Kv2.1 mediates spatial and functional coupling of L-type calcium channels and ryanodine receptors in mammalian neurons

The voltage-gated K+ channel Kv2.1 serves a major structural role in the soma and proximal dendrites of mammalian brain neurons, tethering the plasma membrane (PM) to endoplasmic reticulum (ER). Although Kv2.1 clustering at neuronal ER-PM junctions (EPJs) is tightly regulated and highly conserved, its function remains unclear. By identifying and evaluating proteins in close spatial proximity to Kv2.1-containing EPJs, we discovered that a significant role of Kv2.1 at EPJs is to promote the clustering and functional coupling of PM L-type Ca2+ channels (LTCCs) to ryanodine receptor (RyR) ER Ca2+ release channels. Kv2.1 clustering also unexpectedly enhanced LTCC opening at polarized membrane potentials. This enabled Kv2.1-LTCC-RyR triads to generate localized Ca2+ release events (i.e., Ca2+ sparks) independently of action potentials. Together, these findings uncover a novel mode of LTCC regulation and establish a unique mechanism whereby Kv2.1-associated EPJs provide a molecular platform for localized somatodendritic Ca2+ signals in mammalian brain neurons.

In brain neurons, EPJs occupy approximately 10% of the PM surface area, predominantly within the soma and proximal dendrites (Wu et al., 2017). By electron microscopy, the ER at many neuronal EPJs appears as a micron-diameter, flattened vesicle less than 10 nm from the PM, a structure also called a 'subsurface cistern' (Rosenbluth, 1962;Tao-Cheng, 2018). While the specific functions of neuronal subsurface cisterns remain unclear, in most eukaryotic cells, EPJs represent domains specialized for maintenance of Ca 2+ , lipid, and metabolic homeostasis (Gallo et al., 2016;Chang et al., 2017).
L-type voltage-gated Ca 2+ channels (LTCCs) are prominently expressed in neurons throughout the brain (Catterall, 2011;Zamponi et al., 2015). Their important role in brain is underscored by studies showing genetic variation in the CACNA1C gene encoding Cav1.2, the major voltage-sensing and pore forming a1 subunit expressed in brain, is associated with neurodevelopmental, psychiatric and neurological disorders (Splawski et al., 2004;Ferreira et al., 2008;Bozarth et al., 2018). Given their diverse and crucial roles in neuronal function, LTCCs are subjected to multimodal regulation to ensure their activity is coupled to overall cellular state, especially as related to intracellular [Ca 2+ ] (Lipscombe et al., 2013;Hofmann et al., 2014;Neely and Hidalgo, 2014). In both neurons and non-neuronal cells, Cav1.2-containing LTCCs are clustered at specific sites on the PM where they participate in supramolecular protein complexes that couple LTCC-mediated Ca 2+ entry to specific Ca 2+ signaling pathways (Dai et al., 2009;Rougier and Abriel, 2016). In neurons, LTCCs in dendritic spines participate in a complex whose output contributes to short-and long-term synaptic plasticity (Da Silva et al., 2013;Simms and Zamponi, 2014;Stanika et al., 2015;Wiera et al., 2017). Neocortical and hippocampal pyramidal neurons and dentate granule cells also have substantial LTCC populations in the soma and proximal dendrites (Westenbroek et al., 1990;Hell et al., 1993;Tippens et al., 2008;Berrout and Isokawa, 2009;Marshall et al., 2011;Kramer et al., 2012) representing the 'aspiny' regions (Spruston and McBain, 2007) of these neurons. Many current models of Ca 2+ -dependent activation of transcription factors posit that somatic LTCCs uniquely contribute to transcription factor activation by mediating Ca 2+ influx within specialized and compartmentalized signaling complexes (Wheeler et al., 2008;Ma et al., 2012;Matamales, 2012;Wheeler et al., 2012;Ma et al., 2014;Cohen et al., 2015;Yap and Greenberg, 2018;Wild et al., 2019). However, relatively little research has focused on the molecular mechanisms underlying the spatial and functional compartmentalization of the prominent somatic population of LTCCs compared to those on dendrites and at synapses.
Neuronal somata lack PM compartments analogous to dendritic spines, and fundamental questions remain as to how discrete Ca 2+ signaling events can occur in the absence of such compartmentalization. In many non-neuronal cells, LTCCs are clustered at EPJs that represent specialized microdomains for LTCC-dependent and -independent Ca 2+ signaling (Helle et al., 2013;Lam and Galione, 2013;Henne et al., 2015;Burgoyne et al., 2015;Gallo et al., 2016;Chung et al., 2017;Dickson, 2017). For example, Cav1.2-mediated Ca 2+ entry is spatially and functionally coupled to ER ryanodine receptor (RyR) Ca 2+ release channels at EPJs constituting the cardiomyocyte junctional dyad (Shuja and Colecraft, 2018). Localized Ca 2+ release events (spreading <2 mm from the point of origin) called Ca 2+ sparks arise from clusters of RyRs located in the ER of EPJs and are triggered via local Ca 2+ -induced Ca 2+ release (CICR), a feed-forward phenomenon in which cytosolic Ca 2+ binding to RyRs triggers their opening (Cheng et al., 1993;Cheng and Lederer, 2008). As indicated above, EPJs are abundant on neuronal somata (Wu et al., 2017), and neuronal somata have prominent LTCC-and RyR-mediated CICR (Friel and Tsien, 1992;Isokawa and Alger, 2006;Berrout and Isokawa, 2009). Localized RyR-mediated Ca 2+ release events occur in the somata and proximal dendrites of cultured and acute slice preparations of hippocampal pyramidal neurons (Koizumi et al., 1999;Berrout and Isokawa, 2009;Manita and Ross, 2009;Miyazaki et al., 2012), but a specific molecular structure underlying these events has not been described.
Given the well-characterized spatial and functional coupling of LTCCs and RyRs at EPJs in myocytes and previous observations of somatodendritic clustering of the LTCC Cav1.2 in hippocampal neurons (Westenbroek et al., 1990;Hell et al., 1993), our finding that Kv2.1 clusters are often juxtaposed to RyRs previously led us to hypothesize that Kv2.1 channels cluster with LTCCs to form Ca 2+ 'micro-signaling domains' (Antonucci et al., 2001;Misonou et al., 2005a). More recently, heterologously expressed Kv2.1 and Cav1.2 were found to colocalize in dissociated cultured hippocampal neurons (CHNs) (Fox et al., 2015). However, the spatial association of Kv2.1 with endogenous LTCCs and RyRs in brain neurons has not been determined. Here, we examined the subcellular distribution of Kv2.1, LTCCs, and RyRs in hippocampal neurons and used an unbiased proteomic analysis of brain tissue to identify LTCCs and RyRs as proteins in close spatial proximity to clustered Kv2.1. Using heterologous cells and CHNs, we investigated the impact of Kv2.1 clustering on the spatial coupling and functional properties of LTCCs and RyRs. We also defined how the localization and function of LTCCs and RyRs are affected by the loss of Kv2.1 in mouse CHNs lacking Kv2.1. Together, our findings establish a functional interaction between Kv2.1, LTCCs, and RyRs, reveal a significant influence of Kv2.1 in shaping neuronal LTCC activity, and support a critical role for Kv2.1 in the generation of somatodendritic Ca 2+ signals.

