Light scattering and transmission electron microscopy studies reveal a mechanism for calcium/calmodulin-dependent protein kinase II self-association

Ca²⁺/calmodulin-dependent protein kinase II (CaM-kinase II) is a ubiquitous multifunctional Ser/Thr protein kinase implicated in regulating diverse cellular processes, including gene transcription, calcium homeostasis, receptor function, cytoskeletal alterations, synaptic transmission and synaptic plasticity (Hanson and Schulman 1992; Braun and Schulman 1994). The enzyme is maintained in an inactive state through inhibitory interactions between the autoregulatory domain (LeVine and Sahyoun 1987; Kwiatkowski and King 1989) and both the ATP (Colbran et al. 1989) and substrate-binding domains (Payne et al. 1988). Following Ca²⁺/CaM binding, the enzyme undergoes autophosphorylation, an intersubunit intraholoenzyme reaction (Mukherji and Soderling 1994) that is thought to precede substrate phosphorylation (Kwiatkowski et al. 1988).

CaM-kinase II isolated from forebrain is a holoenzyme composed of 8–12 subunits (450–650 kDa). Although multiple isozymes of CaM-kinase II (α, β, γ and δ) arise from distinct gene products or by alternative splicing (Hanson and Schulman 1992), the α (50 kDa) and β (60 kDa) subunits are the predominant isoforms in forebrain (Bennett et al. 1983). The α isoform is the dominant subunit in forebrain (3 : 1, α : β), producing both mixed subunit holoenzymes (Vallano 1989) and homomers composed entirely of α subunits (Brocke et al. 1999). CaM-kinase II is highly concentrated in brain, representing 1–2% of total cellular protein within the cortex and hippocampus (Erondu and Kennedy 1985). The enzyme is found in both soluble and particulate fractions (Kelly and Vernon 1985; Sahyoun et al. 1985a) and has been reported to comprise as much as 20–40% of the protein in neuronal cytoskeletal preparations (Sahyoun et al. 1985b), and the postsynaptic density (PSD; Miller and Kennedy 1985; Kennedy 1993). The subcellular targeting of CaM-kinase II is a dynamic process (Shen and Meyer 1999; Shen et al. 2000) that is differentially influenced by the α and β isoforms (Shen et al. 1998). Although the mechanism producing the subcellular compartmentalization of CaM-kinase II is unknown, the interaction of the α isoform with synapsin I (Benfenati et al. 1996) and the NR2B subtype of the NMDA receptor (Strack et al. 2000) are blocked by peptides that correspond to the autoregulatory domain of CaM-kinase II, indicating that the catalytic domain may serve a role in subcellular targeting of this isoform.

CaM-kinase II was initially called the major postsynaptic density protein, due to its high content in the PSD; however, this content appears to be dramatically enhanced by ischemia-induced conditions associated with decapitation (Suzuki et al. 1994; Aronowski and Grotta 1996; Margrie and Rostas 1997). The redistribution and inactivation of CaM-kinase II from the cytosolic to the particulate fractions during ischemic conditions is well documented (Kochhar et al. 1991; Aronowski et al. 1992; Hanson et al. 1994; Kolb et al. 1995; Shackelford et al. 1995). The extent of these alterations in CaM-kinase II correlates well to the loss of neuronal function leading to cell death following ischemic damage (Hanson et al. 1994).

Although the PSD has been previously implicated as a target for soluble CaM-kinase II translocation following an ischemic insult (Suzuki et al. 1994; Aronowski and Grotta 1996), a recent study identified 100-nm clusters contaminating PSD preparations using transmission electron microscopy and rotary shadowing (Dosemeci et al. 2000). Biochemical characterization of these clusters demonstrated that they were composed predominantly of αCaM-kinase II. These clusters likely contribute to the content of CaM-kinase II previously ascribed to particulate subcellular structures, such as the PSD. In addition, these authors reported that the presence of CaM-kinase II clusters in cultured neurons was increased following exposure to mitochondrial uncouplers in the absence of glucose (Dosemeci et al. 2000). Exposure of neurons to metabolic poisons likely produce similar intracellular changes as observed during ischemic conditions, including increases in intracellular Ca²⁺ and ATP depletion and possibly decreased pH. During cerebral ischemia, the intracellular pH decreases to 6.2 within minutes following loss of blood flow (Silver and Erecinska 1992) and ATP pools are rapidly depleted (Folbergrova et al. 1990). These changes in ATP and pH are also associated with elevated intracellular Ca²⁺. Thus, although the mechanism leading to the formation of CaM-kinase II clusters is unknown, the formation of these clusters may underlie the observed redistribution of CaM-kinase II from soluble to particulate fractions in neurons following ischemic stress.

