Protein Kinase Mζ Synthesis from a Brain mRNA Encoding an Independent Protein Kinase Cζ Catalytic Domain

Protein kinase Mζ (PKMζ) is a newly described form of PKC that is necessary and sufficient for the maintenance of hippocampal long term potentiation (LTP) and the persistence of memory in Drosophila. PKMζ is the independent catalytic domain of the atypical PKCζ isoform and produces long term effects at synapses because it is persistently active, lacking autoinhibition from the regulatory domain of PKCζ. PKM has been thought of as a proteolytic fragment of PKC. Here we report that brain PKMζ is a new PKC isoform, synthesized from a PKMζ mRNA encoding a PKCζ catalytic domain without a regulatory domain. Multiple ζ-specific antisera show that PKMζ is expressed in rat forebrain as the major form of ζ in the near absence of full-length PKCζ. A PKCζ knockout mouse, in which the regulatory domain was disrupted and catalytic domain spared, still expresses brain PKMζ, indicating that this form of PKM is not a PKCζ proteolytic fragment. Furthermore, the distribution of brain PKMζ does not correlate with PKCζ mRNA but instead with an alternate ζ RNA transcript thought incapable of producing protein. In vitro translation of this RNA, however, generates PKMζ of the same molecular weight as that in brain. Metabolic labeling of hippocampal slices shows increased de novo synthesis of PKMζ in LTP. Because PKMζ is a kinase synthesized in an autonomously active form and is necessary and sufficient for maintaining LTP, it serves as an example of a link coupling gene expression directly to synaptic plasticity.

LTP 1 is a persistent enhancement of synaptic transmission widely studied as a physiological model of memory (1). LTP can be divided into two phases: induction, which triggers the potentiation, and maintenance, which sustains it over time. Many molecules have been implicated in LTP induction, which is initiated by the activation of N-methyl-D-aspartate (NMDA) receptors and involves several protein kinases (2). In contrast, very little is known about the molecular mechanism of maintenance. Recently, however, a specific, autonomously active form of the atypical PKC isozyme (3,4), PKM, has been found both necessary and sufficient for maintaining LTP (5)(6)(7). Overexpression of PKM also prolongs memory in Drosophila melanogaster, suggesting it is part of an evolutionarily conserved molecular mechanism for memory storage (8).
The unique role of PKM in LTP maintenance is due, in part, to its unusual structural and enzymatic properties as an autonomously active kinase. PKM consists of the independent catalytic domain of a PKC isoform (5). PKC isoforms are divided into three classes: conventional, novel, and atypical (reviewed in Refs. 9 -11). Each isoform is a single polypeptide consisting of an N-terminal regulatory domain and a C-terminal catalytic domain linked by a hinge (Fig. 1A, left). The regulatory domain contains binding sites for second messengers and an autoinhibitory pseudosubstrate sequence, which interacts with and blocks the active site of the catalytic domain. Second messengers stimulate PKC by binding to the regulatory domain, translocating the enzyme from cytosol to membrane, and producing a conformational change that releases the autoinhibition. In vitro studies have shown that PKC may then be cleaved at its hinge, permanently removing the regulatory domain to form the independent catalytic domain, PKM (Fig. 1A, left) (12,13). Lacking autoinhibition from a regulatory domain, PKM can persistently phosphorylate substrates in the absence of second messengers. Indeed, once formed in LTP, the autonomous activity of PKM maintains synaptic potentiation for at least several hours (7).
Because PKM is usually thought of as a proteolytic fragment of PKC (12,13), our early work had assumed that PKM was formed from PKC by calpains (5), Ca 2ϩ -dependent proteases that become active during LTP (14) and can generate PKM in vitro (12). Our subsequent studies showed, however, that the increase of PKM in LTP, like the persistence of synaptic potentiation, requires new protein synthesis (6). These results, together with the observation that PKC is the only PKC isozyme with a stable PKM form in hippocampus (5,15), suggested that PKM may be generated by a molecular mechanism unique to .
One possible mechanism for generating PKM might be that a product of protein synthesis regulates PKC proteolysis. To characterize this mechanism the amount of PKC, the putative precursor, needs to be established. Vertebrates, however, express two atypical PKCs, and /; traditional antisera to the C terminus of PKC recognize proteins from both genes (16,17). (PKC (16) and - (17) are human and mouse orthologues, respectively.) Thus the relative amounts of PKC and PKC/ in brain have not been determined.
Alternatively, PKM could be formed by a novel mechanism as a gene product distinct from PKC. A RNA containing a partial C-terminal PKC sequence has been identified (18,19), which is produced by an internal promoter active in certain rat prostate tumor cell lines and in rat brain (20). This alternate RNA, however, is thought not to produce a protein and has been referred to as an untranslatable " pseudogene" RNA (21).
Here we show PKM synthesis from this alternate RNA, which we now call PKM mRNA. Because it is a gene product distinct from PKC, PKM is a new atypical PKC isoform, and the protein synthesis-dependent mechanism for generating PKM in brain may be a core molecular mechanism for the maintenance of LTP.

EXPERIMENTAL PROCEDURES
Experimental animals were used in accordance with the State University of New York Downstate Medical Center Institutional Animal Use Committee.
Antisera Production-Peptides were synthesized by Quality Control Biochemicals (Hopkinton, MA) and correspond to N-terminal, hinge, catalytic, and C-terminal regions of PKC (see Fig. 1A for sequences). The peptides were coupled to bovine serum albumin (Pierce), mixed with Titermax Gold (CytRx Corp., Norcross, GA), and injected intramuscularly into female New Zealand rabbits. After 1-3 boosts at 4-week intervals, the antisera were affinity-purified on peptide conjugated Sulfolink gel columns (Pierce). The C-terminal antibody used for immunoprecipitation was obtained from Santa Cruz Biotechnology (Santa Cruz, CA). The monoclonal antibody was from Transduction Laboratories (Lexington, KY).
