[Frontiers in Bioscience 3, d961-972, September 1, 1998]
Reprints
PubMed
CAVEAT LECTOR




Table of Conents
 Previous Section   Next Section

PHYSIOLOGIC IMPORTANCE OF PROTEIN PHOSPHATASE INHIBITORS

Carey J. Oliver and Shirish Shenolikar

Department of Pharmacology and Cancer Biology, Duke University Medical Center, Durham, North Carolina 27710

Received 8/14/98 Accepted 8/20/98

3. PROTEIN PHOSPHATASE INHIBITORS

3.1. PP1 inhibitors

3.1.1. Inhibitor-1

Inhibitor-1 (I-1) was first identified in rabbit skeletal muscle (24), but is widely expressed in mammalian tissues. This thermostable protein (Mr 18,700) is a potent and specific PP1 inhibitor, Ki = 1.6nM, when it is phosphorylated by PKA (25, 26). Structure-function studies show that it is closely related to DARPP-32, a predominantly neuronal PP1 inhibitor. Studies of I-1 also indicate a complex array of interactions with the PP1 catalytic subunit that together result in enzyme inhibition (27). The essential elements in I-1 that mediate PP1 inhibition include the phosphorylation of threonine35 and a tetrapeptide consensus motif found in many PP1-binding proteins (28). I-1 is also phosphorylated in vivo on serine67 by an unknown protein kinase (29). However, the functional role of this phosphorylation remains unknown.

The role of I-1 in regulating PP1 function has been investigated in many different physiological settings. These include the hormonal control of glycogen metabolism, synaptic plasticity controlled by neurotransmitters, growth of pituitary tumor cells and the control of muscle contraction. Some but not all hormones that elevate intracellular cAMP activate I-1. The precise reasons for this remain unclear but may in itself account for some of the differences in physiological effects of these hormones. In any case, hormones that activate I-1 induce much larger and more rapid changes in signal transduction pathways transduced by PP1 substrates.

3.1.1. 1. I-1 and Glycogen Metabolism

Hormones, like adrenalin, which activate PKA result in the increased phosphorylation and inactivation of glycogen synthase in mammalian skeletal muscle. This activation is not due to the direct action of PKA on glycogen synthase. Although PKA can phosphorylate and inactivate glycogen synthase in vitro, adrenalin promotes the phosphorylation of glycogen synthase at serines that are not PKA targets (30). Moreover, the protein kinase, GSK3, that phosphorylates these serines is itself not activated by PKA. Consequently, it was determined that the mechanism for enhanced glycogen synthase phosphorylation was the PKA-mediated activation of I-1 and the resultant inhibition of PP1 (31). This revealed a novel mechanism for hormone action in that PP1 modulation via I-1 mediated cAMP regulation of proteins that are substrates for kinases other than PKA.

Glycogen synthase is regulated by a pool of PP1 that is bound to the glycogen particle. This association is mediated through the glycogen-targeting subunit(s) of PP1. PKA-mediated phosphorylation of the skeletal muscle G-subunit is thought to cause PP1 dissociation from the G subunit and reduce glycogen synthase phosphatase activity. Thus, cAMP may coordinate PP1 translocation from glycogen with the activation of the cytosolic inhibitor, I-1, to achieve rapid phosphatase inhibition. This would also accommodate the in vitro finding that glycogen-bound PP1 was less sensitive to inhibition by I-1 than free PP1 catalytic subunit. Presence of a common PP1-binding motif in I-1 and the G-subunit also predicted that interaction of these proteins with PP1 may be mutually exclusive. Recently, unpublished studies have begun to question the notion of hormone-induced translocation of PP1 from glycogen to cytosol. The full impact of these findings on the hormonal regulation of glycogen-bound PP1 has not been determined.

Other hormones, such as insulin, result in the dephosphorylation and activation of glycogen synthase. Numerous studies have reported that insulin activates PP1 in insulin-sensitive cells (32). In skeletal muscle, a number of mechanisms may combine to activate the glycogen-bound PP1 in response to insulin. For instance, insulin activates cAMP phosphodiesterase to lower cAMP levels. This would result in I-1 inactivation and in itself promote glycogen synthase dephosphorylation. Others have suggested that insulin actively promotes the dephosphorylation of I-1 by activating an I-1 phosphatase. Finally, phosphorylation of the GM-subunit by an insulin-stimulated protein kinase, most likely PKB, also elevates glycogen synthase phosphatase activity. Combined effects of the reduced kinase and increased PP1 activity promote glycogen synthase dephosphorylation and the storage of excess blood glucose as glycogen in skeletal muscle. A number of GM homologues have been identified in other tissues (33) and unlike the GM-subunit, some of these are not subject to regulation by insulin (34). This potentially makes I-1ís role in hormonal control of glycogen metabolism even more important in these tissues.