Results
Kv2.1 channels spatially associate with LTCCs and RyRs in brain neurons In mature CHNs, endogenous Cav1.2 channels are distributed to PM-localized clusters on the soma and proximal dendrites, distinct from their punctate localization in the more distal postsynaptic compartments that also contain the scaffolding protein PSD-95 (Di Biase et al., 2008) ( Figure 1A). To investigate the spatial relationship between somatic Kv2.1 and Cav1.2 channels, we examined rat CHNs immunolabeled for Kv2.1 and Cav1.2, and also for Kv4.2 channels, which exhibit more uniform PM localization in CHNs than either Kv2.1 or Cav1.2 (Shibata et al., 2003). In CHNs expressing detectable levels of all three immunolabeling signals, presumed to be pyramidal neurons based on their morphological characteristics (Benson et al., 1994;Antonucci et al., 2001;Obermair et al., 2003), we observed clusters of Kv2.1 that were spatially associated with smaller Cav1.2 clusters but not Kv4.2 clusters ( Figure 1B). Triple immunolabeling for Kv2.1, Cav1.2, and RyRs demonstrated that many of the clusters of spatially associated Kv2.1 and Cav1.2 channels were colocalized with RyRs ( Figure 1C). We also observed more prominent spatial overlap of Kv2.1, Cav1.2, and RyR immunolabeling in a subset of CHNs ( Figure 1D). Analysis of Pearson's Correlation Coefficient (PCC) of Cav1.2 and either Kv2.1 or Kv4.2 pixel intensity demonstrated a greater spatial correlation between Cav1.2 and Kv2.1 immunolabeling than that of Cav1.2 and Kv4.2 ( Figure 1E). While the absolute PCC values indicate that the majority of somatic Kv2.1 and Cav1.2 immunolabeling did not co-occur within the same pixels, our data suggested that a subset Cav1.2 channels could be found in close proximity (if not overlapping) with Kv2.1. In support of a spatial association between Kv2.1, Cav1.2, and RyRs, we determined that there was a positive correlation between the PCC of Kv2.1 and Cav1.2, and the PCC of Cav1.2 and RyRs within the same cell ( Figure 1F), suggesting that increased association between Cav1.2 and Kv2.1 was also associated with greater spatial coupling of Cav1.2 to RyRs.
To better evaluate the subcellular distribution of LTCCs relative to Kv2.1 clusters, we next performed super-resolution structured illumination (SIM) imaging of immunolabeled CHNs. These images revealed that Kv2.1 clusters often encompassed smaller clusters of Cav1.2 as well Cav1.3 ( Figure 1G-H). For these super-resolution images, we performed an object-based analysis (rather than a pixel intensity correlation-based measurement such as PCC) to determine whether the localization of somatic Kv2.1 and LTCCs were co-dependent. The approach we used relied on evaluation of the nearest-neighbor distances (NND) of Kv2.1 and Cav1.2 or Cav1.3 cluster centroids and a comparison of these values to the predicted NNDs if Kv2.1 and LTCCs were randomly distributed (Shivanandan et al., 2013;Helmuth et al., 2010). We found that the spatial distributions of somatic Kv2.1 and Cav1.2 puncta significantly correlated (p<0.001 versus the null hypothesis that the spatial distributions of Kv2.1 and Cav1.2 puncta were independent) and could not be recapitulated in images in which their relative positions had been iteratively randomized in silico. We also observed similar expression patterns of endogenous Cav1.3 and RyRs in CHNs, with Cav1.3 clusters spatially associated with RyR clusters ( Figure 1I), in agreement with a recent report (Sahu et al., 2019). Together, these data suggested a spatial correlation between Kv2.1 and LTCCs.
We next evaluated how phosphorylation-dependent bidirectional changes in Kv2.1 clustering influenced the localization of somatic Cav1.2 and RyRs in rat CHNs. One stimulus that results in dephosphorylation of Kv2.1 and dispersal of Kv2.1 clusters in CHNs is acute elevation in intracellular Ca 2+ in response to treatment with the excitatory neurotransmitter glutamate (Misonou et al., 2004;Misonou et al., 2006). In contrast, suppression of neuronal activity with tetrodotoxin (TTX) causes an increase in Kv2.1 phosphorylation and clustering (Cerda and Trimmer, 2011;Romer et al., 2019). We found that glutamate stimulation of CHNs not only reduced Kv2.1 clustering, but also significantly decreased the colocalization between Cav1.2 and RyRs, decreased the size number of Cav1.2 clusters present on the PM, consistent with previous observations that acute Ca 2+ influx results in endocytosis of Cav1.2 channels (Hall et al., 2013). Together, these data show that bidirectional changes in Kv2.1 clustering are coupled to corresponding changes in the spatial distributions of Cav1.2 and RyRs on CHN somata. We next assessed the localization of Kv2.1, Cav1.2, and RyRs in brain sections. Previous immunohistochemical analyses showed that in hippocampal neurons, Cav1.2 localizes to distinct clusters on somata and proximal dendrites (Westenbroek et al., 1990;Hell et al., 1993), a spatial pattern similar to that of Kv2.1 (Trimmer, 1991;Scannevin et al., 1996;Kirizs et al., 2014). Similar to previous observations, in low magnification images of mouse and rat hippocampus, we observed Cav1.2 immunolabeling concentrated in CA1 neuron somata, with increasing labeling in area CA2/CA3 neurons, and greatest labeling in dentate gyrus (DG) granule cell somata and dendrites (Figure 2A-E). In higher magnification confocal images of DG granule cell bodies in rat hippocampus, we found that Kv2.1 clusters tended to colocalize with Cav1.2 clusters ( Figure 2F). The somata of rat CA1 pyramidal neurons exhibited a spatial association of Cav1.2, Kv2.1, and RyR immunolabeling that was qualitatively comparable to that seen in CHNs ( Figure 2G). Similar labeling was observed in high-magnification images of mouse brain sections ( Figure 2H-I). Kv2.2, which also clusters at EPJs through the same mechanism as Kv2.1 (Kirmiz et al., 2018b), similarly colocalized with Cav1.2 immunolabeling in rat CA1 pyramidal cells and DG granule cells ( Crosslinking-based proteomic analyses support that Kv2.1 channels are in close spatial proximity to LTCCs and RyRs in brain neurons We interrogated proteins within the Kv2.1 nano-environment using a crosslinking-and mass spectrometry-based proteomics approach to further determine whether LTCCs and RyRs were in close spatial proximity (having lysine residues within » 12 Å of one another) to Kv2.1. We affinity immunopurified (IPed) Kv2.1 from mouse brain homogenates that were subjected to chemical cross-linking during homogenization. This strategy previously allowed us to identify the ER-resident VAP proteins as Kv2 channel binding partners (Kirmiz et al., 2018a). Importantly, we also performed parallel IPs from brain homogenates prepared from Kv2.1 knockout (KO) mice (Jacobson et al., 2007;Speca et al., 2014) using the same Kv2.1 antibody, to identify proteins IPing in a Kv2.1-independent manner. To further improve the recovery of peptides IPed with Kv2.1, we performed on-bead trypsin digestion, as opposed to the in-gel digestion we had done previously (Kirmiz et al., 2018a). Similar to our earlier findings, enriched in the control Kv2.1 IPs (and absent from the Kv2.1 KO brain IPs) were the VAP isoforms VAPA and VAPB (Table 1). In addition, among the 50 most abundant proteins specifically present in Kv2.1 IPs (i.e., from WT and not Kv2.1 KO brain samples) were numerous proteins involved in Ca 2+ signaling and/or previously reported to localize to neuronal EPJs. These included RyR isoforms RyR2 and RyR3, the LTCC a subunits Cav1.2 and Cav1.3, various Cavb auxiliary subunits of LTCCs, as well as other proteins involved in Ca 2+ signaling and homeostasis (Table 1). Taken together with our imaging analyses, these findings indicate that Kv2.1 is in close spatial proximity to LTCCs and RyRs at EPJs in mouse brain neurons. We note that while Cav1.2 is the predominant LTCC a1 subunit in hippocampus (Hell et al., 1993;Davare et al., 2001;Moosmang et al., 2005;Lacinova et al., 2008;Sinnegger-Brauns et al., 2009), where its localization on neuronal somata overlaps with Kv2.1, it was not as highly represented in these proteomic analyses as was Cav1.3, perhaps as these analyses were performed on whole brain samples.

Kv2.1 organizes the localization of LTCCs
Because our immunolabeling and proteomics results indicated that endogenous Cav1.2 channels spatially associated with clustered Kv2.1 in hippocampal neurons, we investigated how the subcellular localization of Cav1.2 (expressed with the LTCC auxiliary subunits a 2 d 1 and b3) was influenced by the presence of Kv2.1 in heterologous HEK293T cells. HEK293T cells lack endogenous Kv2.1 or Kv2.2 channels (Yu and Kerchner, 1998), and have little to no expression of LTCCs (Berjukow et al., 1996;Geiger et al., 2012). Expression of conducting or nonconducting Kv2 channels in these cells induces EPJ formation (Fox et al., 2015;Bishop et al., 2018;Kirmiz et al., 2018a;Kirmiz et al., 2018b). Using total internal reflection fluorescence (TIRF) microscopy to visualize Cav1.2-GFP expressed in HEK293T cells, we observed small Cav1.2 clusters (average area 0.27 ± 0.24 mm 2 ) adjacent to or overlapping with cortical ER, marked by the general ER marker BFP-SEC61b ( Figure 3A,C). However, in the presence of Kv2.1, the organization of Cav1.2 was dramatically altered, such that Cav1.2 now co-assembled with Kv2.1 into significantly larger clusters (1.05 ± 0.67 mm 2 ) that showed greater colocalization with the ER (as indicated by the PCC of Cav1.2-GFP and BFP-Sec61b) than in the absence of Kv2.1 ( Figure 3B,D-F). To confirm that these large Kv2.1 clusters were present in the PM, we labeled cell surface Kv2.1 with guangxitoxin-633 (GxTX-633), a membrane impermeant, Kv2 channel-specific toxin conjugated to a fluorescent dye (Tilley et al., 2014) ( Figure 3B). The Kv2.1-induced rearrangement of Cav1.2 was accompanied by an increased occurrence of larger Cav1.2 clusters and a reduced occurrence of smaller Cav1.2 clusters, and a nearly linear relationship between the sizes of Cav1.2 and Kv2.1 clusters ( Figure 3F). Kv2.1 spatially associates with Cav1.2 and RyRs in brain neurons. (A) Panels show exemplar images of the hippocampus acquired from a brain section from an adult rat immunolabeled for Kv2.1 (red), Cav1.2, (green) and RyRs (blue), and the merged image (scale bar: 200 mm). (B) As in A but acquired from an adult mouse brain section. (C-E) Summary graphs of normalized mean fluorescence intensity of Kv2.1, Cav1.2, and RyR immunolabeling from ROIs from various laminae within CA1 (s.p.: stratum pyramidale; s.r.: stratum radiatum) and DG (s.g.: stratum granulosum; mo: molecular layer) in WT mouse brain sections. Fluorescence intensity values were normalized to CA1 s.p. for each mouse. Each point corresponds to an individual mouse (one-way ANOVA followed by Dunnett's post-hoc test vs. CA1 s.p.). (C) **p=0.0025, # p=0.0573 (D) *p=0.0408 (E) *p=0.0198 (CA1 s.r.), *p=0.0324 (DG s.g.), *p=0.0107 (DG s.m.) (F) Confocal optical section obtained from the dentate gyrus of a rat brain section immunolabeled for Kv2.1 (red) and Cav1.2 (green) (scale bar: 10 mm). The row below the main panels shows expanded views of immunolabeling in the region marked by the box in the main panels; arrowheads indicate region selected for intensity profile line scan (scale bar: 2 mm). Line scan obtained from inset is shown to the right. (G) Confocal optical section obtained from the pyramidal cell layer of hippocampal area CA1 in a rat brain section immunolabeled for Kv2.1 (red), Cav1.2 (green), and RyRs (blue) (scale bar: 10 mm). The row below the main panels shows expanded view of immunolabeling in the region marked by the box in the main panels (scale bar: 2 mm). (H) As in F but acquired from a mouse brain section. (I) As in G but acquired from a mouse brain section. DOI: https://doi.org/10.7554/eLife.49953.003 The following figure supplement is available for figure 2: Because TIRF microscopy illuminates subcellular structures that can be up to 100 nm away from the PM, we tested whether the observed co-clustering of Cav1.2 with Kv2.1 was occurring within the PM itself. We performed cell surface immunolabeling of intact cells coexpressing Kv2.1 and a Cav1.2 construct possessing an extracellular hemagglutinin epitope tag [Cav1.2-HA, (Obermair et al., 2004). Following cell surface immunolabeling of Cav1.2-HA channels, cells were permeabilized and immunolabeled for total Cav1.2-HA, then imaged by TIRF microscopy ( Figure 3G). Similar to cells expressing fluorescently tagged channels, we found that cell surface Cav1.2-HA also co-clustered with Kv2.1 ( Figure 3H). We also determined that cell surface Cav1.2-HA immunolabeling corresponded to approximately 70% of the total Cav1.2 visible in the TIRF field regardless of Kv2.1 coexpression ( Figure 3I), suggesting that Kv2.1 did not alter the steady-state partitioning of Cav1.2 between PM and intracellular pools. Importantly, similar to results obtained evaluating total Cav1.2, cell surface Cav1.2-HA cluster size was also larger in the presence of Kv2.1, indicating recruitment of cell surface Cav1.2 into larger clusters induced by Kv2.1 ( Figure 3I). We also found that coexpression with the related but distinct Kv1.5 channel did not impact the clustering of cell surface Cav1.2 channels as did coexpression with Kv2.1 (Figure 3-figure supplement 1A). As another measure of the impact of Kv2.1 expression on the organization of PM Cav1.2, we assessed the coefficient of variation (CV: SD/mean) of Cav1.2-HA immunolabeling intensity. The CV is used as a measure of non-uniformity of subcellular distribution, with clustered distributions having high CV values and uniform or diffuse signals having low CV values (Bishop et al., 2015;Jensen et al., 2017;Bishop et al., 2018;Kirmiz et al., 2018a;Kirmiz et al., 2018b). We found that cells coexpressing Kv2.1 had higher CV values for cell surface Cav1.2 than did cells coexpressing Kv1.5 ( Figure 3-figure supplement 1D). Cell surface Kv2.1 labeling also exhibited greater colocalization with cell surface Cav1.2 than did cell surface labeling for Kv1.5 (as indicated by PCC values, Figure 3-figure supplement 1E).
Together, these data demonstrate that clustered but not non-clustered Kv2 channels enhance LTCC clustering and increase their localization to EPJs as a nonconducting function.

