Decoding calcium signals by multifunctional CaM kinase H. SCHULMAN’, P.I. HANSON’ and T. MEYER2 ’ Department of Pharmacology, Stanford UniversitySchool of h&di&e, Stanford, CaMnnia, USA 2Depattment of Cell Biology, Duke University School of Medkzine,Dutbam, North Carvlina, USA Abstract - MultffunctfonalCa2+/calmodulindependentp@ein fdnase (CaM kfnase) is one of the three major protefn kfnases coordinatingceffqiar responses to hormones and nsurotransmitters. lt metdfatesthe action of Ca2+ on neurotransmfttersynthasfs and reieaae, on carbohydrate metabolism and on the cytoskeleton. CaM klnase has structuraUfunctlonal propsrtfes that facflftate fts response to dfstlnctfveattributes of Ca*+ signals whfch often Involve transient increases that span a narrow conosntrattonrange and Inthatare puls8tlle rather than persistent. The fdnase responds to the n8rrow worldng r8nge of Ca2+ signals by the use of calmodulln as the Ca2+ sensor. ft is 8ctfv8ted by tha of calmodulfn to an autolnhlbitory domain that keeps the fdnase fnactfve In the basal state. The tmnsfent nature of the signal is accommodated by olutophosphoryktion of this autofnhff#ory domaln which affows the kfnase to remain partially active after 4modulfn dlssocktes and thereby swftches ft to a ca2’-independM spacfes. The puls8tife n8ture of Ca2’ sfgnals may also be decoded by C8M fdnase. AMqWsphorylatfon traps cafmodulfn on autophosf&o@ated subunfts by greatly raducfng Its off-rate. At hfgh frequency of stfmulatlon, cafmodulfn would remain tr8 durfng the brief interval baleen Ca2+ p!Ied oscillations 8nd each successive rise In C8 + would recruft more c8lmoduffn. This may enable a stimulus frequency dependentactfvatlonof CaM kfnase.

MultifunctionalCazt/caImodulin-dependent protein kinase (CaM kinase) is a prominentmediator of hormones and neurotransmitters that elevate intracellularfree Ca2+. Theenzymeis also refenredto as calmodulin-dependentmultiproteinkinase, type II Ca2tMmodulindependent protein kinase, and Ca2t/cakxhdinde~dent protein kinase II. Several characteristicsof CaM kinase are compatible with its role in mediatinghormoneaction in a varietyof cells. The enzyme has a wide tissue dist-

ribution and is present in all eukaqotic systems examined thus far. Its intracellularlocalization positionsit to respondto Ca2+elevatedas a resultof eitherinflux from the extraa9lularmilieu or release fram intemal stores and thus to respond to diverse signal transductionpathways. Finally,it has a broad substrate specificity that enables it to regulate enzymes involved in many cellular processes, includingcarbohydrate IxWaboliUQ neurotraosmi#er release and resynthesis, cytoskeletal function and




A number of other reviews on ion flux. multifunctional CaM kinase provide perspectives on these and other aspects of the enzyme [l-41. Ca2’ signalling systems have some unique attributes that challenge the regulatory systems designed to decode and mediate their effects. In particular, physiologic Ca2’ concentrations vary over a narrow concentration range so that signal-to-noise detection may be problematic. Ca2’ signals are often transient thereby limiting the period of time when Ca2+dependent processes can be regulated. Finally, intracellular free Ca2+ can change in a pulsatile fashion and with variable amplitude and fmquency. These attributes challenge mediators of Ca2’ action to respond quickly to Ca2’ spikes, to be sufllciently activated during the brief time that Ca2’ is elevated to permit an effective response, and possibly to decode information encoded in the frequency of Ca2+ oscillations. The focus of this review will be on those characteristics of CaM kinase that are consistent with its proposed role in decoding Ca2’ signals. (a) The kinase is markedly activated by small increases in intracellular free Ca2+. (b) It can remain active after its initial stimulation even after free Ca2’ returns to basal levels and can thereby potentiate brief Ca2’ signals. (c) Finally, stimulation of the kinase may produce a read-out that is dependent on the frequency of Ca2+ spikes; CaM kinase may thereby function as a frequency detector. The ability of CaM kinase to decode Ca2’ signals is inherent in its structural design and in the manner by which it is regulated by the binding of Ca2’/calmodulin and by autophosphorylation. Our understanding of this process has come from biochemical and biophysical studies of the native enzyme as well as molecular cloning of the kinase and analysis of site-directed mutants of the kinase. These studies indicate that the kinase can exist in multiple functional states determined by the presence or absence of Ca2’ and by the sites on the kinase that are autophosphorylated (Table 1). At

