Dephosphorylation of Phosphoproteins and Synthetic Phosphopeptides STUDY OF THE SPECIFICITY OF THE POLYCATION-STIMULATED AND MgATP-DEPENDENT PHOSPHORYLASE PHOSPHATASES*

The substrate specificity of different forms of poly- cation-stimulated (PC&, PC&, and PC&) phosphorylase phosphatases and of the catalytic subunit of the MgATP-dependent protein phosphatase from rabbit skeletal muscle was investigated. This was done, with phosphorylase a as the reference substrate, using the synthetic phosphopeptides patterned after the phosphorylated sites of pyruvate kinase (type L) (Arg,-Ala-Se~-(~~P)-Val-Ala (Sd, and its Thr(32P) substitute (T4)), inhibitor-1 (Arg,-Pr~-Thr(~~P)-Pro-Ala (T5), Arg2-Pr~-Thr(~~P)-Pro-Ala (TI), and its S~I-(~,P) substitute (SI)), and some modified phosphopeptides (Argz-Ala- Thr(32P)-Pro-Ala (T2) and Arg,-Pr~-Thr(~~P)-Val-Ala (T3)),

The substrate specificity of different forms of polycation-stimulated (PC&, PC&, and PC&) phosphorylase phosphatases and of the catalytic subunit of the MgATP-dependent protein phosphatase from rabbit skeletal muscle was investigated. This was done, with phosphorylase a as the reference substrate, using the synthetic phosphopeptides patterned after the phosphorylated sites of pyruvate kinase (type L) (Arg,-Ala-Se~-(~~P)-Val-Ala (Sd, and its Thr(32P) substitute (T4)), inhibitor-1 (Arg,-Pr~-Thr(~~P)-Pro-Ala (T5), Arg2-P r~-T h r (~~P ) -P r o -A l a (TI), and its S~I -(~, P ) substitute (SI)), and some modified phosphopeptides (Argz-Ala-Thr(32P)-Pro-Ala (T2) and Arg,-Pr~-Thr(~~P)-Val-Ala (T3)), all phosphorylated by cyclic AMP-dependent protein kinase. In addition, ~asein(Thr-'~P), phosphorylated by casein kinase-2, was also tested. The PCS phosphatases show a striking preference for the T4 configuration, PCSc being the least efficient. The catalytic subunit of the MgATP-dependent phosphatase was almost completely inactive toward all these substrates. As shown for the PCSA phosphatase, and comparing with T4, the two proline residues flanking the Thr(P) in TI and Ts, just as in inhibitor-1, drastically impaired the dephosphorylation by lowering the V,,, and not by affecting the apparent K,. The C-terminal proline (as in T,) by itself represents a highly unfavorable factor in the dephosphorylation. The critical effect of the sequence X-Thr(P)-Pro or Pro-Thr(P)-Pro (TI, TZ, T5, and inhibitor-1) can be overcome by manganese ions. The additional finding that this is not the case with the Pro-Ser(P)-Pro sequence (SI) suggests that the effect of Mn2+ is highly substrate specific. These observations show the considerable importance of the primary structure of the substrate in determining the specificity of the protein phosphatases.
Four protein phosphatases are presumed to play an important role in dephosphorylating the major proteins involved in the control of general metabolism. Based on the enzymedirected regulation of activity they have been classified as MgATP-dependent, polycation-stimulated, Me-dependent protein phosphatases and calcineurin (1). Among these we kundig Wetenschappelijk Onderzoek and by the Ondenoeksfonds K.
* These investigations were supported by the Fonds voor Genees-U. Leuven. 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.