Neuronal Kv2.1 channels functionally associate with endogenous LTCCs and RyRs
Kv2.1 fused to fluorescent proteins such as GFP clusters at neuronal EPJs similar to untagged or endogenous Kv2.1 (Antonucci et al., 2001;Kirmiz et al., 2018a;Kirmiz et al., 2018b). To begin to evaluate Ca 2+ signals at neuronal Kv2.1-associated EPJs, we fused the genetically-encoded Ca 2+ indicator GCaMP3 (derived from GFP) to K + -conducting and -nonconducting Kv2.1 channel isoforms and expressed these constructs in rat CHNs. GCaMP3 has previously been used to study near-membrane Ca 2+ signaling microdomains in astrocytes (Shigetomi et al., 2010), and its higher basal fluorescence relative to newer GCaMP variants facilitated identification of transfected neurons. When expressed in HEK293T cells, GCaMP3-Kv2.1 WT and GCaMP3-Kv2.1 P404W were comparably expressed in surface-localized clusters as reported by both GxTX-633 labeling and GCaMP3 fluorescence ( Figure 4-figure supplement 1). In rat CHNs, GCaMP3-Kv2.1 exhibited clustered localization similar to other fluorescently tagged Kv2.1 isoforms ( Figure 4A) and reported global Ca 2+ spikes, as indicated by the synchronized increase in fluorescence across the PM at sites where the construct was clustered and also in regions with diffuse GCaMP3-Kv2.1 expression ( Figure 4B, Video 1). In   Figure 4 continued on next page addition to synchronized Ca 2+ spikes, we also observed rapid and stochastic Ca 2+ signals occurring at a subset of individual GCaMP3-Kv2.1 clusters within the soma ( Figure 4B-C, Video 1). These Ca 2+ signals were confined to individual clusters such that the fluorescence of adjacent GCaMP3-Kv2.1 clusters < 1 mm from the active clusters remained stable ( Figure 4B, compare regions of interest 2 and 4). We found that Ca 2+ signal amplitude, frequency, and width were insensitive to the K + conductance of the GCaMP3-Kv2.1 reporter, as Ca 2+ signals detected by a K + -impermeable variant of this construct (GCaMP3-Kv2.1 P404W ) showed no difference in any of these parameters relative to GCaMP3-Kv2.1 ( Figure 4D). Next, we assessed the relationship between GCaMP3-Kv2.1 reported Ca 2+ signals and membrane potential (V m ). We performed current clamp experiments to monitor the V m and Ca 2+ signals simultaneously, using the whole-cell perforated patch clamp configuration. Spontaneous action potentials were associated with Ca 2+ spikes, suggesting that these synchronized, large-amplitude Ca 2+ transients reflected Ca 2+ entry through voltage-gated Ca 2+ channels as well as Ca 2+ release through RyRs ( Figure 4E). However, unlike global Ca 2+ spikes, the localized Ca 2+ signals displayed no clear relationship with action potentials or other spontaneous V m fluctuations, similar to previous observations of localized Ca 2+ release events in CA1 pyramidal neurons (Berrout and Isokawa, 2009;Manita and Ross, 2009). We also found that the localized Ca 2+ signals could persist in the presence of TTX ( Together, these observations suggest that the localized Ca 2+ signals arose independently of large, uniform fluctuations in the V m such as those that occur with action potentials. As heterologous expression of Kv2.1 in CHNs is known to result in large Kv2.1 'macroclusters' that recruit RyRs (Antonucci et al., 2001), we next determined whether somatic Ca 2+ signals occurred at native Kv2.1-associated EPJs. For these experiments, we used non-transfected CHNs loaded with the Ca 2+ dye Cal-590 AM and recorded Ca 2+ signals using TIRF microscopy. Using this approach, it was possible to detect spontaneous, localized Ca 2+ release events in the soma that were qualitatively similar to those recorded with GCaMP3-Kv2.1 ( Figure 4F-G, Video 2), although with faster kinetics in fluorescence intensity changes relative to GCaMP3-Kv2.1, likely reflecting differences in the Ca 2+ binding properties of these probes. Post-hoc immunolabeling of these CHNs for Kv2.1, RyRs, and the neuron-specific cytoskeletal protein MAP2 indicated that the observed localized Ca 2+ signals occurred primarily within the soma at sites of colocalized Kv2.1 and RyR clusters ( Figure 4F).
These observations suggested that the Ca 2+ signals observed at neuronal Kv2.1-associated EPJs reflected RyR-generated Ca 2+ sparks. To further assess this possibility, we imaged GCaMP3-Kv2.1-expressing CHNs treated with compounds that modulate LTCC-and RyR-mediated CICR. We found that caffeine, which sensitizes RyRs to cytosolic Ca 2+ , enhanced the frequency of localized Ca 2+ signals ( Figure 5A,B, Video 3). In contrast, depletion of ER Ca 2+ stores with the sarco-/endo-plasmic reticulum Ca 2+ ATPase (SERCA) inhibitor thapsigargin led to an elimination of local Ca 2+ signals ( Figure 5A-B). The functional coupling of dendritic LTCCs and RyRs in hippocampal neurons has previously been demonstrated by the impact of dihydropyridine (DHP) compounds on dendritic Ca 2+ sparks: the LTCC agonist Bay K8644 increased Ca 2+ spark frequency, whereas the LTCC inhibitor nimodipine blocked Ca 2+ sparks (Manita and Ross, 2009). Here, we obtained similar evidence of the involvement of LTCCs in the generation of somatic GCaMP3-Kv2.1 reported Ca 2+ signals. The frequency of these Ca 2+ sparks was enhanced by activation of LTCCs with Bay K8644 ( Figure 5A,B,D, Videos 4 and 5), while they were rapidly inhibited by blockade of LTCCs with nimodipine ( Figure 5A-B). We also performed post-hoc immunolabeling of these imaged CHNs to determine whether the specific GCaMP3-Kv2.1 clusters which exhibited localized Ca 2+ signals were associated with RyRs. Using this approach, we determined that the subset of GCaMP3-Kv2.1 clusters that colocalized with RyRs corresponded to the clusters that produced localized Ca 2+ signals, either spontaneously or in response to the pharmacological modulators caffeine ( Figure 5C) and Bay K8644 ( Figure 5D). We also quantified the relationship between the size of post-hoc immunolabeled RyR clusters and spark frequency and amplitude. Similar to previous observations in vascular smooth muscle (Pritchard et al., 2018) and cardiac muscle cells (Galice et al., 2018), we found that neuronal Ca 2+ spark frequency but not amplitude correlated with RyR cluster size, and that application of the LTCC agonist Bay K8644 steepened this relationship ( Figure 5E). Taken together, these observations demonstrate that Kv2.1-associated EPJs are sites of spontaneous CICR events mediated by LTCCs and RyRs.