Immunoprecipitation-Rat cytosolic hippocampal extract (100 g, 400 g/ml) was added to 150 l of immunoprecipitation (IP) buffer (10 mM Tris-HCl (pH 7.4), 150 mM NaCl, 2 mM EDTA, 1% Nonidet P-40, 0.5% aprotinin, 1% each of phosphatase inhibitor mixtures I and II (Sigma)) and precleared with 10 l of normal goat IgG-agarose conjugate for 1 h at 4°C with nutation. The samples were then centrifuged at 2300 ϫ g for 5 min at 4°C, and the supernatants were incubated with 1.6 g of goat C-terminal antibody (Santa Cruz Biotechnology, Santa Cruz, CA) or 1.6 g of normal goat serum overnight. For peptide competition, the precipitating antibody was preincubated with 8 g of immunizing peptide for 2 h at room temperature. Immunocomplexes were collected by incubation with 20 l of protein G-plus agarose for 6 h, followed by centrifugation, as above. The pellet was washed 3 times with 400 l of IP buffer without aprotinin and boiled at 95°C for 5 min with 40 l of 5ϫ sample buffer.
Calpain Cleavage-A mixture (200 l) containing calpain (1.25 units of either porcine calpain I or II, Calbiochem), baculovirus-Sf9 expressed PKC (1 g), prepared as previously described (7), 20 mM Tris-HCl (pH 7.5), and 2 mM dithiothreitol was prewarmed for 2 min at 20°C, and the proteolytic reaction was started with the addition of 20 l of CaCl 2 (7.5 mM final concentration). After 1 h, the reaction was stopped with the addition of 12.5 l of 25 mM EDTA and placed on ice. Proteolytic fragments were analyzed with C-terminal antiserum by Western blot, as described above.
Total mRNA Isolation-For reverse transcription (RT)-PCR, 60 mg of various tissues were used for the single-step method of total RNA isolation by guanidinium thiocyanate/phenol/chloroform extraction (22). Total RNA sample for RT was treated with DNase I and repurified by the same method. The quality of the total RNA was measured by formaldehyde gel electrophoresis and spectrophotometry. For RNase protection, Northern blotting, and 5Ј-RACE with oligo-capping, the cesium chloride RNA isolation method was used. Briefly, 200 mg of tissues from a 3-month old male Sprague-Dawley rat was homogenized in 2 ml of guanidinium thiocyanate homogenization buffer (4.0 M guanidinium thiocyanate, 0.1 M Tris-HCl, 1% ␤-mercaptoethanol, 0.5% N-laurylsarcosine) for 30 s. The homogenate was centrifuged for 10 min at 3,000 ϫ g. The supernatant was transferred to a cushion of 2 ml of 5.7 M CsCl and 10 mM EDTA (pH 7.5) and centrifuged for 16 h at 81,000 ϫ g. The pellet containing total RNA was resuspended in 300 l of diethyl pyrocarbonate-treated water by heating to 70°C for 10 min, 33 l of 3 M sodium acetate (pH 5.2), and 825 l of 100% ETOH was added, and the sample was stored at Ϫ80°C.
RT-PCR-Total RNA was used to synthesize cDNA with the Super-Script Preamplification System for First Strand cDNA Synthesis kit (Invitrogen). Two hundred ng of cDNA was used in a 100-l final volume PCR. Amplification was for 34 cycles with 94°C for 30 s, 60°C for 1 min, and 72°C for 1 min as cycle parameters, with a final step of 72°C for 10 min. For amplification of PKC and PKM cDNAs, specific forward primers were F 5Ј-CCATGCCCAGCAGGACCACC-3Ј and F 5Ј-CCTTCTATTAGATGCCTGCTCTCC-3Ј, respectively, and R 5Ј-TGAA-GGAAGGTCTACACCATCGTTC-3Ј was the reverse primer for both. As a control we used glyceraldehyde-3-phosphate dehydrogenase primers, F 5Ј-ACATGGTCTACATGTTCC-3Ј and R 5Ј-CAGATCCACAAC-GGAATAC-3Ј.
Northern Blot-Total RNA (30 g) was electrophoresed and transferred to nitrocellulose, rinsed, and UV cross-linked. Digestion with EcoRI-SphI and KpnI-EcoRI gave a 457-and 227-bp specific fragment for PKC and PKM, respectively. The fragments were radiolabeled with 32 P using a Stratagene random octamer protocol (Stratagene Cloning Systems, La Jolla, CA). Hybridization conditions were performed according to instructions for Stratagene QuickHyb hybridization. Blots were developed overnight by film exposure at Ϫ70°C or by Phospho-rImager (Storm 860 gel and blot imaging system, Amersham Biosciences).
Oligo-capping and 5Ј-RACE-Oligo-capping and 5Ј-RACE (24) were performed with a GeneRacer Kit (Invitrogen), using rat hippocampus and kidney RNA as templates, according to the manufacturer's instructions. The GeneRacer Kit 5Ј primer was used as a forward primer and 5Ј-CGCCCAGCCATCATCTCAAACATAAG-3Ј as a reverse primer specific for PKM. For screening of oligo-capping 5Ј-RACE clones, PCR was performed using the GeneRacer 5Ј-nested primer and 5Ј-CTCTGC-CTCTTCAGCACG-3Ј, as a reverse PKM-specific nested primer. The PCR conditions were 30 cycles at 94°C for 30 s, 60°C for 30 s, and 72°C for 30 s, with an initial step of 94°C for 3 min and a final step of 72°C for 10 min. 5Ј-RACE without tobacco alkaline phosphatase yielded no clones.
