In vitro phosphorylation of ciliary dyneins by protein kinases from Paramecium

The ciliary dynein ATPases of Paramecium have been purified and characterized (Doughty, 1979; Travis and Nelson, 1988a; Larsen et al., 1991; Beckwith and Asai, 1993; Walczak et al., 1993b). High-salt extraction of axonemes solubilizes two major ATPase activities, which sediment at 22 S and 12 S. The outer arm dynein has been characterized most thoroughly. It contains three heavy chains (Mr >300,000), three intermediate chains (Mr 81,000 doublet and 65,000 singlet), and eight light chains (Mr 12,000 to 30,000) (Travis and Nelson, 1988a; Larsen et al., 1991; Beckwith and Asai, 1993; Walczak et al., 1993b). In the axoneme and under certain isolation conditions, it is a three-headed species (Schroeder et al., 1990; Larsen et al., 1991); however, under higher-salt conditions, it separates into a two-headed particle, which sediments at 22 S and contains the α- and β-heavy chains, and the γ-heavy chain, which sediments at 12 S (Beckwith and Asai, 1993; Walczak et al., 1993b). Little is known about the inner arm dyneins of Paramecium.

Paramecium responds to stimuli by changing the direction and speed of its motion either by altering the direction of the power stroke or by altering the beat frequency. The second messengers involved in these pathways include Ca²⁺ and cyclic nucleotides. Both cAMP and cGMP increase the swimming velocity of permeable cells reactivated with ATP (reviewed by Bonini et al., 1991). Cyclic AMP and cGMP also increase dynein-like ATPase activity in permeable cells and in motile cilia preparations (Bonini, 1987).

Many of the actions of Ca²⁺ and cyclic nucleotides are believed to be mediated by their respective protein kinases. Cyclic nucleotide-dependent protein kinases PKA and PKG (Miglietta and Nelson, 1988; Hochstrasser and Nelson, 1989; Mason and Nelson, 1989a,b) and Ca²⁺-(CaPK1 and CaPK2) (Gundersen and Nelson, 1987; Son et al., 1993) have been purified from Paramecium. The phosphorylation of several ciliary proteins, including dynein, increases upon addition of Ca²⁺ or cyclic nucleotides to permeabilized cells or cilia (Lewis and Nelson, 1981; Eistetter et al., 1983; Travis and Nelson, 1988b; Hamasaki et al., 1989; Bonini and Nelson, 1990). Phosphorylated dynein isolated after stimulation of axonemes with cAMP and ATP produces faster microtubule motion in an in vitro microtubule gliding assay when compared with unphosphorylated preparations of 22 S dynein (Hamasaki et al., 1991). It thus seems likely that one action of cyclic nucleotides and Ca²⁺ is to stimulate the phosphorylation of proteins in the cilia that regulate dynein activity, thereby altering ciliary motility and cellular behavior.

The biochemical complexity has made it difficult to establish a causal chain from signal to second messenger to protein kinase to phosphoprotein, even with permeabilized cells or isolated cilia that respond to second messengers. We therefore asked whether Paramecium dyneins are targets of phosphorylation in vitro, using sucrose-gradient purified 22 S and 12 S dyneins and partially purified Paramecium kinases. Only cAMP-dependent protein kinase phosphorylated dynein polypeptides. The best substrate for in vitro phosphorylation was a 29 kDa protein found in both 22 S and 12 S dyneins. The factors that influenced phosphorylation were studied. In the process, we detected another, previously uncharacterized, protein kinase in high-salt extracts of axonemes used for dynein preparation.

[γ-³²P]ATP (3000 Ci/mmol) was from Amersham (Arlington Heights, IL). N.N′-methylenebisacrylamide, SDS, molecular mass standards and Coomassie Brilliant Blue stain were from Bio-Rad Laboratories (Richmond, CA). Pepstatin A was from Boehringer-Mannheim (Indianapolis, IN). Intensifying screens were from Dupont. The bovine catalytic subunit of PKA and all other chemicals were from Sigma (St Louis, MO).

