Phosphorylation of linker histones by a protein kinase A-like activity in mitotic nuclei.

Micronuclear linker histones of the ciliated protozoan, Tetrahymena thermophila, are extensively phosphorylated in vivo. Each of these polypeptides, α, β, γ, and δ, contains sites for phosphorylation by cyclic-AMP dependent protein kinase (PKA) but not Cdc2 kinase, and some data have been presented implicating PKA kinase in their phosphorylation in vitro and in vivo (Sweet, M. T., and Allis, C. D. (1993) Chromosoma 102, 637-647; Sweet, M. T., Jones, K., and Allis, C. D. (1996) J. Cell Biol., in press). In this report we have extended these analyses by showing that Cdc2 and PKA kinase are not evenly distributed between micro- and macronuclei. Macronuclei, but not micronuclei, contain a 36-kDa polypeptide that is immunoreactive with p34Cdc2 antibodies. In contrast, a 40-kDa polypeptide is detected with PKA antibodies in micronuclei, that is not detected in macronuclei. In support, extracts from micronuclei, but not macronuclei, contain a kinase activity that resembles some, but not all, characteristics of PKA from other sources. Immunodepletion experiments using anti-PKA antibodies show that a 40-kDa polypeptide can be specifically removed from these extracts with a concomitant loss in kinase activity. Microsequence analyses of δ demonstrate that this linker histone is phosphorylated in vivo on two PKA consensus sequences located in its carboxyl-terminal domain, an optimum PKA consensus sequence, Arg-Lys-Asn-, and a less optimal PKA sequence, Lys-Ser--Val. Collectively, these results suggest that PKA or a PKA-like kinase is responsible for the phosphorylation of linker histone in mitotically dividing micronuclei. In contrast, macronuclei, which divide amitotically, phosphorylate linker histone H1 using a distinct, Cdc2-like kinase.

A variety of kinases and phosphatases have been shown to play an implicit role in the regulation of fundamental cell cycle processes. For example, a family of closely related cyclin-dependent protein kinases is thought to catalyze a series of phosphorylation events associated with cell-cycle progression (for reviews see Refs. [3][4][5][6]. During mitosis, several proteins such as linker histones (7,8), lamins (9), and cytoskeletal proteins (10,11) undergo stage-specific, reversible phosphorylation. The bi-ological consequences of these phosphorylation events are largely unclear.
It has long been suggested that hyperphosphorylation of the linker histone H1 during mitosis is causally linked to mitotic chromosome condensation (reviewed in Ref. 8), although this relationship remains unproven and controversial (discussed in Refs. 12 and 13). It has also been proposed that phosphorylation of linker histone may act as a first-step mechanism to promote transient decondensation of the chromatin fiber, allowing access of specific factors (such as the SMC family of nonhistone proteins; reviewed in Refs. 13 and 14) in a variety of cell cycle-regulated processes including chromosome condensation (15). Mounting evidence has shown that chromosome condensation can occur in the absence of H1 or H1 phosphorylation in vitro (16 -18) and in vivo (12,19).
The ciliated protozoan, Tetrahymena thermophila, provides an ideal model for unraveling complex relationships between H1 phosphorylation, gene expression, and mitotic chromosome condensation. Each vegetative cell contains two types of nuclei, a somatic macronucleus that divides amitotically and a germline micronucleus that divides mitotically. Both nuclei contain linker-associated polypeptides that differ dramatically. Macronuclei, for example, contain a H1 that resembles vertebrate H1s in several properties including growth or division-associated phosphorylation by a Cdc2-like kinase (20). In contrast, micronuclei contains four distinct polypeptides (␣, ␤, ␥, and ␦) that are also phosphorylated in growing or dividing cells (21). If Cdc2 kinase is solely responsible for linker histone phosphorylation and mitotic chromosome condensation in Tetrahymena, ␣, ␤, ␥, and ␦ would be expected to be phosphorylated by this enzyme. However, ␣, ␤, ␥, and ␦ do not contain any obvious recognition sequence for Cdc2 kinase (22), and none of these polypeptides are phosphorylated by this kinase in vitro under conditions where macronuclear H1 is extensively phosphorylated (1). Interestingly, all four of these polypeptides contain at least one canonical phosphorylation site for cAMP-dependent protein kinase (PKA) 1 (22), although none of the in vivo sites of phosphorylation have been identified.
In this study, we have continued to explore the relationship between PKA and linker histone phosphorylation in mitotic micronuclei in Tetrahymena with emphasis on ␦. Immunoblotting data support the notion that PKA (or a PKA-like activity) exists in micronuclei during most stages of the life of the cell, and antibodies against PKA can immunodeplete ␦ kinase activity from micronuclear extracts. Microsequence analyses confirm that ␦ is phosphorylated in vivo on two serine residues embedded in the carboxyl terminus; both sites conform to PKA recognition sites. Collectively, these results suggest that a PKA or a PKA-like kinase is responsible for phosphorylation of ␦ in micronuclei and support the general hypothesis that PKA kinase(s) play an important, and previously unsuspected, role in mitosis.