Clusters formed of predominantly αCaM-kinase II under ATP depleting conditions in neurons (Dosemeci et al. 2000) bare a striking morphological resemblance to clusters of CaM-kinase II formed from purified forebrain enzyme during ischemia-like conditions in vitro (Hudmon et al. 1996). In the latter study, CaM-kinase II holoenzymes were shown to associate with one another to form sedimentable complexes, a process we termed self-association. Self-association was dependent upon ischemia-like conditions of limiting ATP and acidic pH and was accompanied by a time-dependent inactivation of the enzyme (Hudmon et al. 1996). It appeared that self-association could account for both the translocation and inactivation of CaM-kinase II observed during cerebral ischemia (Hudmon et al. 1996).

In the current study, the mechanism and isoform specificity of CaM-kinase II self-association was examined using light scattering and transmission electron microscopy (TEM). The results show that self-association is isozyme specific as recombinant α holoenzymes undergo self-association whereas β holoenzymes do not. Ca²⁺/CaM binding is essential for this process, and peptides mimicking the autoregulatory domain potently block self-association. Consistent with these data is a model for self-association whereby interactions form between the catalytic and regulatory domains of α subunits within different holoenzymes. This proposed mechanism is analogous to the intrasubunit interactions within the holoenzyme that maintain the enzyme in an autoinhibited state. However, subunit interactions leading to self-association would occur between neighboring holoenzymes under the appropriate conditions as defined in the present study.

Experimental procedures

Results

Light-scattering measurements of αCaM-kinase II self-association

Light scattering is sensitive to the shape and size of protein in solution, and was used successfully to monitor the dynamics of glutamine synthetase (a protein similar in mass to CaM-kinase II) polymerization previously (Yanchunas et al. 1994). We adopted similar light-scattering parameters to examine the dynamics of CaM-kinase II self-association. Recombinant α holoenzymes were used in our initial experiments to characterize changes in light scattering under conditions previously shown to induce self-association of forebrain holoenzymes (Hudmon et al. 1996). A typical time course of the reaction of αCaM-kinase II self-association showed a rapid initial increase in light scattering followed by a decline to a constant final value (Fig. 1A). To correlate the changes in light scattering with enzyme solubility using the sedimentation analysis used in our previous study (Hudmon et al. 1996), samples were removed at the indicated times (Fig. 1A, a–e), and the ratio of soluble vs. self-associated enzyme was determined as described in the Methods section. The western blot (Fig. 1B), before addition of Ca²⁺/CaM (point a in the light-scattering trace), showed that the enzyme was initially soluble. During the initial rise in light scattering, immediately after the addition of Ca²⁺/CaM (b), a portion of the enzyme was detected in the pellet. At 0.5 min (c), 2 min (d) and 5 min (e) time points, all detectable enzyme was in the pellet. The migration difference (50–54 kDa) of the α subunit observed in the western blot was characteristic of CaM-kinase II following autophosphorylation under these conditions (Hudmon et al. 1996). Changes in light scattering correspond well to the sedimentation analyses previously used to detect self-association. However, the light-scattering analysis reveals changes in the enzyme complexes (peak decreasing to a constant final value) over time that were not discernable using sedimentation analysis.

Changes in light scattering reflect different stages in the self-association of αCaM-kinase II holoenzymes