GenBank TM accession numbers of the published human and mouse PKM cDNAs obtained by oligo-capping and 5Ј-RACE are as follows: AL514298, AL534124, AL535166, AL535303, AL538613, AL538626, and AL539647 (human), and AU051559, AU067133, AU078878, and BB585441 (mouse). Throughout the text, the first nucleotide in the longest rat PKM mRNA, shown in Fig. 4D, is referred to as the first nucleotide of PKM mRNA.
In Vitro Transcription-Translation-In vitro transcription-translation was performed with the TNT-coupled wheat germ extract system kit, according to the manufacturer's instructions (Promega, Madison, WI). For the T7-PKM clone (sequence 48 -1982), we used a PKM clone (gift from T. Powell, Cleveland Clinic, Cleveland, OH). Briefly, PKM was digested with AatII and XbaI and the pBlueScript SK(Ϫ) vector with HindIII and XbaI. The XbaI sites were ligated, and the HindIII site of the vector was filled in and ligated with a polished AatII site from the PKM. The T7-PKM-(347-1982) clone was created by digesting PKM- (48 -1982) with EcoRI and XbaI and subcloned into pGEM-3Z (Promega, Madison, WI). The SP6-PKM-(587-1982) clone was created by digesting PKM- (48 -1982) with PstI and XbaI, and subcloning into pGEM-3Z (Promega, Madison, WI). The experiment with the T7 clone was performed with linearized plasmid digested with XbaI. The experiment with the SP6 clone was performed both with and without linearization at the XbaI site with identical results.
Hippocampal Slice Electrophysiology-Transverse hippocampal slices (450 m) were prepared with a McIlwain tissue slicer from 16-to 25-day-old Sprague-Dawley rats. Slices were placed in an interface chamber infused initially with a saline solution containing 125 mM NaCl, 2.5 mM KCl, 1.25 mM NaH 2 PO 4 , 26 mM NaHCO 3 , 11 mM glucose, 10 mM MgCl 2 , and 0.5 mM CaCl 2 (pH 7.4), equilibrated with 95% O 2 , 5% CO 2 at 32°C. After 30 min, the divalent ion concentrations were changed to 1.2 mM MgCl 2 and 1.7 mM CaCl 2 (resting buffer) (5). Test stimuli (duration 100 s) were delivered every 15 s through bipolar tungsten electrodes placed across the Schaffer collateral/commissural fibers. Field EPSPs were recorded using glass microelectrodes filled with the saline solution (resistance 2.5-5.0 megohms) and placed in CA1 stratum radiatum. The current intensity of test stimuli (25-50 A) was set to produce one-third maximal EPSPs. Analysis of the initial 10 -50% of the field EPSP slope was performed with Superscope (GW Instruments, Somerville, MA). Following 15 min of stable recordings, LTP was induced with two 100-Hz 1-s tetanic trains, 20 s apart. Control slices received test stimulation for equivalent periods of time as LTP recordings.
Metabolic Labeling of Hippocampal Slices-Following stimulation, hippocampal slices were transferred to a 6-well culture dish in a 32.5°C water bath. Each slice was submerged in a well containing 8 ml of resting buffer, continuously perfused with 95% O 2 , 5% CO 2 , and supplemented with 350 Ci of 35 S-labeled amino acids (PerkinElmer Life Sciences). After incubation for 30 min, slices were quickly frozen by contact with a metal rod cooled to Ϫ55°C. In some experiments, cold methionine (300 M) was added prior to freezing; the results with and without the addition were equivalent. The frozen slices were placed on powdered dry ice and stored at Ϫ80°C.
Each radiolabeled hippocampal slice was then homogenized in 250 l of homogenization buffer (described above) with 0.5% SDS in order to solubilize total protein and centrifuged at 3,000 ϫ g at 4°C for 5 min to remove unhomogenized tissue. Samples containing 50 g of protein and 90 l of IP buffer in a final volume of 290 l were precleared with 15 l of protein G-plus agarose and immunoprecipitated with C-terminal aPKC antiserum, as described above.
Following immunoprecipitation and washing, pellets were homogenized and boiled in 40 l of sample buffer, run on 10% SDS-PAGE, and electroblotted onto nitrocellulose membranes. The membranes were exposed to a PhosphorImager screen for 3 months. The images were analyzed using ImageQuant version 1.2 software for the Macintosh platform, and the intensity of radiolabeled bands was determined by subtracting background values.

PKM Is Expressed Exclusively in Brain-Vertebrates ex-
press two atypical isoforms, PKC and PKC/ (16,17). Because PKC/ is nearly identical to in its C terminus, antisera raised against this region of PKC recognize both atypical isozymes (16) (Fig. 1A, right). We used this antiserum to screen a panel of tissues by Western blot to determine the overall distribution of atypical PKC and PKM forms (Fig. 1B, left). Whereas fulllength aPKC (mass of 72 kDa) was found in many tissues, notably lung, liver, kidney, testis, and brain, atypical PKM (55 kDa) was expressed only in brain. The relative abundance of PKM in neocortex and hippocampus, two forebrain regions, compared with cerebellum, a hindbrain region, is consistent with our previous observation of the preferential rostral to caudal distribution of PKM in brain (15). We next examined the distribution of PKC/ with an antiserum raised against its catalytic domain (Fig. 1B, right), which does not cross-react with PKC (see Fig. 1C, left). Full-length PKC/ was expressed in lung, liver, kidney, testis, and brain, similar to the distribution of total aPKC recognized by the C-terminal antiserum, but only a trace amount of PKM/ was detectable in brain. These results suggest that the major form of atypical PKM in brain is PKM.