Cell Culture and Labeling Conditions
T. thermophila strains CU428 (Chx/Chx-[cy-S]VII) and CU427 (Mpr/ Mpr [6 mp-s VI]) were grown in 1% enriched protease peptone under standard conditions as described previously (23). Previous data had shown that ␦ phosphorylation was maximal in early mating cell cultures, 2-4 h after initiating mating (2). Therefore, 2.5-h mating cultures were utilized in this study. Conjugation was induced according to Bruns and Brussard (24) with modifications described by Allis and Dennison (25). Mating cultures were phosphorylated in vivo by starving cells of each mating type separately at approximately 1-2 ϫ 10 5 cells/ml in 10 mM Tris, pH 7.4, in the presence of [ 32 P]orthophosphate (12.5 mCi/ml) for 2-3 h before adding 2 ϫ growth medium and growing cultures for at least one generation (2-5 ϫ 10 5 cells/ml). Labeled cells were then starved and mated as described above.

Preparation of Nuclei and Nuclear Proteins
Macronuclei and micronuclei were isolated from cells as described by Gorovsky et al. (23), except that the nucleus isolation buffer contained 1 mM iodoacetamide, 1 mM phenylmethylsulfonyl fluoride, and 10 mM sodium butyrate, but not spermidine. Where indicated, purified preparations of micro-and macronuclei were isolated following sedimentation at unit gravity according to Allis and Dennison (25). In order to better preserve labile phosphorylation modifications, chloromercuriphenylsulfonic acid (final concentration 0.1 mM) was added to the cell homogenization buffer. ␦ was partially purified from sulfuric acid extracts of micronuclei by reverse phase (RP)-HPLC using a C8 column (Brownlee) with a linear gradient of 5-90% acetonitrile containing 0.1% trifluoroacetic acid (changing at a rate of 0.45%/min) with a flow rate of 1 ml/min.

Preparation of Nuclear Extracts
Macro-and micronuclei from log-phase growing cells were used to prepare active kinase extracts immediately following isolation. Nuclei were washed once in isolation buffer and then resuspended (at 5 ϫ 10 7 macronuclei/ml and 1 ϫ 10 9 micronuclei/ml) in 10 ϫ phosphorylation buffer (1 M NaCl, 250 mM MOPS, pH 7.4, 100 mM MgCl 2 , 10 mM dithiothreitol, and 0.1% Nonidet P-40). Nuclei were lysed in this buffer for 10 min before a 9-fold volume of cold water was added; the lysate was then vortexed well and then microcentrifuged for 15 min. The clarified supernatant was removed and used immediately or frozen at Ϫ20°C. Extracts prepared in this fashion remained active for kinase activity for up to 1 week at Ϫ20°C.

In Vitro Phosphorylation
Enzymes, HeLa Cdc2 kinase, bovine or Paramecium PKA kinase, or crude Tetrahymena micronuclear extracts (extract from approximately 1 ϫ 10 7 micronuclei/reaction) were added to each reaction immediately prior to the addition of [␥-32 P]ATP (100 -150 units; 1 unit transfers 1 pmol of P i /min), as described previously (1). Incubations proceeded at 30°C for 15 min, and duplicate samples from each reaction were removed, applied to P81 filter paper (Whatman International, Maidstone, United Kingdom), and processed for liquid scintillation counting (26).

Immunoprecipitation
Ten microliters of micronuclear extract was incubated with 5 l of either preimmune serum or immune serum for 2 h with gentle shaking at 4°C. As a negative control, immune serum was added to buffer without nuclear extract. Protein A-Sepharose (15 l) was added to each mixture and the incubation was continued at 4°C for 1 h more. Following centrifugation, the unbound supernatant was removed from each incubation mixture, and 5 l was analyzed in in vitro kinase reactions with and without ␦ as substrate. Immunoblotting analysis was done on an aliquot of the immunoprecipitation supernatant. Protein bound to the Sepharose beads was also analyzed in this fashion after solubilization with Laemmli sample buffer.

Chemical and Enzymatic Cleavages
N-Bromosuccinimide-RP-HPLC-purified ␦ dissolved in 180 ml of 5% acetic acid was cleaved by adding 20 ml of 20 mM N-bromosuccinimide (NBS) freshly dissolved in 5% acetic acid and incubating 2 h at room temperature in the dark (27). The resultant peptides were purified by RP-HPLC using a C18 column with a 0 -90% acetonitrile linear gradient as described above. The single in vivo phosphorylated NBS fragment from ␦, which eluted at approximately 8% acetonitrile, was identified by scintillation counting and gel analysis on 50% acid-urea gels (see below).
