Different Relationship of N- and P/Q-Type Ca2+ Channels to Channel-Interacting Slots in Controlling Neurotransmission at Cultured Hippocampal Synapses

Synaptic transmission at CNS synapses is often mediated by joint actions of multiple Ca2+ channel subtypes, most prominently, P/Q- and N-type. We have proposed that P/Q-type Ca2+ channels saturate type-preferring slots at presynaptic terminals, which impose a ceiling on the synaptic efficacy of the channels. To test for analogous interactions for presynaptic N-type Ca2+ channels, we overexpressed their pore-forming CaV2.2 subunit in cultured mouse hippocampal neurons, recorded excitatory synaptic transmission from transfected cells, and dissected the contributions of N-, P/Q-, and R-type channels with subtype-specific blockers. Overexpression of CaV2.2 did not increase the absolute size of the EPSC even though somatic N-type current was augmented by severalfold. Thus, the strength of neurotransmission is saturated with regard to levels of Ca2+ channel expression for both N-type and P/Q-type channels. Overexpression of Ca2+-impermeable CaV2.2 subunits decreased EPSC size, corroborating competition for channel slots. Striking asymmetries between N- and P/Q-type channels emerged when their relative contributions were compared with channel overexpression. Overexpressed N-type channels could competitively displace P/Q-type channels from P/Q-preferring slots and take over the role of supporting transmission. The converse was not found with overexpression of P/Q-type channels, regardless of their C-terminal domain. We interpret these findings in terms of two different kinds of presynaptic slots at excitatory synapses, one accepting N-type channels but rejecting P/Q-type (N specific) and the other preferring P/Q-type but also accepting N-type (PQ preferring). The interaction between channels and slots governs the respective contributions of multiple channel types to neurotransmission and, in turn, the ability of transmission to respond to various stimulus patterns and neuromodulators.


Introduction
Presynaptic voltage-gated Ca 2ϩ channels (VGCCs) are activated on action potential (AP) invasion to allow rapid influx of extracellular Ca 2ϩ , which, in turn, triggers synaptic vesicle fusion and neurotransmitter release (Katz, 1969). Proximity of Ca 2ϩ channels to the vesicle release machinery is critical for rapid and efficient synaptic transmission (Llinás et al., 1976;Stanley, 1993), suggesting an intimate relationship between the presynaptic Ca 2ϩ channels and the release apparatus. Studies of hippocampal and cerebellar synapses indicate that the vast majority of single release sites are in close proximity to a mixed population of Ca 2ϩ channels that jointly contribute to the local Ca 2ϩ transient that triggers vesicular fusion (Mintz et al., 1995;Reuter, 1995;Reid et al., 1997). Moreover, it is well established that synaptic transmission at most CNS synapses is mediated by joint actions of multiple Ca V 2 subfamily of VGCCs, most prominently, P/Q-and N-type (Luebke et al., 1993;Takahashi and Momiyama, 1993;Wheeler et al., 1994;Sheng et al., 1998;Wu et al., 1999). This raises an important question as to whether P/Q-and N-type Ca 2ϩ channels exhibit subtype specificity in their relationship to the release machinery.
Morphological studies have provided significant insights about the possible existence of presynaptic slots for VGCCs at transmitter release sites (Heuser et al., 1979;Robitaille et al., 1990;Haydon et al., 1994;Harlow et al., 2001). Biochemical studies have identified numerous molecular interactions between VGCCs and their binding partners that play a crucial role in coupling presynaptic VGCC to the release machinery (Sheng et al., 1998;Maximov et al., 1999;Missler et al., 2003;Spafford et al., 2003;Nishimune et al., 2004;Leenders et al., 2008). Although in most cases the putative molecular interaction does not discriminate between N-and P/Q-type channels, some degree of subtype specificity has been reported. For example, the II-III loops of N-and P/Q-type channels vary in their ability to interact with soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) proteins, elements of the synaptic release machinery, and to support synaptic targeting of VGCCs (Mochida et al., 2003). Moreover, the postsynaptic density-95/Discs large/zona occludens-1 (PDZ) domain-binding motif and the proline-rich region at the cytosolic C termini of N-type VGCCs interact with CASK-Mint1-Veli complex, the modular adaptor proteins at the presynaptic active zone (Butz et al., 1998;Maximov et al., 1999). This interaction is both necessary and sufficient for synaptic targeting of N-type channels in cultured hippocampal neurons (Maximov and Bezprozvanny, 2002). On the contrary, the corresponding C-terminal region of P/Q-type Ca 2ϩ channel is not an essential determinant for its subcellular localization, again suggesting the possibility of subtype specificity of VGCC synaptic targeting (Hu et al., 2005).
In a previous study in cultured hippocampal neurons, we demonstrated that endogenous P/Q-type channels fill up their type-preferring "slots" (Cao et al., 2004). The term "slots" refers to binding sites of limited capacity that interact with channel proteins to control their participation in neurotransmission. The channel-slot interaction ultimately plays out at presynaptic terminals, but the term leaves open whether it begins with competition for auxiliary proteins that allow exit from the endoplasmic reticulum (ER) (Saheki and Bargmann, 2009), for proteins involved in axonal sorting and trafficking (Mochida et al., 2003;Ziv and Garner, 2004), or for active zone components at the presynaptic terminal itself (Maximov and Bezprozvanny, 2002). Here, we proceed to explore the question of whether type-specific slots for presynaptic N-type Ca 2ϩ channels coexist along with those that harbor P/Q-type channels. Is neurotransmission operating at a biological ceiling with respect to both channel types? Our results indicate that there exist two different kinds of functional slots for presynaptic N-and P/Q-type Ca 2ϩ channels: one accepting N-type but not P/Q-type (N specific ) and the other accepting N-type channels but preferring P/Q-type (PQ preferring ). Both kinds of slots are filled to capacity under basal conditions. We propose that the channel-slot interactions may play important roles in determining the relative importance of the two kinds of Ca 2ϩ channel and, in turn, the responsiveness of hippocampal synaptic transmission to modulation by firing pattern and G-protein-coupled receptors (GPCRs).
Portions of this manuscript have been published previously in abstract form (Cao and Tsien, 2006).
Cell culture and transfection. Characterization of the nonpermeable Ca V 2.2 subunit was performed in a stable HEK293 cell line expressing ␤ 1C and ␣ 2 ␦ subunits as described previously (Piedras-Renteria et al., 2001). Cells were cotransfected with constructs encoding wild-type or nonpermeable Ca V 2.2 subunit along with the pEGFPC1 plasmid with Lipofectamine 2000. Transfected cells were identified by enhanced green fluorescent protein (EGFP) fluorescence and were recorded 2ϳ3 d after transfection.
Mouse hippocampal neurons were cultured from 1-d-old pups at low density (ϳ5000 cells per coverslip) as described previously (Cao et al., 2004). All animal handling was in accordance with guidelines of the Animal Care and Use Committee. Individual human Ca V 2 construct was cotransfected with pEGFPC1 plasmid (1:1 molar ratio) into 5 or 10 d in vitro (DIV) culture using calcium phosphate method (Cao et al., 2004).
