Erythroid Membrane-bound Protein Kinase Binds to a Membrane Component and Is Regulated by Phosphatidylinositol 4 , 5-Bisphosphate ”

In the erythrocyte, a membrane-bound serinetthreonine protein kinase (a casein kinase) has been shown to phosphorylate a number of membrane proteins, modulating their function. Here we report that the membrane-bound protein kinase binds to membranes by an association with a minor membrane component contained in preparations of glycophorin (possibly a minor glycophorin). The binding of the kinase to glycophorins does not significantly modify kinase activity. However, upon binding, the kinase activity is potently inhibited by phosphatidylinositol 4,5-bisphosphate, and the affinity of the kinase for the glycophorins is increased. Other phospholipids or polyanions such as inositol 1,4,5-trisphosphate or 2,3-diphosphoglycerate do not affect protein kinase activity when the kinase is bound to membranes but do inhibit the solubilized membrane-bound kinase. In the erythrocyte, there is a cytosolic form of the casein kinase which is very similar, having the same molecular weight and substrate specificity as the membranebound casein kinase. The cytosolic casein kinase is inhibited by 2,3-diphosphoglycerate but much less so by glycophorin preparations containing phosphoinosito1 4,5-bisphosphate. When the sequences of both casein kinases were compared by two-dimensional peptide mapping, it was found that the two kinases were very similar but not identical.

In the erythrocyte, a membrane-bound serinetthreonine protein kinase (a casein kinase) has been shown to phosphorylate a number of membrane proteins, modulating their function.
Here we report that the membrane-bound protein kinase binds to membranes by an association with a minor membrane component contained in preparations of glycophorin (possibly a minor glycophorin).
The binding of the kinase to glycophorins does not significantly modify kinase activity. However, upon binding, the kinase activity is potently inhibited by phosphatidylinositol 4,5-bisphosphate, and the affinity of the kinase for the glycophorins is increased. Other phospholipids or polyanions such as inositol 1,4,5-trisphosphate or 2,3-diphosphoglycerate do not affect protein kinase activity when the kinase is bound to membranes but do inhibit the solubilized membrane-bound kinase.
In the erythrocyte, there is a cytosolic form of the casein kinase which is very similar, having the same molecular weight and substrate specificity as the membranebound casein kinase. The cytosolic casein kinase is inhibited by 2,3-diphosphoglycerate but much less so by glycophorin preparations containing phosphoinosi-to1 4,5-bisphosphate.
When the sequences of both casein kinases were compared by two-dimensional peptide mapping, it was found that the two kinases were very similar but not identical.
The turnover of phosphoryl groups on membrane proteins and especially the inositol lipids represents a large fraction of the total metabolic energy expended by the erythrocyte. Phosphorylation of membrane skeletal proteins has been studied in detail, demonstrating that a number of associations are regulated by protein phosphorylation (1). However, at the cellular level there appears to be little convincing evidence that protein phosphorylation has any effect on mature erythrocyte function, such as maintaining the discocyte shape (2). Certainly, ATP is required for maintaining the discoid shape since depletion of ATP leads to the spiculated echinocyte shape (3). However, the only link so far established between cell shape and phosphorylation appears to be phosphorylation of the inositol lipids. Indeed, a reduction in the membrane content of the polyphosphoinositides, in particular phospha-tidylinositol4,5bisphosphate (PIP&' appears to be linked to the formation of the echinocyte shape (4, 5).
In the erythrocyte, the function of the rapid turnover of phosphoryl residues on the inositol lipids is unknown and perplexing.
In other cells, the polyphosphoinositides act as precursors to the second messengers inositol 1,4,5-trisphosphate and 1,2-diacylglycerol, which cause calcium release from internal stores and activate protein kinase C, respectively. In erythrocytes, the polyphosphoinositides do not act as second messenger precursors. Indeed, the cleavage of the polyphosphoinositides by phospholipase C, which is stimulated by calcium, is irreversible since the inositol lipids cannot be synthesized in the erythrocyte. As such, if the turnover of phosphoryl residues on the inositol lipids is important for the normal function of the cell (which appears to be the case), then the polyphosphoinositides must have their impact on the plasma membrane. Examples of regulation of this kind have been identified; in nucleated cells, both profilactin and gelsolin associate with PIP2 on the plasma membrane.