have studied in more detail the structure and regulation of the MgATP-dependent protein phosphatase (2, 3) and the polycation-stimulated phosphorylase phosphatases (4), the only enzymes in mammalian tissue extracts that possess significant activity toward phosphorylase a (5,6), a substrate of choice because of its single phosphorylation site. It is now generally accepted that the free catalytic subunits of these enzymes do not exist as such in vivo. An inactive, activated, and spontaneously active form of the Mg-ATP-dependent protein phosphatase can be recognized (I), while according to their native apparent molecular weights and substrate specificity four different forms of polycation-stimulated (PCS)' phosphorylase phosphatases have been described (4): PCSH~, PC& PC&, and PC&. Considerable information is available on the potential physiological substrates of the MgATP-dependent phosphatase, most of which have Ser-P as the target for dephosphorylation (5); however, this enzyme is also able to dephosphorylate the threonine residue in the subunit modulator protein during the activation process (7) and Thr-11 and Ser-117 of troponin I at similar rates (8). The different forms of polycation-stimulated phosphorylase phosphatases were only recently purified, but it is clear that dephosphorylation affects Thr-P as well as Ser-P, since, for instance, the PCSH enzyme represents a major inhibitor-1 phosphatase (Thr-P) and is, contrary to the other PCS phosphatases and their catalytic subunit, a highly specific deinhibitor phosphatase (9,10) in which Ser-P is involved. The MgATP-dependent and PCS phosphatases have broad and overlapping substrate specificities (5). However, a few substrates such as phosphatase inhibitor-l, pyruvate kinase, the a-subunit of phosphorylase kinase, and casein phosphorylated on threonine are among the preferred targets for the polycation-stimulated phosphorylase phosphatases (4, 5 ) . This remarkable site recognition is not due to the phosphoamino acid residue itself, since this is a threonine in inhibitor-1 and casein, and a serine in the a-subunit of phosphorylase kinase and pyruvate kinase. Hence, the PCS phosphatases are particularly interesting enzymes for substrate specificity studies, since not only two phosphorylated amino acids are involved, but the subunit structure of the PCS phosphatase seems to determine strongly the specificity of the enzyme.
In the present study, synthetic peptides, corresponding to some physiological substrates of protein phosphatases but containing different phosphorylated amino acid residues as well as variations in the surrounding amino acid sequence, ' The abbreviations used are: PCS, polycation stimulated; MES, 2-(N-morpho1ino)ethanesulfonic acid Ser-P, phosphoserine; Thr-P, phosphothreonine.

Specificity of the PCS Phosphoryhe Phosphatases 1061
have been used to shed more light on the primary site requirements of different protein phosphatases. Such an approach also provides useful information on the importance of the molecular integrity of the physiological substrates as well as on the significance of the subunit structure of the phosphatases and on the effects small ligands may exert on the interaction between enzyme and phosphorylated substrates. Some of the present data have been published in a preliminary form (11).
The synthetic peptides were routinely phosphorylated by incubation with 200 PM [ Y -~~P I A T P (300-400 cpm/pmol), 200 mM MES buffer (pH 6.5), 100 mM MgC12, 1.2 units of the catalytic subunit of cyclic AMP-dependent protein kinase, and 0.1 mg/ml synthetic peptide in a final volume of 250 pl. After 2 h, the reaction was terminated and the (-y-32P]ATP removed by addition of acetic acid and application of the sample onto phosphocellulose paper (23). The peptide extraction from the phosphocellulose paper was done using 3-4 ml of 3 M HCl with a yield of 6040% of the total phosphopeptide adsorbed. The HCl was evaporated and removed by washing the samples twice with water. After drying, the phosphopeptides were dissolved in a buffer of 20 mM Tris-HC1, 0.5 mM dithiothreitol, at pH 7.4. The specific radioactivity of 3zP incorporated into the peptides was assumed to be equal to that of [32P]ATP employed, and the evaluation of the 3*P-labeled peptide concentration was made on this assumption. The phosphorylation varied between 60 and 80% for the hexapeptide Sz, already known as the best substrate for the cyclic AMP-dependent protein kinase (191, between 10 and 40% for T4, and between 5 and 20% for all the other peptides employed SI, TI, Tz, T3, and T6 (Table   I).
Protein Phosphutase Assay Procedures-Dephosphorylation of the phosphopeptides and proteins was performed by incubation of 10-pl aliquots of the phosphorylated substrate at pH 7.4 and the concentrations specified in the figures and tables with the different protein phosphatases in a total volume of 20 pl at 30 "C for 10 min unless indicated otherwise. One unit of protein phosphatase was defined as the amount of enzyme which released 1 nmol of [32P]phosphate/min at 30 "C in a 30-pl assay containing 1 mg/ml [32P]phosphorylase. The reaction was terminated by addition of 1.2 ml of 1:l (v/v) isobutyl alcohol-toluene and 0.8 ml of a solution containing 5 mM silicotungstate and 1 mM HzSO,; the 32Pi liberated was extracted as a phosphomolybdic complex and determined as described in Ref. 25.
When a time course of dephosphorylation was followed the total assay volume was 60 pl. The final concentration of the phosphopeptides was kept at 7 p~, while the concentrations of protein phosphatases are expressed as units of phosphorylase a phosphatase activity. At different time intervals 10-p1 aliquots were taken, and the reaction was stopped as described.