Electrophysiology. Transfected neurons were identified by EGFP fluorescence with Nikon TE-2000 microscope. Whole-cell patch-clamp recordings were performed at room temperature with Axopatch 200A or MultiClamp 700B amplifiers (Molecular Devices). pClamp 10 (Molecular Devices) was used to acquire and analyze data. Signals were filtered at 1 kHz and digitized at 10 kHz except for the current-clamp experiments (see below). Recording pipettes had Ͻ3.5 M⍀ resistance. Cell capacitance and series resistance (Ͻ20 M⍀) were constantly monitored throughout recording.
To measure the current-voltage relationships (I-V curves), transfected HEK293 cells were held at Ϫ80 mV and were depolarized from Ϫ60 to ϩ100 mV (10 mV increments) for 35 ms and then repolarized back to Ϫ80 mV. I-V curves in neurons were generated by a ramp depolarization from Ϫ80 to ϩ100 mV at 1.8 mV/ms. Excitatory synaptic transmission. Neurons were recorded between 12 and 18 DIV to study neurotransmission. The recording chamber was perfused with Tyrode's solution (0.5 ml/min) containing the following (in mM): 125 NaCl, 2 KCl, 4 CaCl 2 , 4 MgCl 2 , 25 HEPES, 30 glucose, pH 7.3 with NaOH, 310 mOsm. The pipette solution contained the following (in mM): 100 KCl, 2 EGTA, 4 ATP-Mg, 0.3 GTP-Na, 30 HEPES, 10 Tris-phosphocreatine, 20 U/ml creatine phosphokinase, pH 7.3 with KOH, 290 mOsm. A nonfluorescent neuron in close proximity to the transfected, EGFP-positive neuron was chosen as its possible postsynaptic target. Both neurons were voltage clamped at Ϫ70 mV. The transfected, presynaptic neurons were depolarized to 0 mV for 1 ms to elicit APs once every 10 s. Correspondingly, an EPSC could be recorded at the untransfected, postsynaptic neuron. Peak EPSC amplitude under each pharmacological condition was measured as the average of 10 -20 responses. N-, P/Q-, and R-type Ca 2ϩ channels were blocked by -CTx-GVIA (2 M), -Aga-IVA (1 M), and SNX482 (0.5 M), respectively. When studying paired-pulse modulation, presynaptic neurons were given paired depolarization at 50 ms intervals.
Drug application. Nifedipine (Sigma-Aldrich) was dissolved in 100% ethanol to generate 10 mM stock solution. All peptide Ca 2ϩ channel blockers were purchased from Peptides International, reconstituted in 1 mg/ml cytochrome c (Sigma-Aldrich) at 100ϫ concentration, and stored at Ϫ80°C in aliquots. Before the application of Ca 2ϩ channel blockers, the perfusion to the recording chamber was stopped and cytochrome c was added to the bath to 0.1 mg/ml final concentration. Subsequently, the blockers were added to the bath sequentially and cumulatively as described in Results. In some experiments, NBQX (10 M) and D-AP5 (25 M; both from Tocris) were used to block AMPA and NMDA receptors.
Statistical analysis. All data are reported as mean Ϯ SEM or cumulative percentage histogram. Statistical significance between multiple experimental groups was assessed by ANOVA with post hoc Bonferroni's test. The nonparametric Kolmogorov-Smirnov test was used to compare EPSC amplitudes between two experimental groups.

Results
Long and short isoforms of Ca V 2.1 both occupy type-specific presynaptic slots Both Ca V 2.1 and Ca V 2.2, the pore-forming subunits of P/Q-and N-type Ca 2ϩ channels, undergo extensive alternative splicing at their cytosolic C termini (Soong et al., 2002). Our previous study focused on the "short" isoform of human Ca V 2.1 subunits (Fig.  1A, PQ WT ) and demonstrated that their expression in hippocampal neurons from Ca V 2.1 knock-out mice fully restored the contribution of P/Q-type Ca 2ϩ channels to neurotransmission (Cao et al., 2004). This left open questions about the "long" isoform of Ca V 2.1. Biochemical studies suggest that, unlike the short isoforms, the long isoforms of Ca V 2.1 and Ca V 2.2 subunits ( Fig. 1 A, PQ WT-long and N WT ) use their PDZ domain-binding motifs and the proline-rich regions to interact with CASK-Mint1-Veli complex (Maximov et al., 1999). Because this interaction may be both necessary and sufficient for synaptic targeting of N-type channels (Maximov and Bezprozvanny, 2002), we wondered whether synaptic targeting of P/Q-type channels would also differ according to the nature of their C termini. Accordingly, we directly compared the synaptic effects of overexpressing short and long isoforms of Ca V 2.1 (PQ WT and PQ WT-long ) in cultured hippocampal neurons.
To study synaptic transmission, we used paired patch-clamp recordings from an EGFP-fluorescent, transfected neuron and its untransfected postsynaptic neuron target. Evoking a presynaptic spike in the transfected neuron typically resulted in an EPSC in the postsynaptic neuron ( Fig. 1 B, no blocker). Subtype-specific VGCC blockers -CTx-GVIA (2 M), -Aga-IVA (1 M), and SNX482 (0.5 M) were sequentially and cumulatively introduced into the recording chamber to block N-, P/Q-, and R-type Ca 2ϩ channels, respectively. The addition of each toxin produced a stepwise reduction in EPSC size that was readily quantified ( Fig.  1 B). The EPSC was completely blocked with all three toxins present, as confirmed by the absence of any additional effect of AMPA receptor blocker NBQX (data not shown). This allowed us to calculate the relative contributions of N-, P/Q-, and R-type Ca 2ϩ channels to synaptic transmission. Under this experimental paradigm, we found that the contributions of endogenous N-and P/Q-type channels were each reflected by a ϳ45% reduction of the transmission at synapses expressing EGFP alone (Fig. 1C, EGFP). The contributions of N-and P/Q-type channels to transmission in the PQ WT group were similar to that of EGFP group. Thus, an excess of the short isoform of Ca V 2.1 did not alter the relative contributions of N-and P/Q-type channels to neurotransmission, confirming our previous results (Cao et al., 2004). We went on to express the long isoform, testing the notion that the C-terminal PDZ domain-binding motif of PQ WT-long might confer the ability to displace N-type channels. In fact, overexpressing PQ WT-long did little to alter the pharmacology of transmission (Fig. 1C). The relative contribution of P/Q-type channels was not significantly greater than that in the EGFP or PQ WT groups, nor was the contribution of N-type channels diminished. To verify that PQ WT-long was fully capable of expression and support of neurotransmission at nerve terminals, we transfected cultured hippocampal neurons from Ca V 2.1 knock-out mice with constructs encoding EGFP and PQ WT-long (Fig. 1 D). PQ WT-long channels were able to reconstitute pharmacological responsive-ness to -Aga-IVA, just as previously shown for PQ WT (Cao et al., 2004) (data repeated for ease of comparison in Fig. 1 D). After previous application of -CTx-GVIA, exposure to the P/Q-type channel blocker caused a 50 Ϯ 14% drop in EPSC size in the PQ WT-long -transfected Ca V 2.1 knock-out neurons ( Fig. 1 D), no different ( p Ͼ 0.6) than the 41 Ϯ 4% decrease seen in wild-type mouse neurons with endogenous P/Q-type channels (Fig. 1C, EGFP group). The cumulative distribution of overall EPSC size remained unaltered, regardless of whether long or short forms of Ca V 2.1 were overexpressed ( Fig. 1 E). Given that PQ WT-long , like PQ WT , is fully capable of expression at nerve terminals, it is striking that expression of either long or short forms did not increase synaptic strength.
