Stimulation of Ca²⁺ channel Ca_V1.2 activity by dopamine signaling augments spatial working memory

Abstract

Dopamine drives the neuronal activity and synaptic plasticity required for various forms of learning. It supports short-term working memory through activation of the D₁-like dopamine receptors D₁ and D₅. Here, we found that the L-type Ca²⁺ channel Ca_V1.2 was a critical mediator of D_1/5 signaling in hippocampal pyramidal neurons. In cultured mouse hippocampal neurons, a D_1/5 agonist augmented currents through Ca_V1.2 in the soma and Ca²⁺ influx in the dendrites. This effect was mediated through the second messenger cAMP and cAMP-dependent protein kinase (PKA), which phosphorylated the Ca_V1.2 α₁ subunit at Ser¹⁹²⁸. Ca_V1.2 and D₅ colocalized, suggesting that this signaling was spatially restricted. In mice, D_1/5 agonism facilitated spatial working memory in wild-type but not litter-matched Ca_V1.2 α₁ S1928A knockin animals. These findings identify Ca_V1.2 as a key D_1/5 signaling effector that supports dopamine-driven executive functions in cognition.

Editor’s summary

Dopamine signaling alters the strength of synaptic connections to support various cognitive functions. Man et al. found that dopamine receptors team up with Ca2+ channels in the hippocampus to mediate working memory, which supports problem solving and decision-making. In cultured mouse neurons, activation of dopamine receptors induced the phosphorylation of nearby L-type Ca2+channels by the kinase PKA, which increased Ca2+ influx. Spatial working memory was impaired in mice expressing mutant calcium channels that could not be phosphorylated by PKA. The findings provide deeper insight into the mechanisms of dopamine-supported learning. —Leslie K. Ferrarelli

INTRODUCTION

Dopamine (DA) governs motivation, mood, appetitive, aversive learning, and motor control (1, 2). Novelty stimulates DA release, which is required for reward and avoidance learning, especially when a change in the rules that predict outcome after a conditioned stimulus leads to prediction errors (3–6). Working memory is engaged when animals have to at least briefly remember novel situations for decision-making within seconds to minutes. Consistent with a role of DA in novelty learning, working memory and short-term memory are improved by dopaminergic signaling and are impaired by its inhibition (7–9). Despite the vast literature on DA’s role in learning and memory, relatively little is known about the molecular mechanisms that mediate the effects of DA in behavioral tasks.

DA regulates neuronal excitability, neurotransmitter release, postsynaptic sensitivity to neurotransmitters, synaptic plasticity, and the structural plasticity of dendritic spines (10–14). DA acts through five receptors, D₁ to D₅. D₁ and D₅ are coupled to the trimeric G protein Gα_s, which stimulates adenylyl cyclase (AC) to produce cyclic adenosine 3′,5′-monophosphate (cAMP) and activate the cAMP-dependent protein kinase (PKA). The D₂-like receptors D_2–4 are coupled to Gα_i, which inhibits AC and thereby cAMP production and PKA activity. Hence, DA controls neuronal properties through various effector proteins, including ion channels and neurotransmitter receptors, in different brain regions. For instance, DA modulates currents through N-type, P/Q-type, and L-type Ca²⁺ channels (LTCCs) in striatum (15), prefrontal cortex (PFC) (16, 17), and dorsal root ganglion neurons (18). D₁ interacts with N-type Ca²⁺ channels and N-methyl-D-aspartate (NMDA)–type glutamate receptors (NMDARs) (19) and D₅ with γ-aminobutyric acid type A (GABA_A) receptors (20) to regulate their activities. D₁-like and D₂-like receptors bidirectionally regulate NMDARs by modulating NMDAR currents, surface insertion, and Ca²⁺ influx through NMDARs (21–25). In vivo, some evidence indicates that the D₁-NMDAR interaction is relevant for working memory (7).

DA projections originate from the ventral tegmental area (VTA) and substantia nigra pars compacta (SNc) (1, 2). The VTA sends DA projections to the ventral striatum and the forebrain, constituting the mesolimbic and mesocortial pathways, respectively, which are important for reward processing and reinforcement learning (26). SNc neurons project mainly to the dorsal striatum, constituting the nigrostriatal pathway, which functions in action selection and movement control (1, 27).

The VTA also sends DA projections to the dorsal hippocampus. The hippocampus is essential for spatial learning. It harbors place cells, which fire at specific locations. During spatial learning, the place fields of some place cells rapidly remap in accordance to changes in reward locations (28, 29), which is modulated by VTA activity (30). DA projections from the VTA terminate in the cornu ammonis area 1 (CA1) region in the stratum oriens (SO), stratum pyramidale, and stratum radiatum (SR) in the hippocampus (31, 32). These projections have been implicated in learning and memory in rodents (31–33) and humans (34–36). In rodents, midbrain to dorsal CA1 (dCA1) projections mediate both appetitive and aversive spatial learning (32, 33). The locus coeruleus (LC), which typically sends noradrenergic projections, also contributes to dopaminergic regulation of dCA1 likely by coreleasing DA, which is synthesized in LC neurons as a precursor for norepinephrine (37, 38). In dCA1, catecholaminergic axons from the LC are denser than those from the VTA (39, 40) and are localized to the SO, SR, and stratum lacunosum-moleculare (39). The optical stimulation of the LC-dCA1 pathway enhanced performance in spatial learning tasks including the location of a food reward through D_1/5 signaling in dCA1 (39, 40).

The formation of CA1 place cells during learning of reward locations is mediated by dendritic plateau phases, depolarizations that drive complex dendritic spiking and Ca²⁺ influx (41, 42). These plateau phases can induce, within one or two trials, a synaptic potentiation termed behavioral timescale synaptic plasticity (BTSP). Place cell reorganization during changes in reward locations depends on dopaminergic input from the VTA (43) and the potentially dopaminergic LC input (44). BTSP requires NMDAR and LTCC activity (42). In hippocampal cultures, NMDAR activation can drive dendritic Ca²⁺ plateau phases that closely resemble those in BTSP and are mediated, in part, by LTCCs (45, 46). Collectively, these findings raise the intriguing possibility that the dopaminergic regulation of the LTCC Ca_V1.2 is involved in spatial learning.

Ca_V1.2 fulfills a large spectrum of functions. For instance, several mutations in Ca_V1.2 cause Timothy syndrome, which results in autism-like symptoms, congenital heart disease, dactyl webbing, and immune deficiency (47). Ca_V1.2 has also been implicated in the filopodium formation of invasive cancer cells (48). Ca_V1.2 accounts for ~80% of all LTCCs in the brain (49, 50). It governs the heartbeat, vascular tone, and neuronal functions that include long-term potentiation (51–55), long-term depression (56, 57), neuronal excitability (58, 59), and gene expression (60–66). Analysis of Ca_V1.2 mutant mice implicates this channel in anxiety disorders, depression, and self-injurious behavior (49). Mutations and variations in the CACNA1C gene, which encodes Ca_V1.2, constitute major risk factors for schizophrenia, bipolar disorder, major depression, attention deficit hyperactive disorder, and autism spectrum disorders (47, 67–70). Consistently, inhibitors of LTCCs induce antidepressant effects, whereas LTCC agonists elicit depression-like behavior (71, 72) and self-biting in mice (73).