To confirm that the 55-kDa protein recognized by the Cterminal PKC antiserum is PKM and to determine the level of expression of PKC, we developed a battery of antisera raised against different isozyme-specific epitopes in the Nterminal, hinge, and catalytic domains of PKC (Fig. 1A, right). By Western blot, all the antisera to , but not that to PKC/, recognized baculovirus-overexpressed PKC in Sf9 cells (Fig.  1C, left). In kidney, antisera to both and / detected a 72-kDa band (Fig. 1C, center), indicating the expression of both fulllength aPKCs. In hippocampus, however, the PKC/ and the aPKC C-terminal antisera both recognized a 72-kDa band, but none of the -specific antisera detected full-length PKC (Fig.  1C, right). Overdevelopment of Western blots with the -specific antisera showed only trace amounts of PKC in hippocampus (data not shown). The full-length aPKC recognized by the aPKC C-terminal antisera in hippocampus is thus predominantly PKC/, and the expression of PKC is very low.
In contrast, PKM was abundantly expressed in hippocampus (Fig. 1C, right). The antisera to the hinge, catalytic, and aPKC C-terminal all recognized PKM in the hippocampus, but, as expected, the N-terminal antiserum did not. We confirmed that the protein detected by the aPKC C-terminal antiserum was PKM, by showing that this antiserum immunoprecipitated a 55-kDa protein from hippocampal extracts that was recognized by the -specific catalytic domain antiserum on Western blot (Fig. 1D).
We next determined whether the relative abundance of PKM and the paucity of PKC that we observed in hippocampus was found in other brain regions (Fig. 1E). The N-terminal and catalytic domain antisera (Fig. 1E, left and middle) showed PKC expression in kidney and cerebellum but did not detect PKC in either neocortex or hippocampus. In contrast, the catalytic and aPKC C-terminal antisera (Fig. 1E, middle and right) demonstrated PKM expression in all brain regions. Therefore, PKM and PKC show differential regional expression in brain, with both PKM and PKC in cerebellum, but only PKM in hippocampus and neocortex.
Brain PKM Is Not a Proteolytic Fragment of PKC-The expression of PKM in the near absence of PKC in hippocampus and neocortex suggests that PKM may not be produced by PKC proteolysis. It is possible, however, that the cleavage of PKC in these regions was complete. As our original hypothesis for the formation of PKM was proteolysis by calpain (5), we compared the size of brain PKM with the PKM fragment generated by calpain proteolysis of recombinantly expressed PKC. Both calpains I and II produced PKM fragments that appeared smaller than endogenous brain PKM ( Fig. 2A).
Although these results suggest that calpain proteolysis does not produce PKM of the appropriate size, other proteases might cleave PKC at alternative sites (25). Therefore, to determine whether brain PKM is a proteolytic fragment of PKC, we examined mice in which PKC was eliminated by genetic disruption of its regulatory domain (PKC-reg Ϫ/Ϫ) (26). In mouse, PKC is a single copy locus at chromosome 4, 83.0 centimorgans. In the PKC knockout characterized by Leitges et al. (26), a Lac/Neo cassette was inserted into the sequence encoding amino acids 112-140 of the regulatory domain (exon 5), thereby disrupting the regulatory domain but sparing its catalytic domain. Western blots with the -specific catalytic domain antiserum showed that although the homozygous PKC-reg knockout completely lacked PKC in cerebellum and kidney, PKM in brain was preserved (Fig. 2B). A second knockout line showed identical results (data not shown). Com-FIG. 1. PKM is a brain-specific form of PKC prominent in hippocampus and neocortex. A, left, models of the inactive and active conformations of PKC and the proteolytic formation of PKM. PKC consists of a C-terminal catalytic domain (green) tethered by a hinge (yellow) to an N-terminal regulatory domain (red), which contains an autoinhibitory pseudosubstrate sequence. PKC is maintained in an inactive state by the interaction between the pseudosubstrate and the catalytic site. PKC is activated by second messengers, which produce a conformational change that releases the autoinhibition. PKM is the independent catalytic domain of PKC, usually thought of as a product of proteolysis at the hinge. Because PKM lacks a regulatory domain, it is constitutively active. Right, diagram showing epitopes used to generate antisera in the regulatory, hinge, and catalytic domains of PKC and PKC/. The antiserum to the C terminus of (aPKC C term) recognizes both atypical isoforms. PKC-specific antisera are to the N-terminal (N term), hinge, and catalytic domain. A PKC/-specific antiserum is to its catalytic domain. B, left, pared with wild-type, heterozygous PKC-reg knockout animals (PKC-reg ϩ/Ϫ) expressed diminished levels of PKC, but also no loss of PKM (Fig. 2B). These results demonstrate that brain PKM is expressed in the complete absence of PKC, and therefore proteolysis or other post-translational modification of PKC is not the mechanism of PKM formation.