Together, these observations suggest that both isoforms of P/Q-type channels occupy limited number of presynaptic slots with similar overall efficacy. The C-terminal PDZ domainbinding motif does not alter P/Q-type channel-slot interaction as defined in our previous study. These results are in line with a previous report (Hu et al., 2005) indicating that, unlike N-type channels, P/Q-type channels do not use their C-terminal PDZ domain-binding motif as an essential determinant for synaptic localization.

Overexpression of N-type Ca 2؉ channel significantly increases its contribution to excitatory transmission
The lack of significant reduction in the contribution of N-type channels (Fig. 1C, open bars) to synaptic transmission in PQ WT and PQ WT-long groups implies that endogenous N-type channels likely occupy additional slots, different from those that P/Q-type channels are able to access. This raises several additional questions: for example, if additional slots exist, are they specific to N-type channels, and are they filled to capacity? When N-type channels are overexpressed, can they displace P/Q-type channels from their slots? We approached these questions by expressing human Ca V 2.2 ( Fig. 1 A, N WT ), the pore-forming subunit of N-type Ca 2ϩ channel, in cultured hippocampal neurons. The isoform of N WT we used contains the PDZ domain-binding motif and the proline-rich region at C terminus and can travel to synaptic terminals to mediate neurotransmitter release (Maximov and Bezprozvanny, 2002). We focused on excitatory transmission from neurons overexpressing N WT and dissected contributions of N-, P/Q-, and R-type channels, using the same system of paired patch-clamp recordings. Whereas neurons expressing EGFP alone, PQ WT , or PQ WT-long underwent a ϳ42% reduction of EPSCs with exposure to -CTx-GVIA, transmission from neurons overexpressing N WT behaved a strikingly different way-in this case, N-type channels accounted for ϳ80% of the EPSC amplitude, significantly higher than in any of the other groups ( p Ͻ 0.05). Evidently, the contribution of endogenous N-type channels was not at a biological ceiling, in sharp contrast to P/Q-type channels.
In the preceding experiments, N WT subunits were overexpressed by transfection of low-density hippocampal cultures at 5 DIV, with the goal of achieving high transfection efficiency. Because 5 DIV is about the time of increased synaptogenesis and expression of P/Q-type channels, it seemed plausible that such early overexpression of N WT might affect the development or maturation of synapses, to the detriment of proper assembly of slots and of P/Q-type channel participation. Accordingly, we performed additional experiments wherein overexpression of N WT was postponed to 10 DIV, thus allowing synaptogenesis to be well advanced and endogenous P/Q-type channel expression to approach a plateau before the introduction of excess N WT subunits.
Despite this delay, overexpression of N WT still resulted in an enhanced contribution of N-type channels to excitatory transmission, seen as increase in -CTx-GVIA block from 51 Ϯ 5% (Fig. 1C, EGFP) to 80 Ϯ 8% (Fig. 1C, N WT-10 DIV group, p Ͻ 0.05), very similar to the effect of expressing N WT at 5 DIV. On the other hand, the contribution of P/Q-type Ca 2ϩ channels to transmission decreased from 41 Ϯ 4 to 16 Ϯ 7% when N WT was overexpressed ( Fig. 1C, filled bars, p Ͻ 0.05).
We performed additional tests to determine whether these results can be attributed to a genuine difference between N-and P/Q-type channels in their disposition with respect to the presynaptic release machinery. One possibility to consider is that the excess of N WT channels drives the creation of ectopic release sites, thus expanding the impact of the N-type contribution to transmission by augmenting neurotransmission as a whole. However, we observed that overall EPSC size did not differ between synapses expressing EGFP or N WT subunits ( Fig. 1 E), indicating that N WT overexpression did not enlarge the overall strength of transmission. Evidently, the number of release sites and their individual probability of release (P r ) both remained unaltered. As a separate gauge of P r , we looked for changes in paired-pulse modulation, which veers toward paired-pulse depression rather than facilitation as basal P r rises. The ratio of the EPSC amplitudes elicited by pairs of stimuli 50 ms apart was determined. This paired-pulse ratio (PPR) proved to be no different among the various groups, EGFP (1.16 Ϯ 0.04), PQ WT (1.20 Ϯ 0.08), PQ WT-long (1.14 Ϯ 0.12), and N WT (1.18 Ϯ 0.06) ( p Ͼ 0.6 by ANOVA). The lack of change in pairedpulse modulation supported the idea that P r had remained unaffected by the excess of one or other type of Ca 2ϩ channel.
These data on the fractional contributions of various Ca V 2 channels pointed to a marked asymmetry: overexpressed N-type channels were able to gain dominance over P/Q-type channels but not vice versa. Given that the overall strength of transmission did not increase ( Fig. 1 E), a Figure 1. N-and P/Q-type Ca 2ϩ channels exhibit different relationship to slots in controlling neurotransmission at cultured hippocampal synapses. A, Schematic representations of the C-terminal cytoplasmic domains of human Ca V 2.2 subunit (N WT ) as well as the short and long splice variants of human Ca V 2.1 subunits (PQ WT and PQ WT-long , respectively). The last four amino acids of the C terminus of each subunit are indicated. The lengths of primary sequence are drawn to scale. B, Averaged EPSC traces (5ϳ7 traces at the end of each condition) elicited by neurons expressing EGFP alone (EGFP), or EGFP with human Ca V 2 subunits (PQ WT , PQ WT-long , or N WT , respectively), as well as the effect of toxins on EPSC amplitude. The N-, P/Q-, or R-type Ca 2ϩ channel blockers -CTx-GVIA (2 M), -Aga-IVA (1 M), and SNX482 (0.5 M) were applied sequentially and cumulatively into the bath solution as indicated. C, Relative contributions of N-, P/Q-, and R-type channels to EPSC amplitude in neurons expressing EGFP alone (EGFP, n ϭ 20), or EGFP with human Ca V 2 subunits (PQ WT , PQ WT-long , N WT , and N WT-10 DIV ; n ϭ 16, 7, 23, and 5, respectively). Neurons were transfected at 5 DIV except for the N WT-10 DIV group, which was composed of neurons transfected at 10 DIV. Ca 2ϩ channels were blocked as shown in B. Neither the expression of PQ WT or PQ WT-long altered the relative contributions of N-, P/Q-, and R-type channels to excitatory transmission. However, overexpression of N WT (either starting from 5 or 10 DIV) significantly increased the N-type contribution to synaptic transmission (*p Ͻ 0.05, compared with EGFP group). Correspondingly, the contribution of PQ to transmission was significantly