In the erythrocyte, protein 4.1 requires PIP2 for a high affinity interaction with the membrane (5). In this case, a class of membrane proteins seems to interact specifically with PIPS (6-9), and this results in high affinity binding to protein 4.1 on the plasma membrane. This class of transmembrane glycoproteins, glycophorins, is structurally similar and sequence homologous (10)(11)(12)(13). These transmembrane proteins have properties unique from most integral membrane proteins in that once extracted from membranes they are soluble in aqueous buffers in the absence of detergent. This property results from formation of a very stable protein-phospholipid (PL) micelle structure containing 15-20 glycophorin and 40-70 PL molecules (5-9, 14-16). Previously, it has been demonstrated that this structure retains transmembrane proteins in a biologically active state (5). Finally, when glycophorin-PIP, micelles were covalently linked to Sepharose CL-4B, this was an effective affinity matrix for purifying protein 4.1 The erythrocyte membrane-bound protein kinase (a casein kinase) phosphorylates a number of membrane proteins, modulating their function. These include spectrin, ankyrin, adducin, proteins 3, 4.1, 4.9, and other proteins; the kinase is also autophosphorylated (1,(18)(19)(20)(21). The phosphorylation of ankyrin and protein 4.1 by the membrane-bound protein kinase appears to lower the affinity of their interactions with spectrin, possibly resulting in disassembly of the membrane skeleton (18,19). The phosphorylation of ankyrin reduces its affinity for spectrin tetramers and oligomers (20). Phosphorylation of protein 4.1 reduces the affinity of the spectrinprotein 4.1 association (18). The effect this has on the ability of protein 4.1 to promote spectrin-actin associations is unknown.
Erythrocytes contain a counterpart of the membrane-bound protein kinase which is cytosolic (22,23). The cytosolic casein kinase has an identical molecular mass (33)(34) and has the same substrate specificity (18-24). Indeed, the cytosolic and membrane-bound protein kinases have been used interchangeably when studying the effects of protein kinase phosphorylation on membrane proteins (18,20,21,24). Although the effects of several protein kinases on the function of erythrocyte membrane proteins have been studied, most of these protein kinases have no apparent mechanism for activation intrinsic to the erythrocytes (1). However, the cytosolic casein kinase does appear to be regulated by the concentration of 2,3-DPG (23), an important modulator of erythrocyte function.
In the case of the membrane-bound protein kinase, the mechanism of regulation is not known. Here we report that the membrane-bound protein kinase associates with a membrane component isolated in preparations of glycophorin (possibly a minor glycophorin) and is inhibited by PIP2. The cytosolic protein kinase, although structurally similar, does not bind to membranes and does not interact with glycophorin-PIP*. (a) addition of 4 x concentrated SDS-PAGE sample buffer (28), the solution was then applied to SDS-PAGE; after fixing and staining the gel, the spectrin bands were excised and counted; (b) addition of trichloroacetic acid to 5% and isolation of the 32P-labeled proteins by precipitation onto filter paper according to Tao et al. (19). For quantitation, the gel slices were first solubilized in 0.5 ml of TS-2 tissue solubilizer (RPI) and then counted in a flcounter with 2 ml of scintillation fluid. The 32P on the filter papers was also counted in 2 ml of scintillation fluid. One unit of kinase activity was defined according to Tao et al. (19). Glycophorin Micelles, Liposomes, and Affinity Chromatography-Glycophorin was isolated by the lithium diiodosalicylic-phenol method (29); the intrinsic phospholipids were extracted, and glycophorin was desialated as before (5,7,27). For the chromatographic separation of the glycophorins, the crude lithium diiodosalicylicphenol preparation of the glycophorins was applied to a 2.5 X lOOcm column containing Bio-Gel A-l.5 as described previously (14,30 The IOV pellets were resuspended in isotonic KC1 with 0.01% Triton X-100; 20 pg of spectrin was added as substrate, and phosphorylation was started as above. The total "'P incorporation into spectrin in 15 min was used to determine kinase activity. Incorporation of 'IpP into band 3 and ankyrin was quantitatively the same as into spectrin, indicating that substrate effects are not involved. Cellulose Peptide Mapping of Cytosolic and Membrane-bound Casein Kinase I-Membrane-bound and cytosolic casein kinase were applied to a 7-15s SDS-PAGE, fixed, and stained with Coomassie Blue. The stained bands corresponding to the protein kinases were excised and labeled with ivBI as before (28,32). The electrophoresis and thin-layer chromatography (TLC) were as before (28), but TLC plates were glass backed with a O.l-mm thickness of cellulose (EM reagents). 100 cpm of '""I/dalton of protein molecular mass was loaded/plate.

