of guanylyl cyclase when either M e or Mn2+ was used as cofactor, and the inhibition could not be explained simply by competition for metal ion cofactor (Fig.  5). This result confirms two previous reports that PPi inhibits ROS guanylyl cyclase. One (Hakki and Sitaramayya, 1990) reported competitive inhibition with Ki = 70-100 pM, and the other (Hayashi and Yamazaki, 1991) reported noncompetitive inhibition, with Ki = 30-45 pM. Our results are not extensive enough to distinguish between the two types of inhibition but are consistent with competitive inhibition and a Ki of 100 p~, assuming K,,, = 250 ~L M for GTP (Koch, 1991;Hakki and Sitaramayya, 1990). We have also found that imidodiphosphate can directly inhibit guanylyl cyclase, but with much lower potency. Cyclase was 50% inhibited at 3-4 mM imido-  . 4. Inhibition of PPase by imidodiphosphate and fluoride. A, PPase initial velocity was measured as a function of added imidodiphosphate using [32P]PP, as substrate at three different PPi concentrations: open circles, 0.7 pM; filled circles, 1.5 pM; triangles, 3.0 p~. A value of 1 for V represents 1.0 X units in the 100-p1 assay volume. The three lines drawn were fit by linear least squares and intersect at a value of -1.2 p~ imidodiphosphate, consistent with competitive inhibition and Ki = 1.2 pM. E , PPase activity was measured as a function of added NaF using the colorimetric assay. The maximal activity without Fwas normalized to 100. diphosphate (data not shown). ATP has been reported to inhibit ROS guanylyl cyclase (Sitaramayya et aL, 1991), and we have confirmed this result as well (data not shown), in contrast to another report (Hayashi and Yamazaki, 1991) that ATP is not inhibitory.
Purification of ROS PPase-We have tried a number of different procedures for purification of PPase extracted from ROS and have arrived at a scheme involving: 1) isotonic extraction of homogenized ROS, 2) chromatography on hydroxylapatite, and 3) three rounds of anion exchange HPLC using different gradients or pH conditions each time (Table  I). Procedures used for purification of other PPases including ammonium sulfate precipitation (Irie et al., 1970) and Waminohexyl agarose chromatography (Yoshida et al., 1982) gave very poor recovery of activity. The protein obtained using the procedure outlined in Table I was typically 70%-95% pure, as estimated from Coomassie Blue staining of gels.
In the purest preparations the specific activity was >885 units/mg. This specific activity is comparable to that of PPase purified from pig brain (993 units/mg, Hachimori et al., 1983) which has the highest specific activity reported for any PPase. This high specific activity rules out the possibility that the enzymatic activity observed is due to a minor protein not visible on stained gels. Fig. 6 shows the profiles of PPase activity and absorbance at 280 nm of the third DEAE-HPLC column as well as SDS-polyacrylamide gel electrophoresis of the activity peak fractions. Only one major band on the Coomassie Blue-stained gels correlates well with the relative activity in each fraction, so that we can identify this 36kilodalton band as corresponding to PPase. Traces of a protein with an apparent molecular mass of 65-kDa were occasionally detected in the peak PPase fractions, but its presence or absence did not correlate well with PPase activity, and it is unlikely that such a small amount of protein could account for the observed activity. It may represent two covalently linked PPase subunits. A minor diffuse band migrating just below the main PPase band was also frequently observed. The intensity of this band increased with time after purification, so it is probably a fragment of the PPase subunit produced by proteolytic degradation. For analytical purposes, we chromatographed some of the partially purified PPase on a reverse phase (C-4) column. While most of the enzymatic activity was lost following denaturation by the reverse phase solvents (isopropanol, acetonitrile, and trifluoroacetic acid), enough was recovered upon addition of aqueous buffers to the dried fractions that they could be tested for activity. Only the fractions containing the 36-kDa band had detectable PPase activity (data not shown).
Size of Native PPase-To determine the native size of ROS PPase, sucrose density gradient centrifugation was employed, because PPase appeared to bind to two different HPLC gel filtration columns tried, eluting at a volume consistent with a 45-kDa protein (Superose 12) or a 12-kDa protein (Biosil TSK-250). Yeast PPase, known to be a homodimer of 32-kDa subunits, eluted slightly after the ROS PPase on these columns. Results of the sedimentation velocity experiment are shown in Fig. 7 and indicate a molecular mass of approximately 70 kDa, assuming a globular shape, consistent with ROS PPase being a homodimer of 36-kDa subunits. The distribution of activity in the gradient is broad enough, with enough of a shoulder below the main peak in the gradient, that we cannot rule out the possibility that some higher order aggregates (trimers and tetramers) may be present as well. The mobility of yeast PPase relative to the same standards on the same kind of gradient was consistent with its known molecular weight.

DISCUSSION
While the PPase activity level we observed (up to 2.4 units/ mg soluble protein) in ROS extracts is higher than found in any other mammalian tissue studied, the actual activity inside intact ROS is probably even higher. While our preparations were all derived from frozen retinas, so that a significant fraction of soluble ROS proteins was lost, when PPase activity was measured in ROS prepared from retinas that had not been frozen, and therefore should retain a higher proportion of soluble proteins (Hakki and Sitaramayya, 1990), approximately %fold higher specific activities were found (2.2 units/ mg rhodopsin; the activity of soluble extracts was not reported). Thus bovine rod outer segments contain significantly  higher levels of cytoplasmic PPase than has been found in any other mammalian tissue.
