Report

Functional Stoichiometry and Local Enrichment of Calmodulin Interacting with Ca2+ Channels

See allHide authors and affiliations

Science  16 Apr 2004:
Vol. 304, Issue 5669, pp. 432-435
DOI: 10.1126/science.1093490

Abstract

Calmodulin (CaM) interactions with Ca2+ channels mediate both Ca2+ regulation of channels and local Ca2+ triggering of transcription factors implicated in neuronal memory. Crucial to these functions are the number of CaM molecules (CaMs) regulating each channel, and the number of CaMs privy to the local Ca2+ signal from each channel. To resolve these parameters, we fused L-type Ca2+ channels to single CaM molecules. These chimeric molecules revealed that a single CaM directs L-type channel regulation. Similar fusion molecules were used to estimate the local CaM concentration near Ca2+ channels. This estimate indicates marked enrichment of local CaM, as if a “school” of nearby CaMs were poised to enhance the transduction of local Ca2+ entry into diverse signaling pathways.

The collaboration of Ca2+ channels and calmodulin (CaM) unifies two vital but diverse functions: Ca2+ regulation of channels (16) and local Ca2+ triggering of nuclear transcription (710). For regulation, CaM is a resident Ca2+ sensor that preassociates with channels before Ca2+ entry (1, 2, 11, 12), and Ca2+ binding to the different lobes of CaM can individually activate distinct forms of modulation (5, 13, 14). For transcription, CaM near a channel may be an initial trigger linking local Ca2+ influx to the activation of cyclic AMP response element–binding protein (CREB) (7, 9, 15), an event implicated in neuronal memory (16). For both functions, two critical parameters remain unknown: the number of CaMs regulating each channel and the number of CaMs close enough to sense the local Ca2+ signal emanating from each channel.

Our approach to resolving these parameters was inspired by Ca2+-dependent serinethreonine kinases (CDPKs), which may be plant analogs of mammalian CaM kinases (17). In contrast to the mammalian case, in which CaM binds to and activates CaM kinase, CDPKs contain tandem kinase and CaM-like modules, linked within a single polypeptide. Consequently, such plant kinases themselves bind Ca2+ to induce activity, independently of free CaM. Although vertebrates lack CDPKs, we wondered whether similar fusions between mammalian Ca2+ channels and CaM—in which the CaM/channel ratio is explicitly constrained— could prove useful in gauging the number of CaMs affiliated with channels. As an initial candidate for this approach, we considered L-type Ca2+ channels (CaV1.2), a representative (5) and relatively well characterized (18) member of the Ca2+ channel family. Ca2+-free CaM (apoCaM) preassociates with the PreIQ3–IQ segment of the main subunit of L-type channels (α1C) (12, 19) (Fig. 1A), and subsequent Ca2+-CaM interaction with the IQ region initiates Ca2+-dependent channel inactivation (CDI) (2). Several CaMs may induce CDI, as multiple α1C segments can interact with CaM (1, 3, 12, 1921). In regard to transcription, L-type channel Ca2+ current preferentially increases phosphorylation of nuclear CREB (22), even though other channels support equivalent overall Ca2+ entry (9, 15). Such preferential signaling likely relates to findings that local Ca2+ elevations near L-type channels are the critical Ca2+ signal for CREB activation (7). Moreover, IQ mutations attenuate CREB activation (Fig. 1A) (9), which suggests that CaM(s) situated very close to L-type channels are key transducers of local Ca2+ (18). In fact, CaM translocation to the nucleus may be one messenger linking local Ca2+ to CREB (15). For L-type channels, then, the number of “neighborhood” CaMs clearly presents as a potentially important signaling parameter.