decreased in N WT and N WT-10 DIV groups ( # p Ͻ 0.05, compared with EGFP group). D, Synaptic functionality of PQ WT-long channels, tested in hippocampal neurons from Ca V 2.1-knock-out mice. Expression of exogenous PQ WT-long subunitswascapableofrestoringthe-Aga-IVA-sensitivecomponentofsynapticcurrent(n ϭ 3). Data for Ca V 2.1-knock-out (ko) and PQ WT -transfected neurons from Cao et al. (2004) are included to allow direct comparison. Error bars indicate SEM. E, Overall EPSC amplitude distribution was comparable across all groups (n ϭ 12-70). F, Cumulative distribution of -CTx-GVIAsensitive EPSC amplitude of EGFP, N WT , and PQ WT groups of neurons as in C. The N WT group showed a significantly larger component 4 of N-type-dependent EPSC ( p Ͻ 0.05 vs EGFP group). No significant difference was found between PQ WT and EGFP groups ( p Ͼ 0.4). G, Cumulative distribution of -Aga-IVA-sensitive EPSC amplitude of EGFP, N WT , and PQ WT groups of neurons as in C. The P/Q-component of the EPSC was significantly smaller in the N WT group ( p Ͻ 0.05 vs EGFP group; p Ͻ 0.01 vs PQ WT group). No significant difference was found between PQ WT and EGFP groups ( p Ͼ 0.3). possible scenario is that N-type channels literally replaced P/Q-type channels at synaptic release sites. This led to quantitative predictions about the absolute sizes of N-and P/Q-mediated components of the EPSC. Net increases in the magnitude of the N-type contribution would be expected to take place at the expense of the contribution of P/Q-type channels. Indeed, the absolute size of the N-type channel-mediated EPSC was significantly bigger in the N WT group than that in EGFP ( p Ͻ 0.05) (Fig. 1 F), whereas the amplitude of P/Q-supported EPSCs was correspondingly smaller ( p Ͻ 0.05) (Fig.  1G). These data thereby stand in contrast to overexpression of either PQ WT or PQ WT-long . Neither construct caused a significant change in the amplitude distributions for the total EPSC ( Fig. 1 E, p Ͼ 0.3, compared with EGFP group), just as found with N-type overexpression. But in these cases, the size of N-or P/Q-type channel-mediated components did not differ between PQ WT and EGFP groups ( Fig. 1 F, The same could be said about the N-or P/Q-type components in the PQ WT-long and EGFP groups ( p Ͼ 0.2 and p Ͼ 0.5, respectively), although the sample size in the PQ WT-long group was considerably smaller (n ϭ 7) (data not shown).
Given the steeply nonlinear relationship between Ca 2ϩ influx and presynaptic P r (Rahamimoff and Dodge, 1969), the effect of toxins on synaptic transmission depends on their order of application (Takahashi and Momiyama, 1993;Wheeler et al., 1994). With this in mind, we determined whether the asymmetry between the effects of PQ WT and N WT held true regardless of which channel type was blocked first. Figure 2 A illustrates the outcome of reversing the order of toxin presentation, applying -Aga-IVA before -CTx-GVIA. In this case, -CTx-GVIA blocked 17 Ϯ 3% of the EPSC at wild-type synapses, but 37 Ϯ 6% of the EPSC at synapses expressing N WT , a significant increase ( p Ͻ 0.05) (Fig. 2 A). Conversely, if one performs a direct comparison of the effect of -Aga-IVA in EGFP and N WT groups (temporarily ignoring the PQ WT data) (Fig. 2 A, legend), the fractional P/Q-mediated transmission was 76 Ϯ 4% in the EGFP cells and 60 Ϯ 6% in the N WT group, significantly less ( p Ͻ 0.03). These data provided additional evidence that overexpression of N-type Ca 2ϩ channel significantly increased their contribution to excitatory transmission while concomitantly reducing the participation of P/Q-type. Further tests of the effects of overexpressing N WT or PQ WT on excitatory neurotransmission and Ca 2ϩ channel currents. A, Pharmacological tests with -Aga-IVA applied first. Relative contribution of P/Q-, N-, and R-type channels to EPSC amplitude in neurons expressing EGFP alone (EGFP, n ϭ 19) or EGFP with human Ca V 2 subunits (PQ WT , n ϭ 8; N WT , n ϭ 13). The P/Q-, N-, or R-type Ca 2ϩ channel blockers -Aga-IVA (1 M), -CTx-GVIA (2 M), and SNX482 (0.5 M) were introduced sequentially and cumulatively into the bath solution. Overexpression of PQ WT did not alter the relative contributions of N-, P/Q-, and R-type channels to excitatory transmission. Overexpression of N WT significantly increased the N-type contribution to synaptic transmission (*p Ͻ 0.05, compared with EGFP group). Conversely, a direct comparison of PQ-mediated transmission in EGFP and N WT groups showed that it was significantly decreased by overexpression of N WT ( p Ͻ 0.03). We noted that strict adherence to statistical procedures would preclude such a comparison because one-way ANOVA of the P/Q-mediated transmission in EGFP, PQ WT , and N WT groups failed to show significance. B1, Peak whole-cell Ba 2ϩ current density through P/Q-type Ca 2ϩ channels at ϩ10 mV. P/Q-type component was defined as Ba 2ϩ currents sensitive to the blockade of 1 M -Aga-IVA. Hippocampal neurons were transfected with EGFP alone (n ϭ 19), or EGFP with PQ WT , PQ WT-long , or N WT (n ϭ 6, 13, and 10, respectively). Overexpressing PQ WT or PQ WT-long caused an increase of whole-cell P/Q current density (***p Ͻ 0.001) relative to EGFP group. The level of P/Q current density was similar in PQ WT and PQ WT-long groups. In contrast, overexpressing N WT did not alter P/Q-type current density. B2, Reliability of EGFP expression as a marker of expression of exogenous P/Q-type Ca 2ϩ channels. Cumulative distributions of amplitudes of somatodendritic P/Q-type current density. We tested the null hypothesis that P/Q-type current was no different in neurons transfected with EGFPϩPQ WT-long or EGFPϩPQ WT than that in neurons transfected with EGFP alone. This hypothesis can be rejected even for the smallest observed currents in the PQ WT-long ( p Ͻ 0.01) or PQ WT groups ( p Ͻ 0.001). C1, Peak whole-cell Ba 2ϩ current density through N-type Ca 2ϩ channels at ϩ10 mV, from the same neurons as in B. N-type component was defined as Ba 2ϩ current sensitive to the blockade of 2 M -CTx-GVIA. Overexpression of PQ WT or PQ WT-long did not alter whole-cell N-type current density. On the contrary, overexpression of N WT resulted in a significant increase of N-type current density compared with that of EGFP group (***p Ͻ 0.001). Error bars indicate SEM. C2, Reliability of EGFP expression as a marker of expression of exogenous N-type Ca 2ϩ channels. Cumulative distributions of amplitudes of somatodendritic N-type current density. We tested the null hypothesis that N-type current density was no different in neurons transfected with EGFP and N WT than that in neurons transfected with EGFP alone. This hypothesis can be rejected even for the smallest observed currents in the EGFPϩN WT group ( p Ͻ 0.01).