The Distribution of an Alternate PKM mRNA, Encoding an Independent Catalytic Domain, Correlates with PKM Protein-A novel mechanism for the formation of PKM could be its direct synthesis as a gene product other than PKC. The gene produces two sets of RNAs: a full-length PKC mRNA and a second potential PKM mRNA, previously referred to as Ј (23) or the " pseudogene transcript" (21) (Fig. 3A, top). The 5Ј end of this RNA is a unique sequence not present in PKC mRNA, whereas its 3Ј end is identical to the 3Ј end of PKC mRNA, consisting of a partial regulatory domain and a com-plete hinge, catalytic domain, and 3Ј-UTR (18,20,21,23,27). In contrast to PKC mRNA, however, the 5Ј-terminal of PKM mRNA does not contain an AUG that could initiate translation of the partial regulatory domain. Instead, the open reading frame (ORF) of the kinase sequence in PKM mRNA begins in the hinge and extends to the C terminus of the kinase domain (Fig. 3A, top). To determine whether the patterns of expression of the PKC and PKM mRNAs correlate with the PKC and PKM proteins, we analyzed their distribution by RT-PCR, RNase protection, and Northern blot analysis (Fig. 3).
By using specific forward primers that distinguish between the two mRNAs, RT-PCR analysis showed abundant expression of PKM mRNA in brain but not in non-neural tissues (Fig.  3A, bottom). Only with a higher number of PCR cycles could a small amount of PKM mRNA be detected in kidney (data not shown). In contrast, PKC mRNA was expressed in kidney, Western blot of rat tissues with aPKC C-terminal antiserum shows atypical PKM is specifically expressed in brain. High levels of full-length aPKC are in lung, liver, kidney, testis, and brain. Right, Western blot with the PKC/ catalytic antiserum shows expression of full-length PKC/ in lung, liver, kidney, testis, and brain but only trace amounts of PKM in brain.  2 and 3, respectively, showing most of the PKM was immunoprecipitated. Because of differences in volumes after the immunoprecipitation, less supernatant (S) than pellet (P) protein was loaded on the gel. E, Western blot shows PKM and PKC expression in various brain regions and kidney. Left, N-terminal antiserum shows strong expression of PKC in kidney, low expression in cerebellum, and no expression in hippocampus and neocortex. Center, the catalytic antiserum shows identical results for PKC, but also expression of PKM in brain tissues but not kidney. Right, the aPKC C-terminal antiserum shows aPKC in all tissues and PKM only in brain. All experiments in the figure were performed in triplicate with equivalent results.
FIG. 2. Brain PKM is not formed by proteolysis of PKC. A, Western blot with aPKC C-terminal antiserum shows calpain I and II proteolysis of recombinantly expressed PKC produces a cleaved PKM fragment smaller than endogenous brain PKM. B, PKM is expressed in a PKC knockout mouse in which the regulatory domain was targeted, but the catalytic domain was spared (PKC-reg Ϫ/Ϫ). Western blot with -specific catalytic domain antiserum shows PKC expression in wildtype (PKC-reg ϩ/ϩ) and heterozygous (PKC-reg ϩ/Ϫ) mice in both kidney and cerebellum but not in the knockout (PKC-reg Ϫ/Ϫ). Brain PKM is expressed in all animals. Experiments were performed in triplicate with equivalent results.
lung, testis, and cerebellum but not in neocortex or hippocampus (Fig. 3A, bottom). Thus the distributions of PKC and PKM mRNAs strongly correlate with PKC and PKM pro-teins, with both PKM protein and mRNA the exclusive forms of expressed in forebrain (compare Fig. 1E).
We confirmed and quantified these results using RNase pro- Bottom, PCR products of 5Ј end clones obtained by oligo-capping with 5Ј-RACE of PKM mRNA, showing expression of multiple length 5Ј ends in rat hippocampus (left) and a single length 5Ј-terminal in kidney (right). C, frequency of rat brain PKM mRNA clones obtained with the numbered start sites shown in D. The PKM mRNA clones from kidney all began at start site 3. D, alignment of the complete rat, mouse, and human PKM exon 1Ј with adjacent 5Ј-intronic sequence. Small asterisks denote identical sequences. Colored arrows show the 5Ј-PKM terminal obtained by oligo-capping and 5Ј-RACE for rat (red), mouse (blue), and human (green) PKM mRNAs. Black arrow with large asterisk denotes the start site of PKM mRNA (Ј) from the rat Dunning G prostate tumor cell line (20). A canonical CRE (bar) is conserved in all three species. Humans have a partial duplication of the CRE as an adjacent insert. tection with an antisense probe that protects a 345-nucleotide fragment of PKC mRNA and a 202-nucleotide fragment of PKM mRNA (Fig. 3B, top). Consistent with the RT-PCR analysis, the RNase protection product of PKM mRNA was found only in brain, whereas the protection product of the PKC mRNA was observed in kidney, lung, testis, and cerebellum but not in neocortex or hippocampus (Fig. 3B, center). We determined the relative levels of the two mRNAs by comparing them to an mRNA of the housekeeping gene, rat acidic ribosomal protein (RARP, Fig. 3B, center and bottom). The expression of PKM mRNA in brain was higher than that of PKC mRNA in any tissue examined.
We then determined the sizes of the PKC and PKM mRNAs in different brain regions and kidney by Northern blot, using probes specific to their unique 5Ј ends (Fig. 3C, top). The PKM mRNA is found as a 2.3-and 4.7-kb species in brain but is not expressed in kidney (Fig. 3C, left). These sizes are similar to those reported previously (18,21,27), with the two species possessing different length 3Ј-UTRs and distinct polyadenylation signals (27). PKC mRNA is expressed as a 2.5-and minor 4.8-kb species in kidney and cerebellum but not in neocortex or hippocampus (Fig. 3C, right). Northern blots of mouse tissue showed similar results (data not shown). These results demonstrate that the expression of PKM mRNA strongly correlates with PKM protein.