3.1.1. 2. I-1 and Synaptic Plasticity

Changes in synaptic transmission elicited by prior neuronal activity, have been extensively studied in the hippocampus as a potential model for learning and memory. Activity-dependent enhancement of synaptic transmission is seen as long term potentiation or LTP. Yet other stimuli depress the functions of hippocampal synapses leading to long term depression or LTD. Several protein kinases, including CaM-Kinase II, PKC (35) and fyn (36), have been implicated in LTP. Based on their electrophysiological characteristics, LTD appeared to be the converse of LTP and may therefore involve protein phosphatases. Recent studies established that two protein phosphatases, PP1 and PP2B, acted in tandem to regulate LTD (18). The PP1 inhibitor, I-1, is highly expressed in the hippocampus and functions as the link between PP2B and PP1. This demonstrated a different use of the phosphatase cascade first suppressed to control glycogen metabolism in skeletal muscle.

Interestingly, LTP and LTD are both activated by the excitatory neurotransmitter, glutamate, acting via the second messenger, calcium. So an important question has been how a single second messenger mediated such opposing physiological effects. The answer appears to lie in the different sensitivities of the calcium/calmodulin activated protein phosphatase (PP2B) and kinase (CaM-Kinase II) in neurons. It has been speculated that low calcium levels activate PP2B and induce LTD while higher concentrations of calcium activate CaMKII which is required to trigger LTP. However, LTP and LTD share many common characteristics, suggesting that they also share many signaling components. In this regard, I-1 appears to be involved in both LTP and LTD. PKA inhibitors abolish sustained postsynaptic LTP. These effects were reversed by phosphatase inhibitors, specifically those that target PP1. Subsequently, it was shown that LTP-generating stimuli also activated PKA which in turn activated the endogenous PP1 inhibitor, I-1, in hippocampal neurons (37). Thus, the current model is that LTP-generating stimuli result in the activation of both CaM-Kinase II and calcium/calmodulin-stimulated adenylyl cyclase. The role of cAMP in the LTP pathway is to activate I-1 and suppress PP1 activity, which reverses the functions of CaM-Kinase II. Hence, I-1 functions as the gatekeeper which determines whether the neuron will transition from early to intermediate or late phases of LTP. In contrast to LTP, LTD results from the small influx of calcium, which is unable to activate either CaMKII or adenylyl cyclase, and, the predominant signal transducer for LTD is PP2B or calcineurin. The subcellular organization of signaling molecules may also be an important contributor to the unique specificity of the LTP and LTD responses.

3.1.1. 3. I-1 and Cell Growth

Genetic studies in fungi implicated a critical role for PP1 in the cell cycle, particularly in the transition through mitosis. Further evidence in Xenopus laevis oocytes, which lack the G1 phase of the cell cycle, showed that PP1 inhibitors inhibited cell division or oocyte maturation (38). Oocytes stalled in mitosis show high levels of histone H1 phosphatase activity and are unable to activate the cyclin-dependent protein kinase, Cdc2, or MPF, the maturation promoting factor. Introduction of PP1 inhibitors not only lowered the histone phosphatase activity in these oocyte extracts but also activated MPF (39). This strongly hints at the presence of regulated PP1 inhibitors, like I-1, in the timely activation of MPF and completion of cell division.

Many growth factors activate PP1 (32), but the role for PP1 activation in cell growth remains unknown. In this regard, PP1 was required for growth of the cultured pituitary tumor cell line, GH4D1. cAMP and cyclosporin A, both antiproliferative agents, were shown to enhance the phosphorylation of I-1 in these cells. These studies provided the first evidence for I-1 as a growth regulator and suggested that chronic I-1 activity, induced by PKA activation or calcineurin (PP2B) inhibition blocked cell growth (40). These results also indicated that PP1 should not always be considered as a negative growth regulator (as discussed for RB dephosphorylation) but that it may also provide positive signals for growth by promoting other phases of the cell cycle.