least three autophosphorylation states of the kinase have been identified Each state is characterized by a distinct range of activity between the basal or unstimulated level and the maximal Ca2t-stimulated level. The activity of the kinase in its native or dephosphorylated state, termed CaMK’, ranges from essentially 0% (unstimulated) to 100% (stimulated). Stimulation of such a kinase with Ca2’/ calmodulin leads to autophosphorylation which generates a second phosphorylation state termed CaMKP. The rate of dissociation of cahnodulin from CaM kinase subunits in this autophosphorylation state is reduced by over lOO-foldin either the absence or presence of Ca2’; CaMK’ traps bound cahnodulin even at subthreshold Ca2’ concentrations. Durhig submaximal stimulation, this trapped state may provide a mechanism by which the kinase can recruit cahnodulin. With each successive Ca2’ spike an increasing number of kinase subunits would become fully active, potent- ially enabling the kinase to detect the frequency of Ca2’ spikes. Once calmodulin dissociates tiom a subunit in the CaMKp state, that subunit does not fully deactivate and thus remains partial1 active or autonomous at basal Ca2’. The Ca2Y-independent activity of Cah3Kp ranges from 20-80% of maximal Ca2+stimulated activity, depending on the substrate and assay conditions used. CaMKp remains responsive to Ca2’/ calmodulin which activates it to the same maximal activity as CaMKO, i.e. 100%. When cahnodulin is removed from CaMK’ the kinase undergoes a ‘burst’ of autophosphorylation at sites distinct from the autonomy site t;#enerate the thhd state of phosphorylation, CaMK . This phase of autophosphorylation is activation of GMKpP the activity of CaMK maximal unstimulated level (20-8095) either with or without Ca2t/calmodulin. The molecular basis for this regulation of CaM kinase is described below.

Table 1 Functional states of CaM kinase cah4K” Activity - Ca*’ Activity t Ca*’



caMKp 2040% 100%

Trapped CahN’ 100% 100%

caMKpm 2040% 20-80%


Regulatory domain Catalytic domain

Association domain


Fig. 1 Domain struchnr of CaM kinase isoforms. The approximate position of the amino acids within the regulatory domain of the a isoform is shown. Diffexences between isoforms are indicated by ‘inserts’ of 21 to 34 amino acids relative to the a isofonn. The schematic representation of a kinase subunit illustrates how binding of cabncdulin may activate the enzyme by displacement of an autoinhibitory segment


60 080) which show a broad tissue distribution [S]. Recent electron micrographs suggest that ueuronal CaM kinase consists of a subuuit decamers (540 000 daltons) and p subunitoctamm (480 000 didtons)[9]. The st~cWfunctional domaiosof CaM kinasc are shown schematically in figure 1 15, 6. 81. The cDNA for each isofouu appears to encode all the necessary information for kinase activity and for its activation by Ca2’/cahnodulin. The protein kiuase catalytic domain is encoded in the N-teuninal end of each isofomr (amino acids 1 to 263-271). The regulatory domain contains a calmoduliu-binding domain that overlaps with au autoiubibitory segment and extends from approzhuately amiuo acids 275-314. The mmaining portion of each isoform is termed the oligomerizatioulassociation domain and is responsible for allowing assembly of the kinase into a multimer and perhaps for targetiug the euzyme to various intracellular sites. Electron microscopic analysis suggests that the oligomerization domains of all subunits in a holoenzyme assemble into a single globular structure which is surrounded by 8-10 globular structures coataining the catalytic and regulatory domains of the individual subunits.

Basal and activatedstates of CaM khase Structural features of CaM kinase CaM kinase is conspicuously large for a protein kinase. Biochemical analysis indicates that neuronal CaM kinase consists of thme closely-related subunits which assemble into a multimeric enzyme of 8-10 subunits. Subsequent cloning of the three isoforms from brain identified them as a (478 amino acids: Mr 54 llO), B (542 amino acids; Mr 60 333). and p (527 amino acids; Mr 58 705) [5-71. The p and 8’ subunits are likely generated by alternative splicing and will be referred to collectively as the g subunit [7]. Non-neuronal CaM kinase is also multimeric and appears to be composed of various tissue-specific subunit isoforms that are distinct from c1 and j3. Some of these isofomx may correspond to identified cDNA clones y (527 amino acids; Mr 59 038) and 6 (533 amino acids, Mr

Activation of CaM kinase by Ca2t/calu&ulin is shown schematically in Figure 1. The catalytic domain of CaM kinase has intrinsic activity. but is kept inactive in the holoenzyrne because an autoinhibitory segment of the regulatory domain inhibits the catalytic domain in the basal state. Transition of CaMKOfrom the inactive to the fully activated state by de-inhibiting the catalytic domain is effected by binding of Ca2’/cahnodulin to the cahuodulinbinding segment of the regulatory domain. An assumption of the model in which CaM kinase is activated by de-inhibition is tbat the catalytic fragmeut of the kinase has intrinsic phospho- trauatbrase activity. This is supported by the finding that partial pmteolysis of the kinase produces a catalytic domain with constitutive activity [lO-121. Such activation of soluble CaM kiuase requires prior autophosphorylation [l 1, 121.