Ca_V1.2 forms a signaling complex with the β₂-adrenergic receptor (β₂AR), one of the receptors for the attention-mediating neuromodulator norepinephrine. This complex also contains Gα_s, AC, and PKA (64, 65, 74–80). The stimulation of the β₂AR enhances Ca_V1.2 channel activity in neurons by inducing the phosphorylation of the pore-forming α₁ subunit α₁1.2 at Ser¹⁹²⁸ by PKA in a highly localized manner (54, 55, 75). These findings suggest that Ca_V1.2 is a key target for signaling that is driven by norepinephrine and mediated by the β₂AR-Gα_s-AC/cAMP-PKA signaling cascade.

The phosphorylation of Ca_V1.2 on Ser¹⁹²⁸ is important for various forms of spatial learning (57). DA is another highly prevalent neuromodulator in the brain that acts through D_1/5-Gα_s-AC/cAMP-PKA signaling. Given that Ca_V1.2 forms complexes with Gα_s, AC, and PKA, we hypothesized that DA exerts its effects on hippocampal function by increasing Ca_V1.2 activity. We found that Ca_V1.2 is an important target for signaling by D₁-like receptors in neurons isolated from the hippocampal CA regions. This signaling is likely primarily mediated by D₅ because D₅ resides in close proximity to Ca_V1.2 and is strongly expressed in pyramidal neurons of the CA regions of the dorsal hippocampus (81, 82), where D₁ expression is largely restricted to interneurons (83). Essential to this pathway was the phosphorylation of Ca_V1.2 on Ser¹⁹²⁸, which was requisite for the facilitation of working memory by a D_1/5 agonist during a spatial short-term learning paradigm. These findings provide mechanistic insight into how D_1/5 signaling acts in novelty-driven learning.

RESULTS

D_1/5 signaling potentiates LTCC activity

To test our hypothesis that DA exerts its effects in hippocampal function by stimulating a Ca²⁺ channel, we first tested whether stimulation of D₁-like receptors D_1/5 modulates the activity of LTCCs. Rat hippocampal cultures were prepared from the CA region after the removal of the dentate gyrus (DG) region. The single-channel recording of LTCC activity from pyramidal cells as identified by shape was performed in the cell-attached configuration as before (54, 55, 84). L-type currents were pharmacologically isolated by the inclusion of the N- and P/Q-type channel blocking ω-conotoxins (ω-CTx) GVIA and MVIIC (54, 55, 84). The D_1/5-selective agonist SKF81297 (SKF) placed in the patch pipette caused a twofold increase in overall open-state probability (NPo), which is the product of the number of channels in the patch (N) and the single-channel open probability (Po) of each individual channel (Fig. 1, A to C). The peak ensemble averaged current was slightly increased by SKF placed in the pipette compared with the control, which was not statistically significant (Fig. 1D). The potentiating effect on NPo was mediated by D_1/5 activation because the inclusion of the D_1/5 antagonist SCH23390 (SCH) in the patch pipette along with SKF completely blocked the increase in NPo (Fig. 1E). When neurons were placed in a bath solution containing SKF and subjected to cell-attached recording, NPo was not potentiated compared to control (Fig. 1, A and C). The difference in NPo between SKF in the pipette versus bath was significant (Fig. 1C). These results show that SKF activation of D_1/5 within a cell-attached membrane patch augments NPo of LTCCs within the same patch. That SKF in the bath solution failed to potentiate L-type currents indicates that this potentiation is compartmentalized inside neurons to rather small areas.

Fig. 1. A D_1/5-selective agonist potentiated LTCC activity in hippocampal cultures in the patch pipette but not bath.

(A to E) Cell-attached patch recordings were performed from hippocampal cultures at 10 to 15 DIV. Images in (A) show representative traces from cell-attached recordings, showing 10 consecutive sweeps during depolarization pulses from −80 to 0 mV for 2 s in response to vehicle (control) or SKF81297 (5 μM) in either the patch pipette or the bath. Graphs in (B to D) show ensemble average current traces, NPo of LTCCs, and peak ensemble average current amplitude, respectively, for all recordings for each condition. Graph in (E) shows NPo of LTCCs in hippocampal cultures treated with control or SKF in the pipette as in (A) to (D), alone or with the D_1/5 antagonist SCH23390 (5 μM). Bars are means ± SEM; n = number of neurons recorded in a minimum of N = 5 independent experiments; *P ≤ 0.05 and **P ≤ 0.01 by ANOVA with post hoc Holm-Šidák’s test.

The α₁AR agonist phenylephrine augmented LTCC activity (84). Notably, phenylephrine only potentiated NPo versus control when bath-applied, not when applied inside the recording pipette (fig. S1, A and C). Peak ensemble averaged current showed a trend toward increase by phenylephrine in bath compared with control, although the effect did not reach significance (fig. S1, B and D).

That phenylephrine in the pipette augments LTCC activity when in the bath but not in the pipette indicates that the α₁AR effect requires signaling by one or more components that have to be engaged at or recruited from some distance and are not readily available within the membrane patch or its immediate vicinity, contrasting with the localized D_1/5 signaling. The Gα_q-coupled α₁AR potentiated L-type currents through a cascade involving PKC and the nonreceptor tyrosine kinases Pyk2 and Src, with both tyrosine kinases binding to Ca_V1.2 (84). α₁ARs are coupled through Gα_q to the production of diacylglycerol (DAG) and inositol-1,4,5-trisphosphate (IP₃), which induce Ca²⁺ release from internal stores through IP₃ receptors. That phenylephrine did not potentiate L-type currents when applied inside the patch pipette might be because one or more components of the α₁AR-DAG-IP₃-PKC-Pyk2-Src signaling cascade were not in close enough proximity to the channel and lacking in the membrane patch. The bath effect of phenylephrine also means that the failure of up-regulation of LTCC by bath-applied SKF (Fig. 1) or β₂AR agonists (54, 55, 75) is not due to lack of access of LTCC under the patch to elements that mediate its regulation over a distance.

Potentiation of LTCC activity by D_1/5 signaling requires canonical PKA activation

Typically, D_1/5 signal through the Gα_s-AC-cAMP-PKA pathway. We tested whether this canonical signaling also mediates the potentiation of LTCCs by D_1/5 using two PKA inhibitors that block PKA by different mechanisms—the peptide protein kinase inhibitor (PKI), which is derived from the endogenous protein kinase A inhibitor and binds to the catalytic site of the catalytic PKA subunit, and the inactive cAMP analog Rp–adenosine 3′, 5′-cyclic monophosphorothioate (Rp-cAMPS), which competes with cAMP for binding to the regulatory subunit (85, 86). Preincubation with membrane-permeant 11R-PKI and Rp-cAMPS completely blocked the potentiation of LTCC activity by SKF in the pipette (Fig. 2, A to C). The peak ensemble averages of the current traces showed analogous changes (Fig. 2, B and D). Thus, D_1/5 signals through the canonical cAMP-PKA pathway to modulate LTCC activity.

Fig. 2. D_1/5-induced potentiation of LTCC activity is mediated by canonical PKA signaling.