Paramecium tetraurelia (strain 51S) was grown at 25°C in Soldo’s crude axenic medium (Van Wagtendonk, 1974). Cells were harvested from 25 liter cultures at late log to early stationary phase as described previously (Adoutte et al., 1980). Modifications to this procedure included immobilization in SMEN (0.5 M sucrose, 20 mM MOPS, 2 mM EDTA, 6 mM NaCl, pH 7.5) instead of STEN (0.5 M sucrose, 20 mM Tris, 2 mM EDTA, 6 mM NaCl, pH 7.5), and deciliation by the addition of BaCl₂ instead of CaCl₂ with 0.3 mM phenylmethylsulfonyl fluoride (PMSF) in the deciliation buffer. Cilia were pelleted by centrifugation for 30 minutes at 30,000 g, washed once, and resuspended in MMKE (20 mM MOPS, 1 mM MgCl2, 100 mM KCl, 0.1 mM EDTA, pH 7.5) with protease inhibitors (0.3 mM PMSF, 0.1 trypsin inhibitor units/ml of aprotinin, 2 μg/ml leupeptin, 0.3 mM V-α-p-tosyl-L-arginine methyl ester (TAME), 2 μg/ml pepstatin A) and used immediately for purification.

Dyneins were purified from cilia through sucrose gradients essentially as described previously (Travis and Nelson, 1988a), with modifications as described (Walczak et al., 1993b).

PKA was purified from the high-speed supernatant of cell bodies by chromatography on phenyl-Sepharose followed by red agarose pseudo-affinity chromatography essentially as described previously (Hochstrasser and Nelson, 1989). The catalytic subunit of bovine PKA was purchased from Sigma Chemical Co. (St Louis, MO). PKG was purified from the high-speed supernatant of cell bodies of Paramecium by chromatography on DE-52 followed by cAMP-agarose affinity chromatography as described (Ann, 1991). CaPK1 and CaPK2 were purified from the high-speed supernatant of cell bodies by chromatography on phenyl-Sepharose followed by Mono-Q anion exchange chromatography as described (Gun-dersen and Nelson, 1987; Son et al., 1993). For each of the above kinases, the combination of purification steps chosen resulted in a kinase preparation that was free of contaminating protein kinase activity and free of endogenous protein kinase substrates. The phosphorylation of dynein by PKA was also examined using a partially purified preparation of the ciliary form of the enzyme. Identical results were obtained with ciliary and body forms of the enzyme. Due to the ease of isolating the body enzyme, all experiments described were done with that fraction.

PKA monoclonal antibodies were generated against the cellular form of PKA from Paramecium as described previously (Hochstrasser and Nelson, 1989). PKG monoclonal antibodies were generated against the cellular form of PKG from Paramecium as described previously (White, 1990). CaPK1 and CaPK2 antibodies were generated against both cellular forms of the kinase from Paramecium (Son, 1991). All monoclonal antibodies were used as tissue culture supernatants.

Dyneins were incubated with partially purified protein kinase (10 units) in a 50 pl reaction mixture containing 50 mM MOPS, pH 6.5, 10 mM MgCl2, 100 μM [γ-³²P]ATP (600 Ci/mol); ± 0.2 μM cyclic nucleotide for PKA and PKG. For CaPK1 and CaPK2, the reaction mixture contained 50 mM HEPES, pH 7.2, 0.5 mM EGTA, 5 mM magnesium acetate, ± 0.6 mM CaCl₂. The reaction was initiated by the addition of [γ-³²P]ATP and terminated by precipitation with trichloroacetic acid (TCA). Samples were run on SDS-PAGE and stained with Coomassie Brilliant Blue. The gels were vacuum dried before autoradiography. Autoradiograms were exposed for 1-4 days, in some cases with intensifying screens. For quantitation of phosphate incorporation, the radioactivity in dried gels was counted in the linear range using a Betascope Analyzer (Betagen Corp., Mountain View, CA). A portion of the [γ-³²P]ATP solution used in the reaction was spotted onto nitrocellulose paper and counted for a standard. The amount of protein loaded was determined by protein assays. The stoichiometry of incorporation of phosphate was calculated for a single time point (5 minutes) for 22 S dynein polypeptides. The amount of the specific polypeptide was calculated from reported values in the literature. The amount of heavy chain protein was estimated at 79% of the load (Travis and Nelson, 1988a), and we assumed that two heavy chains were present in this fraction (Walczak et al., 1993b). The amount of 29 kDa protein was estimated at 1.5% of the load (Hamasaki et al., 1991).