The active fragment (30 kD) consists of the N-terminal end of the kinase, from amino acids 1 to 271. If proteolysis is canied out without prior autophosphorylation, a longer fragment is obtained (31 kD) that extends to amino acids 293 or 299. Thus, CaM kinase (1-271) has intrinsic activity but is inhibited in the basal state by an autoinhibitory segment that is contained between amino acids 271 and 299. A similar conclusion was reached by engineering truncated constructs of the c1 subunit [13]. CaM kinase (l-282) was constitutive whereas CaM kinase (l-358) retained sensitivity to calmodulin. The region of the regulatory domain encoding an autoinhibitory segment has been delineated with synthetic peptides and by site-directed mutagenesis. A synthetic peptide corresponding to amino acids 273-302 inhibits CaM kinase in either the presence or absence of cahnodulin and does not bind calmodulin 1141. Such a peptide selectively inhibits CaM kinase without inhibiting other calmodulindependent processes. Determinants at either end of this region contribute significantly to autoinhibition. Without the N-terminal end of this peptide (in peptide 284-302) little inhibition is seen [14]. Alterations in this region of the kinase, e.g. replacement of the N-terminal amino acids His28 and Arg2ss with Asp in the a-CaM kinase, generate a higbly Ca2+-independent enzyme [15]. At the C-terminal end, deletions of amino acids 291-315 or of 304-315 resulted in significant disruption of the autoinhibitory domain and generated a Ca2+-independent kinase [13]. Peptides corresponding to amino acids 281-302 am equally effective inhibitors as longer peptides and are likely to include most of the autoinhibitory segment of the kinase [16, 171. What is the mechanism of this autoinhibition? The autoinhibitory domain may keep the kinase inactive by blocking access to both peptide and ATP binding sites [l&20]. Binding of Ca2’/calmodulin to the 281-309 peptide reduces its inhibitory potency by more than lo-fold suggesting that binding of calmodulin to the kinase disrupts the functioning of the autoinhibitory domain [18, 19, 211. This is consistent with the notion that cahnodulin activates the kinase by perturbing its autoinhibitory domain. Competition for peptide substrate may be localized to the C-terminal end of


the autoinhibitory domain since a peptide corresponding to this region (290-309) weakly inhibits phosphorylation competitively with peptide substrate and non-competitivel&2with Al’P2 116, 18, 191. in a peptide Replacement of His by fi corresponding to amino acids 281-302 significantly reduces potency aud changes tbe kinetics of inhibition from competitive with ATP to noncompetitive with ATP. His282 at the N-terminal end of the autoinhibitory domain may therefore have an important function in blocking access of ATP to its binding site. The general region of calmodulin binding to CaM kinase was predicted based on homology to other calmodulin-binding proteins (Fig, 1). This is supported by studies using synthetic peptides corresponding to the putative calmodulin-binding domain. The studies indicate that the core calmodulinbinding domain is contained in amino acids 296 309 of the 01subunit [16, 221. The last 5 residues are hydrophobic and particularly important for calmodulin binding. The N-terminal end of the cahnodulin-binding domain overlaps with the autoinhibitory segment A synthetic peptide corresponding to amino acids 281-309 binds cahnodulin and inhibits CaM kinase activity [21]. Inhibition is attenuated in the presence of calmodulin. The autoinhibitory segment in the intact enzyme may therefore be able to interact with either the catalytic domain or with calmodulin, but not both simultaneously. Bindin to calmodulin predominates in the presence of Ca9t and the kinase is shifted to its activated state. Half-maximal activation of CaMKO occurs at OS-l.0 pM free Ca2’ (at saturating calmodulin) and at 25-100 nM cahnodulin (at saturating Ca2+>1231. The sensitivity of CaM kinase to cahnodulin is set considerably lower than other cahnodulindependent enzymes which are activated at 1 nM calmodulin or less. Maximal activation occurs when all subunits of a holoenzym have bound calmodulin [241. Calmodulin thus serves as the Ca2+sensor for CaM kinase. The affinity of calmodulhr for Ca2’ and the cooperative nature of Ca2+ binding to calmodulin assures that CaM kinase is inactive at basal physiologic Ca2+concentration (100 III@ while providing it with the sensitivity to be hi hly activat- ed upon cell stimulation that elevates Ca% to OS-l.0 pM.



Ca2+ spikes



Trapped f

Fig. 2 h4odcl of CaM k&se phosphqlation states and their activationby Ca%4modulin and by autop&.qhorylation. The catalytic andFegulntorydoMltroftwoneighboriagsubunitrofaholoemymarcshown. l%etkcpho@orylationsitasintheaiaofamamthc autonomy site @. indicated by circle) and two inhibitocy sitee f15 aud ‘I%?$ indicakd by qua@. Ci& and rqma~ am shown as filled in the phosphoPylatedstate and open in the dephusphotylatcdstak Binding of cabnoduliu (oval shape) activatee each subunit in the basal state (CaMK’) by displacement of an autoiahibitory cxegment. Activation aho permita inbx-Bubunit autopbo@unyWon of an autonomy site when two neighbming &units have cabnodtdia bound. An autophospborJrhtedsubunit of CaImodulin. (CaMKp) is in a highly active happed state while ailmadub ia bound and in an autmlanoua state ai& dkucii Subsequent autophosphorylationof the inhibitory sites produces a capped state (CahXT which is evenmany dephosphoryb~kdto CaMKO

Regulation of CaM kinase by autophosphorylatlon Autophosphmylationof da6 caimudulin

traps bound

Autophosphorylation traps hound calmod& hecause the dissociation rate of cahx&ulin from CaMK’ is markedly slower than its dissociation from CaMKOW]. The dissociation raae of fluorescently-labeled calmodulin at high Ca2+ is reduced by lOOO-fold,from an off-rate of 0.4 s to an off-rate of several hundred seconds. Autophosphorylation

converts CaM kinase from an enzyme with one of the weakest affinities for calmodulin (45 nh4) to au enzyme with one of the highest affinities for calmod& (60 PM). Even when Ca2’ concentration is reduced below the level needed for activation of W kinase (e.g. 100 nM), fluorescentlylabeled cahnod&n remains trapped for a minimum of 10 s (Fig. 2). caz’/calmodulu1 * stimulates xnultiple phosphorylation of CaM N, up to 3-4 moles ofphosphateare&cqoraWpersubunitinthe holoeazyme. Of these, phosphorylation of ‘I’b?s6 is essential for enhanced affinity for calmodulin. Trapping is seen in recombinant cz-CaMkinase but