It is conceivable that the seesaw relationship between contributions of N-and P/Q-type channels could arise because of a competition for a limited supply of auxiliary subunits. To test for this, we went on to examine the effect of overexpressing PQ WT , PQ WT-long , or N WT subunits on whole-cell Ca 2ϩ channel activity ( Fig. 2 B, C). P/Q-and N-type components were defined as Ba 2ϩ currents sensitive to their specific blockers -Aga-IVA (1 M) and -CTx-GVIA (2 M), respectively. In PQ WT or PQ WT-long groups, whole-cell P/Q current densities were approximately fourfold greater than that in the EGFP group (Fig. 2 B1, p Ͻ 0.001), whereas the level of P/Q-current density in PQ WT and PQ WT-long groups was similar (Fig. 2 B1). In contrast, overexpressing PQ WT or PQ WT-long in neurons did not alter the somatodendritic N-type Ca 2ϩ channel activity, as indicated by similar -CTx-GVIA-sensitive current densities in EGFP, PQ WT , and PQ WT-long groups (Fig. 2C1). Neither L-and R-type current density was significantly altered (data not shown). These results indicated that hippocampal neurons were able to synthesize, assemble, and insert both isoforms of P/Q-type channels into somatodendritic membranes at approximately four times their normal level without compromising the activity of other high VGCCs. The same set of data provided reassurance that transfected neurons (as indicated by the EGFP fluorescence) in the PQ WT and PQ WT-long groups consistently expressed exogenous PQ-type channels. In both cases, the smallest P/Q-type current density was no less than the largest P/Q-type current density in neurons expressing EGFP alone (Fig. 2 B2), suggesting that all neurons in the PQ WT and PQ WT-long groups reliably expressed exogenous P/Q-type channels.
In the converse experiment, we found that overexpression of N WT resulted in an approximately threefold increase of -CTx-GVIA-sensitive current density compared with that of EGFP group (Fig.  2C1, p Ͻ 0.001). Once again, the distributions of current were strikingly different; the smallest N-type current density in the N WT group exceeded the largest N-type current density in the EGFP group (Fig. 2C2). Importantly, the magnitude of -Aga-IVA-sensitive current remained unaltered (Fig. 2B1). Based on these comparisons, we conclude that the interplay between N-and P/Q-type channels is not merely a matter of competition for a limited supply of ancillary Ca 2ϩ channel subunits. Instead, N-type channels appear to compete with P/Q-type channels in some another way.

Using nonpermeable Ca V 2.2 subunits to probe for N-type slots
Our experiments uncovered a striking asymmetry between Nand P/Q-type channels. Although even massive overexpression of P/Q-type channels was unable to drive down the synaptic contri- Figure 3. Characterization of nonpermeable human Ca V 2.2 subunits (N imperm ). A, Topology of human Ca V 2.2 subunit. The locations of four concerted glutamate-to-alanine mutations in the pore-forming region of N imperm are indicated. B, Representative AP traces from hippocampal neurons expressing EGFP alone or EGFP with N imperm . Somatic APs were evoked by a depolarizing current injection (0.5 ms, 1 nA). Expression of N imperm did not change AP waveform (N imperm vs EGFP group). Furthermore, blocking both endogenous N-type Ca 2ϩ channels and exogenous N imperm channels with 2 M -CTx-GVIA did not significantly alter the AP shape (N imperm ϩGVIA vs other groups). See Table 1 for comparisons of AP properties. C, The current-voltage relationships (I-V curves) of N WT and N imperm in HEK293 cells (n ϭ 5 in each group). The N imperm group did not show any inward Ba 2ϩ current but displayed an outward current when depolarized beyond 0 mV. D, The I-V curves of the N-type (-CTx-GVIA-sensitive) Ba 2ϩ current in hippocampal neurons expressing EGFP alone (EGFP, n ϭ 6), EGFP with N imperm (N imperm , n ϭ 5), or EGFP with both N imperm and PQ WT subunits (N imperm ϩPQ WT , n ϭ 8). Transfected neurons were voltage clamped at Ϫ80 mV. I-V curves were generated by applying a ramp depolarization from Ϫ80 to ϩ100 mV at 1.8 mV/ms. The -CTx-GVIA-sensitive current densities were obtained by subtraction of I-V relationships in the absence and presence of toxin. Inward currents were consistently seen over the voltage range between Ϫ20 and ϩ20 mV. At more depolarizing voltages, I-V curves from N imperm -expressing neurons exhibited a negative shift of reversal potential and an increase of outward current compared with the I-V curve of EGFP-expressing neurons, as expected for a combination of endogenous N-type channels and exogenous N imperm channels operating in parallel. E, Peak whole-cell Ba 2ϩ current density through P/Q-type Ca 2ϩ channels at ϩ10 mV from hippocampal neurons expressing EGFP alone (EGFP, n ϭ 12), EGFP with N imperm (N imperm , n ϭ 8), or EGFP with both N imperm and PQ WT Ca V 2 subunits (N imperm ϩPQ WT , n ϭ 7). The expression of N imperm channels did not alter the current density from endogenous P/Q-type channels (N imperm vs EGFP groups), nor did it affect the increase of P/Q-current density as a result of overexpression of PQ WT (***p Ͻ 0.001, N imperm ϩPQ WT vs other groups). F, Peak whole-cell Ba 2ϩ current density through N-type Ca 2ϩ channels at Ϫ10 mV from the same neurons in D. At this voltage, the N imperm channels did not exhibit inward or outward current. The expression of N imperm (and further addition of PQ WT ) did not alter the current density from the endogenous N-type Ca 2ϩ channels. Error bars indicate SEM. bution of N-type channels, the converse was not the case: N-type channels were capable of displacing P/Q-type. The finding of an irreducible contribution of N-type channels suggested the existence of slots specific for N-type, impervious to competition from P/Q-type.
To probe further for the possibility of N-type slots, we turned to the strategy of overexpressing Ca 2ϩ -impermeable N-type channels (N imperm ) and testing for dominant-negative effects. Human Ca V 2.2 subunits incapable of Ca 2ϩ -permeation were generated by replacing all four pore glutamates with alanine residues (Fig. 3A), and their effects were tested in recordings of neuronal spikes (Fig. 3B) and synaptic transmission (see below). As a preliminary step, electrophysiological comparisons of N WT and N imperm channels were performed in HEK293 cells (Fig. 3C). Unlike cells expressing N WT , HEK293 cells expressing N imperm failed to display any inward Ba 2ϩ current but only showed an outward current when depolarized beyond 0 mV. The outward current through N imperm channels was blocked by 2 M -CTx-GVIA (data not shown), indicating that N imperm and N WT exhibited similar responsiveness to the toxin blockade.