Changes in the appearance of αCaM-kinase II during the process of self-association were correlated to the light-scattering trace by removing aliquots of the reaction mix and examined by transmission electron microscopy (TEM). Using self-association conditions similar to those in Fig. 1, samples were removed for TEM examination (Fig. 2) during (a) the initial rise (within 20 s) (b) the trailing edge of the peak (45 s), and (c) at a constant value of light scattering (5 min). These time points collected for TEM analysis correspond approximately to points b, c and e on the light-scattering trace in Fig. 1. The initial increase in light scattering appears to be due to the association of individual holoenzymes into loosely packed clusters as shown in Fig. 2(A). Holoenzymes oriented within the cluster so that their ring-like shape 10–20 nm in diameter are visible (arrowheads in inset of micrograph B) and appear similar to the morphology of CaM-kinase II holoenzymes identified previously (Woodgett et al. 1983; Kolodziej et al. 2000). The clusters were sparse on the micrographs and of variable shape and size, precluding a thorough assessment of their average features. For reference, the diameter of the cluster in Fig. 2(A) across its longest dimension was ∼430 nm. Over time, these clusters begin to associate with one another and condense as they adopt a more rounded appearance with sharp outer boundaries (micrographs B and C). A lower magnification micrograph of the 5 min time point shown in micrograph D shows a complex, several microns in diameter, composed of dozens of clusters. The clusters shown in Fig. 2(D) range from 344 nm to 77 nm, with an average diameter across their longest dimension of 182 nm ± 59 nm (mean ± SD; n = 24). Only clusters with distinct boundaries were included in the analysis.

The TEM data provides a morphological explanation for the two phases seen in the light-scattering curves. The initial increase in light scattering occurs as soluble holoenzymes form loosely organized clusters. These clusters condense and associate with one other, forming complexes of several microns in diameter that represent the final scattering value. Thus, the decrease in light scattering leading to the constant final value can be attributed to both a decrease in the total number of clusters in solution (even though they are larger) and/or to the condensation of the clusters.

The concentration dependence of CaM-kinase II self-association

Using self-association conditions similar to those employed in 1, 2, the rate and extent of light scattering was determined over a range of enzyme concentrations (0.07–0.4 µm) (Fig. 3A). The average values (n = 3) for the initial rates are summarized in Fig. 3(B). CaM-kinase II self-association was detected at a protein concentration as low as 0.14 µm (7 µg/mL). As enzyme concentration increased, the light-scattering traces became more complex in shape, exhibiting a transient peak that decreased to a constant value after 1–2 min. The change in initial rate as a function of enzyme concentration (Fig. 3B) is consistent with an intermolecular self-association reaction between holoenzymes.

ΒCaM-kinase II does not undergo self-association

Our experiments indicate that recombinant α holoenzymes mimic the self-association process observed for CaM-kinase II isolated from forebrain composed of both α and β subunits (Hudmon et al. 1996). To test whether self-association was a generalized feature of CaM-kinase II holoenzymes, holoenzymes composed of only β subunits were examined for self-association. βCaM-kinase II did not self-associate when subjected to conditions that produced robust self-association of αCaM-kinase II (Fig. 4).

Increases in pH and ionic strength decrease self-association

Self- association of forebrain CaM-kinase II was dependent on pH (Hudmon et al. 1996). However, in that study, a constant ionic strength was not maintained as pH was varied. Therefore, it was possible that the observed effect was due to changes in ionic strength rather than hydrogen ion concentration.

To examine these two possibilities, the influence of ionic strength on self-association was first examined. The reaction buffer (10 mm PIPES at pH 6.5) was supplemented with KCl to provide increases in ionic strength (either 60, 90, 140, 190 or 240 mm) at three different concentrations of enzyme (0.2, 0.4 and 0.8 µm). The averaged data for final values of light scattering (Fig. 5) indicated that the extent of enzyme self-association was decreased by increasing ionic strength. The half-maximal dependence of ionic strength on self-association was influenced by enzyme concentration and increased from approximately 150 mm to 190 mm and 240 mm at enzyme concentrations of 0.2, 0.4 and 0.8 µm, respectively.

Is the pH sensitivity previously observed for self-association simply a function of ionic strength? To answer this question, the pH dependence of self-association was determined in reactions of a constant ionic strength. The contribution to ionic strength of the MES, PIPES and HEPES buffers at each pH was calculated as described by Ellis and Morrison (1982). The averaged data indicated that at pH values above 7.0, little self-association occured (Fig. 6A). However, at pH values of 7.0 and below, self-association increased with decreasing pH as reported previously when sedimentation was used as the criteria (Hudmon et al. 1996). To confirm the importance of pH in inducing CaM-kinase II self-association, we performed a pH shift experiment in real time using light scattering (Fig. 6B). The enzyme was added to an autophosphorylation reaction mix at pH 7.5, a reaction condition that does not produce self-association. After 30 s, the pH was decreased to 6.5 by the addition of MES/HEPES buffer. The addition of enzyme at 120 s induced a slight change in the baseline of the light scattering trace due predominantly to the presence of detergent in the added sample (Fig. 6B, trace b). A significant increase in light scattering occurs immediately following the drop in pH to 6.5 (trace b, arrow), although the extent of self-association is reduced somewhat relative to standard reactions (compare traces a and b, Fig. 6). Clearly, a drop in pH supports self-association and seems to be a key regulator of the process.