1itCa2+ Rapid spiking


tit Ca2+ it Ca2+

4 Ca2+ .

Ng. 3 Recruitmentof calmoddin may enable molecular potentiationof Ca*’ signals at high stimuli fnquency. The catalytic and regulatory domain of a holoeuzyme are depicted during (3% oscillations, starting with a submaximal stimulus that allows three cahnodulin molecules (blackcircle) to bind three of the subunits. Adivated subuuits any shown with a plue while autopkphorylated subunits am shaded. The initial stimulus leads to the activationof thne subuuitsbut only to the inter-subtit autoplrqkylatioo of the one subunit whose neighbor also bound cabnodulin. Cabuodulin remains trapped to that sub& afkr Ca*’ levels decline and paaicipates in tbe inter-subunit autophosphorylationof two subunits during the next Ca* rise. This feed-forward effect leads to recruitmentof additional cabnodulin with each successive G-I*’rise so that all subunits tm~activated during the last of a series of high frequency stimuli

not in a mutant in which Th?86 is replaced with Ala286[25]. This is the same phosphorylation site that is responsible for switching CaM kinase to a Ca’+-independent enzyme (see below). This site is on the N-terminal side of the calmodulin-binding domain and within the auto- inhibitory region. It is possible that the calmodulin- binding domain of CaMKO is restrained in a conformation with a low afftity for cahuodulin and that positioning a phosphoryl moiety on Thrasa nearby releases this constraint to allow tight binding of calmodulin to cah4Kp. What might be the consequence of calmodulin trapping? An obvious consequence of trapping

would be to extend the activity of the kinase for several seconds after Ca2’ declines to subthreshold levels. Neuronal Ca2’ spikes lasting less than a second would be potentiated. An equally important effect may arise under conditions in which individual Ca2’ rises do not maximahy activate CaM kinase. Tra-ting may then enable the kinase to serve as a Ca spike frequency detector. Submaximal activation would occur, for example, in any cellular compartment in which CaM kinase and other calmodulin-binding proteins exceed the concentration of calmodulin. It may also occur if the (Ca2t)~-calmodulincomplex is limited by Ca2’ oscillations that ate very rapid or of low magnitude.


At low frequency of stimulation each Ca2+ rise would elicit the same submaximal stimulation of the kinase. At high frequency, the inter-spike interval becomes too brief for significant dissociation of trapped cahnodulin and successive stimuli will involve CaM kinase molecules increasingly saturated with calmodulin. Binding of calmodulin would occur onto holoenzymes still retaining calmodulin trapped from a previous stimulus leading to an accumulation of calmodulin and an increase in kinase activity with each successive stimulus (Pig. 3). The efficiency at which bound calmodulin is trapped increases as the number of kinase subunits with bound cahnodulin increases (Meyer et al., unpublished). This likely reflects the cooperative nature of autophosphorylation [26] which may involve inter-subunit autophosphorylation between two neighboring subunits of an oligomer, each bound by calmodulin (Pig. 2). As the proportion of subunits in a holoenzyme with trapped cahnodulin increases, the probability of having calmodulin on proximate subunits and the probability of autophosphorylation and further trapping increases. The cooperativity of trapping may allow the kinase to function as a frequency detector with a threshold frequency beyond which CaM kinase becomes increasingly saturated with calmodulin and highly active 1251.


kinase substrates and is conserved in all the known isoforms of the enzyme. Dephosphorylation of Th?s6 converts the kinase back to the Ca2’dependent species [28,29,3 1 . 6 Phosphorylation of & is necessary and sufficient for the generation of Ca2+-independent activi . Recombinant a CaM kinase, in which Th$ is replaced with Leu or other nonphosphorylatable amino acids by site-dire&d mutagenesis, is incapable of becoming autonomous [37-391. Replacement of other potentially phosphorylated residues in and around the regulatory region with non-phosphorylatable amino acids has no effect on autophosphorylation at Th?16 or on generation of autonomy [37]. Replacement of f16 with As demonstrated that negative charge in place of Ts *’ is alone sufficient to produce a highly Ca2t-independent kinase [15, 381. Thus, autophosphorylationof Thra86may be sufficient and necessary to block deactivation of CaMK’ after dissociation of calmodulin by disrupting the inhibitory function of regulatory domain. Autophosphorylation of Th?86 occurs largely as an intraholoenzyme reaction [29, 401. However, within the holoenzyme the phosphorylation may occur as either an intra- or inter-subunit reaction. Monomers of CaM kinase, generated from recombinant c1 subunits truncated in the association domain have been used to address this question Autophosphorylationof Th?86 switchesthe kinase (Hanson and Schuhnan, unpublished). In these to a Ca2’-independentor autonomousstate monomers, Ca2t/calmodulin-stimulated autophosphorylation of ds6 occurs largely as an interSaitoh and Schwartz [27] first demonstrated that molecular reaction. Within a holoenzyme autophosAp sia CaM kinase is converted to a partially phorylation of Thrasa similarly occurs as an interB Ca +-independent species upon autophosphorylation. molecular, i.e. inter-subunit, reaction. In holoThis seminal report was extended by studies usit! enzymes containing a mixture of both catalytically purified rat brain CaM kinase [28-311. Ca / inactive and active subunits, da6 on the inactive calmodulin stimulates incorporation of approxim- subunits is found to be phosphorylated by the active ately 3-4 moles of phosphate per subunit in the subunits (Pig. 2). holoenzyme. However, phosphorylation of only one The switch to an autonomous CaM kinase has of these sites is neces and sufficient to switch been demonstrated in situ in several cell systems. s%?-independent species. A the enzyme to the Ca Depolarization in the presence of extracellular Ca2+ unique P-thmonine containing phosphopeptide was significantly increases Ca2+ independent CaM found to coincide with the appearance of Ca2’- kinase activity in intact synaptosomes, cerebellar independent kinase activity [32-351. It was granule cells, PC.12 cells, GH3 cells, and acute identified as Th?86 (in the a subuniS Th?*’ in the hippocampal slices [41-461. From 15-5096 of CaM j3 subunit) by dixect peptide sequencing [33-361. kinase can be converted to the autonomous state. The site is within the consensus sequence for CaM Ca2+-independent CaM kinase activity is also