Affinity
Chromatography of the Membrane-bound Protein Kinase on Glycophorin-Affi-Gel15-In the process of studying the phosphorylation of protein 4.1 and its interaction with membranes, it was found that a protein kinase copurified with protein 4.1 (33). In an attempt to separate protein 4.1 and the kinase activity, the isolated protein 4.1 was applied to a glycophorin-Affi-Gel 15 column similar to that described in Fig. 1. The protein kinase was found also to bind to the affinity column. However, this association was not because of an interaction with protein 4.1 since the two proteins could be separated by Sephadex G-100 column; protein 4.1 eluted close to the void volume (-80 kDa) and was completely separated from protein kinase, which eluted with an apparent molecular mass of -35 kDa (results not shown). This suggests that the protein kinase binds directly to the glycophorin column.
To demonstrate more clearly that the protein kinase does not require protein 4.1 for binding to glycophorin-Affi-Gell5, the protein kinase was extracted from intact erythrocyte membranes with 0.5 M NaCl at 0 "C, conditions that do not extract protein 4.1. This extract, containing a large number of proteins, was applied to a glycophorin-Affi-Gel 15 column (see "Experimental Procedures") and eluted with 1 M KCl. When the fractions were assayed for protein kinase activity, using casein or spectrin as substrate, a l-ml column (0.1 mg/ ml of glycophorin) retained 82% of the protein kinase activity when 0.2 unit of protein kinase activity was applied (average of three experiments).
Analysis of fractions by SDS-PAGE showed that four proteins were retained by the column: two bands (100 and 105 kDa) that are the protein adducin (l), a 55-kDa band that was not identified, and a 33-34-kDa doublet. The 33-34-kDa doublet has the same molecular mass as the kinase that copurifies with protein 4.1. A characteristic of many protein kinases is autophosphorylation.
To determine if the 33-34-kDa bands were phosphorylated (perhaps autophosphorylated), an aliquot of each fraction was incubated with 50 pM [-y-"'P]ATP (300 Ci/mol) and applied to a 7-15% acrylamide SDS-PAGE; the autoradiograms of the SDS-PAGE of the 0.5 M NaCl extract from membranes, the flowthrough from the glycophorin column, and the 1 M KC1 elution are shown in Fig. 1. Kinase activity toward other proteins in the 0.5 M NaCl extract is largely removed by the glycophorin column (Fig. 1, lanes A and B). The 1 M KC1 elution shows phosphorylation of only two bands, the lOO-kDa subunit of adducin and a 34-kDa band that corresponds with the protein kinase activity.
A membrane-bound (MB) protein kinase has been purified previously by Tao et al. (19) which has characteristics indistinguishable from those of the protein kinase that binds glycophorins.
To determine if the protein kinase retained by glycophorin-Affi-Gel is similar to the MB protein kinase, we have purified the MB protein kinase by the method of Tao et al. (19). When compared, the protein kinase that binds to glycophorin-Affi-Gel and the MB protein kinase were found to be indistinguishable.
Both kinases are eluted from membranes by high ionic strength, phosphorylate casein and spectrin preferentially, bind to and are eluted from DEAE-cellulose and phosphocellulose at the same ionic strength, are doublets on SDS-PAGE with molecular masses of 33 and 34 kDa, and are autophosphorylated (Ref. 19 and Fig. 1). When the protein kinase is eluted from the glycophorin affinity column and compared with the MB protein kinase by onedimensional peptide mapping (32), the kinases have identical "P-peptide maps (Fig. 1, lanes E-L). By these criteria, the protein kinase that binds to the glycophorin affinity column is indistinguishable from the MB protein kinase isolated by Tao et al. (19) and therefore will be referred to as the MB protein kinase.
These results suggest that the MB protein kinase associates with a glycophorin or a component within the glycoprotein-PL micelle and perhaps binds to the membrane by an association with this same component.