Protein kinase (10 μ l) was assayed in a reaction mixture containing 50 mM MOPS, pH 6.5, 10 mM MgCl₂, 1 mg/ml substrate (casein or histone) and 100 pM [γ-³²P]ATP (50 Ci/mol) for 10 minutes. For PKA and PKG, the reaction mixture also contained 0.2 pM cyclic nucleotide. For CaPK1 and CaPK2, the reaction mixture contained 50 mM HEPES, pH 7.2, 0.5 mM EGTA, 5 mM magnesium acetate, ± 0.6 mM CaCl₂. The reaction was initiated with ATP and terminated by spotting the reaction mixture on Whatman filter paper, which was then dropped into cold TCA. The filters were washed and counted; 1 unit equals 1 pmol of phosphate transferred to substrate/min (Corbin and Reiman, 1974).

Protein was assayed according to Bradford (1976) using bovine serum albumin as a standard. ATPase activity was measured by the method of Lanzetta et al. (1979) with the modifications described by Travis and Nelson (1988a). SDS-PAGE was according to Laemmli (1970) with 10% acrylamide running gels and 3% stacking gels. Immunoblots and ELISAs were as described previously (Walczak et al., 1993b).

We carried out in vitro phosphorylation reactions with Paramecium protein kinases using each dynein as a potential substrate. The kinases tested were PKA, PKG, CaPKI and CaPK2, which were partially purified from Paramecium extracts. Only PKA caused significant phosphorylation of dynein (Fig. 1). This phosphorylation was dependent on the presence of cAMP. The major phosphorylated species was a 29 kDa protein present in both 22 S and 12 S dyneins. In the 22 S dynein fraction, this protein appeared as a doublet. The presence of this doublet in the 22 S fraction was not reproducible; therefore we are uncertain of its significance. The heavy chains of both dyneins were also phosphorylated. Tubulin, a small amount of which contaminates dynein preparations, was phosphorylated by PKA and by every other kinase we tested (data not shown). Several other lower molecular mass proteins were phosphorylated by PKA, but their phosphorylation was not reproducible from preparation to preparation (compare Fig. 1 with Figs 2 and 4). A small but significant amount of cAMP-dependent phosphorylation reproducibly occurred in the absence of exogenous PKA (Fig. 1B). This was probably due to endogenous PKA activity in this fraction, which will be addressed later (see Fig. 5).

Phosphorylation of both 22 S and 12 S dyneins by PKA was rapid, peaking at 5-10 minutes (Fig. 2). From the maximum ³²P incorporation (5 minutes), and assuming that the 29 kDa protein constitutes 1.5% of the total 22 S dynein protein (see Materials and Methods for details of quantitation), the stoichiometry of phosphorylation was calculated to be 1 mol of phosphate incorporated/5 mol of 29 kDa protein in 22 S dynein and 1 phosphate per 20 heavy chains. For each dynein, the ³²P incorporation decreased after 5 minutes. The most rapid loss of label occurred from the 29 kDa phosphoprotein, which by 60 minutes had lost more than 50% of its label.

To examine the cause of the loss of label, we studied the time course of phosphorylation of 22 S dynein in the presence of either NaF, a general phosphatase inhibitor, or vanadate, an inhibitor of dynein ATPase activity. Both NaF and vanadate slowed the loss of label. In the presence of NaF, ³²P incorporation still peaked at 5 minutes but diminished less rapidly after that time than in the control (data not shown). In the presence of vanadate, there was an increase in the total amount of label incorporated. Phosphate incorporation peaked at 20 minutes and did not decrease significantly after that time (data not shown). These results suggest that the loss of label is due to a combination of a phosphatase present in either the dynein or the kinase preparation, and to the decrease in the total amount of [γ-³²P]ATP present due to its hydrolysis by the dynein ATPase. We estimate that less than 10% of the initial amount of [γ-³²P]ATP will remain after 20 minutes due to its hydrolysis by the dynein ATPase.

We compared the phosphorylation of 22 S dynein by the Paramecium kinase with that of the catalytic subunit of bovine PKA. Only the Paramecium enzyme was able to phosphorylate the 29 kDa protein (Fig. 3), although both kinases were equally effective at phosphorylating the heavy chains of 22 S dynein.