(A to D) Cell-attached patch recordings were performed in hippocampal cultures at 10 to 15 DIV. Images in (A) show representative traces from cell-attached recordings, showing 10 consecutive sweeps during depolarization pulses from −80 to 0 mV for 2 s, assessing L-type activity in response to vehicle (control) or SKF81297 (5 μM) in the patch pipette, alone or plus one of two cell-permeant PKA inhibitors in the bath: 11R-PKI (10 μM) or Rp-cAMPS (100 μM). Graphs in (B) show ensemble average traces for all recordings for each condition. Graphs in (C) and (D) show NPo of LTCCs and peak ensemble average current amplitude, respectively, for all recordings for each condition. Bars are means ± SEM; n = number of neurons recorded in a minimum of N = 5 independent experiments; *P ≤ 0.05 and **P ≤ 0.01 by ANOVA with post hoc Holm-Šidák’s test.

Potentiation of LTCC activity by D_1/5 signaling requires phosphorylation of Ca_V1.2 on Ser¹⁹²⁸

The β₂AR enhances NPo of Ca_V1.2 by phosphorylation of Ser¹⁹²⁸ in the pore-forming α₁1.2 subunit of the channel (54, 55). Both β₂AR and D_1/5 are coupled through Gα_s to activation of AC and PKA, which mediates the increase in NPo of Ca_V1.2 by β₂AR signaling. Thus, we evaluated whether D_1/5 signaling induces Ser¹⁹²⁸ phosphorylation and thereby up-regulation of NPo. To monitor the phosphorylation of Ser¹⁹²⁸ (pSer¹⁹²⁸) of Ca_V1.2, wild-type (WT) brain slices were treated with SKF. The CA1 region was dissected and homogenized with 1% Triton X-100. α₁1.2 was immunoprecipitated from the extracts after ultracentrifugation and analyzed by immunoblotting using an antibody that is specific for pSer¹⁹²⁸, followed by probing for total α₁1.2. SKF treatment for 10 min significantly increased pSer¹⁹²⁸ compared with controls by ~70% (Fig. 3, A and B). We conclude that, similar to the β₂AR, D_1/5 signaling induces Ser¹⁹²⁸ phosphorylation.

Fig. 3. D_1/5-induced potentiation of LTCC activity requires phosphorylation of Ser¹⁹²⁸ on the pore-forming α₁-subunit of Ca_V1.2.

(A and B) Western blotting analysis for phosphorylated Ser¹⁹²⁸ Ca_V1.2 immunoprecipitated from acute forebrain slices that had been either untreated (Ctrl) or incubated with SKF81297 (5 μM) for 5 or 10 min. In (A), phospho blots were reprobed for total Ca_V1.2; Rb, rabbit. In (B), optical density of pSer¹⁹²⁸ signals was calculated relative to that for the long form of Ca_V1.2. ****P ≤ 0.0001 by one-way ANOVA with post hoc Holm-Šidák’s test (P = 0.0513, control versus SKF for 5 min). Data are from four mice analyzed in four independent experiments. (C to F) Cell-attached patch recordings were performed in DIV10 to DIV15 hippocampal cultures. Representative traces from cell-attached recordings in (C) show 10 consecutive sweeps during depolarization pulses from −80 to 0 mV for 2 s from WT and S1928A KI neurons treated with vehicle (control) or SKF81297 (5 μM) in the patch pipette. Graph in (D) shows ensemble average traces for all recordings for each condition. Graphs in (E) show NPo for all recordings for each condition. Graphs in (F) show peak ensemble average current amplitude for all recordings for each condition. Bars are means ± SEM; n = number of neurons recorded in a minimum of N = 4 independent experiments. In (E), WT was analyzed by ANOVA with post hoc Holm-Šidák’s test; *P ≤ 0.05; S1928A was analyzed by t test. In (F), both sets were analyzed by unpaired t test.

To test whether SKF potentiates Po of Ca_V1.2 using Ser¹⁹²⁸ phosphorylation, we determined the SKF effect in hippocampal cultures from WT and S1928A knockin (KI) mice, which renders this site phospho-deficient (54, 55, 87) . SKF in the patch pipette significantly increased L-type activity in hippocampal cultures from WT mice but not in S1928A KI mice (Fig. 3, C to E). The peak ensemble average current showed a near significant increase of the current by SKF in WT but not S1928A KI neurons (Fig. 3, D and F). The single-channel currents were completely abrogated by the LTCC blocker isradipine (Fig. 3E), verifying the successful pharmacological isolation of L-type currents. We conclude that Ser¹⁹²⁸ phosphorylation on Ca_V1.2 is required for augmentation of L-type activity by D_1/5 signaling in hippocampal cultures. That the S1928A mutation in Ca_V1.2 completely abrogated this augmentation also suggests that Ca_V1.3, which is also expressed in rodent hippocampal neurons (50, 88), is not a substantial participant in the up-regulation of L-type activity by D_1/5 signaling.

Potentiation of LTCC activity by D_1/5 signaling in dendrites

To test whether D_1/5 signaling augments Ca_V1.2 in dendrites, we loaded pyramidal neurons in hippocampal cultures with the Ca²⁺ sensor Cal-520, AM. Neurons were depolarized by perfusion with recording buffer in which 90 mM Na⁺ was substituted with 90 mM K⁺ (66), which induced Ca²⁺ influx that lasted for roughly 1 to 2 min (Fig. 4, A and B) (65, 66, 89). This Ca²⁺ influx was reduced by about 60% by the LTCC blocker nimodipine (Fig. 4, A to C), as expected given that non–L-type channels and especially N- and P/Q-type channels contribute to some degree to Ca²⁺ influx in dendrites. To block these channels in all subsequent experiments, hippocampal cultures were preincubated with 1 μM highly potent ω-CTx GVIA and MVIIC obtained from natural sources (80). Under these conditions, application of SKF strongly augmented Ca²⁺ influx in WT neurons (Fig. 4, D to F). This augmentation of influx was completely abrogated in S1928A KI neurons (Fig. 4, G to I).

Fig. 4. Augmentation of Ca²⁺ influx into dendrites through Ca_V1.2 by D_1/5 stimulation required phosphorylation of Ca_V1.2 on Ser¹⁹²⁸.

(A to C) Imaging analysis of Ca²⁺ influx induced by bath application of 90 mM K⁺ in DIV14 to DIV18 hippocampal cultures in the absence (CTRL) or presence of nimodipine (Nimo, 10 μM). Representative images in (A) are from before (top) and after (bottom) application of the K⁺ under the indicated condition. Left images show low magnification of neurons (scale bar, 20 μm), and right images show enlargements of the labeled fields (scale bar, 5 μm). Data in (B) show the amalgamated data of all time courses of changes in fluorescence. Data in (C) show the calculated AUCs for all recordings. It took about 40 s for the perfusion solution to reach the neurons. Bars are means ± SEM; n = 11 to 14 neurons from N ≥ 3 independent experiments; *P ≤ 0.05, unpaired t test. (D to H) Imaging analysis as described above for (A) to (C) in WT [(D) to (F)] and S1928A KI [(G) and (H)] neurons, treated with N- and P/Q-type channel blockers of GVIA and MVIIC (1 μM; CTx), alone or with SKF81297 (5 μM). Data are from n = 8 or 9 neurons per condition from ≥3 independent experiments; *P ≤ 0.05, paired t test. (I) Amalgamated data of SKF-induced Ca²⁺ influx, in multiples of control from same neurons, for the experiments in (D) to (H). Bars represent means ± SEM; **P ≤ 0.01, unpaired t test.