Enzyme inactivation prevents self-association

The decrease in light scattering observed in Fig. 6 (compare traces a and b) may have been due to enzyme inactivation under these conditions (Hudmon et al. 1996). Reactions at pH 7.5 do not produce self-association, although a time-dependent inactivation does occur as autophosphorylation increases (Hudmon et al. 1996). To differentiate the impacts of autophosphorylation and enzyme inactivation in preventing self-association, the reaction conditions were modified to inactivate the enzyme in the absence of autophosphorylation. Enzyme inactivation was accomplished by incubating αCaM-kinase II with Ca²⁺/CaM at pH 6.5 for increasing times and then self-association was initiated by the addition of Mg²⁺/ATP. The inset in Fig. 7 shows the time course of inactivation of αCaM-kinase II under these conditions. The light-scattering trace in Fig. 7 demonstrates the consequence of this inactivation on self-association. Self-association is reduced after 30 s (addition of Mg²⁺/ATP at the first arrow) and completely prevented after 2 min (addition of Mg²⁺/ATP at second arrow) of incubation under these conditions. These time points correspond to a 30% and 70% decrease in the initial activity, respectively. These data indicate that inactivation of the enzyme is correlated with prevention of self-association.

Ca²⁺/CaM binding is essential to induce self-association

The binding of Ca²⁺/CaM is thought to disinhibit CaM-kinase II by exposing sites in the catalytic domain for substrate and ATP binding, thus permitting subunits within holoenzymes to undergo both autophosphorylation and substrate phosphorylation. A requirement for Ca²⁺/CaM would suggest that at least one of the partners required for subunit contacts for self-association is not exposed in the native or autoinhibited holoenzyme. In Fig. 8, CaM levels were varied while maintaining saturating Ca²⁺. Ca²⁺/CaM was essential for inducing CaM-kinase II self-association and exhibits a half-maximal value of 0.4 µm CaM. Because of the multisubunit nature of the holoenzyme, a relevant question is how many subunits within each holoenzyme require binding of Ca²⁺/CaM for self-association to occur? At 0.4 µm enzyme, self-association exhibits a strong sensitivity to [CaM] between 0.3 and 2 µm (Fig. 8). The K_D reported for Ca²⁺/CaM binding to CaM-kinase II is approximately 50 nm (Meyer et al. 1992). At 400 nmα subunit and CaM, one would expect roughly 280 nm Ca²⁺/CaM–CaM–kinase II (subunit) complex, with 120 nm free Ca²⁺/CaM and free enzyme. Assuming 12 subunits per holoenzyme (Kolodziej et al. 2000), one would infer that half-maximal self-association requires approximately eight subunits bound with Ca²⁺/CaM. Thus, Ca²⁺/CaM is essential for the initiation of self-association, however, robust self-association is still observed when Ca²⁺/CaM is not saturating the holoenzyme.