Since the 29 kDa phosphoprotein was present in fractions of 22 S and 12 S dyneins, we wanted to know if it was also present elsewhere in the gradient. We assayed for phosphorylation of proteins by exogenous PKA in fractions of the sucrose gradient of a high-salt extract of axonemes (Fig. 4). Phosphorylatable 29 kDa protein peaked with 22 S dynein, but a phosphoprotein of this molecular mass was also found in smaller amounts throughout the gradient. Most of the phosphorylation occurred in the fractions at the top of the sucrose gradient. There was also a significant amount of phosphorylation of many proteins in the presence of ATP but in the absence of exogenous kinase, suggesting the presence of another kinase in these fractions.

Because earlier experiments suggested the presence of endogenous kinases in our sucrose gradient fractions, we wanted to know whether these kinases were associated with any of the dyneins. We assayed all of the sucrose gradient fractions for PKA, PKG, CaPK1 and CaPK2 activity. There was a significant amount of PKA activity at the top of the gradient (Fig. 5), which trailed slightly into the 12 S dynein peak, thus accounting for the small amount of endogenous cAMP-dependent phosphorylation seen in 12 S dynein in Fig. 1. There was a small amount of PKG activity, and no Ca²⁺-stimulated kinase activity was present (data not shown). All of these results were confirmed by immunoblots and ELISAs using antibodies to the above protein kinases.

In addition to PKA, we found a significant amount of second messenger-independent casein kinase activity (Fig.6) in the sucrose gradient. Like PKA, its activity trailed slightly into the 12 S dynein peak. In Paramecium, PKA prefers histone to casein as a substrate for phosphorylation (Mason and Nelson, 1989a,b). Therefore, the number of units of this casein kinase activity was too high to be accounted for by the free catalytic subunit of PKA (compare histone phosphorylating units in the absence of cAMP, Fig. 5, with the number of casein phosphorylating units in Fig. 6). In addition, it is unlikely that this activity results from a constitutively active form of PKG, CaPK1 or CaPK2, because of from the lack of immunological cross-reaction with antibodies to these kinases. This casein kinase is probably responsible for the messenger-independent phosphorylation seen in fractions at the top of the sucrose gradient in Fig. 5. The casein kinase activity will be explored further elsewhere (Walczak et al., 1993a).

Previous studies of the phosphorylation of dyneins involved stimulation of the endogenous protein kinase in either per-meabilized cells or cilia, followed by isolation of dynein (Travis and Nelson, 1988b; Hamasaki et al., 1989; Bonini and Nelson, 1990; Hamasaki et al., 1991). To address the question more directly, we looked at the phosphorylation of purified 22 S and 12 S dyneins by each of the known protein kinases in Paramecium, added exogenously. Consistent with the previous results of Hamasaki et al. (1989), Bonini and Nelson (1990), and Hamasaki et al. (1991), we found that the 29 kDa phosphoprotein is one of the major PKA substrates in dynein fractions. In addition, we found that the heavy chains of both 22 S and 12 S dyneins were substrates of PKA.

In contrast to the results of Travis and Nelson (1988b), we found that neither PKG nor the Ca²⁺-dependent protein kinases phosphorylated any dynein polypeptides. There are several possible explanations for this discrepancy. (1) The cGMP-or Ca²⁺-stimulated phosphorylations that were detected previously may have been the result of a phosphorylation cascade, such that some other kinase served as an intermediate in the reaction. Candidates may be PKA or one of the ciliary casein kinases (Walczak et al., 1993a). (2) The cGMP-dependent substrates in the above studies were often subsets of the cAMP-dependent substrates, and thus may have resulted from stimulation of PKA by cGMP. It is known that the Paramecium PKA can be stimulated by high levels of cGMP (Mason and Nelson, 1989a,b). (3) An unidentified Ca²⁺-dependent kinase may mediate the previously seen Ca²⁺-stimulatable phosphorylation in cilia.

Immunological evidence suggests that CaPK1 and CaPK2 are present in isolated cilia (Son, 1991), but they have not been isolated from cilia yet. CaM kinase, has not been found in Paramecium. Given the large amount of calmodulin present in cilia (Maihle et al., 1981; Walter and Schultz, 1981; Klumpp et al., 1983), if CaM kinase were present in cilia, it would probably be stimulated maximally under the conditions of Ca²⁺-dependent phosphorylation. It could thus be responsible for any Ca²⁺-stimulated phosphorylation of dynein or other ciliary proteins. (4) The highly regular structure of the axoneme may be necessary for the correct recognition of axonemal substrates by protein kinases.