3.1.1. 4. I-1 and Muscle Contraction

The b-adrenergic agonist, isoproterenol, has a positive inotropic effect in the heart. Isoprotenol increases I-1 phosphorylation and results in the inhibition of PP1 activity in the myocardium (41, 42). This decrease in PP1 activity enhances cardiac contractility by preventing the dephosphorylation of proteins such as the Na/K-ATPase, phospholamban, troponin I and voltage sensitive calcium channels which are all involved in maintaining the contractile state of heart muscle (43).

On the other hand, smooth muscle PP1 is tightly bound to myosin and is the primary myosin phosphatase (44). Dephosphorylation of myosin light chains results in the relaxation of smooth muscle. Yet, hormones, such as epinephrine, that increase intracellular cAMP and lead to I-1 phosphorylation, promote the relaxation of smooth muscle. This paradox was resolved by the finding that smooth muscle myosin light chain kinase (MLCK) is phosphorylated at an inhibitory site in response to cAMP. The dephosphorylation of MLCK at this site is also mediated by PP1. Thus, it has been proposed that hormonal activation of I-1 leads to the predicted inhibition of PP1 that acts on MLCK. Thus, MLCK is inactivated and smooth muscle relaxation ensues. Unlike CaM-kinase II which is inactivated by PP1, smooth muscle MLCK, another calmodulin-regulated enzyme, requires PP1 activity to remain in its active state. Thus, I-1 functions in opposing ways to control protein kinases in neurons and smooth muscle. This model also predicts that the myosin-bound PP1, which dephosphorylates myosin light chains to relax smooth muscle, is not regulated by I-1. This may be consistent with the presence of the common PP1-binding motif in I-1 and the 110 kDa myosin-targeting subunit of PP1. Alternately, the interplay between I-1 and the myosin and MLCK phosphatases (both PP1) may modulate the rates of protein phosphorylation and dephosphorylation that set the contractile tone of different smooth muscle beds.

3.1.2. DARPP-32 (dopamine and cAMP-regulated phosphoprotein of apparent Mr 32,000)

DARPP-32 is highly homologous to I-1 near its N-terminus and like I-1, it inhibits PP1 activity only after phosphorylation on threonine34 by PKA (45). DARPP-32 is found mainly in the brain, especially in the basal ganglia (46), but it is also expressed in adipose tissue (47) and to a much lesser extent, in the kidney (48). Recent reports suggest that DARPP-32 is myristoylated at its N-terminus, which mediates its association with membranes. Although originally identified by its enhanced phosphorylation in response to dopamine, DARPP-32 is activated by many hormones and neurotransmitters that modulate cAMP levels (49). Interestingly, some cells in the brain and kidney express both I-1 and DARPP-32. However, DARPP-32 shows a more complex mode of regulation than I-1. For instance, DARPP-32 is phosphorylated in vivo by both PKA and casein kinase I. Phosphorylation by casein kinase I impairs the turnover of phosphate at the activating site, threonine34 (50). Indeed, two different type-2 phosphatases, PP2B acting on threonine34 and PP2C dephosphorylating the casein kinase I sites, may regulate DARPP-32 function. This led to the speculation that the two PP1 inhibitors responded differently to physiological stimuli and may be used in different ways to control PP1 activity in these cells. In brain, dopamine and glutamate have antagonistic effects on the excitability of neurons, possibly mediated by their opposing effects on DARPP-32. Dopamine acting at D1 receptors activates adenylyl cyclase, and through PKA, activates DARPP-32 (45). Glutamate, on the other hand, working through NMDA receptors, reverses DARPP-32 activity. Thus, like I-1, increases in intracellular calcium promote the dephosphorylation of DARPP-32 in the nervous system (51). Unfortunately, unlike I-1, there is only limited information on the physiological role of DARPP-32 in the brain and other tissues.