induced by neurotransmitters,hormones, and growth protect the inhibitory sites while ‘presenting’ Thras6 factors, thus suggesting that the activation of CaM as a substrate for a neighboring subunit in the holokinase is on the pathway of numerous signalling enzyme. Upon dissociation of Ca2+/cahnodulin systems, including those involving phosphatidyl- from CaMK’, the C-terminal end of the autoinhibitinositol (PI) tumover, ligand-gated Ca2’ channels ory domain may be positioned near the phosphoand tyrosine kinase-bearing receptors. For example. transferase site, and Thrao5 and Th?06 then become Ca2+-independentCaIvI kinase activity is enhanced the preferred sites of Ca2t/calmodulin-independent in PC12 cells in response to bradykinin, which autophos horylation in the reaction generating stimulates the PI signalling [43]. Bradykinin causes Cah4KPlFp . an initial Ca2+spike as intracellular Ca2+stores are released in response to IPa. This activates C&l Is CaM kinase a neuronal memory molecule? kinase and increases its Ca2+-independent kinase activity from a baseline of 3% to a stimulated level Several characteristics of CaM kinase suggested that of 22% within 10 s. A significantly elevated auton- it could act as a neuronal memory molecule capable omous activity is seen for several minutes after Ca2+ of sustained activity a&r brief exposure to Ca2’ falls below threshold for activation of the kinase. 128, 51, 521. A ‘threshold’ level of autophosAutophosphorylation may therefore potent&e CaM phorylation would switch each holoenzyme to the kinase activity in situ. ‘on’ state, and as long as this minimal level of phosphate was retained, all subunits in the oligomer Ca2t-independentautophosphorylation blocks would be in the Ca2’-independent state. Since a activationby calmodulin switched kinase also continues to autophosphorylate [28, 30, 34, 471, the kinase should be able to Initial autophosphorylation with Ca2+/calmodulin maintain its ‘on’ state because subunits in the enables the kinase not only to phosphorylate CaMKp state would reverse the action of phosphatsubstrates in the absence of Ca2’/caImodulin but to ases that convert some of them to CaMK”. The continue autophosphorylating itself to generate ‘on’ state could even be extended beyond the C~.IVK~~[28, 30, 34, 471. This phase of autophos- lifetime of the molecule, for example, by insertion phorylation decreases calmodulin binding to the of nascent kinase subunits into holoenzymes in the kinase [48] and caps the activity at the level of its ‘on’ state in place of subunits undergoing natural Ca2+-independent activity [47]. Tryptic phospho- protein turnover. A number of requirements for such a model have peptide mapping demonstrates that Ca2t-independent autophosphorylation occurs at novel sites [34, not been met, however. A recent study suggests 48,49 . The sites am located in peptides containing that there is no cooperativity to the switch [531. A ThrfO51 , Tdy and Serar4 in the calmodulin-binding linear correlation was found between the fraction of domain (of the a subunit) [49]. Site-directed muta- subunits autophosphorylated and the fraction of genesis was used to identify the phosphorylated kinase activity switched to the Ca2t-independent residues as Thrso5, Th?06, and Sera 4 [50]. Auto- state. There was no threshold for the switch at l-4 phosphorylation is functionally redundant, since a subunits per holoenzyme so that subunits in the phosphate moiety on either Thrao5 or Thrao6 is CaMK” state do not appear to contribute to the sufficient to block binding of calmodulin and com- Ca2t-independent activity. A second important requirement of the memory model, maintenance of pletely inhibit Ca2t/cahnodulin-stimulatedactivity. Studies with autoinhibitory domain peptides the autonomous activity by Ca2’-independent autocontaining Thrzsa suggest that this autonomy site is phosphorylation, has also not been demonstrated. only an effective substrate for a constitutive kinase The critical event in propa ation of the ‘on’ state in when calmodulin binds and disrupts the inhibitory the model above is Ca2q/cahnodulin-independent component of the peptide, i.e. it is only a substrate autophos horylation of subunits in the CaMK’ state % when calmodulin is present [18, 19, 211. In the on Thra8. However, while kinase in the ‘on’ state holoenzymc, binding of Ca2t/caImodulin would then is capable of rapid Ca2’-independent autophos-


phorylation, this autophos~knylation converts CaMKp subunits to CaMK’ rather than CaMK’ subunits to CaMK’. Ca2+-independent autophosphorylation does not occur on Thr886 so that CaIvlKp is not generated 134, 47, 49, 501. Autophosphorylation therefore per- mits a potentiation of Ca2’ signals but current evidence only supports a potentiation over seconds and minutes. A modelfor CaM khase regulation by autophosphorylation