The bovine catalytic subunit of PKA does not have the same substrate specificity as the Paramecium enzyme (Hochstrasser, 1989). This is not surprising, since the enzymatic and kinetic properties of the Paramecium enzymes are significantly different from those of the mammalian enzyme (Mason and Nelson, 1989a,b). It is possible that the Paramecium enzymes have evolved to have different substrate specificities for specialized proteins involved in ciliary motility. Chilcote and Johnson (1991) have used the bovine enzyme to phosphorylate the 22 S dynein from Tetrahymena. They obtained identical phosphorylation patterns with an endogenous PKA preparation. The difference between our results and those of Chilcote and Johnson (1991) was surprising to us, given the structural and immunological similarity between Paramecium and ‘Tetrahymena dyneins (Walczak et al., 1993b).

We found that the label of phosphoproteins, including the autophosphorylation of the regulatory subunit of the kinase turned over rapidly under the conditions of our experiments. The turnover was inhibited by NaF, a general phosphatase inhibitor, and by vanadate, an inhibitor of dynein ATPase activity. These results suggest that phosphatase activity is present in either the dynein preparation or the kinase preparation. In a separate experiment, the PKA preparation was used to phosphorylate calmodulin. There was no turnover of label in this preparation (K. Kim and D. L. Nelson, unpublished observations), suggesting that the phosphatase is part of the dynein preparation. Tash and coworkers (1988) have found a Ca²⁺/calmodulin-dependent phosphatase in dynein preparations of sea urchin sperm flagella. This phosphatase is responsible for the dephosphorylation of a 23 kDa light chain of 21 S dynein that is phosphorylated by PKA (Tash and Means, 1989). Stephens and Prior (1992) also reported a turnover of label in their dynein phosphorylation experiments. Given the rapid turnover of the label of the 29 kDa phosphoprotein relative to the turnover of label in the other PKA substrates, it is likely that it is a better substrate of the phosphatase.

We found a 29 kDa phosphoprotein in 22 S dynein and in 12 S dynein. This is in contrast to the results of others, who found a phosphoprotein of this molecular mass located exclusively with 22 S dynein (Travis and Nelson, 1988b; Bonini and Nelson, 1990; Hamasaki et al., 1991). In our experiments, the relative amount of 29 kDa phosphorylation of both 22 S and 12 S dyneins and its distribution between the two dynein fractions varied somewhat between preparations. It is possible that this protein is phosphorylated during the isolation of dynein, and that the extent of phosphorylation varies in different preparations of dynein. Bonini and Nelson (1990) saw some phosphorylation of the 29 kDa phosphoprotein of 22 S dynein in control preparations of permeabilized cells that were incubated with ATP in the absence of cyclic nucleotide or Ca²⁺ stimulation (see Fig. 5B of that paper), suggesting that this is a viable possibility. Alternatively, it is possible that the 29 kDa phosphoprotein is only loosely associated with 22 S dynein, and some of it dissociates from the dynein during purification and is then found in other parts of the sucrose gradient (see discussion below). Finally, the 29 kDa proteins found throughout the gradient may be unrelated to the 29 kDa protein associated with dynein, but coincidentally the same size and also subject to cAMP-dependent phosphorylation.

The amount of the 29 kDa phosphoprotein in 22 S dynein, as judged by protein staining of an overloaded gel, is substoichiometric relative to the other subunits of the enzyme (see also Travis and Nelson, 1988a; Hamasaki et al., 1991). It is clearly associated with this fraction as judged by its cosedimentation with 22 S dynein in sucrose gradients (Travis and Nelson, 1988b; Bonini and Nelson, 1990; Hamasaki et al., 1991), its association by alloaffinity fractionation of dynein (Hamasaki et al., 1991), and the ability of dynein antibodies to immunoprecipitate it (Walczak and Nelson, unpublished observations). In fractions at the top of the sucrose gradient, there is a significant amount of phosphorylation by PKA of small molecular mass proteins, some of which are in the range of 29 kDa and could be this protein. It is possible that the 29 kDa phosphoprotein is easily dissociated from the dynein fraction under the conditions of isolation. If this dissociation occurred, we would expect to find the protein at the top of the sucrose gradient. In a study by Bonini and Nelson (1990), (see Fig. 4B), there was a significant amount of a 29 kDa phosphoprotein remaining in the axonemes after extraction with high salt to solubilize dynein. Very little is known about the inner arm dyneins in Paramecium, but at least some of them are likely to remain with the axoneme under the conditions used to isolate dyneins in this study and in the work of others. It is possible that the 29 kDa phosphoprotein is a component of inner arm dyneins, and thus the majority of it remains unextracted along with the inner arms. The small amount of 29 kDa phosphoprotein seen in the 22 S dynein fraction may be the result of inner arm dyneins sedimenting in the region of 22 S on our sucrose gradients.