Transcription of Brain PKM mRNA from an Internal Promoter within the PKC Gene-PKM mRNA is transcribed by an internal (intronic) promoter within the PKC gene in rat brain and certain rat prostate tumor cell lines (20,23). This promoter initiates transcription at exon 1Ј, which encodes the unique 5Ј sequence of the rat PKM mRNA (20). To determine whether this mechanism is evolutionarily conserved, we examined the organization of the human and mouse PKC genes. Human PKC is a single copy locus at chromosome 1p36 (Fig.  4A). Both the human and mouse PKC genes consist of two clusters of exons: exons 1-4, encoding the 5Ј-UTR and the initial regulatory domain sequence in PKC mRNA, and exons 5-18, encoding the remaining regulatory domain sequence, hinge, catalytic domain, and 3Ј-UTR (Fig. 4A). These two clus-ters are separated by a large intron: 75.7 kb in human and at least 53.4 kb in mouse. As described in the rat (20), we found both the human and mouse PKC genes contain the unique 5Ј sequence of PKM mRNA in a single exon 1Ј nested between the two clusters (Fig. 4A). Furthermore, although predicted to be part of the 5Ј-UTR of PKM mRNA, the sequence of exon 1Ј was highly conserved among the three species (Fig. 4D). Thus, the PKC/PKM gene structure is evolutionarily conserved.
We next examined the promoter region of PKM mRNA by determining the 5Ј terminus of PKM mRNA. Powell et al. (23) had used 5Ј-RACE to sequence PKM mRNA, which may not yield the complete 5Ј terminus. Therefore, we employed the oligo-capping and 5Ј-RACE method that enriches for capped mRNAs, which has been used to determine the transcriptional start site of mRNAs (24). By using a reverse primer specific to PKM mRNA (Fig. 4B, top), we found that the 5Ј termini from all 32 clones obtained from hippocampus by oligo-capping and 5Ј-RACE exhibited unique sequences within exon 1Ј but no PKC sequence (Fig. 4, B-D). This result confirms that PKM mRNA is produced by an internal promoter followed by alternative splicing and not by alternative splicing of PKC mRNA (which would have resulted in a PKC 5Ј-terminal exon in PKM mRNA). Heterogeneity of start sites was observed by oligo-capping 5Ј-RACE for PKM mRNA produced in brain (Fig. 4B, left, and C-D), but only a single start site was observed for the low abundant PKM RNA from kidney (Fig. 4B,  right). Although the 5Ј terminus of the longest rat brain PKM mRNA clone was upstream of the site of transcription initiation reported for the alternate RNA of prostate tumor cells (20), the majority of the brain PKM mRNA start sites were downstream of this site (Fig. 4D). Analysis of the 5Ј ends of the published oligo-capped 5Ј-RACE products from human and mouse PKM mRNAs also showed only PKM exon 1Ј and no PKC sequences and were heterogeneous in length (Fig. 4D). Therefore, transcription from a dedicated internal promoter within the PKC gene is the primary mechanism for PKM mRNA formation in all three species.
Analysis of the promoter region of the PKM mRNA showed a conserved canonical cAMP-response element (CRE) (Fig. 4D),  1-4) has the same molecular weight as endogenous brain PKM from hippocampal homogenate (lane 6). Experiments were performed in triplicate with equivalent results. first identified in the rat (20). We find the human PKM promoter contains an additional partial CRE duplication in an insert immediately 3Ј to the canonical CRE. Other putative transcription factor binding sites found in the PKM mRNA promoter region in all three species include those for nuclear factor-B (NF-B, Ϫ74 to the longest rat PKM transcript) and CCAAT/enhancer-binding protein (C/EBP, Ϫ272).
In Vitro Translation of PKM mRNA-Although the AUG in the hinge of PKM mRNA (corresponding to methionine 184 of PKC (27)) is within a putative Kozak sequence (28), a previous attempt to translate PKM mRNA did not produce detectable protein in the rabbit reticulocyte system (21), a result which we confirmed (data not shown). However, the sequence 5Ј to the kinase ORF is long (595 bp) and contains 6 short ORFs, which may serve to inhibit, and thus regulate, the translation of the kinase ORF (Fig. 5, top). We therefore examined PKM mRNA expression in the wheat germ in vitro translation system, which may contain fewer inhibitory factors, and varied the length of the 5Ј-UTR (Fig. 5, bottom). PKM mRNA with a full-length 5Ј-UTR (lanes 1 and 2) expressed PKM, as shown by both autoradiography of [ 35 S]methionine/cysteine-labeled proteins and Western blot with aPKC C-terminal antiserum. Consistent with a role in translational regulation, truncation of the 5Ј-UTR sequence containing the first four short ORFs increased PKM expression (lane 3). Interestingly, additional shortening of the 5Ј-UTR to within 9 nucleotides of the putative start AUG, which removes the two short ORFs with the strongest Kozak sequences, did not further significantly increase expression (lane 4). In contrast to the short PKM fragment produced by calpain cleavage of PKC ( Fig. 2A), PKM produced by PKM mRNA was identical in size to endogenous brain PKM (Fig. 5, lane 6).