To the ability of glycophorin preparations retaining intrinsic PL to compete with inside-out erythrocyte membrane vesicles for kinase binding was measured. To assay for the MB protein kinase, spectrin or casein was used as substrate. Spectrin and casein are substrates that are specifically phosphorylated by the MB protein kinase, not by CAMP-dependent protein kinase, which is also known to be associated with human erythrocyte membranes (for review, see Ref. 1). Indeed, the only protein kinase associated with erythrocyte membranes which phosphorylates spectrin is the MB protein kinase; thus, spectrin phosphorylation is a selective method to quantitate MB protein kinase.
Competition of MB protein kinase binding to IOVs by glycophorin micelles was measured by combining IOVs and micelles; after an incubation (see "Experimental Procedures"), the IOVs were sedimented through a 10% sucrose cushion. Using this method, the IOVs were separated completely from the glycophorin-PL micelles as demonstrated previously (5). Separation was also checked by applying the IOVs to an SDS-PAGE and staining with silver. The desialated glycophorin stains very intensely with silver and migrates with a different apparent molecular weight than does native glycophorin.
By this criteria, no measurable micellar glycophorin was retained by the IOVs. The amount of MB protein kinase retained by IOVs was assayed by dissolving the IOV pellet in buffered 0.01% Triton X-100 and then determining kinase activity toward spectrin. The addition of Triton X-100 to the IOVs had two effects. (a) The membranes were solubilized, and thus kinase was completely accessible to substrate. (b) Addition of Triton X-100 to membranes increased IOV-associated casein kinase activity toward spectrin. Addition of Triton X-100 appears to eliminate inhibition of kinase activity due to membrane binding. This cannot be explained by exposure of trapped kinase by Triton X-loo-induced lysis of IOVs since glycophorin-intrinsic PL micelles can completely remove all MB protein kinase activity from the IOVs (Fig. 2). As demonstrated in Fig. 2, increasing the concentration of the glycophorin micelles containing intrinsic PL eluted greater than 95% of the MB protein kinase activity from the IOVs. To demonstrate that the kinase activity was in the supernatant, the supernatants containing glycophorin micelles were assayed after addition of Triton X-100 to the supernatant.
In a representative experiment (see Fig. 2), glycophorin micelles with intrinsic PL (25 pg/ml) eluted 93% of the MB protein kinase activity from IOVs (500 fig/ml). Upon addition of Triton X-100,68% of the protein kinase activity was recovered in the supernatant. This experiment demonstrates that the majority of the MB protein kinase has been eluted from the membrane and was recovered in the supernatant.
Further, this experiment also suggests that the intact structure of both IOVs and glycophorin-PL micelles is required for inhibition of kinase activity. To determine the role played by intrinsic PLs retained by isolated glycophorin micelles, the PLs were extracted, and both glycophorin micelles and the extracted PLs were tested for the ability to elute MB protein kinase activity from the membrane.
Glycophorin micelles lacking phospholipid were found to compete with MB protein kinase binding but much less effectively than glycophorin containing intrinsic phospholipid (Fig. 2). The phospholipid fraction, containing mainly phosphatidylserine (PS), phosphatidylinositol4-phosphate (PIP), and PIP? at a 4.2:1:1.2 M ratio (6-9), was resuspended in buffer, sonicated, and combined with IOVs. These phospholipids depleted IOVs of 50% of the MB protein kinase activity only at 33 pM PIP,. This amount of PIP2 is equivalent to that contained in 1 mg/ml of glycophorin retaining intrinsic This suggests that the MB protein kinase requires both glycophorin and PIP2 for a high affinity association.
Since intrinsic PL in the glycophorin micelle enhanced the affinity of MB protein kinase for glycophorin, micellar glycophorin was reconstituted with a variety of PLs to determine if a specific PL was required. The glycophorin-PL micelles were then tested for their ability to deplete the MB protein kinase content of membranes.
As demonstrated in Fig. 2, glycophorin-PL micelles must contain PIP2 to be effective at depleting membranes of MB protein kinase activity. However, for maximal affinity, the glycophorin-PL micelles must have PIP2 and in addition l-4 mol of a negatively charged PL (PS or PA). However, PA and PS only enhanced the effect of PIP2; alone they are not effective at inhibiting MB protein kinase binding to IOVs.