D₅ resides in close proximity to Ca_V1.2

Our cell-attached recordings suggest close proximity of Ca_V1.2 to the D₁ or D₅ receptor that regulates Po. To test this notion, we performed a proximity ligation assay (PLA), which identifies protein pairs localized within 40 nm from each other (90, 91). Our hippocampal cultures were prepared mostly from the CA1–3 regions, where expression of D₁ is much lower than that of D₅. In rodent dCA1, D₁ expression is restricted to interneurons (83), whereas D₅ is strongly expressed in pyramidal neurons (81, 82). Thus, we tested whether D₅ is near Ca_V1.2 to mediate the spatially restricted regulation induced by SKF. The specificity and use of our Ca_V1.2 antibody in PLA are well established (92, 93). The specificity of the D₅ antibody had been demonstrated earlier (94) and was further validated by immunolabeling of human embryonic kidney (HEK) 293T cells exogenously expressing rat D₅, rat D₁, or human D₅ (fig. S2, A to D). This antibody was specific for rat D₅ and did not label rat D₁ or human D₅. Consistent with earlier work (50, 75), standard immunofluorescent labeling of Ca_V1.2 showed a strong, punctate labeling pattern on the somata of rat hippocampal cultures, with weaker labeling of neuronal processes (fig. S3, A and B). Colabeling of D₅ showed a very similar pattern that overlapped very well with the punctate Ca_V1.2 signals.

We systematically performed PLA analysis for Ca_V1.2 with D₅ in comparison with two other proteins present on the plasma membrane of hippocampal cultures, the Na⁺,K⁺-ATPase (Na⁺- and K⁺-dependent ATPase) Na⁺/K⁺ transporting ATPase alpha 1 (ATP1A) and the K⁺ channel K_V4.2 (fig. S3, C and D). On the basis of the maximal binding capacity (B_max) values obtained from saturating binding curves with specific radioligands, the Na⁺,K⁺-ATPase is one of the most prevalent proteins at the cell surface. B_max for its specific ligand ouabain is 15 to 30 pMol/mg total protein in dissociated neurons including hippocampal cultures (95–97). B_max for the LTCC-specific ligand isradipine (PN200–110) is 0.1 to 0.3 pMol/mg protein in the hippocampus and other brain regions (98, 99) and potentially even lower (100). B_max for the D₁/D₅-specific ligand SCH39166 is 0.02 pMol/mg protein in cortical neurons (101) and, less relevant, 0.1 to 0.45 pMol/mg protein in striatum (102), which has by far the highest density of dopaminergic receptors in the brain (103). Accordingly, the abundance of LTCCs and D₁/D₅ is two orders of magnitude below that of the Na⁺,K⁺-ATPase.

The PLA labeling density of D₅ with Ca_V1.2 at somata was much higher than that with ATP1A and that of Ca_V1.2 with K_V4.2 (Fig. 5, A and B). Two different antibodies against the α1_1.2 subunit of Ca_V1.2 were used as a positive PLA control, which indicated, as predicted, a high puncta density for the Ca_V1.2-Ca_V1.2 pair that was significantly greater than that for the Ca_V1.2-D₅ pair. Negative controls with only one primary antibody present showed negligible amounts of signals (Fig. 5, A and B). Quantification by counting PLA puncta per cell showed similar results (Fig. 5C). Accordingly, D₅ and Ca_V1.2 reside within 40 nm or less to each other, which allows localized signaling.

Fig. 5. D₅ resided in close proximity to Ca_V1.2 in hippocampal neurons.

(A) Representative images showing detection of protein proximity (<40 nm) on DIV13 hippocampal neurons by PLA. The signals are black/white inverted, showing PLA signals as black dots on a white background. Antibodies used against the protein pairs of interest are shown at the top left corner of each panel. Positive control (Pos ctrl) was performed with two different antibodies against the α₁ subunit of Ca_V1.2. Negative controls were performed by pairing the highly prevalent ATP1A and the voltage-gated K⁺ channel K_V4.2 with D₅ and Ca_V1.2, respectively, and by testing the D₅ and the Ca_V1.2 antibody with omission of the other antibody. Scale bar, 20 μM. (B and C) Number of PLA puncta from cells described in (A), analyzed as per micrometer (B) and per cell (C). Only signals on the cell soma were analyzed. Bars represent means ± SEM; n = number of neurons imaged in N = 4 independent experiments, except for the negative control groups, which are from two independent experiments. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001 by ANOVA with post hoc Šidák’s test.

D_1/5 signaling enhanced spatial working memory by Ser¹⁹²⁸ phosphorylation

D_1/5 signaling has been implicated in mediating spatial navigation in the hippocampus (31–33, 39, 40) and in short-term and working memory in general (7–9) and, particularly, in the working memory version of the Morris water maze (MWM) (8). This working memory task in the MWM was used to specifically investigate the role of D_1/5-mediated regulation of Ca_V1.2 in spatial working memory (Fig. 6A). Litter-matched WT and S1928A KI mice were tested at 18 to 24 months of age because SKF significantly improved learning at that age for rats in the MWM (104) and in a working memory task (105). In addition, the partial D_1/5 agonist SKF improved spatial learning in the Barnes circular maze in 18-month-old mice (106). In the working memory version of the MWM, mice are challenged every day to learn a new location of a hidden platform, its location changing day to day but remaining the same during the four trials on each specific day. The mice used in this test had undergone a battery of behavioral tests including open field, novel object recognition, elevated plus maze, and full training in the standard MWM for 7 days, followed by a 10-day break and then 4-day reversal training, as described (57). Accordingly, these mice had thoroughly been habituated to behavioral testing and had learned the standard MWM as adults. We waited at least 3 months after the mice learned the standard MWM before testing WT and litter-matched S1928A KI mice in the working memory version of the MWM.

Fig. 6. Working memory and its augmentation by SKF81297 was impaired in S1928A KI mice.

(A) Working memory was analyzed in 18- to 24-month-old WT and litter-matched S1928A KI mice in the MWM with daily shifting locations of entry points and platform. Saline or SKF81297 (SKF; 3 mg/kg) was intraperitoneally injected 15 min before the test. Violin plots show path efficiency of individual mice in trial 1 compared with trial 4. N = 31 WT and 30 KI mice; ***P ≤ 0.001, unpaired t test. (B) Working memory was analyzed in 20- to 22-month-old WT and litter-matched S1928A KI mice in the Y-maze novel arm paradigm. DI indicates preference. N = 12 WT and 12 KI mice; *P ≤ 0.05, unpaired t test.

Mice were injected with either saline or SKF (3 mg/kg), a dose that had improved memory in the MWM in 24-month-old Long-Evans rats (104). We primarily monitored path efficiency, which is the quotient of the length of the path a mouse takes from the starting point to the platform divided by the straight distance between the two. Determining latency to reach the platform has proven less informative, partly because the mice reduce their swim speed with each trial and the KI mice were somewhat slower (fig. S4). Violin plots of path efficiency for trials 1 and 4 suggested that overall WT and KI mice learned the platform location over the four trials after saline and SKF administration (Fig. 6A). A t test confirmed this impression for all four test groups (table S1). To test our hypothesis that S1928A KI mice have a weaker response to SKF than WT mice, we analyzed path efficiency in the fourth trial by analysis of variance (ANOVA). Accordingly, learning was significantly impaired in KI compared with WT mice (Table 1). There was a significant interaction between genotype and drug treatment (Table 1), reflecting that SKF improved learning more strongly in WT than KI mice. That the overall drug effect for the WT and KI mice cohorts together did not reach significance (Table 1) reflects that SKF did not improve learning in the KI mice at all (Fig. 6A). This observation indicates that Ser¹⁹²⁸ phosphorylation is required for SKF to improve working memory. No significant differences were observed for sex when this variable was added to the ANOVA analysis (table S2).