Endoproteinase Lys-C-In vivo phosphorylated NBS peptide from ␦, isolated as described above, or in vitro phosphorylated ␦ or a synthetic ␦ peptide was cleaved by mixing endoproteinase Lys-C (Boehringer Mannheim) with peptide at 1:100 (w/w) in 0.1 M ammonium bicarbonate buffer, pH 9.0, and incubating at 37°C for 1-2 h. A second equal aliquot of enzyme was then added to the reaction mixture and incubation continued overnight (28). Where indicated, peptides were purified by cation-exchange HPLC using a polyCAT A column (PolyLC Inc.) with a 0 -0.3 M NaClO 4 linear gradient (in a 10 mM phosphate buffer, pH 6.5) increasing at a rate of 0.03 M NaClO 4 /min. Peptides were also purified by elution with water from acid-urea 50% polyacrylamide gels (29) as described previously (30). Due to the small size and positive charge of several of the ␦ peptides analyzed in this study, carboxymethylcellulose membrane was chosen for best retention of peptides (31). Peptides were eluted from carboxymethylcellulose membranes for sequencing, as described previously (31). Small peptides were also analyzed by thin layer chromatography (TLC) using a phospho-chromatography buffer (37% n-butanol, 25% pyridine, and 7% glacial acetic acid) as described by Boyle et al. (32).

Preparation of Synthetic Peptides and Sequencing
Peptides used in this study (see Fig. 2 for details) were synthesized by the solid phase procedure on a peptide synthesizer (model 430A; Applied Biosystems Inc., Foster City, CA). Peptides were cleaved from the resin and analyzed by RP-HPLC, amino acid analysis (Pico Tag system; Waters Associates, Milford, MA). Amino acid sequence analysis (model 477A protein sequencing system; Applied Biosystems Inc.) was used to confirm the sequence of peptides and to identify sites of serine phosphorylation on purified peptides.

Electrophoresis and Immunoblots
One-dimensional sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis systems have been described previously (33,34). Acidurea acrylamide gel electrophoresis of small peptides has also been described previously (29). Immunoblot analyses of SDS-polyacrylamide gels were done as described previously (35). Balanced protein loads of samples were ensured by staining parallel gels of equivalently loaded samples and by staining immunoblots directly with Ponceau red stain. Immunoreactivity was detected by alkaline phosphatase-conjugated secondary antibodies or by chemiluminesence and autoradiography as indicated. Polyclonal antiserum against the catalytic subunit of PKA was prepared as described (36). Antibodies against yeast p34 cdc2 were a generous gift from the Beach laboratory.

Unexpected Partitioning of Cdc2 and PKA between Microand Macronuclei-
The unique partitioning of mitosis-related events to micronuclei along with a specialized set of linkerassociated polypeptides, ␣, ␤, ␥, and ␦, raises the question which kinase is responsible for their phosphorylation in vivo. If, as suggested by our earlier studies (1), PKA or a PKA-like activity is responsible for micronuclear linker histone phosphorylation, we reasoned that the catalytic subunit of this kinase should be present in micronuclei. When an antibody raised against the catalytic subunit of PKA from Paramecium is used to probe a blot of total nuclear protein from extensively purified micro-and macronuclei, a single, strongly immunoreactive band with an apparent molecular mass of 40 kDa is observed in lanes containing micronuclear protein (Fig. 1A). Unexpectedly, at all stages of the life cycle examined (growing, starved, and 2 h mating), macronuclei contain little or no detectable PKA catalytic subunit.
Opposite results are obtained when identical samples are probed with antibodies raised against yeast recombinant p34 cdc2 (Fig. 1B). In agreement with our previous studies (20), a 36-kDa polypeptide is detected in macronuclear samples when probed with an antibody-generated yeast p34 cdc2 . Surprisingly, this polypeptide is not detected in micronuclei isolated from the same cells. Thus, macronuclei and micronuclei are distinguished not only by distinct, non-overlapping linker histones, but also by the kinases that might be responsible for phosphorylating them.
Micronuclear Extracts Phosphorylate Synthetic ␦ Peptide in Vitro-Previous experiments suggested that ␦, the smallest of the micronuclear linker polypeptides, was phosphorylated in vivo on at least one serine located in the carboxyl-terminal third of the protein (1). It was suggested that PKA or a PKAlike kinase was responsible for this phosphorylation since the catalytic subunit of bovine PKA also phosphorylated the same CNBr-generated peptide in vitro. As shown in Fig. 2A, three putative sites for PKA phosphorylation are contained in the carboxyl terminus of ␦. Two serines matching the most stringent PKA consensus sequence (Arg-Arg/Lys-Xxx-Ser, underlined) are shown as black boxes; one serine contained within a less stringent consensus sequence (Arg-Xxx-Xxx-Ser) is indicated as an open box. These two classes of PKA consensus motifs are utilized in the majority of the PKA target sequences (37).