LTP Increases de Novo Synthesis of PKM-Finally, we ex-amined the synthesis of PKM in LTP. LTP in the CA1 region of hippocampal slices was induced by two 100-Hz 1-s trains, 20 s apart, and followed for 10 min to ensure stable synaptic potentiation (Fig. 6, A and B). The tetanized slices and control slices that received test stimulation for equivalent periods were then transferred to a bath containing [ 35 S]methionine/cysteine for 30 min. In parallel experiments we found that this manipulation did not disturb synaptic responses, because slices that were returned to the recording chamber after incubation showed stable field excitatory postsynaptic potential (EPSP) responses (data not shown). After labeling, the slices were homogenized, PKM immunoprecipitated with C-terminal antiserum, and new synthesis determined by incorporation of label into immunoprecipitated PKM. The level of de novo synthesis of PKM in tetanized slices was 246 Ϯ 75% of that in control slices that received test stimulation alone (set at 100%, n ϭ 6, p Ͻ 0.01, paired t test, Fig. 6, C and D). Thus there is a large increase in PKM synthesis during LTP. DISCUSSION PKM, a New Brain-specific Atypical PKC Isoform-PKM was first described by Nishizuka and colleagues (13) as a constitutively active proteolytic fragment of PKC. Because proteolysis is irreversible, PKM formation was recognized early on as a potential mechanism for the persistent activation of PKC, in contrast to its transient activation by lipid second messengers, which are usually rapidly metabolized (12). Although an attractive mechanism to sustain long term cellular functions (9), physiological PKM formation was not detected for many years.
In a 1993 study of multiple PKC isozymes in LTP, however, we observed a PKM form specific to the atypical PKC isozyme in rat hippocampus (5). This PKM was not generated artifac- tually during homogenization (5) and was, moreover, the specific form of PKC to persistently increase in LTP maintenance (5,6). Because proteolysis was the only mechanism known to produce PKM, our initial studies had assumed that PKM was generated by cleavage of PKC. We subsequently observed, however, that PKM formation required new protein synthesis during LTP, a result that was not easily explained by proteolysis (6).
Here we have characterized the mechanism of brain PKM formation, which is a novel protein synthesis-dependent pathway for the persistent activation of a kinase. We first determined the relationship between PKM and PKC by examining their distribution in brain and other tissues with newly designed -specific antisera. PKM, but not PKC, was found exclusively in brain. The only other PKC isozymes expressed specifically in nervous tissue are the conventional PKC␥ (29) and a form of the novel PKC (30). PKM and PKC were also differentially expressed in distinct brain regions, with PKM the major form in forebrain. This pattern of expression suggested that PKC and PKM might not have a precursor-product relationship. Analysis of PKC regulatory domain knockout mice confirmed that brain PKM was expressed in the complete absence of full-length PKC.
These results indicated that brain PKM was not formed by a proteolytic mechanism but perhaps as a distinct gene product. The gene produces two RNAs (18,23,27) from separate dedicated promoters (20). The first promoter produces PKC mRNA, and the second, an internal promoter within the PKC gene, produces an RNA in certain rat prostate tumor cell lines and in brain (20). This alternate RNA was thought to be incapable of translation (21). We observed, however, that the distribution of PKC mRNA and this alternate RNA correlated well with PKC and PKM proteins, respectively. Furthermore, although the alternate RNA did not produce detectable protein in the rabbit reticulocyte translation system, in agreement with a previous study (21), in vitro translation in the wheat germ translation system produced PKM that was identical in size to endogenous brain PKM, demonstrating that the alternate RNA was PKM mRNA (Fig. 5). The long 5Ј-UTR of PKM mRNA, which contains multiple short ORFs, may have contributed to the difficulty in observing translation in all in vitro systems. Indeed, truncation of the PKM mRNA 5Ј terminus that eliminated the first four short ORFs greatly increased the translation of the message. Brain PKM and PKC are therefore distinct gene products and, by the standard nosology of the PKC family (9 -11), are separate PKC isoforms. Alternative names for PKC and brain PKM would be PKCI and PKCII, respectively.
PKM from PKM mRNA-PKM formation from its own mRNA allows for mechanisms of regulation not possible by proteolysis. Separate mRNAs, for example, permit the PKC and PKM isoforms to be segregated in distinct cell types. This may be important because a constitutively active PKM, lacking regulation by second messengers, could have a dominant effect over full-length PKC, disrupting the important cellular functions of the latter that include insulin and growth factor signaling (3,4,31,32). The catalytic domains of PKC and PKC/, however, are extremely similar (16,17), and PKM might also have a dominant effect over PKC/. This suggests that the two atypical isoforms might be differentially compartmentalized within neurons.
Although not as common a mechanism as alternative splicing, separate dedicated promoters have been found to increase the diversity of products from several genes, including two other PKCs, Caenorhabditis elegans PKC1 (33) and mouse PKC (34). In addition to providing a mechanism for tissue-specific expression, separate promoters may allow different sets of transcription factors to regulate each mRNA. The promoter region of PKM mRNA contains several putative binding sites for activity-dependent transcription factors. The presence of a conserved canonical CRE in the mouse, rat, and human PKM promoter suggests regulation by the transcription factor CREB (cAMP-response element binding protein), which has been implicated in memory formation in a variety of species (35). CREB and PKM may thus lie in the same signaling pathway, a result consistent with experiments showing that overexpression of CREB (36) and PKM (8) produce similar memory enhancement phenotypes in Drosophila. Interestingly, the human PKM mRNA promoter has an additional partial CRE insert that might enhance the binding of CREB and the activity-dependent transcription of PKM mRNA. The PKM mRNA promoter region also contains putative binding sites for other transcription factors implicated in memory, including Nf-B and C/EBP (37)(38)(39).