Generation of the different activity states of CaM kinase by autophosphorylation is depicted using two neighboring subunits in a holoenzyme (Fig. 2). Prior to activation by Ca2’+/calmodulin,the autoinhibitory domain of each subunit maintains basal catalytic function very low. Subunits in a holo are activated individually by the binding of enzP Ca kalmodulin. With some probability, Ca2’/ calmodulin will bind to the two neighboring subunits, and avail their catalytic site to protein substrates and ATP by displacing the autoinhibitory domain. Calmodulin masks the two inhibitory sites (Thrso5 and Thrsc6) so they cannot be phosphorylated in its presence. Inter-subunit autophosphorylation then proceeds when an activated subunit is presented with I%?*’ from the adjacent subunit by the binding of calmodulin. In this example, only one subunit becomes autophos horylated during a brief exposure to elevated Ca2P. The subunit that was not autophosphorylated during the Ca2’ transient, rapidly deactivates as calmodulin dissociates, similar to many calmodulin-dependent enzymes. By contrast, the autophosphorylated sub unit remains maximally active while calmodulin is trapped, and retains Ca2t-independent activity after calmodulin slowly dissociates. An autophosphorylated subunit is therefore maximally potentiated while calmodulin is trapped and partially potentiated when calmodulin dissociates. It is intriguing to consider that calmodulin trapping under conditions in which individual stimuli are submaximal allows for accumulation of calmodulin with successive stimuli and kinase activation that is dependent on the frequency of stimulation. When trapped calmodulin dissociates, the calmodulin-binding domain is vacated allowing the now exposed Th?


and Th? to be rapidly autophosphorylated in a Ca2+-independentreaction (not shown). This blocks rebinding of &nod&u, capping the activity of that subunit at the level of its autonomous aetivity. Dephosphorylation eventually deactivates the kinase. CaM kinase appears to be elegantly designed to detect chauges in Ca2’ as well as to accommodate the unique features of the Ca2+ signalling system Biochemical studies will no doubt continue to provide us with unexpected properties of the kinase which will be tested in intact neuronal systems.

References 1. Hanson PI. Schubnan H. (1992) Neumnal Ca~/cabnodulindependentprotein kiuases. Annu. Rev. B&hem., 61. 559-601. 2. Rostas JAP. Dmddcy PR. (1992) Distributionof multiple forms of cakiumWmodulin stimulatedprotein kinaseII in

brain.J. Neumcban., Inprers. 3. Dunkky PR (1992) Autophospboqlatioa of neumna! cakkmWmodulin-stimulated protein kinase II. Mol. Neurobiol, In press. 4. Colbma RJ. Soderling TR. (1990) Caki~admodulinindependentautopGphcryiation sites ofcakium/ calmcdulin-dependentpmtein kinase II. Studies on the effect of phosphorylationof tireon& 305/306 and serine 3 14 on calmodulin binding using synthetic peptidea.J. Biol. Chem., 265,11213-11219. 5. Lin CR. Kapikff MS. DurgerianS. et al. (1987) Mokcubu cloning of a brain-specificcakium/calmodulin-depe&ut proteinkinase. Pmt. NatLAcad. Sci., USA, 84,5%2-5966. 6. Bennett MK. Kennedy MB. (1987) Deducedgrimary stmctureof the fJsubunit of brain type II Ca /calmodulindepemknt proteinkinase determ&d by mokcuhu cloning. Proc. NaU. Acad. Sci., USA, 84,1794-1798. 7. Bulleit RF. Bennett MK. Molloy Ss. Hurky JB. Kezmedy MB. (1988) Consecvedand variable regions in the subunits of braintype II Ca~/calmodulin-dependeat proteinkinase. Neuron, 1.63-72. 8. TobimatsuT. Fujisawa H. (1989) Tissue-specifk expmasionoffourtypesofratcalmodG+pe&ntprotein kinase II mRNAs. J. Biol. Ckm, 264. 17907-17912. 9. Kauaseki T. lkeuchi Y. Sugiuca H. Yamauchi T. (1991) struchua1feaaues of Ca+abnodulin-dependependent protein kinase II revealed by electron microscopy. 1. Cell Biol., 115, 1049-1060. 10. Levine H. III Sabyouo NE..(1987) Chamcterizationof a soluble Mr 30,000 catalytic &agmeatof the neumnal cabnodulia-deptit protein kinase II. Eur. J. Biochem., 168,481~486. 11. Kwiatkowaki AP. Kiug MM. (1989) Autopbo+orylatioa of thetypeIIcabnodulin~tproteinkinaseisessential for fobrmation of a protbolytici+agmentwith catalytic



