In order to determine whether macro-or micronuclei contain kinase activities capable of phosphorylating PKA or Cdc2 substrates, a salt extract was prepared from each type of nuclei and tested with either of two contrasting peptides. One peptide, containing the three putative PKA sites described above, was synthesized for use as a model PKA substrate (␦ peptide, Fig.  2A); a second peptide, containing two consensus sequences for Cdc2 kinase from macronuclear H1 (H1 peptide, Fig. 2B), was synthesized for use as a model Cdc2 substrate. In vitro phosphorylation with macro-and micronuclear extracts and purified bovine catalytic PKA are shown in Fig. 3. Significant incorporation of [ 32 P]phosphate is observed when the synthetic ␦ peptide is used as a substrate with either crude micronuclear extract or purified bovine catalytic PKA subunit. In contrast, the synthetic H1 peptide is not a good substrate for either of these kinase reactions, although this peptide is an excellent substrate with purified preparations of human Cdc2 kinase (data not shown). This result is consistent with our inability to detect p34 cdc2 in micronuclei (Fig. 1B).
To determine if the PKA activities assayed in Fig. 3 display properties characteristic of PKA in higher organisms, parallel studies were repeated in the presence of the well known, competitive inhibitor of mammalian PKAs (PKI; Ref. 38). Unlike bovine PKA, which is clearly inhibited by this peptide, neither purified Paramecium (39) nor crude Tetrahymena PKA activity is inhibited by 5 mM PKI (Fig. 3). These data suggest that ciliate PKAs as well as PKAs from other sources (40) differ significantly from PKA in higher organisms. Another hallmark of PKA is its dependence on cAMP. While cAMP reproducibly increases the kinase activity of the crude micronuclear extract  1 and 2), starved (lanes 3 and 4), and 2 h mating (lanes 5 and 6) cells were purified by sedimentation at unit gravity. Total protein extracts were resolved on a 12% SDS-acrylamide gel, transferred to nitrocellulose and probed with either antibodies against Paramecium PKA catalytic subunit (A) or yeast p34 cdc2 (B). In agreement with the published results of Hochstrasser and Nelson (57) and Roth et al. (20), bands at 40 and 36 kDa were observed in the blots shown in A and B, respectively. No other cross-reacting bands were observed. Samples used in this blot were identical to those used by Stargell et al. (55) and shown to be highly purified, as indicated by the failure to detect macronuclear type H1 in any of the micronuclear preparations. by about 20%, it does not show a significant requirement for cAMP. The significance of this result is not clear, but may be due to the high salt extraction conditions used to prepare the extract.
Immunodepletion of Kinase Activity from Micronuclear Extract-Antibodies against the Paramecium catalytic subunit were used to precipitate PKA from crude micronuclear extracts (Fig. 4A). A 40-kDa polypeptide is precipitated with the anti-PKA (lane 2), but this peptide is not precipitated with preimmune serum (lane 1) or when extract is not added to the immune reaction (lane 3). Supernatants from these precipitation were then assayed for kinase activity as described above. Preimmune sera has little effect on the kinase activity of the supernatant, while the supernatant from the immune reaction shows an 80% decrease in kinase activity (Fig. 4B). These immunodepletion data are consistent with the suggestion that PKA or a PKA-like kinase is responsible for ␦ phosphorylation in micronuclei.
Mapping of Phosphorylation Sites in ␦-The amino acid sequence of the carboxyl terminus of ␦ suggests that there are three potential phosphorylation sites that could be utilized by PKA. Shown in Fig. 5 (top) are the expected peptides resulting from a limit digestion of synthetic ␦ peptide with lysine-specific (Lys-C) endoproteinase. If ␦ is phosphorylated by PKA using optimal PKA recognition motifs, we predicted that two 32 Plabeled peptides would be produced following Lys-C endoproteinase digestion, one with a single labeled serine, Asn-Ser-Thr-Ser-Lys, and a second with two potential phosphoserines, Arg-Arg-Ser-Ser-Ser-Lys (see the serines in black and white boxes in Fig. 5).