3.1.2. 1. DARPP-32 and Na/K-ATPase

In the kidney (48) as well as in the brain (45), activation of the D1 receptors by dopamine results in the blockade of Na+/K+-ATPase activity which in turn causes vasodilation and increased natriuresis. DARPP-32 immunoreactivity has been shown in the thick ascending limb of the loop of Henle (52). Use of DARPP-32 phosphopeptides showed that PP1 inhibition increased the phosphorylation of Na+/K+-ATPase. This resulted in increased Na+ excretion as the Na+/K+-ATPase no longer pumps Na+ from the urine back into the blood stream. Vasodilation occurs as water flows into the kidney to maintain the ion gradient (53). Thus, DARPP-32 appears to be a critical component of the control of salt balance in the mammalian kidney.

3.1.2. 2. DARPP-32, lipogenesis and lipolysis

DARPP-32, not I-1, is present in 3T3-L1 adipocytes and in these cells, it is required for adipogenesis. Active DARPP-32 maintains low PP1 activity in differentiated fat cells and this is essential for insulin to stimulate PP1 activity and facilitate triglyceride biosynthesis (34). These studies suggest that, like I-1 in other tissues, changes in DARPP-32 activity may be an important avenue for insulin signaling. cAMP-mediated activation of DARPP-32 may be critical for amplifying the phosphorylation events that lead to activation of hormone-sensitive lipase and enhanced lipolysis.

3.1.3. Inhibitor-2

Inhibitor-2 (I-2) was isolated, along with I-1, as a heat-stable protein from skeletal muscle extracts and specifically inhibits PP1 activity with a Ki = 3.1 nM (24). Unlike I-1 or DARPP-32, I-2 does not need to be phosphorylated to inhibit PP1. I-2 forms a stable and inactive complex with the PP1 catalytic subunit. This inactive complex was isolated as an ATP-Mg-dependent phosphatase which, as the name implies, was inactive until incubated with ATP-Mg. Activation of the latent complex is accompanied by the phosphorylation of I-2 on threonine72 by GSK-3. I-2 is also phosphorylaed on three serines by casein kinase II (CK2). Phosphorylation by CK2 does not alter I-2 activity but greatly facilitates the subsequent phosphorylation by GSK-3 (54, 55). The activation cycle for the ATP-Mg-dependent phosphatase is complicated and somewhat controversial. The fundamental aspect of this cycle is that I-2 phosphorylation by GSK-3 promotes a conformational change in the PP1/I-2 complex, which does not by itself activate the enzyme. The slow autodephosphorylation of I-2 correlates best with the increase in PP1 activity. Throughout this process, I-2 remains bound to PP1 and in a longer time frame the complex relaxes back to its original inactive conformation. Recent studies show that multiple domains mediate the rapid and reversible inhibition of PP1 and the slower inactivation to the stable complex that can be reactivated by GSK-3 (56). Interestingly, the active PP1/I-2 complex can itself be inhibited by addition of exogenous I-2. Against this background of complex interactions of I-2 with PP1, its role as a PP1 inhibitor has been questioned (57).

Recent in vitro studies suggest that I-2 interacts with denatured PP1 to promote a rapid and effective refolding of this protein and yield an active enzyme. In support of this, recombinant PP1, which differs from native PP1 in significant ways, behaves much more like the native enzyme after incubation with I-2 and reactivation by GSK-3 (58, 59). This has led to the proposal that I-2 is a chaperone for PP1. This hypothesis may also be consistent with preliminary reports that overexpression of I-2 in mammalian cells does not inhibit cellular PP1 activity but in fact elevates the cellular content of PP1 catalytic subunit. Studies with Glc8, the yeast homologue of I-2, came to a more complex conclusion, suggesting that under different circumstances, Glc8 was either an inhibitor or an "activator" of the yeast PP1 (60). So, despite extensive studies the physiological functions of I-2 remain elusive.

Perhaps the most intriguing property of I-2 is that its protein and mRNA levels fluctuate during the cell cycle, peaking twice, at S phase and mitosis (5). Recent experiments used I-2 fused to the green flourescent protein (GFP) to show that I-2 was cytosolic during G1 and translocated to the nucleus in S phase (61). Moreover, phosphorylation at four sites was necessary for nuclear translocation as mutations at any one serine prevented nuclear entry. These studies also identified a putative nuclear localization sequence in I-2 and showed that mutations of two lysines in this sequence abolished nuclear localization of I-2. Changes in the phosphorylation of I-2 during G1 and S phase also argued for a role of I-2 phosphorylation in its subcellular localization. As I-2 localization failed to correlate with PP1 distribution in cells, this has also raised questions about I-2ís role as a PP1 regulator, but other functions for I-2 have not been identified.