RESULTS
The time course of dephosphorylation of phosphopeptides S p and T4, differing only in the nature of the phosphorylated amino acid, by PC& phosphatase is illustrated in Fig. 1A. It clearly shows the importance of the phosphorylated amino acid on phosphatase activity, since T4 is dephosphorylated at  The concentration, calculated on the basis of (32P)phosphate, was 7 BM for the synthetic phosphopeptides, 1 PM for inhibitor-1, and 2 p~d for casein. Dephosphorylation with the release of less than 1 pmol. min". ml" is indicated as not detectable (ND). MgATP-dependent phosphatase (active catalytic subunit) a rapid rate, reaching 50% within 20 min while dephosphorylation of its phosphoseryl counterpart S2 is almost negligible.
To determine the influence of the amino acid sequence on the efficiency of dephosphorylation by PCSH phosphatase, other synthetic phosphopeptides have been tested (Fig. 1B). Apparently, the replacement of the two amino acids alanine and valine flanking the phosphoamino acid with two proline residues drastically lowered the PCSH phosphatase activity (compare TI with T, and SI with SJ. It should be stressed that to obtain significant dephosphorylation of the phosphoseryl peptides the concentration of the enzyme is 10-fold higher than in the experiment shown in Fig. 1A. The effects of structural modifications of the peptide on the dephosphorylation rate by PC& and PC& phosphatase and by the catalytic subunits of the PCS and the MgATPdependent phosphatases are summarized in Table 11, where the dephosphorylation rates of (32P-Thr)casein and (32P-Thr)inhibitor-1 are also reported. Two main conclusions can be drawn: 1) PCSH phosphatase was the enzyme displaying the highest activity toward all substrates tested; and 2) phosphopeptide T, was by far the best substrate for all the PCS phosphatases, but particularly for PCSH phosphatase. Moreover, the dephosphorylation rate of all phosphopeptides dropped to very low levels with the PCSc phosphatase and became undetectable with the catalytic subunit of MgATPdependent protein phosphatase.
From both Fig. 1 and Table 11, it appears that a structural modification which dramatically impairs the dephosphorylation efficiency is the presence of two proline residues around the Thr-P in Tq such as in TI and T5 as well as in inhibitor- Table I11 for the PCSH phosphatase, such an impairment is accounted for by a much lower V, , , for the

Kinetic constants of the PCS~phosphutase with inhibitor-1 and different synthetic substrates, phosphorylated by cyclic AMP-dependent protein kinase
The kinetic parameters were calculated by double reciprocal plots from initial rates. The dephosphorylation by PC& phosphatase (specific activity, 150 units/mg with phosphorylase a as the substrate) was carried out in the presence of 1 unit/ml (Sz), 1 unit/ml (TI, Tg, and inhibitor-1). and 0.1 unit/ml (TI). peptides TI and T5 and for inhibitor-1 in comparison with the V,,, for T,. Interestingly, the K,,, and VmaX values for the TI, Tg, and inhibitor-1 were quite comparable. Which of the proline residues adjacent to the phosphothreonine represents the unfavorable factor on the phosphatase activity has also been studied. Table I1 shows that the single proline C-terminal to phosphothreonine (as in T2) caused the major loss of phosphatase activity, while the enzyme activity was practically not affected by the presence of the N-terminal proline (as in T3). The structural requirements for threonyl peptide phosphorylation by cyclic AMP-dependent protein kinase (19) and dephosphorylation by PCSH phosphatase were not completely identical; while prolyl residues influenced both processes in a quite comparable manner, the length of the N-terminal basic stretch and the overall structural integrity of inhibitor-1 dramatically improved the phosphorylation efficiency while having no appreciable effect on the dephosphorylation rate.