We first considered L-type channel fusions to mutant CaM (CaMMUT) (Fig. 1A), rendered insensitive to Ca2+ by mutations in all four EF-hand domains (1). Because CaMMUT acts as a dominant-negative to eliminate CDI through occupancy of the channel preassociation site (1, 2, 23), the lack of CDI in channel-CaMMUT chimeras would provide an obvious initial indication that a fused CaM participates in CDI. Before fusion to CaMMUT, L-type channels showed strong CDI. Ca2+ currents decayed faster than Ba2+ currents (Fig. 1B, left); and the fraction of peak Ca2+ current remaining after 300-ms depolarization (r300) showed a classic, ∪-shaped dependence on test-pulse voltage (red circles). The corresponding Ba2+ relation (black circles) declined less, which reflected a slower, voltage-dependent mechanism (1, 23). The difference between relations (f, Fig. 1B, bottom) thus quantifies pure CDI. By contrast, when CaMMUT was fused to the main α1C subunit through polyglycine linkers, CDI was completely eliminated (Fig. 1B, G2 and G12), which suggests that preassociation of a single CaMMUT abolishes Ca2+ regulation of channels.

Still, it was possible that the linkage of CaMMUT caused nonspecific disruption of CDI. We therefore tested whether fusion of wild-type CaM (CaMWT) would permit normal CDI. If fusion of CaMMUT were to induce a generalized breakdown of CDI, fusion of CaMWT would be unlikely to rescue CDI. By contrast, channel fusions to CaMWT, using linkers of 4 to 12 glycines, gave rise to CDI (Fig. 1, C and D) indistinguishable from that of control channels (Fig. 1B).

In itself, however, the presence of CDI in channels fused to CaMWT did not explicitly exclude the collaboration of additional CaMs (endogenous) in the overall CDI mechanism. One possibility was that endogenous CaMs preassociated at site(s) distinct from that for the fused CaM and that Ca2+-driven movement of these secondary preassociated CaMs also contributed to CDI (Fig. 2A). We therefore examined the effects of overexpressing CaMMUT as a separate molecule. When native L-type channels are coexpressed with CaMMUT, CDI is abolished owing to the occlusion of preassociation site(s) (1, 2, 23). In contrast, when CaMMUT was coexpressed with channels fused to CaMWT1CΔ-G4-CaMWT), CDI was fully maintained (Figs. 2B and 1E). Thus, no secondary preassociated CaMs appear to participate in CDI. To test for the only remaining form of collaboration, that of free endogenous CaM to participate in CDI (Fig. 2C), we coexpressed channels fused to CaMWT1CΔ-G12-CaMWT) with PreIQ3–IQ, a 73–amino acid channel segment that binds both apoCaM and Ca2+-CaM with high affinity (12). Although PreIQ3–IQ functions as a “sink” for free CaM, CDI was unperturbed (Figs. 2D and 1E). Hence, one CaM appears to be both necessary and sufficient to produce CDI of its associated channel.

Fusions of channels and CaM also proved useful to examine the local concentration of endogenous CaM near channels (Fig. 3A). When linkers are short (Fig. 1B, G2 and G12), the effective local concentration of CaMMUT at the channel is enormous (Fig. 3A, left), and CaMMUT completely usurps the channel preassociation locus with consequent loss of CDI. Conversely, if linkers were lengthened sufficiently, the local concentration of CaMMUT would approach zero (Fig. 3A, right); endogenous CaM would dominate at the preassociation site and would yield normal CDI. The characteristic linker lengths for a transition between these extremes can be used to estimate the local concentration of endogenous CaM. In fact, when linkers were elongated beyond those in Fig. 1B, CDI progressively reemerged (Fig. 3B, G24 to G72).