APs in both cell bodies and axons of excitatory neurons in hippocampus are dominated by large currents generated by Na ϩ and K ϩ channels (Schmidt-Hieber et al., 2008). Activation of VGCCs lags behind that of Na ϩ channels (Bean, 2007), leading to opening of only a small fraction of Ca 2ϩ channels for the narrow APs of hippocampal neurons (King and Meriney, 2005). To be sure that the outward current through N imperm channels would not exert confounding effects on AP generation, we expressed N imperm subunits along with EGFP or EGFP alone in hippocampal culture and measured somatic APs evoked by a 0.5 ms, 1 nA depolarizing current injection in transfected neurons. Expression of N imperm did not change AP properties (Fig. 3B); neither amplitude, half-width, or AP waveform were altered to any significant degree (Table 1). That key properties of spiking in the presynaptic neuron remained essentially unchanged was as expected if Na ϩ and K ϩ currents dwarfed currents through Ca 2ϩ channels (either inward or outward). This is a worst-case analysis of what would occur in axons, in which Na ϩ channels are denser than in cell bodies and would be even more dominant over Ca 2ϩ channels. In an additional test, we exposed neurons expressing N imperm to -CTx-GVIA (2 M), which, as already mentioned, acutely blocks outward as well as inward currents through N-type Ca 2ϩ channels (McCleskey et al., 1987). The N-type specific blocker did not significantly alter AP properties (Fig. 3B, Table 1). This result further confirmed that N-type currents exerted little or no influence on AP shape in cultured hippocampal neurons.
As additional control experiments, we turned back to hippocampal neurons to find out whether N imperm interferes with the whole-cell Ba 2ϩ current through the endogenous N-and P/Qtype Ca 2ϩ channels. N-type current was isolated as the -CTx-GVIA-sensitive component of membrane currents evoked by a ramp depolarization from Ϫ80 to ϩ100 mV. Hippocampal neurons expressing EGFP alone displayed the I-V relationship expected for endogenous N-type channels, first activating near Ϫ20 mV, a milder depolarization than required in HEK293 cells, possibly because of differences in surface charge. In neurons expressing Ca V 2.2 subunits rendered incapable of Ca 2ϩ permeation, the -CTx-GVIA-sensitive component consisted of clear inward currents at weaker depolarization levels, giving way to pronounced outward currents beyond ϩ20 mV. This was just as expected for the summation of mostly inward currents carried by endogenous N-type channels (Fig. 3D, green symbols) and solely outward currents supported by N imperm channels (Fig. 3C, maroon symbols). Additional expression of PQ WT channels (N imperm ϩ PQ WT ) left the composite N-type I-V curve unchanged, suggesting that all three channel types operate independently and in parallel.
Further analysis of P/Q-type and N-type currents is documented in Figure 3, E and F. The -Aga-IVA-sensitive P/Q-type current density, gauged at a test potential of ϩ10 mV, was similar in N imperm -expressing neurons to that of the control group, confirming that the number of somatodendritic P/Q-type channels remained unaltered (Fig. 3E). In contrast, P/Q-type current was increased approximately fivefold by additional expression of PQ WT . We also assessed the magnitudes of -CTx-GVIAsensitive current through endogenous N-type channels (Fig. 3F ), using a test potential of Ϫ10 mV to avoid the outward current through N imperm channels (Fig. 3D). Neither overexpression of N imperm nor that of N imperm ϩPQ WT in combination caused a significant alteration in currents through endogenous N-type Ca 2ϩ channels.

Effects of N imperm channels on synaptic transmission
Having confirmed that endogenous N-and P/Q-type channels display normal activity in neurons expressing N imperm , we proceeded to test how N imperm affects the relative contribution of Nand P/Q-type channels to synaptic transmission as well as overall synaptic strength (Fig. 4). The expression of N imperm channels significantly reduced the size of EPSCs (Fig. 4 A, C,D, p Ͻ 0.05, N imperm vs EGFP group), indicating that the nonpermeable N-type channels were properly expressed in the presynaptic neuron and somehow competed with endogenous channels. We again measured PPR as an indication of P r changes. The ratio of the EPSC amplitudes elicited by pairs of stimuli 50 ms apart was significantly increased for presynaptic partners expressing N imperm (1.40 Ϯ 0.07) compared with neurons expressing EGFP alone (1.16 Ϯ 0.04; p Ͻ 0.05) (Fig. 4 E), suggesting that impermeant N-type channels reduce P r at presynaptic terminals. However, the clear reduction in synaptic strength was not accompanied by a shift toward P/Q-dominated transmission (Fig. 4 B, N imperm vs EGFP group). The implication is that N imperm channels displaced both endogenous N-and P/Q-type channels from slots that they would otherwise occupy, leaving their relative contributions unchanged while diminishing transmission overall.
To substantiate this interpretation, we coexpressed PQ WT along with N imperm , on the idea that this should ensure that P/Qpreferring slots are filled up with permeable P/Q-type channels, while leaving N imperm to compete with endogenous N-type channels. Indeed, we observed that coexpression of PQ WT and N imperm resulted in a marked reduction of the N-type contribution to synaptic transmission, from 51 Ϯ 4 to 21 Ϯ 3% (Fig. 4 B, open bars) (N imperm ϩPQ WT vs EGFP group, p Ͻ 0.05). Correspondingly, the contribution of P/Q-type Ca 2ϩ channels to transmission increased from 41 Ϯ 4 to 70 Ϯ 4% when PQ WT and N imperm were overexpressed (Fig. 4 B, filled bars) ( p Ͻ 0.05). At the same time, the absolute strength of transmission was no less than the EGFP control (Fig. 4C,D), nor did values of PPR differ be- tween the two groups (Fig. 4 E). Evidently, the excess of Ca V 2.1 subunits protected P/Q-preferring slots from being occupied by N imperm . However, N imperm replaced some endogenous N-type channels and prevented them from coupling to release machinery.

No spare N-type-specific slots
To test whether N-type-specific slots are normally filled to capacity by endogenous N-type channels, we compared normal transmission with that found when N WT and PQ WT subunits were both coexpressed in great excess of normal levels. The PQ-preferring slots would be filled up by PQ WT and/or the endogenous P/Qtype channels. If there were spare N-type specific slots, overexpression of N WT would still increase the relative contribution of N-type channels to synaptic transmission; if not, N WT would only displace endogenous N-type channels and the percentage of N-supported transmission would remain unaltered. Indeed, expression of extra PQ WT reduced the -CTx-GVIA-sensitive fraction from 81 Ϯ 8% back to 48 Ϯ 5%, not significantly different from the control (EGFP group) and in clear contrast to the findings with overexpression of N WT alone (Fig. 5A, N WT ϩ PQ WT vs N WT ). Remarkably, overall synaptic strength was not increased (Fig.  5B), despite the fact that whole-cell P/Qand N-type currents both increased severalfold (Fig. 5C,D). The values of PPR were not different either: EGFP (1.16 Ϯ 0.04), N WT (1.18 Ϯ 0.06), and N WT ϩPQ WT (1.16 Ϯ 0.07). These data indicated that, like endogenous P/Q-type channels, endogenous N-type channels fully engaged the presynaptic slots preferentially available to them. The apparent contribution of R-type channels remained relatively minor under the conditions of these experiments.