To test this prediction, synthetic ␦ peptide was phosphorylated in vitro by bovine PKA catalytic subunit or the crude Tetrahymena micronuclear extract and subjected to a limit digestion with Lys-C endoproteinase. In addition, RP-HPLCpurified ␦ was phosphorylated in vitro and digested under identical conditions. Phosphopeptides resulting from these digestion were then resolved on a long (30 cm) 50% acid-urea acrylamide gel and identified by autoradiography (Fig. 5). Regardless of the source of PKA, micronuclear extracts or purified PKA from Paramecium (data not shown) or bovine heart, the phosphopeptide maps produced are essentially identical when assayed in this fashion. In addition, these kinases produce identical phosphopeptide maps when either synthetic ␦ peptide (Fig. 5, lanes 2 and 3) or intact ␦ (data not shown) are used as in vitro substrates.
A comparison of in vivo versus in vitro phosphorylated peptides was then performed to determine whether in vitro phosphorylation of ␦ peptide produces a similar map as ␦ phosphorylated in vivo. ␦, labeled in vivo with [ 32 P]orthophosphate, was recovered by RP-HPLC and cleaved with Lys-C endoproteinase. Digestion products were then analyzed as above. These results demonstrate that the in vivo phosphopeptide map of ␦ (Fig. 5, lane 1) is comparable with, but not identical to, that of ␦ phosphorylated in vitro. Interestingly, not all of the peptides are phosphorylated to the same extent. In all cases tested, in vitro and in vivo, peptide c is labeled to the greatest extent. Peptide a (variable in different digests, see below) is phosphorylated to a lesser extent, while peptide(s) b, which resolve to varying extents, is typically poorly phosphorylated. To confirm that each of the above peptides with similar acid-urea polyacrylamide gel mobilities are indeed identical peptides, the bands were excised from the gel and analyzed by thin layer chromatography. In all cases phosphopeptides with the same electrophoretic mobility on the acid-urea gel had the same mobility on the thin layer chromatography plates, and those with different electrophoretic mobilities also differed in chromatographic mobility (data not shown).
PKA Site Utilization on ␦ Is Ordered in Vivo-Phosphorylated isoforms of linker histones migrate with a reduced mobility on SDS gels (for example, see Ref. 41). Previous experiments suggested that, like macronuclear H1, the mobility of ␦ on an SDS gel is reduced by phosphorylation and that ␦ is maximally phosphorylated during early stages of the sexual pathway, conjugation (2). As shown in Fig. 6A, three distinct bands of ␦ are well resolved when HPLC-purified ␦ from 2.5-h mating cells is analyzed in long SDS gels by staining and autoradiography. Densitometric analysis of the stained bands and corresponding autoradiogram indicate that the relative specific activity of the slowest migrating band (labeled S) was approximately twice that of the intermediate mobility band (M) (data not shown), suggesting that ␦ is phosphorylated at two distinct sites during this stage of the life cycle (see below). In contrast, the fastest band (F) is not phosphorylated and most likely represents the dephosphorylated isoform of ␦.
In order to determine if phosphorylation of ␦ occurs in a random or ordered fashion, the labeled bands resolved in Fig.  6A were excised and the individual isoforms extracted. Each band was then separately cleaved with Lys-C endoproteinase, and the digestion products were resolved in a high resolution acid-urea gel. The resulting phosphopeptide map clearly documents a non-random pattern of phosphorylation site utilization in the ␦ isoforms resolved by SDS-gel electrophoresis. Consistent with the idea that the band labeled M is a monophosphorylated isoform of ␦, a single labeled phosphopeptide is observed after Lys-C digestion of isoform M (Fig. 6B, peptide c). In contrast, the slowest migrating species (band S), a putative diphosphorylated isoform of ␦, produces two labeled peptides (b and c) with similar relative mobility to the peptides labeled b and c in Fig. 5.
Identification of Two in Vivo Phosphorylation Sites in ␦-To rigorously determine the phosphorylation sites utilized in ␦, ␦, labeled in vivo with [ 32 P]orthophosphate, was cleaved with NBS (see Fig. 7A) and the resultant single phosphopeptide was purified by RP-HPLC. Direct microsequence analysis verified that this peptide was indeed the indicated carboxyl-terminal portion of ␦ (the length of the arrow in Fig. 7A corresponds to the amino acid sequence obtained). During each cycle of microsequencing, an aliquot of each cleaved phenylthiohydantoinderivative was collected and the presence of 32 P was directly determined by scintillation counting. As anticipated, the first indicated serine, contained within the optimal PKA consensus sequence Arg-Lys-Asn-Ser, is phosphorylated, yielding counts twice that of background (see Fig. 7B). Although sequencing progressed through the second indicated PKA consensus sequence, no additional site of phosphorylation could be detected in this approach.