activity. hnplications for long-term synaptic potention. Biochemistry, 28,5380538X Yamagata Y. Cxemik AJ. Gmengard P. (1991) Active catalytic fragment of Ca%xilmod”En-dependent protein kinase II. Purifiiation, chamcterixation, and structural analysis. J. BioL C&m., 266.15391-15397. Hagiwara T. Oh&o S. Yamauchi T. (1991) Studies on the mgulatoty domain of Ca*+/calmodulin-dependent protein kinase II by expression of mutated cDNAs in .?Ischsrfchia coil. J. Biol. Chem., 266,16401-16408. Malinow R. Schubnan H. Tsien RW. (1989) Inhibition of postsynaptic PKC or CaMKII blocks induction but not expression of LTP. Science, 245862-865, Waldma~ R. Hanson PI. Schulman H. (1990) MultifunctionaI Ca%almoduliu-dependent protein kinase made C!a*+independent for functional studies. Biochemistry, 29,1679-1684. Payne ME. Fong Y-L. Ono T. et al. (1988) Calcium!calmodulin-dependent protein kinase II. mation of distinct cahuodulin binding and inhibitory domains. J. BioI. C&m. 263,7190-7195. Nichols RA. Sihra TS. Cmrnik AJ. Naim AC. Gmengard P. (1990) Calcium/calmodulin-dependent protein kinase II increases glutamate and noradrenahne release from synaptosomes. Natum, 343.647-651. Colbran RJ. Fang Y-L. Schwomr CM. SoderIing TR. (1988) Regulatory interactions of the cahnoduIin-binding, inhibitory, and autophosphorylation domains of Ca*+/cabnodulindependent protein kinase II. J. Biol. Chem., 263.18145-18151. Colbran RJ. Schworer CM. Hashimoto Y. et al. (1989) Calcium/calmoduIin-dependent protein kinase II. Biochem J.. 258,313-325. Smith MK. Colbran RJ. B&key DA. Soderling TR. (1992) Functional determinants in the autoinhibitory domain of calcium/cahnodulindependent protein kinase II. Role of His282 and multiple basic residues. J. Biol. Chem., 267, 1761-1768. Kelly PI’. Weinherger RP. Waxham MN. (1988) Active site-directed inhibition of Ca*+/cahuodu1in-dependent protein kinase type II by a bifunctional calmodulin-binding peptide. Pmt. Natl. Acad. Sci., USA, 85.4991-4995. Hanley RM. Means AR. Kemp BE. Shenolikar S. (1988) Myping of cahnodulin-binding domain of Ca +/cabnodulin-dependent protein kinasc II from rat brain B&hem. Biophys. Res. Commun.. 152.122-128. Schulman H. (1988) lbe multifunctional Ca~/calmodulin-dependent protein kinase. In Advances in Second Messenger and Phosphoprotein Research. 22.39-l 12. Katoh T. Fujisawa H. (1991) Autoactivation of cahnodulindependent protein kinase II by autophosphorylation. J. Biol. Chen~, 266.3039-3044. Meyer T. Hanson PI. Stryer L. Schulman H. (1992) Cahnodulin trapping by calcium-calmoduhn-dependent protein kinase. Science, 256.1199-1202. LeVine H. III Sahyoun NE. Cuatmcasas P. (1986) Binding of cabnodulin to the neuronal cytoskektal protein kinase type II ccoperatively stimulates autcphosphorylation. Prcc. Natl. Acad. Sci., USA, 83.2253-2257. Saitoh T. Schwartz JH. (1985) Phosphorylationdependent














8ubcelluhu translocation of a Ca%almodulin-dependent protein kinase produces an autonomous enxyme in AJ?1ysgX neutons. J. Cell Biol., 100, 835-842. Miller SB. Kennedy MB. (1986) Re&tion of brain type II Ca*+/caImodulin-dependent~mtein kinase by autophosphorylatiom A Ca +-ttiggemd molecular switch. Ceil, 44,861-870. Lai Y. Nahn AC. Gmengatd P. (1986) Autophosphorylation reversibly reguIates the Ca%ahnodulin-dependence of Ca2t/cabnodulin-dependent protein kinase II. Proc. NatL Acad. Sci., USA, 83.4253-4257. Lou LL. Lloyd SJ. Schubnan H. (1986) Activation of the multifunctional Ca%ahnodulin-dependent protein kinase by autophosphorylation: ATP modulates production of an autonomous enxyme. Proc. Natl. Acad. Sci., USA, 83, 9497-9501. Schworer CM. Colbran RJ. Soderling TR. (1986) Reversible generation of a Ca2t-i&pendent form of Ca*+(calmodulin)-dependent protein kinase II by an autophosphoryhuion mechanism. J. Biol. Chem., 261, 8581-8584. Lai Y. Naim AC. Gomlick F. Greengard P. (1987) Ca*+/r&nodulin-dependent protein kinase Ik Identigcation of autophosphorylation sites msponsible for generation of Ca*+/cahnodu1in-independence. Proc. Natl. Acad. Sci., USA, 84,5710-5714. Schwomr CM. Colbmn RJ. Keefer JR. Soderhng TR. (1988) Ca*+/calmodulin-dependent protein kinase Il. IdentitIcation of a regulatory autophosphorylation site adjacent to the inhibitory and calmodulin-binding domains. J. Biol. Chem., 263,13486-13489. Lou LL. Schubnan H. (1989) Distinct autophosphorylation sites sequentially produce autonomy and inhibition of the multiRmctiona1Ca~/cahnodulin-dependent protein kin-. J. Neurosci., 9,2020-2032. Miller SO. Patton BL. Kennedy MB. (1988) Sequences of autophosphorylation sites in neuronal type II CaM kinase that control Ca%-independent activity. Neuron, 1.593604. Thiel G. Cxemik AJ. Gonlick FS. Naim AC. Greengard P. (1988) Csz’lcalmodulin-depen&nt protein kinase II: Identifilcation of thmonhm-286 as the autophcephorylation site in the alpha subunh associated with the gene-mtion of Ca*‘-independent activity. Proc. NatI. Acad Sci. USA, 85, 6337-6341. Hanson PI. Kapiloff MS. Lou LL. Rosenfeld MG. Schubnan H. (1989) Expmssion of a multifunctional Ca*+/calmoduIin-dependent protein kinase and mutational analysis of its automgtdation. Neuron, 3,59-70. Fong Y-L. Taylor WL. Means AR. Soderling TR. (1989) Studies of the regulatory mechanism of Ca*+/caImodulin-dependent protein kinase II. Mutation of thmonine 286 to ahmine and aspaitate. J. Biol. Chem., 264, 16759-16763. Waxham MN. Aronowski J. Westgate SA. Kelly PT. (1990) Mutagenesis of Thr-286 in monomeric Ca2+/cabnodulin-dependent protein kinase II eliminates Cazt/calmodulin-independent activity. Pmt. Natl. Acad. Sci, USA, 87.1273-1277. Kumt J. Schulman H. (1985) Mechanism of autoplmsplmrylation of the mult&nctional