Table 1.

Type III repeated-measures ANOVA for trial 4 logit-transformed outcome.

Effect	Sum square	df	F value	P value
Genotype	16.752	1	8.5840	0.0048**
Drug	3.125	1	2.5887	0.1130
Genotype:drug	6.867	1	5.6881	0.0203*

^*P ≤ 0.05.

^**P ≤ 0.01.

Analysis of variance (ANOVA) to assess the effect of genotype and the interaction between genotype and drug treatment on the performance of WT and S1928A KI mice in the fourth trial of the MWM. df, degrees of freedom.

The role of this signaling mechanism in working memory was further evaluated in the novel arm Y-maze test. Mice were first exposed for 10 min to a Y-shaped maze with one of the three arms closed. After a 4-min break, the mice were put back into the maze with all three arms now open. Mice naturally show a preference to explore the newly open arm, which requires that mice remember which arm they had explored before (57). The discrimination index (DI) is defined as the quotient of time spent in the novel arm divided by the sum of time spent in the originally open arms plus novel arm. It serves as a readout for the preference for the novel arm and thereby for working memory. Consistent with previous work (57), WT showed a tendency for preference for the novel arm upon saline injection, which was augmented by SKF in a statistically significant manner (Fig. 6B). KI mice did not show any preference for the novel arm upon saline injection and only a tendency upon SKF injection, which was statistically not different from saline (Fig. 6B). We conclude that phosphorylation of Ca_V1.2 on Ser¹⁹²⁸ plays a central role in the improvement of working memory in the MWM and Y maze upon stimulation of D_1/5 signaling.

DISCUSSION

Although D₁-like receptors potentiate L-type currents in chromaffin cells (107) and the striatum (15, 108, 109), we know very little about the role of D_1/5 receptors in regulating LTCCs in the hippocampus, a brain area that is critical for spatial learning. In the striatum, the prevalent LTCC is Ca_V1.3, suggesting that it mediates the up-regulation of L-type currents by D_1/5. In this study, D_1/5 augmented Ca_V1.2 currents in hippocampal cultures through phosphorylation of Ca_V1.2 on Ser¹⁹²⁸ by PKA, implicating specifically Ca_V1.2.

In our experiments, a D_1/5 agonist potentiated L-type currents when present inside, but not outside, the cell-attached recording pipettes. Accordingly, signaling from D_1/5 to LTCCs is tightly compartmentalized, an arrangement not realized in earlier studies (15, 107). In rats and monkeys, D₅ is strongly expressed in the CA1 to CA3 regions and the granule cell layer of DG (81, 110). Although D₁ is also expressed in granule cells in the DG, within the CA1 to CA3, it is only expressed in GABAergic neurons (83) and is not detectable in pyramidal cells (82, 83, 111). We removed the DG before culturing to focus on D₅ signaling and limit the potential contribution of D₁ in our experiments. Consistent with the spatially restricted D₅ signaling, proximity of D₅ to Ca_V1.2 within a 40-nm region was supported by PLA (90). Such close proximity augments the speed, efficacy, and selectivity of signaling by cAMP/PKA (112, 113).

That up-regulation of Ca_V1.2 requires widespread stimulation for α₁AR signaling but localized stimulation for D_1/5 and β₂AR signaling could mean that α₁AR signaling is geared toward integration of agonist availability over a much wider spatial dimension and time frame than D_1/5 and β₂AR signaling and the latter geared toward more spatially and temporally restricted regulation of Ca_V1.2. Given that the α₁AR is thought to have a higher affinity for norepinephrine than the β₂AR (114, 115), the α₁AR might be more tuned toward tonic signaling, whereas β₂AR and D_1/5 might be tuned more toward phasic signaling by DA.

D_1/5 signaling has been implicated in novelty-stimulated learning and working memory tasks (3, 5, 7–9). Because Ca_V1.2 expression in the forebrain was important for spatial learning and memory (53) and S1928A KI mice showed a modest but significant impairment in acquisition in spatial reference memory in the MWM (57), we hypothesized that Ser¹⁹²⁸ phosphorylation downstream of D_1/5 signaling is important in the early phases of learning and specifically working memory. In earlier work, D_1/5 agonists improved spatial learning especially in rodents in the MWM (104) and the Barnes maze (106) and augmented working memory (105). Thus, we tested whether SKF augments working memory and short-term learning in a modified version of the MWM in WT and S1928A KI mice. The D_1/5 agonist SKF enabled WT but not S1928A KI mice to more quickly learn how to use a more direct path in their travel to reach the safety platform. Similarly, in a less stressful Y-maze test, SKF improved working memory only in WT mice.

D_1/5 stimulation specifically in the hippocampus potentiated learning in different types of spatial tasks (31, 32, 39, 40). Our work is consistent with enhancement of spatial learning and especially working memory by D_1/5 signaling. That SKF is less effective in S1928A KI mice indicates that D_1/5 acts to promote working memory at least in part through Ser¹⁹²⁸ phosphorylation. Deficits in spatial learning and working memory have been observed in both D₅ knockout mice (116, 117) and D₁ knockout mice (118–120), implicating both receptors. It is plausible that the two DA receptors play varying roles in different brain regions. D₅-Ca_V1.2 signaling in pyramidal neurons in hippocampal CA regions is likely important for working memory because (i) pyramidal neurons in CA1–3 express mostly D₅ not D₁, (ii) D₅ augments Ca_V1.2 currents in these neurons, and (iii) spatial memory manifests itself first in these neurons (121).

Spatial working memory depends on the hippocampus and PFC, likely supported by synchrony between the regions. Specifically, the coherence of theta (122) and gamma oscillations (123) between them, phase-locking of PFC neuronal firing to hippocampal theta oscillation (124), and cross-frequency coupling between hippocampal theta and PFC gamma oscillations (125) have been suggested to support spatial working memory. Because SKF was given systemically, the effect of D_1/5 stimulation in the hippocampus versus PFC was not segregated. Also, although the D₅-Ca_V1.2 Ser¹⁹²⁸ signaling cascade was demonstrated in hippocampal neurons, this cascade may also function in other brain regions, including the PFC. The CA1 and PFC have been ascribed different roles that support spatial working memory: CA1 is thought to encode alternative choices and PFC the upcoming choice (126). It remains to be established where exactly the D₅-Ca_V1.2 Ser¹⁹²⁸ signaling cascade is specifically needed in either one of these two brain regions and other brain regions required for spatial working memory. Nonetheless, our findings that enhancement of spatial working memory by SKF is abrogated in Ca_V1.2 S1928A KI mice show the requirement of D_1/5 signaling during this task.

Although it is quite possible, if not likely, that the D_1/5-PKA-Ca_V1.2 Ser¹⁹²⁸ signaling cascade is also important during working memory at a young age, we demonstrated its function specifically in aged mice. Multiple studies have found reduced dopaminergic transmission in aged compared with young animals and humans, manifested as a reduction in dopaminergic fibers and altered DA synthesis (105, 127, 128), and in dopaminergic receptors in the hippocampus, cortex, and striatum (129–131). Dopaminergic modulation of working memory follows an inverted U-shaped curve in rodents, primates, and humans (132–134). Thus, in aged animals in which DA transmission is reduced, a manipulation to strongly increase DA signaling beyond what might be helpful in healthy young animals may be especially effective in improving cognitive functions.