A second approach utilized peptides generated from Lys-C endoproteinase digestion of in vivo phosphorylated ␦, which were resolved by electrophoresis on acid-urea gels shown pre-viously. Resulting phosphopeptides (labeled a, b, and c in Figs. 5 and 6) were excised from the gel using the autoradiogram as a template, eluted with water, and bound directly to carboxymethylcellulose membrane for sequencing (31). Phosphopeptide a was identified as Asn-Ser-Thr-Ser-Lys by microsequencing (see arrow labeled a in Fig. 7A) and, as expected, the first Ser residue in this peptide is well labeled with 32 P (Fig. 7B). Thus, in agreement with the results obtained with the NBS peptide, serine 201, contained within an optimum PKA consensus site, is confirmed as an in vivo phosphorylation site by this independent method.
Unfortunately, the other in vivo labeled peptides indicated with bracket b and arrow c in Fig. 5, were not be sequenced despite several attempts. However, because in vivo and in vitro phosphorylated samples generate identical phosphopeptides (Fig. 5) another approach was utilized to determine the identity of the b and c peptides. In vitro phosphorylated ␦ peptide was used as a model peptide substrate to confirm the sequences of the remaining in vivo phosphorylated peptides. The catalytic subunit of bovine PKA was chosen for these analyses because of its purity and availability and because the peptide map obtained with this kinase was identical to that obtained with the other PKAs and, more importantly, to in vivo phosphorylated ␦. In vitro labeled ␦ peptide was cleaved with Lys-C endoproteinase and the peptides so generated were isolated by acid-urea gel electrophoresis and processed as described above. Peptide c sequenced as a partial digest product containing the sequence Arg-Lys-Asn-Ser-Thr-Ser-Lys (see Fig. 7A, arrow c). The fact that the intensity of 32 P labeling of peptides a and c varied from preparation to preparation could be explained as variability in the extent of Lys-C endoproteinase digestion (see Fig. 7A). Moreover, the first serine of this in vitro labeled peptide was determined to be the only phosphorylated residue in this peptide. This result demonstrates agreement between in vitro phosphorylation of ␦ peptide by bovine PKA and in vivo phosphorylation of ␦. Interestingly, phosphopeptide c, phosphorylated at serine 201, is the same peptide as the single peptide observed in the digest of bands M and S in Fig. 6B. Collectively, these results strengthen our overall conclusion that serine 201 is a major site of in vivo phosphorylation in ␦.
Unfortunately the b peptide(s) was either not retained or eluted well from the carboxymethylcellulose membrane used to immobilize peptides for microsequencing. For that reason, the b peptide was recovered by cation-exchange HPLC. An early eluting radiolabeled peptide was recovered from this column and analyzed by acid-urea gel electrophoresis to confirm its identity as a b peptide (data not shown). Unexpectedly, upon microsequencing, this peptide was identified as Gly-Lys-Ser-Ser-Val-Ser-Lys, the extreme carboxyl-terminal Lys-C peptide (arrow b, Fig. 7A), with the second serine in this peptide being phosphorylated (Fig. 7B). Since this peptide was phosphorylated in vivo and in vitro with a wide range of PKAs, including bovine, Paramecium (data not shown), and Tetrahymena (Fig.  5), our data suggest that serine 215, two amino acids removed from a single lysine, can be phosphorylated by a wide range of authentic PKAs. Interestingly, peptide b, the slowest migrating of these peptides, is detected only in band S (Fig. 6B). Collectively, these results suggest that PKA first phosphorylates ␦ at serine 201 contained in the optimum consensus sequence Arg-Lys-Asn-Ser and then becomes a maximally phosphorylated, diphosphorylated isoform utilizing serine 215 contained in the less optimal PKA consensus sequence Lys-Ser-Ser-Val.