Kv2.1 augments LTCC and RyR2-mediated CICR reconstituted in HEK293T cells
We next asked how Kv2.1-induced clustering of LTCCs would impact RyR-mediated Ca 2+ release in HEK293T cells. For these experiments, we expressed Kv2.1 along with Cav1.2, the LTCC auxiliary subunits a 2 d 1 and b3, RyR2, and the STAC1 adaptor protein, an approach similar to that previously used to recapitulate Cav1.1-and RyR1-mediated Ca 2+ release in HEK293T cells (Perni et al., 2017). We found that in the presence of these auxiliary subunits, Kv2.1, Cav1.2, and RyR2 could spatially associate in HEK293T cells ( Figure 6A). Thus, the spatial association of Kv2.1, Cav1.2, and RyRs seen in neurons could be recapitulated in HEK293T cells. To detect Ca 2+ release events, we performed TIRF microscopy of cells loaded with the Ca 2+ -sensitive dye Cal-590 AM. Although it was not possible to establish whether a cell expressed all transfected constructs, we observed spontaneous Ca 2+ release events in a subset of cells ( Figure 6B,E) that were not seen in untransfected HEK293T cells and focused our analysis on cells that exhibited this phenotype. These spontaneous Ca 2+ release events were rapidly blocked by the RyR inhibitor tetracaine ( Figure 6G, Video 6), suggesting that they reflected CICR mediated by RyRs. Expressing Kv2.1 in these cells resulted in enhanced spark frequency and amplitude ( Figure 6C To better understand the mechanism underlying the influence of Kv2.1 on these reconstituted Ca 2+ sparks, we next compared how they were affected by the nonconducting Kv2.1 P404W and the non-clustering Kv2.1 S586A point mutants ( Figure 6I). By using these Kv2.1 isoforms, we determined that there was an interplay between both Kv2.1 K + conductance and clustering on Ca 2+ sparks reconstituted in HEK293T cells. Expression of Kv2.1 channels capable of clustered EPJ formation (i. e., Kv2.1 WT and Kv2.1 P404W ) increased spark frequency, whereas non-clustering Kv2.1 S586A did not ( Figure 6J). Interestingly, we found that spark amplitude was enhanced by K + -conducting Kv2.1 WT but not Kv2.1 P404W ( Figure 6J), suggesting that while Kv2.1-mediated clustering alone was sufficient to impact spark frequency, K + conductance was required to impact the amplitude of reconstituted Ca 2+ sparks in HEK293T cells. We hypothesize that the high input resistance of HEK293T cells relative to CHNs, the latter of which possess numerous endogenous ionic conductances (including native Kv2.1 channels), enabled K + conductance through Kv2.1 WT and Kv2.1 S586A to promote V m hyperpolarization in HEK293T cells, maintaining a greater electrochemical driving force for extracellular Ca 2+ and also promoting recovery of Cav1.2 channels from voltage-dependent inactivation. In conclusion, these observations indicate that Kv2.1-mediated clustering promotes the functional coupling of Cav1.2 and RyRs. Different colors indicate spatially distinct ROIs within the same neuron. Dashed lines indicate approximate maximum amplitude for localized Ca 2+ signals as opposed to the larger amplitude, synchronized global Ca 2+ transients that exceed the dashed lines. (B) Summary data of the amplitude and frequency of all sparks recorded from CHNs treated with pharmacological probes of CICR. Each point corresponds to a single cell (**p=0.0013 vs. control; ****p<0.0001 vs. control; {}: no Ca 2+ sparks detected; one-way ANOVA followed by Dunnett's test). (C) Image of rat CHN transfected with GCaMP3-Kv2.1 and treated with caffeine, followed by post-hoc immunolabeling for RyRs (scale bar: 10 mm). Numbered arrows indicate ROIs where localized Ca 2+ signals were detected (ROIs 1-3) or not detected (ROI 4). ROI fluorescence traces are shown in lower panel; note lack of spontaneous Ca 2+ signals at ROI 4 despite its proximity to ROI 3, which displays prominent spontaneous Ca 2+ release. (D) As in panel A, except CHN was treated with 500 nM Bay K8644 to induce spontaneous Ca 2+ signals (scale bar: 10 mm). (E) Plot of individual RyR cluster size (determined from post-hoc immunolabeling) versus its spark amplitude (left panel) or frequency (right panel) reported by GCaMP3-Kv2.1 fluorescence in control (black symbols) or Bay K8644-treated (red symbols) cells. Each point corresponds to an individual RyR cluster (n = data from 4 cells [control]  Having demonstrated a spatial and functional association of Kv2.1, LTCCs, and RyRs in hippocampal neurons that could be reconstituted in HEK293T cells, we next investigated whether clustering by Kv2.1 influenced the Cav1.2-mediated LTCC activity. As physical interactions between adjacent LTCCs promote enhanced LTCC activity (reducing the membrane voltage threshold for channel opening and elevating channel open probability) (Navedo et al., 2005;Dixon et al., 2012;Moreno et al., 2016), we reasoned that this functional property of Cav1.2 might be enhanced by Kv2.1-induced clustering. To test this possibility, we obtained whole-cell patch-clamp recordings from HEK293T cells transfected with Cav1.2 and the clustered but non-K + conducting Kv2.1 P404W point mutant, which allowed us to measure Ca 2+ currents (I Ca ) in the absence of the very large outward K + currents produced by Kv2.1 WT . Consistent with an influence of Cav1.2 spatial organization on its activity, we found that expression of Cav1.2 with Kv2.1 P404W enhanced peak I Ca as compared to cells expressing Cav1.2 alone ( Figure 7A-B). Analysis of the conductance-voltage (G-V) relationship also showed an influence of Kv2.1 on the V m threshold for Cav1.2 opening, with currents produced by Cav1.2 activating at more negative voltages in the presence of Kv2.1 P404W than those produced by Cav1.2 alone, with no effect on steady-state inactivation ( Figure 7C). However, we did observe a greater reduction in the fraction of peak current remaining after 250 ms of depolarization (r 250 ), ( Figure 7D) shift in Cav1.2 opening, similar to results obtained without STAC1 (Figure 7-figure supplement 1A-C). However, in the presence of STAC1, which substantially reduces CDI in Cav1.2 (Campiglio et al., 2018), Cav1.2 r 250 values were comparable between control cells and cells expressing Kv2.1 (Figure 7-figure supplement 1D). Measurement of Ca 2+ -induced fluorescence increases in cells loaded with the Ca 2+ -sensitive dye Rhod-2 via the patch pipette also revealed an enhancing effect of Kv2.1 P404W on Cav1.2-mediated Ca 2+ influx ( Figure 7E). Similarly, HEK293T cells loaded with the Ca 2+ dye Fluo-4 and expressing Cav1.2 and either Kv2.1 WT or Kv2.1 P404W displayed greater K + -depolarization induced Ca 2+ influx than control cells ( Figure 7F-G), further supporting that K + -conducting as well as -nonconducting isoforms of Kv2.1 augment Cav1.2 activity. Ion channel activity can be described by the product of the number of channels present in the PM (n), the channel's unitary conductance (i), and the open probability of these channels (P o ), such that the whole-cell current I can be described by the relationship I = nP o i. Thus, the enhancement of Cav1.2 activity observed in the presence of Kv2.1 could be caused by an effect on any one or more of these parameters. To better understand the underlying mechanism, we acquired gating and ionic tail currents from the same cell. Depolarization-induced voltage sensor movement in activating voltage-gated channels produces a gating current (Q on ) that is proportional to the number of channels present in the PM (n). Repolarization-induced ionic tail currents (I tail ) reveal overall channel activity (I). Changes in one or both can be used to infer whether it is 'n' versus some combination of 'P o ' and/or 'i' that yield changes in total channel activity. We used nitrendipine, a DHP LTCC gating inhibitor, to pharmacologically isolate Cav1.2 Q on when the V m was stepped to the I Ca reversal potential, and to measure I tail elicited by returning to the À70 mV holding potential ( Figure 7H). Nitrendipine-sensitive Q on values produced by Cav1.2 alone were comparable to those measured in the presence of Kv2.1, indicating that the increased I Ca in cells coexpressing Kv2.1 was not associated with an increase in the number of PM Cav1.2 channels ( Figure 7I). However, the nitrendipine-sensitive I tail was significantly greater in the presence of Kv2.1, demonstrating that the open probability and/or conductance of Cav1.2 was increased when coexpressed with Kv2.1. As comparable Q on values (i.e., Cav1.2 voltage sensor movement) produced a larger I tail in the presence of Kv2.1, these data taken together with the altered G-V curve shown in Figure 7C suggest that the Kv2.1-dependent increase in I Ca apparently came from enhanced Cav1.2 voltage sensor coupling to channel opening.
We next tested how Kv2.1 impacted the spontaneous opening of Cav1.2 at hyperpolarized V m values. We used an optical approach to measure single Cav1.2 channel activity by recording Cav1.2mediated Ca 2+ sparklets, local elevations in intracellular Ca 2+ produced by the opening of a single or small cluster of Cav1.2 channels (Cheng and Lederer, 2008). In addition to providing the singlechannel activity of all active Cav1.2 channels present in the TIRF footprint, this approach had the additional benefit of revealing where in the PM the active channels were localized. We recorded Ca 2+ sparklets at a holding potential of À70 mV in the presence of 20 mM external Ca 2+ to increase the driving force for Ca 2+ influx. In control cells expressing Cav1.2 alone, we observed occasional sparklets whose activity was enhanced by the application of the LTCC agonist Bay K8644 ( Figure 8A-B,H-I, Video 7). In contrast, control cells coexpressing Cav1.2 and DsRed-Kv2.1 P404W displayed significantly more sparklets than control cells expressing Cav1.2 alone and a higher level of basal activity as measured by nP s , where n is the number of quantal levels of Cav1.2 opening and P s is the probability of sparklet occurrence ( Figure 8C-E,I,J,L, Video 8). This observation is consistent with our ionic tail current data indicating that Cav1.2 open probability was enhanced in the presence of Kv2.1. Interestingly, Bay K8644 treatment did not result in any further enhancement in nP s in cells expressing DsRed-Kv2.1 P404W ( Figure 8J), suggesting that DsRed-Kv2.1 P404W coexpression may result in near-maximal activation of Cav1.2-mediated sparklets at this hyperpolarized membrane potential. Many sparklets occurred near clusters of Kv2.1 channels, and we found that the nearest-neighbor distance (NND) of individual sparklet sites was significantly reduced in the presence of Kv2.1 ( Figure 8K). Taken together, these findings demonstrate that Kv2.1 enhanced the open probability of Cav1.2 channels and increased the proximity of spontaneously active Cav1.2 channels to each other.
Given the impact of Kv2.1 coexpression on LTCC activity in heterologous HEK293T cells, we next asked whether LTCC currents were altered in CHNs lacking Kv2.1. For these experiments, we chose to record from CHNs as opposed to acutely dissociated neurons. Although the absence of extensive processes in acutely dissociated neurons enables much better control of the V m than in arborized neurons, we reasoned based on the loss of Kv2.1 clustering upon dissociation in other cell types expressing clustered Kv2.1 (PC12, MDCK, and HEK293 cells; J.S. Trimmer, unpublished observations), and that endogenous Kv2.1 clusters in CHNs are sensitive to changes in intracellular Ca 2+ and metabolism (Misonou et al., 2005b), that acute dissociation would disrupt the clustered localization of Kv2.1, potentially concealing LTCC regulation by Kv2.1 clustering. To improve somatic voltage clamp, we used recording solutions lacking Na + and containing Cs + and Ba 2+ (which block K + channels; Ba 2+ also permeates voltage-gated Ca 2+ channels) to increase membrane impedance. We focused our analyses of electrophysiological recordings on repolarization-induced tail currents after activation of channels by a depolarizing prepulse, rather than measurement of currents induced by depolarizing voltage steps that can be distorted due to space clamp limitations (e.g., see Milescu et al., 2010). Similar to the impact of Kv2.1 on LTCCs in HEK293T cells, whole-cell Ba 2+ currents (I Ba ) at +10 mV ( Figure 9A), as well as LTCC tail currents ( Figure 9B,C) were larger in CHNs from WT mice than those measured in Kv2.1 KO CHNs ( Figure 9A-C). To isolate the LTCC component of I Ba , we applied the LTCC gating inhibitor nimodipine (10 mM), and found that the reduced I Ba observed in Kv2.1 KO CHNs ( Figure 9A-C) was primarily due to a reduction in the nimodipinesensitive component of the current (Figure 9A,B,E), with no apparent difference in the nimodipineresistant current ( Figure 9A,B,D). We also examined nimodipine-sensitive gating and ionic tail currents when the V m was stepped to the I Ba reversal potential and found that while Q on was not significantly different between control and Kv2.1 KO CHNs, peak I tail was reduced in Kv2.1 KO CHNs ( Figure 9F-G). The data in Figures 7 and 8 (from exogenously expressed channels in HEK293T cells) and Figure 9 (from endogenously expressed channels in CHNs) show that Kv2.1 enhances neuronal LTCC activity and suggest that the underlying mechanism in both experimental systems involves enhanced coupling efficiency between LTCC voltage-sensor movement and channel opening due to Kv2.1-mediated clustering.

Kv2.1 promotes spatial coupling of LTCCs and RyRs
Given that Kv2.1-mediated clustering impacts the spatial distribution of Cav1.2 in coexpressing HEK293T cells, we next examined whether loss of Kv2.1 was associated with changes in the expression and localization of Cav1.2. We first performed immunolabeling of hippocampal neurons in brain sections from adult control and Kv2.1 KO mouse littermates. We have previously determined that the anatomic structure of mouse brains lacking Kv2.1 is comparable to controls, and there do not appear to be compensatory changes in the expression of other Kv channels tested (Speca et al., 2014). Here, we confirmed that immunolabeling for somatodendritic Kv2.2 and dendritic Kv4.2 channels was similar in WT and Kv2.1 KO hippocampus ( Figure 10A-C). However, Cav1.2 labeling was increased in pyramidal neurons in area CA1 in Kv2.1 KO brain sections, both within the cell bodies and in the apical dendrites ( Figure 10C). These results suggest that in adult mice lacking functional Kv2.1 channels, Cav1.2 expression may be elevated, potentially as a compensatory mechanism to overcome reduced Cav1.2 channel function.
To obtain more detailed individual cell information, we next investigated how the loss of endogenous Kv2.1 influenced the localization and function of LTCCs and RyRs in WT and Kv2.1 KO CHNs.
To determine whether Kv2.1 channels regulate the localization of somatodendritic Cav1.2 and/or RyRs, we first analyzed the size and morphology of immunolabeled Cav1.2 and RyR clusters in WT and Kv2.1 KO mouse CHNs ( Figure 11A-B). We found that compared to WT CHNs, Kv2.1 KO CHNs had reduced colocalization between Cav1.2 clusters and RyR, decreased size of RyR clusters, and increased distance between Cav1.2 clusters ( Figure 11C-F). However, unlike the increased Cav1.2 immunolabeling found in adult Kv2.1 KO mouse brain sections, we found that the number of Cav1.2 clusters per mm 2 of somatic membrane did not differ between WT and Kv2.1 KO CHNs. These observations suggests that while compensatory changes in Cav1.2 expression did not occur in cultured Kv2.1 KO CHNs after approximately two weeks in vitro as it did in adult brain neurons in vivo, the presence of Kv2.1 promoted the spatial coupling of Cav1.2 to RyRs, consistent with our results in HEK293T cells.
Finally, to evaluate how impaired Cav1.2 and RyR spatial coupling in Kv2.1 KO CHNs affected spontaneous CICR events or sparks, we imaged Cal-590-loaded cells using TIRF microscopy. Similar to rat CHNs, we observed spontaneous sparks in WT mouse CHNs that were associated with Kv2.1, Cav1.2, and RyR clusters identified by post-hoc immunolabeling ( Figure 11G). Consistent with the reduced colocalization of Cav1.2 and RyRs in Kv2.1 KO CHNs, we found that loss of Kv2.1 was associated with a significant reduction in spark frequency relative to WT control CHNs ( Figure 11H). Taken together, these findings demonstrate that Kv2.1 channels promote the spatial and functional association of endogenous Cav1.2 and RyRs in neurons, as well as the corresponding exogenous channels in HEK293T cells.