Decreased DA transmission has been implicated in multiple neuropsychiatric disorders, including Alzheimer’s disease (135) and schizophrenia (136). Ca_V1.2 is a major risk factor for schizophrenia, major depression, and other mental diseases (67, 68, 70). Thus, reduced activation of the D₅-Ca_V1.2 Ser¹⁹²⁸ cascade may underlie cognitive dysfunction in these disorders. Conversely, therapeutic activation of the D₅-Ca_V1.2 Ser¹⁹²⁸ cascade may be leveraged in the future to improve cognitive function in disorders caused by reduced DA transmission.

MATERIALS AND METHODS

Animals

All procedures followed National Institutes of Health (NIH) guidelines and had been approved by the Institutional Animal Care and Use Committee at University of California, Davis (protocol numbers 20673, 22403, and 23965). Sprague-Dawley rat embryonic day 18 (E18) embryos were used to prepare dissociated neuronal cultures from hippocampi and ~3-month-old rats for tissue extracts. Rats were obtained from Harlan, Charles River Laboratories, and Envigo. Hippocampal cultures were also prepared from postnatal P0 and P1 pups from S1928A KI mice (87) and WT mice of the same genetic background (50% C57bl/6J and 50% 129). WT and S1928A KI mice were also used in the working memory version of the MWM and the Y-maze task.

Reagents and antibodies

R(+)-SKF81297 hydrobromide (catalog no. S179), R(−)-phenylephrine (P6126), and Rp-cAMPS triethylammonium salt (A165) were from Sigma-Aldrich. 11R-PKI (RRRRRRRRRRRGFIASGRTGRRNAI), which carried 11 arginine residues rendering it membrane permeant, was custom-synthesized by WM Keck Biotechnology Resource Center, Yale University (New Haven, CT) (137, 138). ω-CTx GVIA and MVIIC were obtained as synthetic peptides (China Peptides) or from biological sources (gift by B. Olivera and S. Espino, University of Utah; Ca²⁺ imaging). PLA was performed using Duolink In Situ Detection Reagents Orange (DUO92007), anti-rabbit PLUS probe (DUO92002), anti-mouse MINUS probe (DUO92004), and anti-goat MINUS probe (DUO82006, all from Sigma-Aldrich). Antibodies used for immunolabeling and PLA are the following: rabbit anti-FP1 against the loop between domains II and II in the central α₁1.2 subunit of Ca_V1.2 produced by our laboratory (139, 140) and monoclonal mouse anti-Ca_V1.2 made against the membrane proximal region of the C terminus of α₁1.2 (UC Davis/NIH NeuroMab Facility, clone N263/31, RRID: AB_2877476), anti-D₅ (MAB5292, MilliporeSigma), anti-ATP1A (ABclonal, A0643, RRID: AB_2757312), anti-Kv4.2 (N-15, Santa Cruz Biotechnology, sc-11680, RRID: AB_650411), and anti-MAP2B (BD Transduction Laboratories, 610460, RRID: AB_397833). Secondary antibodies (all from Invitrogen Molecular Probes) for immunolabeling were goat anti-rabbit-AF555 (A21121) for anti-FP1 and anti-ATP1a; goat anti-mouse IgG₁ (A21121) for D₅, donkey-anti-goat-AF488 (A11055) for Kv4.2, and goat anti-mouse IgG₁-AF647 (A21240) for MAP2B. Chemiluminescent substrates used were Immobilon Classico (Millipore) and SuperSignal West Femto (Thermo Fisher Scientific). Other reagents were from the usual suppliers and of standard quality.

Culture of primary hippocampal neurons

Hippocampal neurons were cultured from E18 embryos (both male and female) from Sprague-Dawley rats or in the case of mice, from P0 or P1 mouse pups (both male and female). The hippocampal CA region was excised from the brains of embryos in ice-cold Hank’s buffer (H2387, Sigma-Aldrich) with 10 mM Hepes, NaHCO₃ (0.35 g/liter), and gentamicin (5 μg/ml) and digested in papain (0.78 mg/ml; 10108014001, Roche Diagnostics) in 5 ml of the same buffer at 37°C for 30 min. Digested hippocampal tissue was washed with hippocampal culture medium twice and triturated in medium. The hippocampal culture medium used for washes, trituration, and culturing of neurons consists of 1× B-27 supplement, 1× GlutaMAX, 5% fetal bovine serum, and gentamicin (1 μg/ml) in neurobasal medium (all from Gibco). A total of 15,000 neurons were plated per well in 24-well plates on coverslips coated with polyornithine and laminin (Corning) and cultured in an incubator at 37°C and 5% CO₂. 5-Fluoro-2′-deoxyuridine and uridine (both 10 μM and both from Sigma) were added around 7 days in vitro (DIV7) to block the growth of glial cells.

Single-channel recording from hippocampal neurons

Cell-attached single-channel recording was performed from hippocampal cultures at DIV10 to DIV15 on an Olympus IX50 inverted microscope as before (54, 55, 84). The membrane potential was clamped at ~0 mV using a high K⁺ external solution containing 145 mM KCl, 10 mM NaCl, 10 mM Hepes, and 30 mM D-glucose (pH 7.4 with NaOH, 325 to 330 mOsM). The internal solution contained 110 mM BaCl₂, 20 mM tetraethylammonium (TEA)–Cl, 10 mM Hepes, 500 nM BayK 8644 (Tocris), and 1 μM each of ω-CTx GVIA and MVIIC (China Peptides) (pH 7.2 with TEA-OH, 325 to 330 mOsM). Pipettes with a resistance of 3.5 to 5.5 megohms were used. The D_1/5 agonist SKF (5 μM) or α₁AR agonist phenylephrine (10 μM) was added either to the patch pipette or external solution. For D_1/5 blockade, SCH23390 (5 μM) was used in the pipette solution. For inhibition of PKA activity, neurons were preincubated with cell-permeant PKA inhibitors 11R-PKI (10 μM) or Rp-cAMPS (triethylammonium salt, 100 μM) in culture medium for 10 min before the experiment, and the inhibitors were present in the external solution during recording. The measurement of currents started within 10 min of coverslips having been placed into the recording chamber. Currents were obtained at 100 kHz and low-pass filtered at 2 kHz using an Axopatch 200B amplifier, and data were digitized using Digidata 1440A (Axon Instruments). Currents were elicited by step depolarizations for 2 s per sweep from a holding potential of the patch at −80 to 0 mV with a start-to-start interval of 7 s. Usually, 100 sweeps were recorded per cell, and data were only analyzed if more than 70 sweeps were recorded. We used the single-channel search event detection algorithm of Clampfit 10.7.0.3 to determine open probabilities (NPo). Ensemble average traces were computed by averaging all sweeps from one neuron and averaging the averaged traces from all neurons in each group.