The effect of Mn2+ and polycations such as protamine on the activity of the phosphatases toward phosphopeptides was also investigated. Table IV shows that Mn2+ increased the phosphatase activity in all cases, but especially toward the phosphopeptides which, because of the presence of proline, either C-terminal or on both sides of the phosphoamino acid, were very poor substrates for all the protein phosphatases tested (e.g. T1, Tz, and TS) (4-%fold increase of the phosphatase activity). In particular, Mn2+ was required to obtain detectable dephosphorylation of phosphopeptides by the catalytic subunit of the MgATP-dependent phosphatase, which otherwise was inactive on these substrates (see also Table 11). These data are consistent with the finding (2) that the active catalytic subunit of the MgATP-dependent phosphatase was able to dephosphorylate inhibitor-1 only in the presence of Mn2+. These results also suggest that, in the case of PCSH phosphatase, a stereospecific interaction between Mn2+ and the sequence Pro-Thr(P)-Pro or at least X-Thr(P)-Pro in the short peptides as well as in inhibitor-1, was responsible for optimizing the phosphatase reaction. The peptide dephosphorylation by all the PCS phosphatases was not affected by protamines (not shown), but these results obtained with small peptides do not exclude the enzyme-directed polycation stimulation seen with phosphorylase a and inhibitor-1.

DISCUSSION
Seven small synthetic phosphopeptides partially related to the phosphorylation site of inhibitor-1 and pyruvate kinase (Table I), phosphorylated by cyclic AMP-dependent protein kinase, have been investigated as substrates for different forms of polycation-stimulated (PCS) phosphorylase phosphatases (PCSH, PC&, and PC&) and the catalytic subunit of the MgATP-dependent protein phosphatase. The results obtained support the following main conclusions: 1) unlike the catalytic subunit of the MgATP-dependent phosphatase, the PCS phosphatases readily dephosphorylated the same peptides with kinetic constants comparable with those of protein substrates; 2) PCSH and PC& phosphatases exhibited

TABLE IV Enzyme activity of the PCSH-and the MgATP-dependent phosphatase toward the SerC2P)-and Thr(JZP)-containing peptides and inhibitor-1 (IJ in the presence or absence of Mn2+ @mol. min" . ml")
For the experimental conditions see Table 11. ND, not determined.  Ts   14  21  121  161  20  171  14  60  511  674  1025  1496  30  114  61  187   ND  5  5  26  ND  29  34  ND   222  14  ND  12 sphoryluse Phosphatases 1063 a remarkable preference for the phosphothreonyl peptide Argz-Ala-Thr(P)-Val-Ala (T4) over its phosphoseryl derivative; 3) the primary structure of the phosphopeptide played a considerable role in determining the phosphatase activity; and 4) Mn2+ ions dramatically improved the dephosphorylation efficiency of just those peptides exactly reproducing the phosphorylation site of phosphatase inhibitor-1. The first conclusion can be drawn from Table 11, showing that while the PCS phosphatases were more or less active toward all the phosphopeptides, the catalytic subunit of the MgATP-dependent phosphatase (at equivalent phosphorylase phosphatase activity) was practically unable to dephosphorylate any one of the phosphopeptides tested. These data support the conclusion that the structural features of our synthetic phosphopeptides were very unfavorable for the MgATP-dependent phosphatase activity. In particular, the negative results with peptides reproducing the phosphorylated site of inhibitor-1 support the concept that this protein as such was not a physiological substrate of the MgATP-dependent protein phosphatase. It is possible that an unfavorable feature shared by inhibitor-1 and all our peptides was the presence of arginine residues starting from the second residue on the N-terminal side of the phosphorylated amino acid. Actually, such a feature cannot be found in the majority of known physiological substrates of the MgATP-dependent phosphatase (5). So Argz-Ala-Ser(P)-Val-Ala (Sz), reproducing the phosphorylated site of rat liver pyruvate kinase (type L), and the homologous threonyl-substituted Arg,-Ala-Thr(P)-Val-Ala (T4) hexapeptides could be used to discriminate between the catalytic subunits of the MgATP-dependent and the PCS phosphatases (PCSc), since under comparable conditions they were dephosphorylated to an appreciable extent only by the latter enzyme.

Phosphatases
Among the different forms of the polycation-stimulated phosphorylase phosphatases, PCSc phosphatase displayed the lowest activity toward the phosphosubstrates tested (see Table 11). Even if the PCS phosphatases share a similar substrate specificity, the highest enzyme activity was shown by the PCSH phosphatase, while the threonyl peptide T4 was the preferred substrate for all PCS phosphatases. The PCSH phosphatase could be resolved by Mono Q fast protein liquid chromatography into two enzyme forms, PCS", and PCSH, phosphatase. Both forms were active toward the deinhibitor protein, the latter containing a 55-kDa subunit in addition to the 62-and 35-kDa components (4). However, no differences were observed with the phosphopeptides using these two enzyme preparations (data not shown). Hence, all the experiments described were done with a PCSH phosphatase preparation.