The phosphorylation of dynein light chains has been observed in many organisms (reviewed by Stephens and Stommel, 1989; Tash, 1989), but the phosphorylation of dynein heavy chains has been reported only in flagella (Piperno and Luck, 1981; Tash and Means, 1989; Stephens and Prior, 1992). Chilcote and Johnson (1991) found that the phosphorylation of ciliary dynein heavy chains of Tetrahymena occurred only after protein denaturation. It is a possibility that the phosphorylation of our dynein heavy chains is due to some denaturation of the protein, despite our efforts to avoid this by carrying out the purification quickly and in the presence of a wide variety of protease inhibitors.

We do not find an increase in the ATPase activity or the microtubule-stimulated ATPase activity upon dynein phosphorylation by PKA (data not shown). It is possible that this is because of the less than optimal stoichiometry of phosphate incorporation. Alternatively, ATPase activity may not reflect dynein’s physiological activity. Hamasaki et al. (1991) found a 40% increase in the microtubule-gliding rate of phosphorylated preparations of dynein compared to unphosphorylated preparations even though they also saw no increase in the ATPase activity of these fractions. Stephens and Prior (1992) and Chilcote and Johnson (1991) reported no change in ATPase activity of their phosphorylated dyneins.

Most of the ciliary kinases are soluble or membrane-associated (Miglietta and Nelson, 1988; Mason and Nelson, 1989a,b). A portion of PKA is axonemal (M. Hochstrasser et al., unpublished data); it is released by the same high-salt treatment that releases dynein. It is interesting that this is the only kinase that phosphorylates dynein. The other kinase in the high-salt extract of axonemes, CKA (Walczak et al., 1993a), may also be able to phosphorylate dynein, but this could not be tested at this time. There is a large amount of messenger-independent phosphorylation in fractions at the top of the sucrose gradient and a messenger-independent casein kinase in those fractions. It is likely that the casein kinase is responsible for phosphorylating a large number of those substrates.

There are a large number of protein kinases in the cilia, thus it is unlikely that a single phosphorylation event is responsible for the changes in motility that occur in response to various stimuli. Rather, it seems more likely that it is the integration of a signal transduction cascade that results in these changes. Dey and Brokaw (1991) have shown that in Ciona sperm the activation of motility requires cAMP-dependent protein kinase and a soluble factor from ciliary extracts. In addition, activation of motility is completely inhibited by a random copolymer of glutamate and tyrosine, which also inhibits protein phosphorylation in sperm. This random copolymer has been shown to be a potent inhibitor of tyrosine kinases (Braun et al., 1984), although Dey and Brokaw (1991) were unable to show any tyrosine phosphorylation. Alternatively, it has been shown that this random copolymer is an inhibitor of casein kinase II (Meggio and Pinna, 1989). It is possible that it is a casein kinase and not a tyrosine kinase that is important in this transduction cascade.

Previous work has shown cAMP-dependent phosphorylation of the 29 kDa phosphoprotein in permeabilized cells and in isolated cilia. We have shown that it is the direct substrate only of PKA. It is likely that this phosphoprotein plays a role in the regulation of ciliary motility. The identity of this protein and its regulation by phosphorylation and dephosphorylation may provide insights into the mechanism of the regulation of motility.

We thank Gail Carlson and Kwanghee Kim for several protein kinase preparations, for useful suggestions during the course of this work, and for critical reading of this manuscript. We also thank Joan Peterson for critical reading of this work and Brook Soltvedt for editorial comments. This work was supported by grant GM34906 to D.L.N. and a departmental fellowship to C.E.W.