Discussion
The findings in this study support a new model for the formation of Ca 2+ signaling microdomains at EPJs and the local control of Ca 2+ release from these structures. In this model, neuronal EPJs are Ca 2+ signaling microdomains in which Cav1.2 and RyRs are brought into close proximity by Kv2.1mediated clustering, forming a specialized somatic complex for the generation of localized Ca 2+ signals by these Ca 2+ channels ( Figure 11I). We propose that a nonconducting function of PM Kv2.1 channels is to not only anchor the ER to the PM via a direct interaction with ER VAP proteins (Johnson et al., 2018;Kirmiz et al., 2018a), but also to promote the organization of Cav1.2 channels into clusters in direct apposition to nearby ER-localized RyRs. Our data indicate that Kv2.1mediated clustering also increases the activity of Cav1.2. The enhanced spontaneous openings of Cav1.2 channels at negative potentials is evidenced by the increased frequency of sparklets, which allow a small amount of Ca 2+ to enter the cell at EPJs, activating nearby RyRs by the mechanism of CICR. The resulting Ca 2+ sparks occur independently of action potentials. Thus, our model proposes the molecular architecture of a protein complex ( Figure 11I) underlying the localized somatodendritic Ca 2+ signals previously observed in brain neurons (Berrout and Isokawa, 2009;Manita and Ross, 2009), and suggests a mechanism whereby Kv2.1 modulates these Ca 2+ signals by simultaneously promoting the spatial association of Cav1.2 channels with RyRs and increasing their activity to trigger CICR.

Kv2 channels dynamically cluster LTCCs
A key finding in this study is that endogenous LTCCs colocalize with clustered Kv2.1 in brain neurons, a finding supported by our crosslinking-based proteomic analyses showing that they exist in close spatial proximity. Moreover, colocalization of LTCCs with Kv2.1 could be reconstituted in heterologous cells, a property that required Kv2.1's ability to cluster at EPJs but was separable from its voltage-gated K + channel function. The Kv2.1-mediated association of Cav1.2 with EPJs appears to be dynamically regulated and sensitive to neuronal activity, as acute dispersal of Kv2.1 clusters in CHNs by glutamate stimulation reduced Cav1.2 association with RyRs, whereas suppression of neuronal activity with TTX (which increases Kv2.1 phosphorylation and clustering) enhanced spatial coupling of Cav1.2 and RyRs. In addition, Kv2.1 expression in heterologous cells simultaneously enhanced the size of LTCC clusters and recruited LTCCs to Kv2.1-mediated EPJs where they more functionally coupled to RyRs to generate sparks. Consistent with this, we found that the spatial and functional coupling of somatic Cav1.2 channels to RyRs was reduced in Kv2.1 KO CHNs. Together, these findings indicate that LTCCs are recruited to Kv2.1-associated EPJs, a property we found was not shared by the T-type Ca 2+ channel Cav3.1. Moreover, the co-purification of several Cavb auxiliary subunits, which associate with LTCCs but not T-Type Ca 2+ channels such as Cav3.1 (Fang and Colecraft, 2011), by IP of Kv2.1 from crosslinked brain samples further suggests a specific spatial interaction of LTCCs with Kv2.1. While like other plasma membrane proteins LTCCs can also exhibit stochastic clustering (Sato et al., 2019), numerous proteins have been identified that promote clustering of LTCCs in dendritic spines, including AKAP15 (Marshall et al., 2011) and PDZ domain-containing proteins . The absence of these known LTCC clustering proteins from our proteomic analyses of proteins in close spatial proximity to Kv2.1, and our observation that expression of Kv2.1 increases Cav1.2 clustering in heterologous HEK293T cells, suggests that the proteins mediating Cav1.2 clustering in dendritic spines and at somatic EPJs may be distinct. We note that while our studies support that Kv2.1 coexpression leads to enhanced clustering of PM Cav1.2, our data do not allow us to distinguish whether this occurs through clustering of Cav1.2 channels already in the PM, or through other mechanisms, such as enhanced fusion of Cav1.2-containing endocytic vesicles that support enhanced clustering of Cav1.2 upon its reappearance in the PM, and that also leads to enhanced Cav1.2 clustering and cooperative gating (Ghosh et al., 2018).
Although the molecular mechanism of Kv2.1 recruitment to EPJs is now established, and occurs via its phosphorylation-dependent interaction with VAPs (Johnson et al., 2018;Kirmiz et al., 2018a), the precise molecular mechanism that underlies how LTCCs and RyRs are recruited to these sites is not yet clear. However, our data show that PM Cav1.2 organization was not impacted by coexpression of the clustering-and EPJ formation-deficient Kv2.1 S586A mutant as it was by Kv2.1 WT and the nonconducting Kv2.1 P404W point mutant. Additionally, Kv2.1 S586A was unable to enhance Cav1.2-and RyR-mediated sparks reconstituted in HEK293T cells, unlike these clustering-competent Kv2.1 isoforms. These findings support that Kv2.1 clustering and induction of EPJs is necessary for its spatial association with LTCCs.
It has been reported that LTCCs can also be recruited to the EPJs formed in HEK293T cells upon heterologous expression of junctophilin-2 (Perni et al., 2017), an ER-localized protein critical for bridging the PM to the ER in myocytes (Jiang et al., 2016). This is consistent with a model whereby tethering of LTCCs at or near Kv2-associated EPJs could be mediated by an intermediary recruited to Kv2.1-mediated EPJs, perhaps even one of the proteins identified in our proteomics analyses Video 7. Ca 2+ sparklets in a control HEK293T cell expressing Cav1.2. Stack of TIRF images acquired at approximately 100 Hz of a single HEK293T cell transfected with Cav1.2, PKCa, and auxiliary subunits and loaded with Fluo-5F via the patch pipette, before (left) and after (right) application of 500 nM Bay K8644. Each pixel has been normalized to its minimum pixel intensity (i.e., F/F min ). DOI: https://doi.org/10.7554/eLife.49953.024 Video 8. Ca 2+ sparklets in a HEK293T cell coexpressing Cav1.2 and Kv2.1. Stack of TIRF images acquired at approximately 33 Hz of a single HEK293T cell transfected with Cav1.2, DsRed-Kv2.1 P404W , PKCa, and auxiliary subunits and loaded with Fluo-5F via the patch pipette, before (left) and after (right) application of 500 nM Bay K8644. Each pixel has been normalized to its minimum pixel intensity (i.e., F/F min ). DOI: https://doi.org/10.7554/eLife.49953.025 (although note a different study did not observe this effect of junctophilin-2 coexpression [Dixon et al., 2015]). We note that these proteomics analyses have the potential to identify any proteins with lysine residues in close spatial proximity ( » 12 Å ) to those in Kv2.1, making them amenable to being crosslinked to Kv2.1 by DSP, and do not require their direct association. Moreover, the crosslinking reaction could potentially yield 'daisy-chained' protein linkages of spatially adjacent proteins. While any such crosslinked protein chain would need to ultimately connect back to Kv2.1 to be immunopurified, every protein present in the purified sample need not be in close spatial proximity to Kv2.1 itself. Our observation that immunolabeling of endogenous Cav1.2 and Cav1.3 channels often appeared adjacent to rather than co-occurring with Kv2.1 (e.g., see Figure 1H) may also indicate that there is an indirect interaction between Kv2.1 and LTCCs. However, it remains possible that PM Kv2s and LTCCs associate through a direct intermolecular interaction. Any domains on Kv2.1 contributing to the spatial association with LTCCs, whether it occurs via direct or indirect interaction, would likely be conserved in Kv2.2, as we found that both Kv2 channel paralogs similarly impacted LTCC cluster size and localization. It is unlikely that RyRs are directly recruited to EPJs by Kv2 channels, as RyR clusters persist in CHNs exposed to treatments that disperse Kv2.1 clusters (Misonou et al., 2005a) and while reduced in size in CA1 pyramidal neurons in the double Kv2.1/ Kv2.2 knockout (Kirmiz et al., 2018a), in general RyR clusters persist in neurons in the brains of mice lacking Kv2 channels Kirmiz et al., 2018a). Further experiments are needed to determine the molecular mechanisms and direct protein-protein interactions that result in the spatial association of these proteins at neuronal EPJs.