Since the glycophorin-PL micelles were effective at competing with membranes for MB protein kinase binding, the glycophorin-PL micelles were assayed for their effect on MB protein kinase activity. The results show that the catalytic activity of the MB protein kinase is inhibited by glycophorin micelles containing intrinsic PL (Fig. 3). As with inhibition of binding to membranes, extraction of the intrinsic PL from the glycophorin destroyed the ability of the glycophorin micelles to inhibit kinase activity. Further, the extracted intrinsic PLs and glycophorins without PL did not inhibit protein kinase activity. When micellar glycophorin was reconstituted with PL by detergent dialysis (5), inhibition of MB protein kinase activity was also recovered but only when PIP:! was reconstituted into the glycophorin micelles (Table I)  was, as mol %, 75% PC, 20% PS, and 3% PA, with or without added PIP, (Fig. 4) activity. This inhibition is dependent upon the presence of PIP2 in the liposomes; other PLs did not elicit this effect. Further, both the glycophorins and PIP, must be present in liposomes for inhibition of MB protein kinase activity. PIP2 alone in liposomes does not inhibit the MB protein kinase, even at high concentrations of liposomes or at high mol % of PIP, in liposomes (up to 5 mol %).
Glycophorin preparations contain at least four membrane proteins (10-12, 14, 28, 29, 32). To determine if a specific component was required for the interaction with the MB protein kinase, the glycophorin preparation was fractionated by gel permeation chromatography (14,29). The resulting pooled fractions (Fig. 5) were reconstituted with PIP*, PS, and PA as above and assayed for their effectiveness at inhibiting the MB protein kinase. Fractions C and D containing the minor glycophorins B, C, and D and other components in lesser amounts were the most effective at inhibiting MB protein kinase activity. Although this experiment does not conclusively identify the membrane component that interacts with the MB protein kinase, it does demonstrate that the MB protein kinase interacts with a specific component in the preparation.
Glycophorin A does not appear to interact with the kinase since it is the major component of fraction A but does not inhibit kinase activity effectively.
The erythrocyte has two forms of casein kinase: a membrane-bound and a cytosolic form. Both protein kinases have very similar properties.
They have an identical molecular mass by SDS-PAGE of 33-34 kDa, both are eluted from ion exchange columns at the same salt concentration, and both appear to have the same substrate specificity (18-24). The difference is that one protein kinase is membrane associated, and the other is soluble. The erythrocyte cytosolic casein kinase has been shown to be inhibited by 2,3-DPG with a K, of 4.6 mM (23 2,3-DPG with an Iso of l-2 mM but only when the kinase is solubilized from membranes; when associated with membranes, it is not inhibited by 2,3-DPG. Inositoll,4,5-trisphosphate also inhibited the solubilized MB protein kinase but at concentrations higher than observed in cells (Is0 of 0.2 mM). Inositol 1,4,5-trisphosphate, like 2,3-DPG, did not inhibit the MB protein kinase when the kinase was bound to membranes (results not shown). These results suggest that the cytosolic and MB protein kinases are functionally similar and can be regulated by some of the same modulators. However, as shown in Fig. 6, purified cytosolic casein kinase was not inhibited by glycophorin-PIP2 micelles as effectively as was the MB protein kinase. To determine how similar or dissimilar the cytosolic and MB casein kinases are, both kinases were isolated and their sequences compared by two-dimensional "'1 peptide mapping. Both tryptic and chymotryptic peptide maps were done on two different preparations of both the MB and cytosolic casein kinases. The chymotryptic peptide maps of the cytosolic and MB protein kinase are shown in Fig. 7. Clearly the two protein kinases are very similar; indeed, the cytosolic protein kinase has all of the peptides found in the MB protein kinase. However, for both tryptic and chymotryptic peptide maps, the cytosolic protein kinase has additional peptides that are not contained within the peptide map of the MB protein kinase.