3.1.3. 1. I-2 and Sperm Motility

The only physiological role proposed for I-2 is in the control of sperm motility. The testis-specific isoform, PP1g2, forms an inactive complex with I-2. The increased PP1 activity seen in nonmotile immature sperm was accredited to an elevation in GSK-3 activity which activates the PP1/I-2 complex. Incubation of immature sperm with the phosphatase inhibitors, okadaic acid and calyculin A, induced motility (62) suggesting that I-2 inhibits PP1 activity in mature mammalian sperm to facilitate their motility.

3.1.4. NIPP-1 (nuclear inhibitor of PP1)

NIPP-1 was originally identified as two PP1 inhibitory polypeptides of 16 and 18 kDa in the particulate fraction of bovine thymus nuclei (63). Subsequent studies showed that the full length NIPP-1 cDNA encoded a protein of 38.5 kDa that was extremely sensitive to proteolysis. NIPP-1 is perhaps the most potent PP1 inhibitor thus far identified with a Ki in the picomolar range. The protein is largely nuclear in its localization and is ubiquitously expressed in mammalian tissues (64). Although NIPP-1, like I-1 and DARPP-32, possesses the consensus PP1-binding motif, it is more like I-2 in its action, being a more potent phosphatase inhibitor in the dephosphorylated state. Moreover, NIPP-1, like I-2, forms a stable, inactive complex with PP1. Following the phosphorylation of NIPP-1 by PKA on a site within the PP1-binding motif, itís activity as a PP1 inhibitor is dramatically reduced.

NIPP-1 is phosphorylated at serine-178 and serineĖ199, by PKA and on threonine-161 and serine-204 by CK2 (65). Both kinases reduce the affinity of NIPP-1 for PP1 and increase phosphatase activity. The effects of the two kinases in activating the NIPP-1/PP1 complex are synergistic, but do not cause dissociation of the complex. It has been speculated that NIPP-1 phosphorylation prevents its reassociation with PP1 (64).

3.1.4.1. NIPP-1 and RNA Metabolism

No clear physiological role for NIPP-1 has been identified but one clue comes from the observation that NIPP-1 contains a RNA-binding motif also found in the bacterial RNA processing enzyme, Ard-1. The extensive sequence homology between mammalian Ard1 and the carboxy-terminus of NIPP-1 suggests that the two proteins are derived from the same gene through alternate splicing (64). Pursuant to this, recombinant NIPP-1 was shown to bind to RNA and its C-terminal fragment was shown to possess endonuclease activity. While the RNA-binding was unaffected by the presence of PP1, the full-length NIPP-1 did not degrade RNA (66). Whether NIPP-1 mediates PP1ís association with RNA is not certain, but other PP1-binding proteins (67) have been implicated in PP1ís function to control spliceosome assembly and RNA splicing. At this time, the physiological role of NIPP-1 as a PP1 inhibitor has not been established.

3.1.5. RIPP-1 (ribosomal inhibitor of PP1)

RIPP-1 is a 23 kDa basic polypeptide that is complexed with PP1 in rat liver ribosomes. In vitro, RIPP-1 is a potent inhibitor of PP1 (Ki = 20 nM) with some substrates, phosphorylase a and myelin basic protein, but a much poorer inhibitor (Ki = 400nM) with other substrates, histone IIA and casein. In addition, RIPP-1 seems to be able to inactivate PP1. Incubation of PP1 with RIPP-1 for 45 minutes at 25į C converted PP1 into a less active enzyme that could not be reactivated by proteolytic degradation of RIPP-1. This inactivation of PP1 is reminiscent of I-2 except that following phosphorylation by GSK-3, I-2ís effects are reversed. At this time, no mechanism for reversal of PP1 inactivation by RIPP-1 has been demonstrated. Few experiments have been done thus far to elucidate RIPP-1ís physiologic role. However, RIPP-1 inhibits PP1-mediated dephosphorylation of ribosomal S6, a component of the 40 S ribosomal subunit and suggests a role in the control of protein synthesis (68).