Alternative promoters also provide each mRNA with a distinct 5Ј-UTR that may allow for differential mechanisms of translational regulation. Exon 1Ј, which contributes the unique sequence to the PKM 5Ј-UTR, is strongly conserved in mouse, rat, and human, and results in a 5Ј terminus that is much longer than that of PKC mRNA (Fig. 4). Exon 1Ј may thus contribute to translational regulation specific to PKM mRNA. Indeed, this sequence of 5Ј-UTR, rather than the shorter sequence shared by the two mRNAs, was the major inhibitory constraint on PKM translation (Fig. 5). By oligocapping and 5Ј-RACE, the 5Ј terminus of brain PKM mRNA appears to be FIG. 7. Model of LTP maintenance by PKM synthesis. Our previous work has shown that activation by the excitatory neurotransmitter glutamate (Glu) of the NMDA receptor (NMDAR) during tetanus triggers the induction of PKM formation (5,6), and that PKM phosphorylation potentiates ␣-amino-3-hydroxy-5-methyl-4-isoxazolepropionate receptor (AMPAR)-mediated synaptic transmission in LTP maintenance (7). We now show the formation of PKM in LTP is through increased de novo protein synthesis from a PKM mRNA, produced by a dedicated internal promoter within the PKC gene that is nested between clusters of regulatory domain exons (reg, red) and catalytic domain exons (cat, green). Potential activity-dependent regulatory mechanisms for PKM formation include enhanced transcription by CREB and increased local translation. heterogeneous, consisting of different lengths of exon 1Ј. This suggests multiple transcriptional start sites, which is characteristic of transcription without a TATA box (40 -43). Although this heterogeneity could possibly be due to artifacts of the oligocapping and 5Ј-RACE method, the 5Ј terminus of the low abundant PKM mRNA in kidney, obtained in parallel experiments, was only a single length (Fig. 4B). The functional significance of the apparent heterogeneity of brain PKM mRNA is not clear, but one possibility is that variation in the length of the 5Ј-UTR regulates the translational efficiency of the message. Different patterns of transcription factors may produce these distinct PKM mRNA sizes, providing a mechanism to regulate the capacity of a neuron to translate PKM in response to synaptic stimulation.
Implications for LTP Maintenance-Our present findings have important implications for the molecular mechanism of LTP. Our original model for PKM formation was based upon observations with C-terminal antisera (5), which recognize both aPKC isoforms, and /. We found a sequential activation of aPKC in LTP, rapid translocation of full-length aPKC in induction followed by a sustained increase of PKM in maintenance, and we therefore assumed that the two kinases were related as proteolytic precursor and product (5). Whereas a proteolytic mechanism may be important for PKM formation in other forms of synaptic plasticity (44,45), our current results indicate that the two aPKCs activated in LTP induction and maintenance are distinct isoforms: PKC/, the full-length aPKC expressed in hippocampus, rapidly translocates during induction, whereas PKM is synthesized de novo from its own mRNA in maintenance (Fig. 7).
Future work will be required to determine the relative contributions of translational and transcriptional regulation to PKM synthesis in LTP. However, because the size of the PKC gene encoding PKM is ϳ100 kb and the rate of transcription is thought to be 1-2 kb/min (46), new transcription of PKM mRNA may require many minutes to hours. Therefore, the de novo synthesis of PKM we observed 10 -40 min after tetanization in hippocampal slices (Fig. 6, C and D) is likely due to increased translation of pre-existing PKM mRNA. Release from the translational block mediated by the long 5Ј-UTR of the PKM mRNA by rapamycin-sensitive pathways or an internal ribosomal entry site (47) are potential mechanisms for the increase in PKM synthesis that we observed.
PKM, a Cognitive Kinase for Long Term Synaptic Memory Storage-Autonomously active kinases have long been attractive candidates for maintaining long term memory because they provide a cogent molecular mechanism for the persistence of synaptic plasticity (48 -50). When transiently stimulated by second messengers, several kinases can produce a short term enhancement of synaptic transmission (49). Following the intense synaptic stimulation that induces long term synaptic plasticity, however, a few serine-threonine kinases may also be converted from second messenger-dependent into second messenger-independent forms through persistent post-translational modifications. Two well known examples in LTP are the autophosphorylation of Ca 2ϩ /calmodulin-dependent protein kinase II (CaMKII) (48) and PKC oxidation (51). Both modifications result in a conformational change that displaces the autoinhibitory pseudosubstrate of the enzymes, rendering the kinases autonomously active. Because they may then persistently enhance synaptic transmission, these autonomously active forms might serve as molecular memory stores and thus have been called "cognitive kinases" (49).
However, because most cognitive kinases are generated by post-translational modifications of second messenger-dependent enzymes, their persistent action is fundamentally limited by protein turnover, because they are eventually replaced by their non-autonomously active precursors (52). It is therefore generally assumed that cognitive kinases serve mainly in the non-protein synthesis-dependent, early decremental phase of LTP, but serve only a transitional role in the late phase of LTP that requires new protein synthesis and is thought to mediate long term memory. Consistent with this view, inhibitors of CaMKII and conventional/novel PKCs block LTP induction but do not reverse LTP maintenance when applied after potentiation has been stably established (7,53).
PKM, however, has a structure and function different from other cognitive kinases that allows it to couple new protein synthesis directly to the mechanism of synaptic enhancement. PKM is not produced by a post-translational modification of a second messenger-dependent enzyme but is synthesized de novo by transcriptional and translational mechanisms that completely eliminate the autoinhibitory pseudosubstrate of PKC from its catalytic domain (Fig. 7). Thus in contrast to other cognitive kinases, PKM is an autonomous kinase that is generated and maintained by new protein synthesis. Furthermore, the persistent increase in PKM activity, unlike that of CaMKII and conventional/novel PKCs, is critical for maintaining the late phase of LTP, as demonstrated by the reversal of established LTP by PKM inhibitors (7). Thus the synthesis of PKM provides the missing link between gene expression and the mechanism of synaptic enhancement in LTP. These properties may allow PKM to serve as a molecular substrate of long term memory.