Interestingly, the second "expected" optimum consensus sequence Arg-Arg-Ser-Ser-Ser is not utilized for phosphorylation in vivo or in vitro with any of the substrates that we have tested. However, we found it is possible to phosphorylate this FIG. 6. Non-random site utilization between mono-and diphosphorylated ␦ isoforms phosphorylated in vivo. A, in vivo labeled ␦ was electrophoresed in a 30-cm-long 10% SDS-polyacrylamide gel and resolved into three bands denoted slow (S), medium (M), and fast (F). The identity of each of the three bands as ␦ was confirmed by immunoblotting using a general anti-␦ antibody (data not shown). The corresponding 32 P autoradiograph of this gel is shown immediately to the right. Densitometric analyses of the stained gel and autoradiograph suggested that the relative specific activity of the slow isoform was twice that of the medium isoform; no 32 P label is associated with the fastest migrating isoform even when the autoradiograph is intentionally overexposed. The two bands labeled S and M were then excised from the gel and cleaved with Lys-C endoproteinase, and digestion products were resolved on an short (10 cm) acid-urea gel. An autoradiograph of this gel is shown in B. The relative mobility of Lys-C endoproteinase peptides labeled b and c (with arrows) are equivalent to those in Fig. 5 but appear different due to the smaller size of the gel used. In this experiment, no radiolabel was observed at the position of peptide a. optimal PKA consensus sequence when the neighboring sequence of the ␦ peptide is altered. Omission of a single serine (serine 205, see Fig. 7A; a mistake made during the original synthesis of the ␦ peptide) that lies between the two optimum PKA consensus sites changes the Lys-C phosphopeptide map of ␦ significantly. Using this "mutated" ␦ peptide, phosphopeptides a and c are not observed and a new, faster migrating phosphopeptide with the sequence Ser-Arg-Arg-Ser-Ser-Ser-Lys is obtained (data not shown). As expected for a PKA optimal consensus motif, microsequence analysis shows that the second serine following the second arginine is phosphorylated (data not shown). The above peptide is phosphorylated when micronuclear extract, Paramecium PKA, or bovine PKA is used as a source of kinase with the altered peptide substrate, and thus as expected, the second "expected" optimal PKA consensus motif is a good in vitro substrate for most PKAs. However, unexpectedly, the Arg-Arg-Ser-Ser-Ser (Ser-209 is underlined) motif is not utilized when Ser-205 is present in the ␦ model peptide substrate. These results suggest that conformation of ␦ plays an important role in determining which sites of phosphorylation are utilized in ␦ in vitro and in vivo. DISCUSSION One of the more unexpected results to emerge from this study is the clear and non-overlapping partitioning of PKA and Cdc2 kinase catalytic subunits between micro-and macronu-clei of Tetrahymena, respectively. Several lines of evidence suggest that PKA or a PKA-like kinase, but not Cdc2, is responsible for phosphorylation of ␦ (and likely other linker histones) in mitotic micronuclei. First, micronuclei, but not macronuclei, contain a 40-kDa polypeptide that is immunoreactive with antibodies raised against the catalytic subunit of Paramecium PKA. Given the clear evidence documenting the regulation of transcription by phosphorylation (42) and, in particular, the involvement of PKA in a subset of these phosphorylation events (43,44), the complete absence of PKA catalytic subunit in transcriptionally active macronuclei is striking and unexpected. Equally remarkable is the absence of p34 cdc2 in micronuclei, given the suspected role of this kinase in a variety of mitosis-associated events (3).
Second, micronuclear extracts and more purified PKAs from Paramecium and bovine heart produced essentially identical in vitro phosphopeptide maps with two in vitro substrates, ␦ or a synthetic ␦ peptide spanning all known potential PKA phosphorylation sites in ␦ (1). Importantly, phosphopeptide maps obtained with all three sources of PKA are comparable with that obtained with in vivo phosphorylated ␦. Third, antibodies against the PKA catalytic subunit immunoprecipitated a 40-kDa protein from micronuclear extracts, and this immune serum, but not preimmune serum, depleted kinase activity from crude micronuclear extracts. Finally, two in vivo sites of phos- ސ‬ asterisk denotes the position of the serine residue that was accidentally omitted during the synthesis of the ␦ peptide (see text for details). Arrows correspond to results obtained by microsequence analyses in B such that the length of each arrow corresponds directly to the appropriate amino acids confirmed by microsequencing. The method used for generating each of the phosphorylated peptides used in the microsequence analyses is as follows: NBS peptide, reverse phase-purified NBS cleavage fragment of in vivo phosphorylated ␦; peptide a, Lys-C endoproteinase-generated peptide of in vivo phosphorylated ␦ that was purified by acid-urea gel electrophoresis, excised and eluted; peptide b, Lys-C endoproteinase-generated peptide of in vivo phosphorylated ␦ that was purified by acid-urea gel electrophoresis as well as in vitro phosphorylated ␦ peptide that was purified by cation-exchange chromatography; peptide c, Lys-C endoproteinase-generated peptide of in vitro phosphorylated ␦ peptide that was purified by elution from an acid-urea gel. B, histograms of direct measurements of [ 32 P] in the phenylthiohydantoin-derivatives during microsequencing. NBS and Lys-C endoproteinase peptides were generated as described above. As indicated by the length of the arrow in A above, essentially the entire in vivo labeled NBS peptide from ␦ was sequenced (from Gly-188 to Ser-217). However, only [ 32 P] counts/min data between Thr-196 and Arg-206 are plotted; no other counts/min above background levels were detected. phorylation in ␦ have been identified. Both sites conform with recognition sites for PKA (37), although other kinases with similar sequence requirements have been reported (45). Collectively, these results lend strong support to the idea that PKA or a PKA-like kinase is present in micronuclei and that this activity is at least partially responsible for phosphorylation of ␦ in vivo.