Peptide substrates and peptide inhibitors block self-association

These data indicate that self-association requires that subunits be activated by Ca²⁺/CaM-binding, possibly by exposing catalytic domains through disruption of their interactions with regulatory domains. This hypothesis was explored using synthetic peptides that bind to and occupy the catalytic cleft following activation. Self-association was first assessed in the presence of increasing concentrations of the peptide substrate autocamtide-2. The sequence of autocamtide-2 was derived from the autoregulatory domain surrounding Thr²⁸⁶ and has a K_m of 0.4 µm (Ishida and Fujisawa 1995). Increasing concentrations of autocamtide-2 decreased the extent of self-association in a dose-dependent fashion (Fig. 9). Approximately 1.6 µm autocamtide-2 was required for half-maximal inhibition of self-association. Autocamtide-2 is phosphorylated by the enzyme and phosphorylation would decrease its binding affinity, thus decreasing the peptide's inhibitory potency. Therefore, we also examined the dose-dependent effects of a similar peptide, AIP, which contains an Ala substituted for the phosphorylated Thr residue (Ishida and Fujisawa 1995). As anticipated, AIP was a more effective inhibitor of self-association than autocamtide-2, exhibiting half-maximal inhibition at 0.2 µm. In contrast, the peptide substrate syntide-2 (K_m = 8.5 µm (Ishida and Fujisawa 1995) did not prevent self-association at concentrations as high as 100 µm (data not shown). The fact that syntide-2 cannot block self-association indicates: (i) that the inhibition seen with autocamtide-2 and AIP is specific, and (ii) that autocamtide-2 and AIP bind to the catalytic domain in a distinct manner from syntide-2. These data strongly support a role for the catalytic domain in self-association, but not the substrate binding site. One obvious target for binding to the catalytic domain is the autoregulatory domain. Specifically, residues around Thr²⁸⁶ in the autoregulatory domain are implicated by these peptide competition studies. In summary, we propose a model for self-association based partly on the mechanism for autoinhibition; interactions between the catalytic domain of one subunit with the autoregulatory domain in another subunit are formed between neighboring holoenzymes in the presence of Ca²⁺/CaM to produce self-association.

Discussion

Using electron microscopy and light scattering, we observed that self-association of CaM-kinase II in vitro is a multiphasic process. Electron microscopy showed that holoenzymes initially bind to each other to form loosely packed clusters that condense and undergo further association over time. The TEM images also indicate that self-association does not involve a destruction of the enzyme's ring-like morphology. The threshold for detecting self-association was approximately 0.14 µmαCaM-kinase II under standard reaction conditions, indicating that this process occurs at remarkably low protein concentration. The typical enzyme concentration (0.4 µm) utilized in our experiments is well below the estimated concentration of 19–37 µm for αCaM-kinase II in forebrain homogenates (Kennedy et al. 1990). Increases in ionic strength decreased self-association indicating that the protein interactions between holoenzymes are, in part, ionic in nature. Increases in enzyme concentration decreased the sensitivity to ionic strength, with robust self-association still observed at enzyme concentrations 20–40 fold below levels reported for brain at physiological ionic strength.

In addition to requiring Ca²⁺/CaM, self-association was highly dependent on reaction pH. At a pH of 7.5 little self-association was evident, but as the pH decreased to values of 7.0 and below, the extent of self-association increased. Thus, pH appears to be a critical triggering step that determines whether interholoenzyme subunit contacts form and/or are stabilized following enzyme activation. The protonation state of His²⁸² in the regulatory domain of CaM-kinase II is sensitive to pH in a range reported here that produced self-association (Smith et al. 1992; Brickey et al. 1994). Protonation of His²⁸² led to an increased exposure of the ATP binding pocket within the catalytic domain and, interestingly, this effect required the previous binding of Ca²⁺/CaM (Brickey et al. 1994). Apparently, His²⁸² resides in a sequestered environment. Upon Ca²⁺/CaM binding, protonation of His²⁸² leads to exposure of the catalytic domain by disrupting autoregulatory interactions (Brickey et al. 1994). The pH sensitivity and location of His²⁸² in the autoregulatory domain make this residue a potential trigger in regulating self-association, possibly by relieving steric constraints and exposing domains in the autoregulatory and/or catalytic regions that are not accessible at pH values above 7.0.

Synthetic peptides mimicking the autoregulatory domain (autocamtide-2 and AIP) were potent inhibitors of self-association implicating the catalytic domain as a binding partner for the self-association reaction. Specifically, these experiments implicate the region of the catalytic domain that is normally inhibited by the autoregulatory domain, presumably residues around Thr²⁸⁶, as the interholoenzyme subunit partner to produce self-association (see Fig. 10). The syntide-2 peptide showed no inhibitory effect even at 50-fold over the half maximal concentration observed for AC-2, indicating that the substrate binding pocket is not directly involved in self-association. Ishida and Fujisawa (1995) described similar findings on the ability of these peptides to differentially prevent enzyme inactivation. AIP and autocamtide-2, but not syntide-2, protected CaM-kinase II from time-dependent inactivation following activation by Ca²⁺/CaM in the absence of ATP and autophosphorylation. Presumably autocamtide-2/AIP and syntide-2 interact with the catalytic domain in unique ways to explain their distinct effects in both inactivation and self-association. The peptide competition experiments reported by Ishida and Fujisawa (1995) suggest mechanistic similarities to the self-association reaction observed here between holoenzymes. Inactivation of the enzyme by Ca²⁺/CaM treatment prior to a pH drop prevented self-association. It would appear that exposing the enzyme to Ca²⁺/CaM in the absence of ATP, produces subunit interactions or structural changes within the holoenzyme that compete for binding partners involved in the self-association reaction.