It seems likely that the very high level of PPase activity in ROS is related to their extraordinarily high levels of cGMP metabolism (Fig. 8). Even in the dark, production of 29 p~ cGMP s-' could lead in less than 1 min to accumulation of sufficient PPi to inhibit guanylyl cyclase. However, the ROS PPase activity measured in vitro corresponds to hydrolysis inside ROS of 3-9 mM PPi s-* at saturating [MgPPi] (assuming 3 mM rhodopsin) or to a PPi half-life of 77-350 ps under substrate-limited conditions. This should be more than enough activity to keep up with guanylyl cyclase, even when cGMP production increases 4-fold in the light.
With such high levels of PPase activity, it seems unlikely that PPi or PPase plays a regulatory role unless under some conditions the activity in vivo is much less than estimated by extrapolation from in vitro measurements. One possible mechanism for lowering PPase activity is inhibition by Ca2+, which Cyclase \ GTP -cGMP + PPi would allow for stimulation of PPase along with guanylyl cyclase when Ca2+ levels are reduced after exposure to light. Preliminary studies of the effects of M e and PPi on Ca2+ inhibition2 suggest that the mechanism for inhibition may be similar to that found with yeast PPase (Ridlington and Butler, 1972;Moe and Butler, 1972b). In that case the principal inhibitory species is actually CaPPi, which binds tightly to PPase (& = 7 nM for CaPPase and K d = 100 nM for Mg-PPase), and Ca2+ and CaPPi appear to compete with Mg2+ and MgPPi, respectively, for binding to sites essential in catalysis. While our results are fairly consistent with the effects of Ca2+ on yeast PPase and on rat mitochondrial soluble PPase (Davidson and Halestrap, 1989), our observation of inhibition at submicromolar Ca2+ levels contrasts with other reports for mammalian PPase (Baykov et Yoshida et al., 1982;Hachimori et al., 1983) that indicated much higher concentrations of Ca2+ were needed for inhibition. Although it is possible that there are species-and tissuespecific differences in Ca2+ sensitivity, it is also possible that the discrepancy is due to the fact that previous reports were based on total Ca2+ added rather than on free Ca2+. The similarity of the Ca" sensitivity of PPase in bovine liver and ROS extracts argues against any large tissue-specific differences in calcium sensitivity. Because the inhibitory potency of Ca2+ is a function of [PP,] and [Mg2+], further studies of the quantitative dependence of Ca2+ inhibition on the concentrations of these ions, as well as better estimates of their free concentrations in ROS, will be necessary to determine whether Ca2+ regulation of PPase has any functional significance in ROS.
An unresolved question is whether the ROS PPase is specific for photoreceptors or some larger set of cell types, or is instead identical to a single type of cytoplasmic PPase expressed in all bovine tissues. A number of the properties of the bovine ROS enzyme are somewhat different from those previously reported for mammalian PPases, including its lower K,,, and its greater Ca2+ sensitivity. It is not clear, however, to what extent these differences reflect species differences or different assay methods, as opposed to properties unique to PPase expressed in ROS. Our studies on PPase in other bovine tissues suggest that, while levels of activity expressed vary widely, differences in kinetic properties (e.g. K,,, and Ca2+ sensitivity) if any, may be fairly subtle. One study (Baykov et al., 1989) on bovine heart mitochondrial PPase indicated that its dependence on [PPi] and its inhibition by [Ca"] appear to be quite different from those of the ROS enzyme. There have been reports from three groups Z. Yang and T. G. Wensel, unpublished results. (Thuillier, 1978;Fisher et al., 1974a;Pynes and Younathan, 1967) that the subunits of human erythrocyte PPase are 20-23 kDa, implying significant structural differences between this enzyme and other eukaryotic PPases, all of which have been reported to have subunits of 30-36 kDa. Using the technique of enzyme assays in starch gels, humans and a number of other mammals have been found to have multiple (two to six) electrophoretically distinct PPase isoforms (Fisher et al., 1974b). This technique apparently did not detect a distinct mitochondrial enzyme in humans, although a distinct nuclear gene encoding mitochondrial PPase is presumably present in animals as it is in Saccharomyces cerevisiae (Lundin et al., 1991). It seems likely that mammals in general have more than one functionally important isoform of PPase.
We have detected multiple chromatographically distinct forms of PPase in extracts of total bovine retina but have not yet determined their cellular origins nor compared them functionally to the ROS enzyme. While we have been able to obtain partial amino acid sequence data for one of these isoforms (see our companion study, Yang and Wensel(1992)), the amino terminus of the ROS enzyme was found to be blocked, and we have not yet been able to obtain sufficient amounts of peptides cleaved from it for microsequencing and comparison to other PPase sequences. Therefore it remains to be determined whether there are functionally distinct molecular forms of PPase in the bovine retina or simply a single molecular species which is degraded into multiple products during isolation.