Kv2.1-dependent potentiation of Cav1.2 currents
Given their prominent physiological role, the regulation of LTCCs is extensive and multimodal (Lipscombe et al., 2013;Hofmann et al., 2014;Neely and Hidalgo, 2014). The mechanisms involved in the modulation of LTCC function include post-translational modification (e.g., phosphorylation), as well as changes in the expression of the subunits (principal a1, and auxiliary Cavb and a2d) that together comprise the quaternary structure of an LTCC (Catterall, 2011;Zamponi et al., 2015). We have recently demonstrated a novel mechanism for regulating Cav1.2-(and Cav1.3-) containing LTCCs, whereby LTCCs function differently when clustered due to their clustering-dependent cooperative gating (Dixon et al., 2012;Dixon et al., 2015;Moreno et al., 2016). Thus, LTCC activity is sensitive to its spatial organization in the PM, influenced by its proximity to adjacent LTCCs (Navedo et al., 2005;Navedo et al., 2010;Dixon et al., 2012;Moreno et al., 2016) and also to its localization to specific neuronal compartments (Hall et al., 2013;Tseng et al., 2017). In neurons,  such regulation likely acts to ensure that Cav1.2 is most active when properly targeted to specific subcellular domains and less active when outside these regions. Here, we show that the subcellular localization and activity of somatic Cav1.2 channels are influenced by Kv2.1, which increases both Cav1.2 clustering and its opening at polarized V m values. At least two other proteins, a-actinin (Hall et al., 2013) and densin-180 (Wang et al., 2017), exert a similar dual regulation on neuronal Cav1.2, by promoting its localization to dendritic spines and enhancing its activity at these sites. Neither of these proteins was identified in our proteomic analyses of proteins in close spatial proximity to Kv2.1, further suggesting that Cav1.2 complexes in dendritic spines and at somatic EPJs may be distinct. The reduced whole-cell LTCC currents and impaired association of somatic Cav1.2 with RyRs in Kv2.1 KO CHNs suggests that Kv2.1 serves this dual targeting/modulation function for LTCCs within the soma and proximal dendrites.
In both CHNs and HEK293T cells, currents resulting from the opening of endogenous and exogenous Cav1.2 channels, respectively, are increased in the presence of Kv2.1. In HEK293T cells, Cav1.2 channels coexpressed with clustered Kv2.1 are activated at more polarized V m values relative to those produced by Cav1.2 alone. Moreover, coexpression of Cav1.2 with nonconducting Kv2.1 increases the frequency of spontaneous Cav1.2 channel openings in HEK293T cells as reflected in an increased frequency of Ca 2+ sparklets. The Kv2.1-dependent increase in whole-cell Cav1.2 current amplitude in both HEK293T cells and CHNs occurs without an apparent change in the number of Cav1.2 channels present on the PM, as total Cav1.2 on-gating charges were unaltered by coexpression with Kv2.1. Instead, it appears that coupling of Cav1.2 voltage sensor movement to channel opening is enhanced in the presence of Kv2.1. What is the molecular mechanism underlying this effect on Cav1.2 channel opening? We suggest three possibilities. First, the increase in I Ca and leftward shift in the voltage-dependence of activation that we observed upon coexpression of Kv2.1 in HEK293T cells are similar to those observed during optogenetic induction of Cav1.2 channel oligomerization (Navedo et al., 2010;Dixon et al., 2012). Thus, one possible mechanism is that Kv2.1induced clustering at EPJs increases the probability of physical interactions between Cav1.2 channels, which promotes their cooperative gating. A second possibility is that Kv2.1 functions as an auxiliary voltage sensor for Cav1.2 channels, perhaps through a direct intermolecular interaction of the two channels. However, the apparent localization of many Cav1.2 clusters adjacent to rather than directly overlapping with Kv2.1 clusters in CHNs (e.g., see Figure 1B,H) suggests that although these proteins associate in close spatial proximity, there may not be a direct interaction between individual Kv2.1 and Cav1.2 channels.
A third potential explanation for the Kv2.1-mediated increase in Cav1.2 channel activity is that Cav1.2 is modulated by signaling molecules that it encounters when recruited to EPJs by Kv2.1. It is well established that phosphorylation of Cav1.2 is a major mechanism to regulate its activity. Phosphorylation by protein kinase A (PKA) increases Ca 2+ influx through Cav1.2, enhancing CICR . Another candidate which might impact Cav1.2 at EPJs is Ca 2+/ calmodulindependent protein kinase II (CaMKII), which has also been shown to interact with Kv2.1 (McCord and Aizenman, 2013). Enhanced Cav1.2 opening at polarized V m values and increased open probability are produced by both PKA- (Tsien et al., 1986;Bers and Perez-Reyes, 1999) and CaMKII- (Erxleben et al., 2006;Blaich et al., 2010) dependent phosphorylation of Cav1.2. Moreover, given the well-established association of RyRs with PKA and CaMKII (Zalk et al., 2007), it is conceivable that RyRs, Cav1.2, and Kv2.1 are substrates of these protein kinases at somatic EPJs. A recent study showed that in dendritic EPJs adjacent to spines, Cav1.2 is inhibited through a direct interaction with the ER-localized protein stromal interaction molecule 1 (STIM1) in a negative feedback response to Cav1.2-and RyR-mediated CICR . As such the Kv2.1-mediated localization of Cav1.2 at EPJs may bring it in close proximity to numerous regulatory molecules, at least a subset of which should also be expressed in HEK293T cells as these also exhibit prominent effects of Kv2.1 clustering on Cav1.2 activity.

Properties of Ca 2+ sparks at Kv2.1-associated EPJs
The results presented here indicate that Ca 2+ sparks occurring at Kv2.1-associated EPJs were triggered primarily by Ca 2+ influx through LTCCs initiating the opening of juxtaposed RyRs. Accordingly, Ca 2+ spark frequency increased when neurons were exposed to Cav1.2 channel agonists and decreased by blockade of LTCCs. Loss of Kv2.1 expression was also associated with a decrease in Ca 2+ spark frequency, likely because of decreased spatial association of Cav1.2 and RyRs, decreased RyR cluster size, and decreased LTCC currents.
As Kv2.1 clusters intrinsically represent EPJs by nature of their formation via an interaction with ER-resident VAPs (Johnson et al., 2018;Kirmiz et al., 2018a;Kirmiz et al., 2018b), our observation that localized Ca 2+ signals occurred only at a subset of GCaMP3-Kv2.1 clusters suggests that only a subset of Kv2.1-associated EPJs possess the molecular machinery required to generate these Ca 2+ signals. As identified in electron micrographs, EPJs represent a major class of membrane contact sites in brain neurons, where > 10% of the somatic PM may be engaged in EPJs (Wu et al., 2017). In addition to those formed by the Kv2.1:VAP association (Johnson et al., 2018;Kirmiz et al., 2018a;Kirmiz et al., 2018b), EPJs can be organized by a set of ER membrane proteins that bind PM phospholipids (Henne et al., 2015;Gallo et al., 2016). Experiments in heterologous cells exogenously expressing these ER-PM junction components show these ER tethers can also participate in ER-PM junctions formed by Kv2.1-VAP association (Kirmiz et al., 2018b). However, the relationship of the Kv2.1, LTCC and RyR-containing EPJs described here to those formed by these ER tethers in brain neurons and other cells that endogenously express these proteins is not known.
Our findings indicate that Kv2.1-mediated somatodendritic EPJs provide a molecular platform to elevate local Ca 2+ at individual EPJs without an increase in global Ca 2+ , but that can also contribute to global, action potential-induced increases in cytoplasmic Ca 2+ . These results reinforce previous observations (Berrout and Isokawa, 2009;Manita and Ross, 2009;Miyazaki et al., 2012;Miyazaki and Ross, 2013) that hippocampal neurons possess the molecular machinery to produce spontaneous local elevations in somatodendritic Ca 2+ that could potentially impact a wide variety of signaling pathways. That sparks can occur independently in neighboring Kv2.1-containing EPJs suggests a mechanism for compartmentalized Ca 2+ signaling in the aspiny regions of neurons (somata, proximal dendrites, axon initial segment) in which Kv2.1 clusters are located. One specific role identified for Ca 2+ signals produced by somatic RyR receptors at EPJs is in cartwheel cells (inhibitory interneurons found in the dorsal cochlear nucleus), where they trigger rapid gating of BK Ca 2+ -activated K + channels to control electrical excitability (Irie and Trussell, 2017). While this mode of BK channel activation has not been observed in CA1 pyramidal neurons (Ross, 2012), somatic LTCC-and RyRmediated Ca 2+ release has recently been demonstrated to activate KCa3.1 channels in hippocampal neurons, reducing spike frequency (Sahu et al., 2019). Sparks at Kv2.1-associated EPJs might also influence electrical activity in pyramidal cells through Ca 2+ -sensitive enzymes that modify ion channel function, such as protein kinases and phosphatases that influence their phosphorylation state (Misonou et al., 2004). In addition, a role for somatic Ca 2+ sparks has been identified in DRG neurons, where they promote non-synaptic exocytosis of ATP-loaded secretory vesicles (Ouyang et al., 2005). Whether Ca 2+ entry mediated by LTCCs and RyRs at Kv2.1-associated EPJs impacts secretory vesicle exocytosis in brain neurons or other cell types will need to be investigated in future studies.

Potential impact on downstream signaling pathways
Somatodendritic LTCCs are preferentially coupled to activation of signaling pathways resulting in changes in gene expression (Wheeler et al., 2012;Wild et al., 2019). In sympathetic neurons, local Ca 2+ influx through LTCCs rather than bulk elevation of intracellular Ca 2+ efficiently activates the transcription factor cAMP response element-binding protein (CREB) (Wheeler et al., 2008) through a mechanism that involves a signaling complex containing components of a PM-to-nucleus Ca 2+ shuttle Ma et al., 2014;Cohen et al., 2015). Moreover, somatic LTCCs play a unique role in the Ca 2+ influx that leads to activation of the NFAT transcription factor (Wild et al., 2019). The results presented here suggest that Kv2.1-mediated organization and regulation of somatic LTCCs provides a molecular mechanism to control local Ca 2+ influx and serve as an organizer of Ca 2+ signaling microdomains. Previous work from us (Misonou et al., 2004;Misonou et al., 2005b) and others (Mulholland et al., 2008;Aras et al., 2009) has shown that acute ischemic or depolarizing events lead to Ca 2+ -dependent dispersal of Kv2.1 clusters and hyperpolarize its V m activation threshold, potentially as a homeostatic mechanism to reduce neuronal activity and Ca 2+ overload that can lead to excitotoxicity. In our experiments here, we determined that Kv2.1-mediated clustering was associated with enhanced functional coupling of Cav1.2 and RyRs, as well as increased activation of Cav1.2 at polarized V m values. Therefore, the Ca 2+ -dependent dispersal of Kv2.1 clusters and the resulting dissociation of Cav1.2 and RyRs may represent a negative feedback loop to limit excessive increases in cytoplasmic Ca 2+ . By decreasing LTCC-and RyR-mediated CICR, dispersal of Kv2.1 clusters may help to curb excessive accumulation of intracellular Ca 2+ , which inappropriately activates signaling pathways contributing to neuronal damage or death (Dirnagl et al., 1999). Activity-dependent declustering of Kv2.1 may also help to reduce currents conducted by LTCCs, both through increased activation of hyperpolarizing Kv2.1 currents at polarized V m (opposing activation of voltage-gated Ca 2+ channels) and also through limiting Cav1.2 activity by altering its spatial organization in the PM. Our findings may also contribute to an understanding of the pathogenic mechanisms underlying mutations in Kv2.1 predicted to selectively disrupt the PRC domain required for Kv2.1 clustering (de Kovel et al., 2017).
Overall, the findings presented here identify a molecular structure underlying the spontaneous somatodendritic Ca 2+ signals previously observed in hippocampal pyramidal neurons. While our live cell experiments were primarily confined to CHNs cultured for 9-15 days in vitro, our data indicate that the spatial association of Kv2.1, Cav1.2, and RyRs is preserved in intact adult mouse and rat brains and can be recapitulated in heterologous cells. Moreover, somatodendritic Ca 2+ sparks have been observed in acute hippocampal slices obtained from rats aged P3-P80 (Miyazaki et al., 2012), suggesting that these Ca 2+ release events serve functional roles that emerge early in pyramidal neuron development and continue beyond this period. Although it is unclear whether spontaneous Ca 2+ sparks serve a specific function at their site of generation, or if they instead reflect stochastic events whose primary impact lies in their group behavior (i.e., through modulation of bulk cytosolic Ca 2+ ), the results described here have relevance to obtaining a better understanding of their generation as well as their downstream effects.