Our data suggest, however, that several differences exist between ciliate and mammalian PKAs. Unlike bovine PKA, neither Paramecium nor crude Tetrahymena PKA activity is inhibited by PKI. These results are in agreement with the findings that Paramecium and Dictyostelium PKAs, respectively, are poorly inhibited by this peptide (39,40). Although, the catalytic core of the PKA is highly conserved between most species (Ref. 46; reviewed in Ref. 47), sequence differences in the recognition site between isomers of the PKA subunit in yeast and mammalian cGMP-dependent protein kinase reduce these enzymes' affinities for the Kemptide substrate and PKI (48,49).
Our data strongly suggest that phosphorylation of ␦ by PKA is influenced by both protein sequence and conformation. Studies using model substrates with mammalian PKAs have suggested that substrates containing a pair of basic residues (usually arginines) one to three amino acids removed from the site of phosphorylation are generally favored over substrates with a single basic residue (37). Based upon this general rule, two putative sites of phosphorylation by PKA were predicted from the sequence of the carboxyl terminus of ␦ (1).
Our data demonstrate, however, that the above prediction is upheld for only one of these sites. Serine 201, contained within the optimal consensus sequence Arg-Lys-Asn-Ser, is the exclusive site of monophosphorylation within ␦ in vivo and is a highly preferred site in vitro. Unexpectedly, a second serine 209 contained within an optimal PKA recognition motif, Arg-Arg-Ser-Ser-Ser, is not utilized in vivo or in vitro. However, PKA, regardless of source, did utilize serine 209 when a synthetic peptide spanning this region, but missing serine 205, was used a substrate. Since this sequence is not utilized in vivo or in vitro when the correct ␦ peptide is used as substrate, we favor the interpretation that some aspect of ␦'s conformation, possibly altered by the absence of serine 205, lowers the efficiency that this site is utilized. Surprisingly, a second serine 215, embedded within the sequence Lys-Ser-Ser-Val is phosphorylated when ␦ is phosphorylated in vitro or in vivo (site b, Fig. 6A). This less optimal PKA phosphorylation site contains a hydrophobic valine one amino acid after the phosphorylation site, a requirement shown to contribute to efficient substrate binding with mammalian PKA (50). These data support our contention that the overall structure of ␦ plays an important role in site utilization by PKA in vivo.
Previous studies have indicated that micronuclear linker histones are extensively phosphorylated in growing and mating cells, but not in starved cells (21). Our results (Fig. 1) demonstrate that PKA catalytic subunit is detected in micronuclei from growing, starved, and young mating cells, and at first glance, it appears that the amount of the 40-kDa subunit does not change appreciably with the physiological state of the cell. However, the crystal structure of PKA (50) demonstrated the importance of post-translational phosphorylation (at Thr-197) and suggested possible roles of phosphorylation in promoting PKA activation (reviewed in Ref. 47). Whether Tetrahymena PKA undergoes physiologically regulated modification during its life cycle is unknown.
Recent data suggest that an increasingly important mechanism in the role of phosphorylation events is the subcellular location of kinase or phosphatase subunits (reviewed in Ref. 51). For example, activation of PKA also requires a well known, cyclic AMP-dependent dissociation of catalytic subunit from an inhibitory regulatory subunit. Studies of fluorescence-tagged PKA injected into mammalian cells found that upon dissociation the free catalytic subunit moved from the cytoplasm to the nucleus while the regulatory subunit remained in the cytoplasm (52,53). These data suggest the possibility that the activity of the catalytic subunit in the micronucleus may be modulated by a regulatory subunit that is selectively targeted to micronuclei during select stages of the life cycle. Along this line, the R subunit of PKA has been detected in Dictyostelium nuclei, and, interestingly, its amount in nuclei increases upon differentiation (54).
Although the biological function of linker histone phosphorylation is presently unknown, it has been proposed that it may facilitate chromatin decondensation facilitating factor access to DNA (15). Condensed chromatin, in contrast, is thought to be stabilized by dephosphorylated linker histone isoforms. Consistent with this model is the recent finding that phosphorylation of ␦ in micronuclei is linked (temporally and spatially) to a transient period of chromatin decondensation and transcriptional activation during meiotic prophase (2). We hypothesize that phosphorylation of the two carboxyl-terminal PKA sites identified in ␦ in this study acts to destabilize or decondense micronuclear chromatin, thereby facilitating the binding of factors required for activation including unique histone variants such as hv1 (55). Moreover, Lamb and co-workers (56) have demonstrated that inhibition of PKA in living mammalian cells results in rapid chromatin condensation at all phases of the cell cycle. Based on these results, we speculate that phosphorylation of PKA sites in mammalian H1s may play a previously unsuspected role in chromatin decondensation.