Before quantifying these effects, we considered two trivial explanations for the recurrence of CDI. First, the longer linkers might have become progressively susceptible to proteolysis, yielding graded restoration of CDI. When we used antibodies against CaM, immunoblots of cells expressing channel-CaMMUT chimeras excluded this possibility (Fig. 3C). No intermediate bands for CaMMUT-containing fragments were present, excluding proteolysis. A second trivial explanation comes by explicitly considering the anticipated strength of CDI (f, Fig. 1B). $Math$(1) where f0 is the CDI strength for wild-type channels, and the bracketed terms are the respective local concentrations of endogenous CaM and linked CaMMUT at the preassociation site. Most relevant for the trivial case is the α term, specifying the ratio of dissociation constants for wild-type CaMlocal/endo and CaMMUT/linked. If CaMMUT/linked preassociates with lower affinity than CaMlocal/endo (α ≪ 1), then slight linker elongation, equivalent to small reduction of [CaMMUT/linked], could produce the re-emergence of CDI. However, live-cell, two-hybrid fluorescence resonance energy transfer (FRET) analysis (12) of apoCaM or CaMMUT binding to a major preassociation segment of the α1C subunit (IQ) yielded indistinguishable dissociation constants (Fig. 3D) (11). Moreover, L-type channels fused to CaMWT by G72 linkers showed CDI equivalent to that of control channels (Fig. 1D); this outcome renders unlikely the possibility that the long linkers themselves hindered the ability of fused CaM to preassociate with channels (fig. S2). These results argue that α ≈ 1.

We could then apply a classic formula developed by Flory (24), for polymer-chain statistics, to specify the local (molar) concentration of CaMMUT/linked fused to the channel. $Math$ $Math$ $Math$(2) where there are n glycines in the linker, r is the distance (in Å) from the linker origin to the preassociation site, and there are 2.4 glycines per statistical segment measuring 8.6 Å in length. Equations 1 and 2 can be combined to fit CDI strength (f) as a function of n (Fig. 3E), by using two free parameters (r and [CaMlocal/endo]). This procedure estimates the local endogenous CaM to be ∼2.5 mM, and the preassociation locus to be <10 Å (r) from the linker origin (Fig. 3E and fig. S3). This indication of strongly enriched local CaM seems robust, as fits based on measurements of free intracellular [CaM] in these cells (∼50 nM) (2527) diverge from our data by orders of magnitude (Fig. 3E). Also, the suggestion of increased local [CaM] is insensitive to the particular theory used for data analysis. Even if linked CaM was uniformly distributed in a sphere, estimated [CaMlocal/endo] would still be ∼1.3 mM (fig. S4). Thus, the concentration of local endogenous CaM near channels appears greater by several orders of magnitude than that for free cytoplasmic CaM.

Our experiments establish two critical parameters for channel regulation and signaling. That one CaM is both necessary and sufficient for CDI constitutes a simplifying advance for establishing a molecular mechanism of CaM-Ca2+ channel regulation. In general, the approach of linking regulatory modules to reveal functional stoichiometry is a valuable adjunct to structural biology. Even if structures resolved multiple CaMs per channel (28), the number of functionally important CaMs would still be indeterminate without a complementary strategy as illustrated here. Second, our results suggest that numerous CaMs are positioned near each channel, within earshot of the local Ca2+ signal. Assuming a hemispheric CaM domain with a 400 Å diameter (7, 29), a 2.5 mM concentration entails a local cadre of ∼25 free CaM molecules. Alternatively, the enrichment of effective local CaM concentration could reflect direct transfer of CaM from multiple nearby CaM buffer sites to the preassociation locus (30). In either case, the local Ca2+ plume of one channel need not signal through a solitary local CaM (fig. S5A). Instead, it is likely that the local Ca2+ influx of one channel can activate multiple CaMs, either to diversify the consequences of channel activity by triggering different signaling pathways (fig. S5B) (9, 10, 15) or to afford parallel amplification of a single type of pathway (fig. S5C). In the CaM translocation hypothesis, wherein local Ca2+ drives CaM to the nucleus and thereby activates CREB (15), the amplification scenario suggests that CaMs hailing from L-type channels could supply a substantial fraction of the nucleus-bound contingent.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S5

References

View Abstract