Animals
All procedures involving rats and mice were approved by the University of California, Davis Institutional Animal Care and Use Committee and performed in accordance with the NIH Guide for the Care and Use of Laboratory Animals. Animals were maintained under standard light-dark cycles and allowed to feed and drink ad libitum. Sprague-Dawley rats were used for immunolabeling experiments and as a source of hippocampal neurons for primary culture. Kv2.1 KO mice (RRID:IMSR_MGI: 3806050) (Jacobson et al., 2007;Speca et al., 2014) were generated from breeding of Kcnb1 +/mice that had been backcrossed on the C57/BL6J background (RRID:IMSR_JAX:000664). Littermates were used when available. Adult male (mice and rats) and female (rats) were used in immunohistochemistry experiments; adult male and female mice were used in proteomics; P0-P1 mouse littermates were used as a source of hippocampal neurons for primary culture. Experiments using CHNs were performed using neuronal cultures obtained from pooling neurons from animals of both sexes (rats and mice) and also cultures in which individual pups were grouped by sex after visual inspection (mice).

HEK293T cell culture
HEK293T cells were obtained from ATCC (Cat# CRL-3216). The accompanying Certificate of Analysis shows species determination was performed and yielded the expected results. HEK293T cells were further validated by short terminal repeat (STR) analysis. Cells were tested for mycoplasma contamination using the MycoAlert Mycoplasma Detection Kit (Lonza Catalog#: LT07-318). HEK293T cells were maintained in Dulbecco's modified Eagle's medium (Gibco Cat# 11995065) supplemented with 10% Fetal Clone III (HyClone Cat# SH30109.03), 1% penicillin/streptomycin, and 1x GlutaMAX (Ther-moFisher Cat# 35050061) in a humidified incubator at 37˚C and 5% CO 2 . Cells were transiently transfected using Lipofectamine 2000 following the manufacturer's protocol, in DMEM without supplements, then returned to regular growth medium 4 hr after transfection. 20-24 hr later, cells were passaged to obtain single cells on glass bottom dishes (MatTek Cat# P35G-1.5-14 C) or microscope cover glasses (VWR Cat# 48366-227) coated with poly-L-lysine. Cells were then used for experiments approximately 15 hr after being passaged.

Immunolabeling of cells
Immunolabeling of CHNs and HEK293T cells was performed as described previously (Kirmiz et al., 2018a;Kirmiz et al., 2018b). CHNs were fixed in ice cold 4% (wt/vol) formaldehyde (freshly prepared from paraformaldehyde, Fisher Cat# O4042) in phosphate buffered saline (PBS, Sigma Cat #P3813) supplemented with 4% (wt/vol) sucrose (Sigma Cat# S9378), pH 7.4, for 15 min at 4˚C. HEK293T cells were fixed in 3.2% formaldehyde (freshly prepared from paraformaldehyde) and 0.1% glutaraldehyde (Ted Pella, Inc, Cat# 18426) prepared in PBS pH 7.4, for 20 min at room temperature    Kv2.         Table 2). Following 3 Â 5 min washes in blotto-T, cells were incubated with mouse IgG subclass-and/or species-specific Alexa-conjugated fluorescent secondary antibodies (Invitrogen) diluted in blotto-T for 45 min, then washed 3 Â 5 min in PBS. Cover glasses were mounted on microscope slides with Prolong Gold mounting medium (ThermoFisher Cat # P36930) according to the manufacturer's instructions. For cell surface immunolabeling of HEK293T cells, cells were fixed and quenched in sodium borohydride as described above, followed by 3 Â 10 min washes in PBS without Triton X-100, blocked for 1 hr in blotto-T without Triton X-100, then incubated for 2 hr in primary antibodies diluted in blotto-T without Triton X-100. Cells were then washed 3 Â 10 min in PBS without Triton X-100, followed by fixation of surface antibody with 1% formaldehyde prepared in PBS for 15 min. Cells were then washed 3 Â 5 min in PBS and processed for immunolabeling of cellular proteins as described above. For TIRF imaging of fixed cells, cover glasses were mounted in PBS onto glass depression slides. Unless otherwise stated, optical sections were acquired with an AxioCam MRm digital camera installed on a Zeiss AxioImager M2 microscope or with an AxioCam HRm digital camera installed on a Zeiss AxioObserver Z1 microscope with a 63Â/1.40 NA Plan-Apochromat oil immersion objective and an ApoTome coupled to Axiovision software (Zeiss, Oberkochen, Germany). Confocal images were acquired using a Zeiss LSM880 confocal laser scanning microscope equipped with an Airyscan detection unit and a Plan-Apochromat 63Â/1.40 NA Oil DIC M27 objective. Structured illumination microscopy (N-SIM) images were acquired with a Hamamatsu ORCA-ERCCD camera on a SIM/widefield capable Nikon Eclipse Ti microscope with an EXFO X-Cite metal halide light source and a 100 Â PlanApo TIRF/1.49 objective. Colocalization and morphological analyses of Cav1.2 and RyRs in CHNs was performed using Fiji (NIH). For the colocalization analyses, an ROI was drawn around the soma of a neuron and PCC values were collected using the Coloc2 plugin. All intensity measurements were collected using Fiji. All intensity measurements reported in line scans are normalized to the maximum intensity measurement. Measurements of cluster sizes were performed essentially as previously described (Kirmiz et al., 2018a;Kirmiz et al., 2018b). Images were subjected to rolling ball background subtraction and subsequently converted into a binary mask by thresholding. Cluster sizes were measured using the 'analyze particles' feature of Fiji; nearest neighbor distances were calculated from cluster centroid values using the nearest neighbor distance plugin in Fiji. The spatial distributions of Kv2.1 and Cav1.2 puncta were analyzed using the Interaction Analysis function that is part of the MosaicSuite plugin for Fiji. For presentation, images were exported as TIFFs and linearly scaled for min/max intensity and flattened as RGB TIFFs in Photoshop (Adobe).

Immunolabeling of brain sections
Following administration of pentobarbital to induce deep anesthesia, animals were transcardially perfused with 4% formaldehyde (freshly prepared from paraformaldehyde) in 0.1 M sodium phosphate buffer pH 7.4 (0.1 M PB). Sagittal brain sections (30 mm thick) were prepared and immunolabeled using free-floating methods as detailed previously (Rhodes et al., 2004;Speca et al., 2014;Bishop et al., 2015;Palacio et al., 2017). Sections were permeabilized and blocked in 0.1 M PB containing 10% goat serum and 0.3% Triton X-100 (vehicle) for 1 hr at RT, then incubated overnight at 4˚C in primary antibodies ( Table 2) diluted in vehicle. After 4 Â 5 min washes in 0.1 M PB, sections were incubated with mouse IgG subclass-and/or species-specific Alexa-conjugated fluorescent secondary antibodies (Invitrogen) and Hoechst 33258 DNA stain diluted in vehicle at RT for 1 hr. After 2 Â 5 min washes in 0.1 M PB followed by a single 5 min wash in 0.05 M PB, sections were mounted and air dried onto gelatin-coated microscope slides, treated with 0.05% Sudan Black (EM Sciences) in 70% ethanol for 2 min (Schnell et al., 1999). Samples were then washed extensively in water and mounted with Prolong Gold (ThermoFisher Cat # P36930). Images of brain sections were taken using the same exposure time to compare the signal intensity directly using an AxioCam HRm high-resolution CCD camera installed on an AxioObserver Z1 microscope with a 10Â/0.5 NA lens, and an Apo-Tome coupled to Axiovision software, version 4.8.2.0 (Zeiss, Oberkochen, Germany). Labeling intensity within stratum pyramidale and stratum radiatum of hippocampal area CA1 was measured using a rectangular region of interest (ROI) of approximately 35 mm x 185 mm. Labeling intensity within stratum granulosum and the inner third of stratum moleculare of the dentate gyrus (DG) was measured using a rectangular ROI of approximately 48 mm x 200 mm. To maintain consistency between samples, the average pixel intensity values of ROIs from CA1 were acquired near the border of CA1 and CA2, and those from DG were obtained near the center of the dorsal/suprapyramidal blade of the DG. Signal intensity values from all immunolabels and of Hoechst dye were measured from the same ROI. Background levels for individual labels were measured from no primary controls for each animal and subtracted from ROI values. High magnification confocal images of rat and mouse hippocampus were acquired using a Zeiss LSM880 confocal laser scanning microscope equipped with an Airyscan detection unit and a Plan-Apochromat 63Â/1.40 NA Oil DIC M27 objective.

Immunopurification of Kv2.1 and proteomics
Crosslinked mouse brain samples for immunopurification were prepared as previously described (Kirmiz et al., 2018a). Excised brains were homogenized over ice in a Dounce homogenizer containing 5 mL of ice-cold homogenization and crosslinking buffer (in mM): 320 sucrose, 5 NaPO 4 , pH 7.4, supplemented with 100 NaF, 1 PMSF, protease inhibitors, and 1 DSP (Lomant's reagent, Thermo-Fisher Cat# 22585). Following a 1 hr incubation on ice, DSP was quenched with 20 mM Tris, pH 7.4 (JT Baker Cat# 4109-01 [Tris base]; and 4103-01 [Tris-HCl]). 2 mL of this homogenate was then added to an equal volume of ice-cold 2x radioimmunoprecipitation assay (RIPA) buffer (final concentrations): 1% (vol/vol) TX-100, 0.5% (wt/vol) deoxycholate, 0.1% (wt/vol) SDS, 150 NaCl, 50 Tris, pH 8.0 and incubated on a tube rotator at 4˚C for 30 min. Insoluble material was then pelleted by centrifugation at 12,000 Â g for 10 min at 4˚C. The supernatant was incubated overnight at 4˚C with the anti-Kv2.1 rabbit polyclonal antibody KC (Trimmer, 1991). Following this incubation, we added 100 mL of magnetic protein G beads (ThermoFisher Cat# 10004D) and incubated the samples on a tube rotator at 4˚C for 1 hr. Beads were then washed 6x following capture on a magnet in ice-cold 1x RIPA buffer, followed by four washes in 50 mM ammonium bicarbonate (pH 7.4). Proteins captured on magnetic beads were digested with 1.5 mg/mL trypsin (Promega Cat# V5111) in 50 mM ammonium bicarbonate overnight at 37˚C. The eluate was then lyophilized and resuspended in 0.1% trifluoroacetic acid in 60% acetonitrile. Proteomic profiling was performed at the University of California, Davis Proteomics Facility. Tryptic peptide fragments were analyzed by LC-MS/MS on a Thermo Scientific Q Exactive Plus Orbitrap Mass spectrometer in conjunction with a Proxeon Easy-nLC II HPLC (Thermo Scientific) and Proxeon nanospray source. Digested peptides were loaded onto a 100 mm x 25 mm Magic C18 100 Å 5U reverse phase trap where they were desalted online, then separated using a 75 mm x 150 mm Magic C18 200 Å 3U reverse phase column. Peptides were eluted using a 60 min gradient at a flow rate of 300 nL per min. An MS survey scan was obtained for the m/z range 350-1600; tandem MS spectra were acquired using a top 15 method, where the top 15 ions in the MS spectrum were subjected to HCD (High Energy Collisional Dissociation). Precursor ion selection was performed using a mass window of 1.6 m/z, and normalized collision energy of 27% was used for fragmentation. A 15 s duration was used for the dynamic exclusion. MS/MS spectra were extracted and charge state deconvoluted by Proteome Discoverer (Thermo Scientific). MS/MS samples were then analyzed using X! Tandem (The GPM, thegpm.org; version Alanine (2017. 2. 1.4)). X! Tandem compared acquired spectra against the UniProt Mouse database (May 2017, 103089 entries), the cRAP database of common proteomic contaminants (www.thegpm.org/crap; 114 entries), the ADAR2 catalytic domain sequence, plus an equal number of reverse protein sequences assuming the digestion enzyme trypsin. X! Tandem was searched with a fragment ion mass tolerance of 20 ppm and a parent ion tolerance of 20 ppm. Variable modifications specified in X! Tandem included deamidation of asparagine and glutamine, oxidation of methionine and tryptophan, sulfone of methionine, tryptophan oxidation to formylkynurenin of tryptophan and acetylation of the N-terminus. Scaffold (version Scaf-fold_4.8.4, Proteome Software Inc, Portland, OR) was used to validate tandem MS-based peptide and protein identifications. X! Tandem identifications were accepted if they possessed -Log (Expect Scores) scores of greater than 2.0 with a mass accuracy of 5 ppm. Protein identifications were accepted if they contained at least two identified peptides. The threshold for peptide acceptance was greater than 95% probability. Data in Table 1 are presented as spectral counts over three independent experiments, normalized to spectral counts for Kv2.1 peptides returned in each experiment.