DISCUSSION
The molecular basis for modulation of MB protein kinase activity appears to be 2-fold. (a) There appears to be an interaction with a component within the glycophorin preparation, perhaps a minor glycophorin. (b) There is a ternary interaction with PIP, which is of higher affinity and inhibits kinase activity. This interaction is enhanced by other negatively charged PLs such as PS or PA. The glycophorins, when isolated, retain intrinsic PLs, and the PL composition in such glycophorin micelles is selective for PS, PIP, and PIP2. There is evidence that the PL composition of glycophorin micelles results from an association of glycophorins with PS and PIP, in the membrane (6-9). Liposomes containing PC, PS, PA, and PIP2 but without glycophorin did not inhibit the MB protein kinase; however, PIP, micelles alone do weakly inhibit kinase activity. This inhibition requires a much higher concentration of PIP2 than is required in the presence of the glycophorins. Since the glycophorins are known to interact with acidic PLs, specifically PS and PIP, (5-9), mechanistically, a glycophorin or a similar protein within the glycophorin preparation may serve to bind the MB protein kinase, PIP2, and PS, bringing the components into spatial proximity.
Both the cytosolic and solubilized MB protein kinases are inhibited by 2,3-DPG, suggesting a functional similarity. The stereochemistries of the phosphate residues on the 2-and 3hydroxyls of 2,3-DPG and of the phosphate residues on the 4-and 5-hydroxyls of the myo-inositol ring of PIP2 are identical. Since this is the case, a functional feature that is intrinsic to both the membrane-bound and the cytosolic pro- and dissimilar peptides. T TLC tein kinases is regulation by 2 adjacent phosphate residues of the correct stereochemistry. The association of the MB protein kinase with a membrane component may simply potentiate this regulation.
Since 2,3-DPG, inositol 1,4,5trisphosphate, and PIP? inhibit the MB casein kinase, an association with a membrane component (possibly a glycophorin) could place the protein kinase close to the bilayer in an environment with a high local concentration of PIP,. This may also explain why glycophorin micelles with intrinsic PL or reconstituted with PIPp and PS are such potent inhibitors of MB protein kinase activity. In this structure, unlike the membrane, the protein kinase associates with a component that spatially is very close to PIP1. In the membrane, the local concentration of PIP, is likely not as high, thus the binding affinity and inhibition of activity may be lower than that observed for the interaction with the glycophorin micelle. However, other associations not yet defined could also orient the protein kinase such that regulation of protein kinase activity by PIP2 in the membrane is more pronounced.
The ""I-peptide maps show that the cytosolic and MB protein kinases are structurally very similar. Indeed, the cytosolic protein kinase has all of the peptides found in the MB protein kinase as well as additional peptides that do not have counterparts within the MB protein kinase. Since the peptide maps are prepared by excising the purified protein kinase bands from a SDS-PAGE, there are three mechanisms by which this could occur: (a) a difference in sequence between the cytosolic and the membrane-bound protein kinase; (b) post-translational modification of the cytosolic protein kinase; or (c) the cytosolic protein kinase may have a protein of identical molecular mass which copurifies with it, from 7375 B electrophoresis which the additional peptides arise. The peptide maps demonstrate that the MB kinase is a subset of peptides derived from the cytosolic kinase. This suggests that all of the tyrosine-containing sequence found in the MB kinase is also found in the cytosolic kinase. Thus, if the two kinases differ in sequence and since they have an identical molecular mass, then a region of the MB protein kinase sequence would be required to have no tyrosine (silent to ""I-peptide mapping). To obtain the peptide map that we have for the cytosolic protein kinase, the silent region in the MB kinase would have to be replaced by a sequence in the cytosolic protein kinase which would have tyrosines; this would result in new peptides. Such a sequence change is possible but seems unlikely.
Post-translational modification of the cytosolic kinase, such as phosphorylation, would also result in peptides unique from the MB kinase. Both the MB protein kinase and the cytosolic protein kinase are autophosphorylated (18-24); whether or not they are phosphorylated in vivo is not known. However, phosphorylation of the protein kinase at one or more sites could explain the differences and similarities between the two kinases.
The third possibility is that there is a protein of identical molecular mass which copurifies with the cytosolic protein kinase, possibly forming an association. Intuitively this seems unlikely but cannot be ruled out. Indeed, statistically a 33-kDa protein should contain about 16 tyrosines (assuming 1 tyrosine in 20 amino acids); this should give rise to about 16 9-containing peptides. A low exposure of the peptide map of the MB protein kinase shows about 16 peptides. The map shown is at higher exposure and contains about 30 peptides