Discussion
This study defined some key features of the slots that determine which kinds of presynaptic Ca 2ϩ channels trigger excitatory neurotransmission in cultured hippocampal neurons. The functional roles of N-and P/Q-type channels were probed by overexpression of both permeable and impermeable channels. Although some of the results fulfilled predictions arising from our previous work (Cao et al., 2004), most of our findings were unexpected. In summary, (1) neurotransmission is normally saturated with regard to Ca 2ϩ channel expression and cannot be increased by strong overexpression of exogenous Nand P/Q-type channels; (2) N-and P/Qtype Ca 2ϩ channels differ markedly in their relationships to slots that impose a ceiling on their presynaptic roles; (3) P/Q-type Ca 2ϩ channels maximally fill a subset of slots (PQ preferring ) that restrict their contribution to transmission; they can be displaced from the PQ preferring slots by overexpression of N-type Ca 2ϩ channels, Figure 4. Effects of N imperm channels on synaptic transmission. A, Averaged EPSC traces (5ϳ7 traces at the end of each condition) elicited by neurons expressing EGFP alone (EGFP), or EGFP with human Ca V 2 subunits (N imperm , N imperm ϩPQ WT , and PQ WT , respectively) as well as the effect of toxins on EPSC amplitude. Neurons were exposed to the N-, P/Q-, or R-type Ca 2ϩ channel blockers -CTx-GVIA (2 M), -Aga-IVA (1 M), and SNX482 (0.5 M) sequentially and cumulatively as indicated. B,RelativecontributionsofN-,P/Q-,andR-typechannelstoEPSC amplitude in neurons expressing EGFP alone (EGFP, n ϭ 20), or EGFP with human Ca V 2 subunits (N imperm , N imperm ϩPQ WT , and PQ WT ; n ϭ 10, 10, and 18, respectively). Ca 2ϩ channels were blocked as shown in A. Overexpression of N imperm did not alter the relative contributions of N-, P/Q-, and R-type channels to excitatory transmission. However, coexpression of PQ WT and N imperm resulted in a significant reduction of N-type contribution to synaptic transmission (*p Ͻ 0.05, N imperm ϩPQ WT vs EGFP group,). Correspondingly, the contribution of P/Q-type channels to transmission was significantly increased in N imperm ϩPQ WT group ( # p Ͻ 0.05, compared with EGFP group). C, Cumulative distribution of overall EPSC amplitude at synapses expressing EGFP alone (n ϭ 70), or EGFP with human Ca V 2 subunits (N imperm , n ϭ 22; N imperm ϩPQ WT , n ϭ 18; PQ WT , n ϭ 26). Presynaptic expression of N imperm channels alone significantly reduced the size of EPSCs ( p Ͻ 0.05, N imperm vs EGFP group). In contrast, the synaptic strength was well maintained when PQ WT and N imperm were coexpressed at presynaptic terminals ( p Ͼ 0.17, compared with EGFP group). D, Ensemble average of overall EPSC amplitude from the same groups in C. E, PPR at synapses expressing EGFP alone or EGFP with various human Ca V 2 subunits (n ϭ 12-70). Pairs of EPSCs were elicited by stimuli 50 ms apart, and the ratio of the EPSC amplitudes was determined. PPR was significantly increased at synapses expressing N imperm compared with that at synapses expressing EGFP alone (*p Ͻ 0.05, one-way ANOVA post hoc Bonferroni). PPR values were not altered in other experimental groups. Error bars indicate SEM. but put back by additionally overexpressing P/Q-type channels; (4) N-type Ca 2ϩ channels normally occupy an additional set of slots (N specific ) that cannot be populated by P/Q-type channels; these slots are maximally occupied by N-type channels.
Our findings lead to significantly different predictions for how deletion of P/Q-or N-type channels would affect global synaptic strength and readjustment of the burden on the channels that remain. Deletion of P/Q-type channels should merely shift the burden of transmission to N-type channels, which can replace P/Q at PQ preferring slots. Indeed, at the calyx of Held, EPSC size is preserved relative to wild-type at Ca V 2.1 knockout synapses (Inchauspe et al., 2004(Inchauspe et al., , 2007Ishikawa et al., 2005). Conversely, deletion of N-type channels should weaken neurotransmission because the remaining P/Q-type channels would not interact with N selective slots and would thus fail to compensate fully. Indeed, in Ca V 2.2 knock-out mice, basal excitatory transmission is substantially reduced (Jeon et al., 2007). Thus, observations with channel overexpression and deletion are mutually consistent in supporting the hypothesis in Figure 5E. The central feature is the mutual interaction of (1) functionally different kinds of Ca V 2 channels and (2) multiple, operationally distinct kinds of slots. The scheme features a pronounced asymmetry. PQ preferring slots are not wholly selective, displaying greater proclivity for P/Q-type channels but also accepting N-type channels if these are particularly abundant or if P/Q-type channels are absent. On the other hand, N specific slots reject P/Q-type channels even when these are greatly overexpressed or N-type channels are eliminated.

Functional implications of multiple types of Ca 2؉ channel slots
The existence of two kinds of slot might offer functional advantages if the two types of channel differ substantially in how they are regulated or how they trigger transmitter release. Studies at large nerve terminals indicate that presynaptic P/Qand N-type channels are fairly similar in their voltage and time dependence of activation and deactivation (Li et al., 2007). Ca 2ϩ currents supported by P/Q-type channels are more apt to facilitate during a train of brief stimuli (Inchauspe et al., 2004(Inchauspe et al., , 2007Ishikawa et al., 2005). N-type channels are more susceptible to modulation by GPCRs for neurohormones (Bourinet et al., 1996;Currie and Fox, 1997;Zhou et al., 2003;Cao and Tsien, 2005). Because Ca V 2 family members differ in responsiveness to repetitive activity and neuromodulators, controlling their contribution to presynaptic release provides a powerful way to delimit the range of regulation. By putting a ceiling on the abundance of P/Q-type channels at the active zone, PQ preferring slots could cap the degree of Ca 2ϩ channel facilitation while setting a lower bound on the extent of neuromodulation, both important aspects of presynaptic regulation. Patterns of stimulation and delivery of neuromodulators vary from moment to moment, but their range of impact could be predetermined by the makeup of presynaptic slots. Figure 5. N-type-specific slots are filled to capacity by endogenous N-type channels. A, Relative contributions of N-, P/Q-, and R-type channels to EPSC amplitude in neurons expressing EGFP alone (EGFP, n ϭ 20), or EGFP with human Ca V 2 subunits (N WT , n ϭ 23; N WT ϩPQ WT , n ϭ 11). The N-, P/Q-, or R-type Ca 2ϩ channel blockers -CTx-GVIA (2 M), -Aga-IVA (1 M), and SNX482 (0.5 M) were applied sequentially and cumulatively into the bath solution. In the presence of extra PQ WT , overexpression of N WT could no longer increase the N contribution to synaptic transmission, indicating that N-and P/Q-type channels compete for a common subset of slots (*p Ͻ 0.05, N WT vs EGFP group; no significant difference between N WT ϩPQ WT and EGFP groups). B, Cumulative distribution of overall EPSC amplitude at synapses expressing EGFP alone (n ϭ 70), or EGFP with human Ca V 2 subunits (N WT , n ϭ 61; N WT ϩPQ WT , n ϭ 18). The synaptic strength was comparable across all groups ( p Ͼ 0.33, Kolmogorov-Smirnov test). C, Peak whole-cell Ba 2ϩ current density through P/Q-type Ca 2ϩ channels at ϩ10 mV from hippocampal neurons expressing EGFP alone (EGFP, nϭ19), or EGFP with human Ca V 2 subunits (N WT ,nϭ10; N WT ϩPQ WT ,nϭ7). Overexpression of PQ WT , either alone or with N WT , resulted in significant increase of whole-cell P/Q current density (***p Ͻ 0.001, relative to EGFP group). The magnitude of P/Q-current density was not significantly different between N WT ϩPQ WT (50.1 Ϯ 3.8 pA/pF) and the PQ WT groups (58.0 Ϯ 13.3 pA/pF) (Fig. 2B) ( p Ͼ 0.5). Overexpressing N WT in neurons did not alter P/Q-type current density. D, Peak whole-cell Ba 2ϩ current density through N-type Ca 2ϩ channels at ϩ10 mV from the same neurons in C. Overexpression of N WT , either alone or with PQ WT , significant increased the whole-cell N-type current density (**p Ͻ 0.01, relative to EGFP group). The magnitude of N-type current density was comparable between the N WT (32.8 Ϯ 3.3 pA/pF) and the N WT ϩPQ WT (35.1 Ϯ 4.7 pA/pF; p Ͼ 0.6) groups. Error bars indicate SEM. E, Different relationship of N-and P/Q-type Ca 2ϩ channels to slots in controlling neurotransmission at cultured hippocampal synapses. Our data indicate that there are at least two different kinds of presynaptic slots at excitatory hippocampal synapses. The N specific slots accept N-type channels selectively but reject P/Q-type Ca 2ϩ channels. The PQ preferring slots prefer P/Q-type Ca 2ϩ channels but also accept N-type when P/Q-type channels are missing or N-type are overexpressed. Note that terminology for slots does not consider R-type channels.