Ca2%&nodulin-dependent protein kinase. J. Biol. Chem., 260,6427-6433. Gorelick FS. Wang JKT. Lai Y. Naim AC. Greengard P. (1988) Autophosphorylation and activation of Ca2+/cabnoduEn-dependent protein kinase II in intact nerve terminals. J. Biol. Chem., 263,17209-17212. Fultun~ta K. Rich DP. Soderling2~ (1989) Generation of the Ca -independent form of Ca /cabnodulin-dependent protein kinase II in cerebellar granule cells. J. Biol. Chem., 264.21830-21836. MacNicol M. Jefferson AB. Schulman H. (1990) Ca2+/cabnodulin kinase is activated by the phosphatidylinositol signaling pathway and becomes Ca2’-independent in PC12 cells. J. Biol. Chem., 265, 18055-18058. Jefferson AB. Tmvis SM. Schuhnan H. (1991) Activation of multifunctional Ca2%almodulin-dependent protein kinase in GH3 cells. J. Biol. Chem., 266, 1484-1490. Ocorr KA. Schuhnan H. (1991) Activation of multifunctional Ca2+/calmodulin-dependent protein kinase in intact hippocampal slices. Neuron, 6,907-914. Molloy SS. Kennedy MB. (1991) Autophosphorylation of type II Ca2%ahnodulin-dependent protein kinase in cultures of postnatal rat hippocampal slices. Pmt. Natl. Acad. Sci.. USA, 88,4756-4760. Hashimoto Y. Schworer CM. Cobran RJ. Soderling TR. (1987) Autophosphmylation of Ca2~/cahncdulii-dependent protein l&se II. Bffects on total and Ca2’-independent activities and kinetic parameters. J. Biol. Cbem.. 262, 8051-8055.


48. Lickteig R. Shermlikar S. Denuer L. Kelly PT. (1988) Regulation of Ca~+/calmodulin-depatdeJtt pm&n kinaae II by Ca2+/calmcdulin-i&per&t auu&osphorylation. J. Biol. Chem., 263.1923219239. 49. Patton BL. Miller X3. Kenuedy MB. (1990) Activation of type II cakiumlcahnodulindependent protein hinase by Ca2%ahnodulin is inhibited by autophoaphoryIatkn of thmonine within the calmodulin-binding domain. J. Biol. Chem., 265,1120411212. 50. Hanson PI. Schuhnan H. (1992) Inhibitory . * au~+phosphoryktton of multdimcuonal Ca /calmodulin-dependent protein kinase analyzed by site-directed mutagenesis. J. Biol. Chem., In press. 5 1. Lisman JE. (1985) A meohanism for memory stooge insensitive to mokcuhu tumoverz A bistable autophosphorylating kinase. Proc. Natl. Acad. Sci. USA, 82. 3055-3057. 52. Lisman J. Goldring M. (1988) Evaluation of a model of term memory based on the proper&s of the lo?!Ca +/calmoduE&ependent protein kinase. J. Physiol. (Paris), 83.187-197. 53. Ikeda A. Okuno S. Fujisawa H. (1991) Studies on the generation of Ca2+/c&nodulii-independent activity of cabnodulin-dependent protein kinase II by autophosphory1ation. Autothiophosphorylation of the enzyme. J. Biol. Chem., 266.11582 11588. Please send reprint requests to : Dr H. Schuhnan, Department of Pharmacology, Stanford University School of Medicine, Stanford, CA 94305-5332. USA

Decoding calcium signals by multifunctional CaM kinase.

Multifunctional Ca2+/calmodulin-dependent protein kinase (CaM kinase) is one of the three major protein kinases coordinating cellular responses to hor...
1MB Sizes 0 Downloads 0 Views