We have shown that recombinant α holoenzymes undergo self-association while β holoenzymes do not. This result suggests that the ratio of α and β subunits within the same holoenzyme, as found in CaM-kinase II preparations from rat brain (Vallano 1989; Brocke et al. 1999), would influence their propensity for self-association. In support of this possibility, CaM-kinase II isolated from the cerebellum (predominantly β subunits) is less susceptible to the process of self-association compared with the forebrain enzyme (predominantly α subunits) (Hudmon and Waxham, unpublished observations). The three-dimensional structure of αCaM-kinase II reveals that the enzyme is dodecameric and that the catalytic and regulatory domains reside in foot-like processes extending out from a central core on tethers (Kolodziej et al. 2000). Inter-holoenzyme contacts between the catalytic/autoregulatory foot-like processes could produce the lattices of holoenzymes visible by electron microscopy. We have determined that the overall three-dimensional architecture of βCaM-kinase II and αCaM-kinase II are similar (manuscript in preparation), indicating that a difference in their subunit organization or content (both are dodecamers) cannot explain the difference in their capacity for self-association.

The relative hydrophobicity, predicted secondary structure, and charge distribution are conserved throughout much of the amino acid sequence of the α and β subunits (Bulleit et al. 1988; Kolb et al. 1998). Variations in their structures are primarily attributed to two variable inserts that make-up the ‘unique’ domain of the β subunit (Bulleit et al. 1988). This domain may be responsible for preventing the self-association of βCaM-kinase II. Interestingly, the region analogous to the variable domain of β in splice variants of Drosophila CaM-kinase II appears to interact with the catalytic domain (GuptaRoy and Griffith 1996). It is possible that additional contacts are established between the variable and catalytic domains that prevent exposure of sites necessary for self-association in the β holoenzymes.

CaM-kinase II autophosphorylation and substrate phosphorylation are stringently coupled to Ca²⁺/CaM-binding. Binding produces a conformational change that disrupts contacts between the catalytic and autoregulatory domains, exposing the catalytic domain for activation and the autoregulatory domain for autophosphorylation. It is thought that autophosphorylation is required to fully disinhibit the enzyme (Colbran et al. 1989) and it leads to the production of Ca²⁺/CaM-independent activity. Based on these findings and our present data, Fig. 10 presents a model for self-association, whereby subunit interactions between holoenzymes underlie this process. Specifically, the autoregulatory domain of a subunit in one holoenzyme stably interacts with the catalytic domain of a subunit in a neighboring holoenzyme under conditions of low pH. Due to the multisubunit nature of the CaM-kinase II holoenzyme, additional subunits can also undergo these interactions with the same or other holoenzymes leading to the clusters seen in the EM. The basic interactions of this model are similar to the autoinhibition that normally maintains each subunit in an inactive state within the holoenzyme. Predictions from this model are directly supported by the present data. First, Ca²⁺/CaM-binding would be essential for exposing subunit contacts necessary for self-association. Second, the formation of these contacts between holoenzymes would be blocked by synthetic peptides that mimic the autoregulatory domain.

It is of interest to note that autophosphorylation is an intersubunit intraholoenzyme reaction mechanism (Mukherji and Soderling 1994). Autophosphorylation occurs when a Ca²⁺/CaM-bound subunit presents Thr²⁸⁶ within the autoregulatory domain to a neighboring Ca²⁺/CaM-bound subunit catalytic domain (Rich and Schulman 1998). This mechanism appears to predispose subunits within the holoenzyme to make intersubunit contacts, albeit transiently. Thus, the intersubunit–interholoenzyme interactions that form the basis of our model are a combination of the native autoregulatory interactions (intraholoenzyme–intrasubunit) as well as the autophosphorylation reaction (intraholoenzyme–intersubunit).