PCSH phosphatase is known to affect both threonyl and seryl residues (e.g. in inhibitor-1 and the deinhibitor protein, respectively) (4). Therefore, the dephosphorylation of the synthetic peptides T, and T5 (which reproduced the phosphorylation site of inhibitor-1) and phosphoserine, as well as phosphothreonine peptides varying in the nature of the two residues flanking the phosphorylated amino acid, was investigated. The unusual sequence Pro-Thr(P)-Pro found in inhibitor-1 and reproduced in Tl and T5 was very unfavorable for PCSH phosphatase activity, since both phosphopeptides were dephosphorylated much more slowly than T4, which lacked the two proline residues on both sides of Thr(P). The decrease in V, . , , without obvious change in K,, was due to the C-terminal rather than the N-terminal proline, as shown by the fast dephosphorylation of T, compared with T, (Table   11). Interestingly, the slow dephosphorylation of the sequence Pro-Thr(P)-Pro was not improved when intact inhibitor-1 specificity of the PCS Phc was the substrate (Tables I1 and IV), indicating that the integrity of the substrate molecule could not overcome the negative effect due to the C-terminal Pro. Mn2+ ions in general improved the PCSH phosphatase activity, but the effect was most pronounced with the phosphosubstrates which had the critical sequence X-Thr-Pro or Pro-Thr-Pro. The additional finding that the phosphatase activity with peptide SI, having two proline residues surrounding the Ser(P), was not significantly increased indicated that the stimulatory effect of Mn2+ was substrate specific and required a phosphothreonyl residue. The activity of the MgATP-dependent phosphatase in the presence of Mn2+ was more evident with T, than with TI, T5, or inhibitor-1. This could point to different kinds of interaction between the two types of enzyme and the substrates when Mn2+ is present. Using a cardiac protein phosphatase (consisting apparently of 88% PCSc phosphatase and 12% catalytic subunit of the MgATP-dependent phosphatase) Shacter-Noiman and Chock (26) also noticed a complete Mn2+ dependence for the substrate Leu-Arg-Arg-Ala-Ser(P)-Leu-Gly in contrast with the slight Mn2+ stimulation they observed with Leu-Arg-Arg-Ala-Ser(P)-Val-Ala-Gln-Leu. This would suggest that dephosphorylation of a serine-containing peptide could also be stimulated by Mn2+. However, the K, values for these peptides were about 1 order of magnitude higher than those we report now. This difference could be caused by extending the primary structure of S p on both sides or to the use of the catalytic subunit instead of the holoenzymes, which in our studies show a much higher activity. Titanji et al. (27) observed a decrease in K , by extending Leu& with Gln-Leu to the C-terminal, using an ill-defined catalytic subunit preparation. These observations would, therefore, point to the higher specificity of the holoenzymes compared with the catalytic subunit(s).
It would be tempting to assume that a prolyl residue adjacent to the C terminus and to a lesser extent the N terminus of phosphothreonine represents a natural device for restricting the specificity of PCS phosphatases. Dephosphorylation of some of its potential substrates could then be conditioned by effectors such as Mn2+ or by endogenous compounds capable of mimicking the effect of this cation. A prolyl residue adjacent to the C-terminal side of the target one has been shown to prevent also the fast phosphorylation of synthetic peptides by cyclic AMP-dependent protein kinase (19). However, such a hindrance is obviously overcome in the overall structure of inhibitor-1, which was phosphorylated much more readily than the hexa-and octapeptides reproducing its phosphorylation site. This was not so for the PCS phosphatases, which showed comparable kinetic constants with the phosphopeptides TI and T5 as well as with inhibitor-1 (Table 111). Cyclic AMP-dependent protein kinase obviously prefers seryl peptides over similar threonyl derivatives (19), whereas the PCS phosphatases exhibited a remarkable preference for the phosphothreonyl over the phosphoseryl peptides ( Fig. 1 and Table 11). This finding and the ability of PCSH and PC& phosphatase to dephosphorylate (32P-Thr)casein phosphorylated by casein kinase-2 are intriguingly reminiscent of protein phosphatase-T, a phospho-sphorylase Phosphatases casein phosphatase isolated from rat liver (28,29). Although their chromatographic behavior on DEAE-cellulose, under different experimental conditions, was not the same, the possible identification of protein phosphatase-T with one of the forms of PCS phosphorylase phosphatase should be considered.