Live cell imaging
HEK293T cells transfected with RyR2-YFP, LTCC a1 subunit (Cav1.2 or Cav1.3s), Cavb3, Cava 2 d 1 , STAC1, and empty vector control (pcDNA3) or DsRed-Kv2.1 P404W plasmids in a 1.5:1:0.5:0.5:0.25:1 ratio were seeded to glass bottom dishes (MatTek Cat# P35G-1.5-14 C) approximately 15 hr prior o recording. Total internal reflection fluorescence (TIRF) and widefield microscopy imaging of HEK293T cells and DIV9-10 (transfected with GCaMP3-Kv2.1) or DIV14-21 (loaded with Cal-590 AM) CHNs cultured on glass-bottom dishes was performed in KRB at 37˚C as previously described (Kirmiz et al., 2018a;Kirmiz et al., 2018b). For imaging of cells loaded with Ca 2+ -sensitive dye, cells were first incubated in regular culture medium to which had been added 1.5 mM Cal-590 AM (AAT Bioquest Cat# 20510) for 45 min or Fluo-4 AM (Invitrogen Cat# F14201) for 25 min at 37˚C. Dye-containing medium was then aspirated, followed by two washes in KRB which had been warmed to 37˚C. Cells were then incubated in KRB for an additional 30 min at 37˚C prior to imaging. Caffeine (Sigma Cat# C0750), thapsigargin (Millipore Cat# 586005), nimodipine (Alomone Cat# N-150), Bay K8644 (Alomone Cat# B-350), and tetracaine (Sigma Cat# T7508) were dissolved in warm KRB at 2x the final concentration and added to cells during imaging by pipette. For GxTX-633 labeling of cells, cells were incubated in 300 nM GxTX-633 dissolved in KRB supplemented with 0.1% BSA for 20 min at 37˚C, followed by a single wash with KRB. Images were acquired on a Nikon Eclipse Ti TIRF/widefield microscope equipped with an Andor iXon EMCCD camera and a Nikon LUA4 laser launch with 405, 488, 561, and 647 nm lasers, using a 100Â/1.49 NA PlanApo TIRF objective and NIS Elements software. For post-hoc immunolabeling of CHNs, the dish orientation and location of the imaged cell was recorded, after which the CHNs were fixed in ice-cold 4% formaldehyde/4% sucrose in PBS, pH 7.4, and processed for immunolabeling as described above. Recorded CHNs were identified on the basis of expression of GCaMP3-Kv2.1 and/or neurite morphology revealed by immunolabeling for MAP2. Acquired image stacks were processed and analyzed using Fiji; we used the Fiji plugin xySpark (Steele and Steele, 2014) for automated spark detection and analysis.
Electrophysiology HEK293T cells transfected with Cav1.2-GFP, Cavb3, Cava 2 d 1 , and empty vector control (pcDNA3) or DsRed-Kv2.1 P404W plasmids in a 1:0.5:0.5:1 ratio were seeded to microscope cover glasses (Fisher Cat# 12-545-102) approximately 15 hr prior to recording to obtain single cells. Coexpression of Cav1.2 and Kv2.1 P404W in HEK293T cells was apparently cytotoxic and thus necessitated seeding of cells at a higher density to obtain viable single cells as compared to control cells expressing Cav1.2 alone. HEK293T cells were patched in an external solution of modified Krebs-Ringer buffer (KRB) containing (in mM): 146 NaCl, 4.7 KCl, 2.5 CaCl 2 , 0.6 MgSO 4 , 1.6 NaHCO 3 , 0.15 NaH 2 PO 4 , 8 glucose, 20 HEPES, pH 7.4, approximately 330 mOsm. Transfected cells were identified by the presence of GFP and DsRed expression. I Ca was recorded in transfected cells using the whole-cell voltage clamp patch configuration using fire-polished borosilicate pipettes that had resistances of 2-3 MW when filled with an internal solution containing (in mM): 125 Cs-methanesulfonate, 10 TEA-Cl, 1 MgCl 2 , 0.3 Na 2 -GTP, 13 phosphocreatine-(di)Tris, 5 Mg. ATP, 5 EGTA, 10 HEPES, adjusted to pH 7.22 with CsOH, approximately 320 mOsm. Currents were sampled at 20 kHz and low-pass-filtered at 2 kHz using an Axopatch 200B amplifier, and acquired using pClamp 10.2 software (Molecular Devices, Sunnyvale, CA). All experiments were performed at room temperature (22-25˚C). Pipette capacitance was compensated using the amplifier, and capacitance and ohmic leak were subtracted online using a P/5 protocol. Current-voltage (I-V) relationships were obtained approximately three minutes after obtaining the whole-cell configuration by subjecting cells to a series of 300 ms depolarizing pulses from the holding potential of À70 mV to test potentials ranging from À60 to +100 mV in 10 mV increments. The voltage dependence of G/G max was obtained from the recorded currents by converting them to conductances (G) using the equation G = I Ca /(test pulse potential -E rev(Ca) ), plotting the normalized values (G/G max ) versus the test potential, and fitting them to a Boltzmann function. Steady-state inactivation was measured by subjecting cells to a series of 2500 ms conditioning prepulses from the holding potential to potentials ranging from À60 to +100 mV, returning to the À70 mV holding potential for 5 ms, then measuring the peak current elicited by a 300 ms step to the À20 mV test potential. Data were analyzed and plotted using Prism software (Graphpad Software Inc, San Diego, CA). For experiments in which depolarization-induced increases in Ca 2+ -sensitive dye were measured, we included 0.2 mM Rhod-2 (AAT Bioquest Cat# 21068) in the patch pipette solution. Images were acquired at 10 Hz using a through-the-lens TIRF microscope built around an Olympus IX-70 inverted microscope equipped with an oil-immersion ApoN 60Â/1.49 NA TIRF objective and an Andor iXON CCD camera using TILLvisION imaging software (TILL Photonics, FEI, Hillsboro, OR).
To measure gating and ionic tail currents, we first determined the reversal potential for I Ca from the I-V relationship obtained using the I-V protocol described above. Gating currents were then measured by applying a series of depolarizing steps from the holding potential (À70 mV) to potentials ± 5 mV of the reversal potential in 1 mV increments. Currents were sampled at a frequency of 25 kHz and low-pass filtered at 2 kHz. We first obtained recordings in cells perfused with KRB alone, then obtained recordings from the same cell after it had been perfused for two minutes with KRB containing 1 mM nitrendipine (Alomone Cat# N-155). To isolate gating currents and I tail produced by Cav1.2, we subtracted currents measured in the presence of nitrendipine from those measured in KRB alone. The on-gating charge (Q on ) was then obtained from these records by integrating the gating current within approximately 2 ms of a depolarizing step to the reversal potential, and maximal I tail amplitudes were measured upon repolarization to the holding potential.
Somatic whole-cell patch clamp recordings were acquired from WT and Kv2.1 KO mouse CHNs cultured on microscope cover glasses after 15-16 DIV. Pyramidal neurons were selected based upon their morphological characteristics (Benson et al., 1994). Patch pipettes were fashioned and filled with intracellular recording solution as described above. After establishing the whole-cell configuration in KRB, the bath solution was exchanged with an extracellular recording buffer containing (in mM): 135 NMDG, 30 TEA-Cl, 5 BaCl 2 , 8 glucose, 20 HEPES, adjusted to pH 7.4 with HCl. Series resistance was 9.9 ± 0.9 (WT) and 10.4 ± 0.9 (Kv2.1 KO) MW (p=0.694, Student's t-test) (before compensation); cell capacitance was 52.9 ± 4.8 (WT) and 58.4 ± 4.0 (Kv2.1 KO) pF (p=0.789, Student's ttest). Prior to recording, cell capacitance was canceled, and series resistance was partially (60-70%) compensated. Recordings of LTCC ionic and gating currents were then performed as described for HEK293T cells. We used 10 mM nimodipine to isolate the contribution of LTCCs to the measured currents.
For simultaneous measurement of the V m and Ca 2+ sparks, rat CHNs transfected with GCaMP3-Kv2.1 were recorded using the whole-cell perforated patch clamp configuration. CHNs were patched in KRB using pipettes filled with a solution containing (in mM): 135 K-gluconate, 15 KCl, 5 NaCl, 1 MgCl 2 , 0.1 EGTA, 10 HEPES, pH adjusted to 7.22 using KOH, and amphotericin B (Millipore Cat# 171375) dissolved in DMSO and added at a final concentration of approximately 50 mg/mL. Upon obtaining a GW seal, the amplifier was switched to the current clamp mode to record spontaneous fluctuations in the V m . Measurement of the V m (sampled at 25 kHz) and widefield image acquisition (acquired at 5 Hz) were triggered simultaneously using the same microscope described above.

Experimental design and statistical analysis
For all data sets presented in this study for which statistical analyses were performed, measurements were imported into GraphPad Prism and Microsoft Excel for presentation and statistical analysis. Reported values are mean ± SEM, unless stated otherwise. Exact p-values are reported in each figure or figure legend. Paired data sets were compared using a Student's t-test if the data passed a normality test; a non-parametric test was used otherwise.Proteomics on brain samples were collected from three independent sets of age-matched male wild-type and Kv2.1 KO adult mice. For experiments involving HEK293T cells and CHNs, at least two independent cultures were used for experimentation; the number of samples (n) indicates the number of cells analyzed and is noted in each figure or figure legend. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication. Ethics Animal experimentation: This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All of the animals were handled according to approved institutional animal care and use committee (IACUC) protocols (#20485 and #21265) of the University of California, Davis. All perfusions were performed under sodium pentobarbital anesthesia, and every effort was made to minimize suffering. Data availability All data generated or analysed during this study are included in the manuscript and supporting files.