3.1.6. CPI17 (C-kinase activated PP1 inhibitor, apparent Mr 17,000)

CPI17 is the newest member of the PP1 inhibitor family. It was isolated from porcine aorta smooth muscle as a 17 kDa PKC substrate that potently inhibits PP1 activity (69). The mRNA for CPI17 is expressed exclusively in smooth muscle, such as aorta and bladder, but not in skeletal muscle or non-muscle tissues. When phosphorylated by PKC, CPI17 shows a Ki = 0.18 nM for PP1 and a Ki = 1.3 mM in its dephosphorylated state. An unusual aspect of CPI17 is that it inhibits the myosin-bound PP1 complex as well as the free PP1 catalytic subunit (70). This implies that it functions differently from I-1 and DARPP-32 and suggests an important role for CPI17 in the control of smooth muscle contractility.

3.2. PP2A inhibitors

3.2.1. I1PP2A

I1PP2A was purified from bovine kidney as a 30 kDa heat stable protein that specifically inhibited PP2A with a Ki = 30 nM (17). Subsequent cloning of this protein showed it to be the bovine homologue of the human putative histocompatibility leukocyte antigen class II associated protein-1 (PHAP-1) (71). PHAP-1, a protein of unknown function, was thought to be involved in the immune response as it is bound to the C-terminal region of the DR2a chain of MHC class II receptors (72). I1PP2A is highly acidic at its C-terminus and it has been hypothesized that this region mediates PP2A inhibition as the other PP2A inhibitor, I2PP2A, also has a highly acidic tail (71, 73). Little else is known about I1PP2A in terms of PP2A regulation, but it is interesting to note that I1PP2A has recently been identified as a target for granzyme A. T-lymphocytes trigger apoptotic death in target cells by releasing cytotoxic granules, containing perforin, a pore-forming protein, and several granzymes, which are serine proteases. A putative substrate trapping mutant of granzyme A identified I1PP2A and I2PP2A as binding proteins (74). Thus, it has been inferred that degradation of PP2A inhibitors may be an important signal for apoptotic cell death.

3.2.2. I2PP2A

I2PP2A was also identified as a heat stable PP2A inhibitor from bovine kidney (17). I2PP2A is a 39 kDa protein that is a homologue of the human SETa protein, also called PHAP-II. I2PP2A inhibits PP2A with a Ki = 2 nM (73). Preliminary studies show that I2PP2A is phosphorylated in vivo at two serines near its N-terminus, but the functional consequences of these phosphorylations remain unknown. Subcellular localization shows that I2PP2A is largely nuclear. A chromosomal translocation that leads to the fusion of a nuclear porin, NUP214, with I2PP2A has been linked to a variety of leukemias and may provide new clues to the physiological functions of I2PP2A .

3.3. Endogenous PP2B Inhibitors

PP2B is the recognized target of the two immunosuppressive drugs, cyclosporin and FK506, commonly used in organ transplantation. While cyclosporin only inhibits PP2B in conjunction with a variety of cyclophilins, recent studies suggest that PP2B associates with cyclophilin in the absence of drug (20). This has led to the speculation that cyclophilins and other immunophilins may be natural regulators of PP2B. Recently, a direct endogenous inhibitor of PP2B has been identified. Cain is a 240 kDa protein which inhibits PP2B with a Ki = 0.44 uM, making it the most potent endogenous inhibitor of PP2B so far. Co-localization studies show that both calcineurin and cain are found in the brain exclusively in neurons. In addition, it has been shown that cain binds to PP2B at a site distinct from that of FK506/FKBP12 binding, thus, identifying a novel regulatory site on PP2B (21). PP2B also associates with the PKA-binding protein, AKAP79 (22) and this association reduces the phosphatase activity. Whether this qualifies AKAP79 as a phosphatase inhibitor is not yet clear, but fragments of AKAP79 that mediate PP2B binding have been shown to suppress PP2B functions in intact cells. Finally, PP2B associates constitutively with its T-cell substrate, NFAT. It is now well established that PP2B dephosphorylates NFAT and thereby facilitates its entry into the nucleus to initiate gene transcription. Introduction of the PP2B-binding region of NFAT into cells inhibited the transcription of cytokine genes (23). Thus, one scenario is that NFAT association restricts PP2B functions solely to regulate NFAT or, alternately, additional signals are required to activate the NFAT-bound PP2B and promote rapid dephosphorylation of the transcription factor. Future studies will shed more light on the role of PP2B-binding proteins in the function of this phosphatase.