A further prediction is that the state of Thr²⁸⁶ autophosphorylation should alter the enzyme's susceptibility to self-association. The specific prediction is that stoichiometric autophosphorylation of Thr²⁸⁶ would prevent self-association (Hudmon et al. 1996). We previously reported that autophosphorylation of forebrain CaM-kinase II alters the enzyme's susceptibility to undergo self-association. This conclusion was drawn from experiments where ATP concentrations around the K_m for the enzyme (10 µm) induced self-association, whereas, ATP levels well above this level (500 µm) prevented self-association (Hudmon et al. 1996). We hypothesize that ATP concentration influences the rate of Thr²⁸⁶ autophosphorylation during activation; high ATP maximizes Thr²⁸⁶ autophosphorylation and subsequently minimizes self-association. Thus, as shown in Fig. 10, the process of autophosphorylation would compete with that of self-association, with ATP concentration influencing the outcome of this competition by determining the extent of Thr²⁸⁶ autophosphorylation.

Although somewhat different in size, the enzyme complexes formed following CaM-kinase II self-association in vitro appear remarkably similar in morphology to the CaM-kinase II clusters recently described by Dosemeci et al. (2000) following metabolic poisoning of the neurons. We propose that the mechanism described here for producing clusters in vitro is also responsible for the formation of clusters in neurons. Dosemeci et al. (2000) postulated that one possible function for the formation of these clusters is to sequester enzyme activity during conditions of metabolic stress. The self-association mechanism that we propose would provide a mechanism for inactivation because the interactions described would inactivate the enzyme by a form of intersubunit inhibition between holoenzymes.

Is in vitro self-association, as described here, and the formation of enzyme clusters in neurons as observed by Dosemeci et al. (2000) a potentially reversible process? It currently is unknown whether altering parameters such as pH, ATP and Ca²⁺ concentration can dissociate clusters of self-associated enzyme, although such experiments are underway. In addition, the possible reversal of cluster formation in neurons was not addressed in the study by Dosemeci et al. (2000). Ischemic-induced translocation of soluble enzyme into the particulate fraction is reversible in vivo if the period of ischemia is kept brief (5 min in the rat model of global ischemia; Aronowski et al. 1992). Thus, following reperfusion, a mechanism must exist in neurons that can reverse the soluble to particlulate translocation detected following ischemia. In fact, the recovery of soluble CaM-kinase II may be crucial for neuronal survival, because the ability to restore a pool of soluble active enzyme is correlated with the extent of neuronal survival in ischemic-sensitive regions of the brain, such as the hippocampus (Aronowski et al. 1992).

An interesting parallel exists between the mechanisms of self-association reported here and the proposed mechanism for CaM-kinase II association with other proteins. Peptides resembling the autoregulatory domain of CaM-kinase II that prevented self-association inhibit the binding of αCaM-kinase II to both the NR2B subunit of the NMDA receptor (Strack et al. 2000) and synapsin I (Benfenati et al. 1996). A reasonable hypothesis is that αCaM-kinase II self-association is an extension of this conserved targeting mechanism of the α isoform with other proteins. The CaM-kinase II clusters (Dosemeci et al. 2000) that form in neurons are then a consequence of enzyme activation and exposure of a physiological targeting mechanism to pathological conditions; in the case of ischemia, this involves sustained intracellular calcium, reduced pH and reduced ATP levels. Our results demonstrating that synthetic peptide inhibitors potently block self-association could offer both a reagent to dissect the role of CaM-kinase II during ischemic damage as well as afford new interventions for neuronal protection. Finally, the role of Thr²⁸⁶ autophosphorylation in regulating self-association is intriguing, and mutagenesis studies currently are underway to examine the role of ATP and Thr²⁸⁶ autophosphorylation in preventing self-association.

Acknowledgements

The authors acknowledge Steve Kolodziej for help in preparing the figure of the electron micrographs. We also recognize Sheela Singla and Dr Ulli Bayer for critical discussions and comments on the manuscript. This study was supported by NIH grants NS26086 (MNW) and HL42886 (JKS). AH and SAK were supported in part by NIH training grant NS07373. SJK was supported in part by the University of Texas Health Science Center at Houston MD/PhD Program.