Implications for neuronal development and disease
The concept of multiple kinds of slots could be related to the staggered onsets of Ca 2ϩ channel expression during neuronal development, N-type channels appearing in embryonic synapses, whereas P/Q-type emerges much later in development. Although some synapses wind up relying almost completely on P/Q-type, not N-type (Iwasaki et al., 2000), many CNS synapses do not undergo a complete N-to P/Q-type switch. For example, the P/Q-type component of transmission flattens off at ϳ50% at synapses in the dorsal horn (Iwasaki et al., 2000) and at somewhat higher levels at hippocampal synapses (Scholz and Miller, 1995). It may be advantageous that the channel that expresses earlier (N-type) acts as a universal donor that can fully support transmission in the absence of the other. Conversely, the delayed but abundant expression of P/Q-type channels allows them to capture a subset of the slots. Their inability to compete for N selective slots helps curb their contribution, while also safeguarding the participation of N-type channels and the G-protein-mediated modulation of Ca 2ϩ influx. Regulating the number of slots may fine-tune neurotransmission more reliably than grading the production of the channels themselves. Ca V 2 channels support vital functions other than neurotransmission, including regulation of transcription and other forms of Ca 2ϩdependent signaling in somatodendritic regions. Deployment of slots allows local control of presynaptic channel abundance.

Possible structural features to account for distinctions between N-and P/Q-type channels
Our findings allow us to refine the concept of slots beyond its initial, inclusive definition as binding sites of limited capacity that interact with channel proteins to control their participation in neurotransmission. This deliberately left open whether the interaction is initiated before proteins can exit from the ER, or at checkpoints upstream of axonal trafficking, or at synaptic sites themselves. The present results show that whole-cell Ca 2ϩ channel currents are greatly increased by overexpression of N-or P/Q-type pore-forming subunits, whereas synaptic strength has reached a ceiling. This narrows down the possible checkpoints for mammalian Ca V 2 channels, excluding one at the exit of proteins from the ER (Saheki and Bargmann, 2009), which would jointly restrict surface expression in cell body and axon.
Recognition of the distinctness of PQ preferring and N selective slots provides fresh perspective on previous biochemical findings aimed at understanding how N-and P/Q-type channels are anchored at the active zone. Maximov et al. (1999) initially focused on common features of long forms of Ca V 2.2 (here N WT ) and Ca V 2.1 (here PQ WT-long ), reasoning that the most important molecular determinants would be shared by both N-and P/Q-type channels. This led to description of how specific portions of a C-terminal domain of Ca V 2.2 interacted with the modular adaptor proteins Mint1 and CASK. The corresponding channel fragment disrupted channel clustering at synaptic sites, exerted a dominant-negative effect on synaptic function, and drove a chimera with a marker protein to synaptic sites (Maximov and Bezprozvanny, 2002). Although a strongly homologous region was present in the PQ WT-long , this domain proved to be entirely dispensable for proper synaptic localization of Ca V 2.1 (Hu et al., 2005). Likewise, inclusion or exclusion of the same domain had no impact on neurotransmission (Fig. 1). The disparity between the structural basis of N-and P/Q-type channel anchoring would be mysterious on the initial premise that both shared similar molecular determinants but is readily interpretable in light of our findings that N-type channels can access binding sites not avail-able to P/Q-type. Recognition of the existence of an N selective slot directs attention to proteins that bind selectively to N-type and not P/Q-type channels. Accordingly, a possible element of N selective slots is the MAP1A light chain 2, which binds to a subdomain within the C-terminal of Ca V 2.2 with low homology to Ca V 2.1 (Leenders et al., 2008). Similar lines of reasoning prompt a refined search for the structural basis of PQ preferring slots. Results from our electrophysiological study (Fig. 1), and parallel cytochemical findings of Hu et al. (2005), suggest that the PQ preferring slot interaction does not involve the distal C-terminal domain of Ca V 2.1 encoded by exon 47. Attention must now shift to protein-protein interactions not specific to P/Q-type channels, but shared with N-type. An obvious candidate is the synprint site, a relatively well conserved region of the II-III loop of Ca V 2.1 and Ca V 2.2 subunits, originally defined by interaction with various SNARE proteins (Catterall, 1999;Mochida et al., 2003). Skepticism about the involvement of synprint sites has arisen from study of transmission in invertebrate species whose Ca V 2 channels lack synprint sites yet support neurotransmission (Spafford et al., 2003). However, these findings could be reinterpreted in terms of Ca V 2 interactions with slots analogous to the N specific slots described here. In mammalian systems, deletion of the synprint site in Ca V 2.1 reduced the effectiveness of P/Q-mediated neurotransmission, and insertion of N-or P/Q-synprint sites into L-type channels was sufficient to establish L-mediated synaptic transmission (Mochida et al., 2003). Likewise, splice variants of mammalian N-type channels lacking the synprint region are strongly impaired in presynaptic clustering (Szabo et al., 2006). Together with our finding that P/Q-and N-type channels compete for a common subset of slots (Fig. 5), these data suggest that both types use synprint sites for interactions with PQ preferring slots, whereas N-type channels can do without synprint sites for binding to N selective slots but rely instead on determinants near their C termini. Our interpretations may reconcile multiple studies but will require additional testing with new channel constructs.