Slice preparation, immunoprecipitation, and immunoblotting

Seven- to 20-week-old mice were decapitated, and brains were placed into ice-cold slicing artificial cerebrospinal fluid (slicing ACSF; 10 mM NaCl, 230 mM sucrose, 26 mM NaHCO₂, 1.2 mM KH2PO₄, 2.5 mM KCl, 1 mM CaCl₂, 1.3 mM MgSO₄, and 10 mM D-glucose, 290 to 300 mOsm/kg, saturated with 95% O₂ and 5% CO₂; final pH 7.3). One-third of the rostral and caudal portions of the brain was trimmed off. Forebrain slices (400-μm thick) containing hippocampus were prepared with a vibratome (Leica VT 1000A) and equilibrated in ACSF (126 mM NaCl, 26 mM NaHCO₂, 1.2 mM KH2PO₄, 2.5 mM KCl, 1 mM CaCl₂, 1.3 mM MgSO₄, and 10 mM D-glucose, 290 to 300 mOsm/kg, saturated with 95% O₂ and 5% CO₂; final pH 7.3) for 1 hour at 32°C before use.

Slices were either untreated (incubated in ACSF saturated with 95% O₂ and 5% CO₂) or incubated with SKF (5 μM) for 5 or 10 min in ACSF saturated with 95% O₂ and 5% CO₂ and homogenized by trituration with insulin syringes in a 10-fold excess (volume/weight) of immunoprecipitation (IP) buffer [1% Triton X-100, 150 mM NaCl, 10 mM EDTA, 5 mM EGTA, and 50 mM tris-HCl (pH 7.4)] containing protease inhibitors [pepstatin A (10 μg/ml), leupeptin (1 μg/ml), aprotinin (2 μg/ml), and 200 nM phenylmethylsulfonyl fluoride, the latter being added immediately before homogenization and again at the start of IP], and phosphatase inhibitors (4 μM microcystin LR, 1 mM p-nitrophenyl phosphate, 25 mM Na-pyrophosphate, and 25 mM NaF) (77, 141). Nonsolubilized material was removed by ultracentrifugation (250,000g for 30 min) before IP (overnight at 4°C) with Ca_V1.2 antibody FP1 or rabbit control IgG (Thermo Fisher Scientific) using equal amounts of antibody (4 μg per sample) and protein A Sepharose. After three washes with tris-buffered saline (TBS), samples underwent SDS–polyacrylamide gel electrophoresis and were transferred onto polyvinylidene difluoride membranes. Membranes were blocked with 5% nonfat dry milk in TBS plus 0.1% Tween 20 (TBST), incubated overnight at 4°C with primary antibody against pSer¹⁹²⁸ [1:100; antibody raised and purified in our laboratory (139, 140)] in 5% nonfat dry milk in TBST, washed three times with TBST, incubated for 1 hour at room temperature with horseradish peroxidase (HRP)–conjugated secondary antibody (1:10,000 in 5% nonfat dry milk in TBST), and washed with TBST three times each for 15 min before detection of HRP signals with enhanced chemiluminescence (ECL) or ECL plus chemiluminescence reagents by film exposure. Multiple exposures of increasing length ensured that signals were in the linear range, as described (141). For reprobing, blots were stripped with 62.5 mM tris-Cl at pH 6.8, 100 mM β-mercaptoethanol, and 2% SDS at 50°C for 20 min, followed by washing with TBST three times each for 15 min. Blots were reprobed with Ca_V1.2 antibody FP1 and HRP-conjugated antibody. Signals from pSer¹⁹²⁸ were normalized to Ca_V1.2 signals after quantification by densitometry.

Ca²⁺ imaging in hippocampal neurons

DIV14 to DIV18 hippocampal cultures were treated for 1 hour with 1 μM ω-CTx GVIA and ω-CTx MVIIC. Neurons were preloaded with 5 μM calcium probe Cal-520, AM (21130, AAT Bioquest) for 5 min, followed by 10 min of washing in Tyrode buffer [150 mM NaCl, 5 mM KCl, 10 mM Hepes, 10 mM glucose, 2 mM MgCl₂, 2 mM CaCl₂ (pH 7.3 with NaOH; 310 mOsm)]. Neurons were visualized on an inverted microscope iX81 (Olympus), and image acquisition was performed on camera iXon Ultra (Andor) by recording one image every 5 s. Depolarization of neurons was induced by switching the perfusion to high K⁺ Tyrode [90 mM KCl, 50 mM NaCl, 10 mM Hepes, 10 mM glucose, 2 mM MgCl₂, and 2 mM CaCl₂ (pH 7.3 with KOH; 310 mOsm)] for 50 s. Dendritic mean fluorescence intensity was calculated by subtracting the background signal from the mean intensity of the region of interest using Fiji software (ImageJ). Relative fluorescence over time was obtained by normalizing the fluorescence intensity of each image to the baseline (average of the five images before induction). The area under the curve (AUC) was calculated above the baseline level of the curve (set to 1) using Prism 10 software (GraphPad). SKF-dependent Ca²⁺ entrance was calculated by (AUC_SKF-AUC_CTx)/AUC_CTx. One outliner was detected with the ROUT method in the S1928A panel and removed from analysis.

Immunolabeling

Pilot immunolabeling experiments were performed to determine the optimal concentration of antibodies to be used in PLA (figs. S2 and S3). Rat hippocampal cultures on 12-mm glass coverslips (DIV13 to DIV15) were placed on a parafilm-lined culture dish and gently rinsed three times with phosphate-buffered saline (PBS) pre-equilibrated in a 37°C cell culture incubator. Hippocampal cultures were fixed with 4% paraformaldehyde in 4% sucrose solution for 20 min at room temperature (RT), followed by quenching using 100 mM glycine in PBS for 20 min at RT. Hippocampal cultures were washed twice with PBS, 3 min each at RT with gentle shaking, followed by permeabilization with 0.1% Triton X-100 in PBS for 20 min at RT. Blocking was performed using Duolink blocking solution (DUO82007, MilliporeSigma) for 30 min at RT. Antibodies were diluted in the Duolink antibody diluent (DUO82008, MilliporeSigma) at the following concentrations: 1:50 Ca_V1.2 antibody, 1:50 anti-D₅, 1:100 anti-ATP1A, 1:400 anti-K_V4.2, and 1:500 anti-MAP2B. Hippocampal cultures were incubated with the antibodies overnight at 4°C and washed at RT with PBS three times each for 5 min, with gentle shaking. Secondary antibodies used were diluted at 1:1000 in the antibody diluent and incubated with the fixed hippocampal cultures for 1 hour at RT, followed by three washes with PBS each of 5 min. Coverslips were mounted with the Duolink mounting medium (DUO82040, MilliporeSigma) or with Prolong Diamond mounting medium (P36965, Invitrogen). Images were acquired with a Zeiss LSM 710 confocal microscope as Z-stacks and are shown as maximum intensity projections.

To validate the specificity of the anti-D₅ antibody, human and rat D₅ and rat D₁ were ectopically expressed in HEK293T/17 (American Type Culture Collection CRL-11268, RRID: CVCL_1926, from human) cells under the control of cytomegalovirus promoter. Human D₅, but not rat D₅ and rat D₁, was fused with green fluorescent protein (GFP) (Fig. 2, A to D). For visualization of transfection, rat D₅ and rat D₁ were cotransfected with pEGFP-N1 for enhanced EGFP (EGFP) expression (fig. S2, A to C). Two days after transfection, cells were fixed and immunolabeled as described above, except that anti-D₅ was used at 1:100.

Proximity ligation assay

PLA was performed using Duolink In Situ Detection Reagents Orange (DUO92007), anti-rabbit PLUS probe (DUO92002), anti-mouse MINUS probe (DUO92004), and anti-goat MINUS probe (DUO82006, all from Sigma-Aldrich). Hippocampal cultures were fixed on DIV13. The labeling procedure was identical to that performed in the pilot immunolabeling experiments, up to primary antibody incubation. The mouse monocolonal Ca_V1.2 antibody made against the membrane proximal region of the C terminus of α₁1.2 (UC Davis/NIH NeuroMab Facility, clone N263/31) was used at 1:200 as a positive control with FP1 antibody. Antibody pairs were diluted in the blocking solution at identical concentrations as in the immunolabeling experiments and incubated with the neurons overnight at 4°C. Hippocampal cultures were washed three times for 5 min with PBS. “+” and “−“ probe solutions were prepared and incubated with the neurons in a preheated humidified chamber for 1 hour at 37°C. Hippocampal cultures were then washed twice each for 5 min with 1× wash buffer A at RT. Ligase solution was prepared and incubated in the humidified chamber for 30 min at 37°C. After two washes each of 5 min with 1× wash buffer A at RT, a polymerase solution was added, and the hippocampal cultures were incubated at 37°C for 100 min in dark. Hippocampal cultures were then washed with 1× wash buffer B twice each for 10 min at RT, followed by a wash with 0.01× wash buffer B for 1 min. Coverslips were then mounted onto glass slides with Duolink mounting medium, stored at −20°C, and imaged within 1 week. Imaging was performed with a Zeiss LSM 710 confocal microscope or a Leica TCS SP8 X confocal microscope. PLA signals were imaged from the neuronal cell body by acquiring Z-stacks of thickness around 4 μm consisting of 12 slices using laser line 405 nm and a Cy-3 filter. The analysis of PLA signals was performed in ImageJ using the plugin 3D Objects Counter. Image stacks were loaded onto ImageJ, and the area of the cell body was selected using a bright-field image acquired along with the image stack. The number of objects in the stack in the selected area was counted in three dimensions for all image stacks using the same parameter settings (size filter: 5 to max voxels; threshold = 55).

Working memory in the MWM

Experimenters were blinded with respect to genotype of the mice. The MWM task (142, 143) was modified to evaluate spatial working memory (8, 143, 144). Mice were between 18 and 24 months old. Mice were extensively handled for at least 4 weeks and habituated to injections with at least five saline injections before any injection experiments. The MWM consisted of a water-filled round, seamless pool with a diameter of 1.2 m and water depth of 30 cm. Signs that were about 30 cm by 30 cm high and wide showed different white symbols on a black background and were mounted on the rim of the pool at the four cardinal directions to serve as conspicuous spatial cues. Nontoxic white paint was used to make the water opaque. A hidden platform was submerged (about 1 cm under the water surface) at the centroid of the testing quadrant. During the working memory–MWM procedure, mice were placed at a fixed start point on the perimeter of the maze with the platform submerged in a fixed quadrant. Mice were then allowed to swim until they found the platform or 60 s had elapsed. Mice that did not find the escape platform in the allotted time were placed on the platform for 10 s. If mice left the platform before 10 s had elapsed, then they were returned to the platform and held there for 10 s. Upon trial completion, mice were moved to a warm box filled with absorbent material for 30 s. Mice were then returned to their original start point and allowed to repeat their search for the platform. Mice repeated this procedure for a total of four trials per day before they were returned to their home cages. The experiment lasted for 6 or 9 days, during which all WT and Ser¹⁹²⁸Ala KI mice underwent the same training and drug treatment schedule. On days 1 to 3 (or days 1 and 2 for a second independent cohort), mice were familiarized with the training routine of four trials per day and daily changing platform location, and no injection was given to the mice. Mice received intraperitoneal injections of saline on days 4 to 6 (days 3 and 4 for cohort 2), and SKF (3 mg/kg) was dissolved in saline on days 7 to 9 (days 5 and 6 for cohort 2) 15 to 30 min before the first trial of each mouse. The water temperature was 27° to 29°C. The start points were at one of the four cardinal directions of the pool. The locations of the starting point and the platform remained unchanged for all four trials within the same day but changed across consecutive days of testing working memory. The order of the individual mice for the trials remained the same each day. Trials were performed between 10:00 a.m. and 4:00 p.m. for the first cohort and between 6:00 a.m. and 12:00 p.m. for the second cohort.

The trials were recorded, and the path of mice was tracked by the software AnyMaze for the first cohort and EthoVision XT (Noldus) for the second cohort. Parameters including latency to goal, path efficiency, and swim speed were extracted by the respective software. Path efficiency was defined as the straight-line distance between the first position and the last position in the trial divided by the total distance traveled by the mouse during the trial.

Y-maze novel arm recognition

Experimenters were blinded with respect to genotype of the mice, which were randomly assigned to test conditions. Mice were placed into an opaque, three-arm Y-shaped maze containing distinct intra-maze visual cues at the end of each arm. For the initial training phase, animals were allowed to explore two of the three arms for 10 min, whereas access to the third (novel arm) was blocked using an opaque separator. The starting arm, set of open arms, and cue placements were fixed across both training and test phases for each individual mouse but randomized between mice to control for arm preference and cue bias. The available arms were rotated across animals to avoid consecutive mice encountering the same novel arm. Between testing days, the arms of the maze were rotated 180°. These measures were taken to control for any unintentional spatial arm bias in relation to the environment. After initial exploration in the training phase, mice were returned to their home cage for an intertrial interval of 4 min. During this interval, the arm separator was removed and the maze floor was wiped with 70% ethanol to eliminate odor cues. Mice were reintroduced to the same starting arm as in the initial training phase and allowed to freely explore all three arms for 5 min. Mouse movement and arm entries were recorded and analyzed using EthoVision XT (Noldus). The primary outcome measure was the DI, calculated as DI = Time_Novel/(Time_Novel + Time_Familiar), which represents the cumulative time spent in the novel and familiar (other) arm, during the test phase.

Statistical analysis

All data were presented as means ± SEM. Statistical analyses were performed using Prism 5, 7, or 10 (GraphPad). For analysis of single channel data (Figs. 1 to 3) and Ca²⁺ imaging (Fig. 4), outliers were identified using the ROUT method in Prism 7 and 10, respectively. The statistical significance for NPo was analyzed using one-way ANOVA with post hoc Holm-Šidák’s test. For analysis of Ca²⁺ transients, the different pairs of groups were analyzed by either paired or unpaired t test. For analysis of the PLA assay (Fig. 5), statistical significance was analyzed using one-way ANOVA with post hoc Šidák’s test. For analysis of working memory in the MWM (Fig. 6, A to D), the statistical significance between trials 1 and 4 was analyzed by a one-sided paired t test. To examine whether the effect of drug treatment (SKF versus saline) on path efficiency in trial 4 differed by genotype (WT versus KI), we conducted a type III mixed-design ANOVA using the afex package in R. The outcome variable was the logit-transformed path efficiency score (logit Outcome), which was then averaged across repeated experimental sessions for each mouse within each drug condition (three repeated trials for each condition for cohort 1 and two repeated trials for cohort 2). In this design, drug was treated as a within-subject factor, given that each mouse received both saline and SKF treatments, allowing comparisons of drug effects within the same individuals. Genotype was treated as a between-subject factor, because each mouse belonged exclusively to either the WT or KI group.

Supplementary Material

The PDF file includes:

Figs. S1 to S4

Tables S1 and S2

Other Supplementary Material for this manuscript includes the following:

MDAR Reproducibility Checklist

Data and materials availability:

All data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.