Single-molecule imaging of the functional crosstalk between surface NMDA and dopamine D1 receptors

Significance

Dopamine receptor signaling in the brain participates in memory encoding through the regulation of glutamatergic signaling. Here we provide evidence that single dopamine D1 receptors are highly dynamic at the surface of hippocampal neurons. In addition, these receptors, together with glutamatergic NMDA receptors, form surface clusters in the vicinity of glutamate synapses, providing a strategically located reservoir pool from which they can be laterally redistributed during synaptic adaptations. The plasma membrane and receptor dynamics thus appear as an important level of the glutamate–dopamine interplay.

Abstract

Dopamine is a powerful modulator of glutamatergic neurotransmission and NMDA receptor-dependent synaptic plasticity. Although several intracellular cascades participating in this functional dialogue have been identified over the last few decades, the molecular crosstalk between surface dopamine and glutamate NMDA receptor (NMDAR) signaling still remains poorly understood. Using a combination of single-molecule detection imaging and electrophysiology in live hippocampal neurons, we demonstrate here that dopamine D1 receptors (D1Rs) and NMDARs form dynamic surface clusters in the vicinity of glutamate synapses. Strikingly, D1R activation or D1R/NMDAR direct interaction disruption decreases the size of these clusters, increases NMDAR synaptic content through a fast lateral redistribution of the receptors, and favors long-term synaptic potentiation. Together, these data demonstrate the presence of dynamic D1R/NMDAR perisynaptic reservoirs favoring a rapid and bidirectional surface crosstalk between receptors and set the plasma membrane as the primary stage of the dopamine–glutamate interplay.

Hippocampal dopaminergic neuromodulation participates in several cognitive functions including novelty detection and long-term memory storage (1, 2). As a consequence, impairments in hippocampal neuromodulatory transmission affect synaptic plasticity at glutamatergic synapses, prevent learning and memory formation, and have been proposed to be a cellular substrate for neurodevelopmental psychiatric disorders such as schizophrenia (3). In the hippocampus and cortex, pyramidal neurons express mostly dopamine D1 and D5 receptors along their dendritic tree (4–6). Their recruitment affects the trafficking and surface expression of glutamate NMDA receptors (NMDARs), two processes that are essential for excitatory neurotransmission and synaptic plasticity. Indeed, activating dopamine D1 receptors (D1Rs) promotes the surface expression and function of NMDAR and thereby favors the long-term potentiation of excitatory glutamate synapses (7–10). Reciprocally, the activation of NMDAR modulates D1R surface expression and signaling (11). The bidirectional dialogue between dopamine and glutamate NMDAR-associated signaling thus involves changes in membrane receptor content and trafficking.

Although this functional interaction is usually considered as relying on intracellular protein kinase signaling cascades (7, 10, 12), physical interactions between D1R and NMDAR at the plasma membrane were recently reported to stabilize laterally diffusing surface D1R in spines, modulate D1R- and NMDAR-mediated signaling, and influence working memory (13–16). Thus, direct interactions between these receptors could contribute to the regulation of their surface distributions and play a major role in the dopamine–glutamate interplay (15, 17). In particular, because the regulation of NMDAR synaptic content involves surface diffusion processes in and out of synaptic and extrasynaptic compartments (18), the possibility emerges that dopamine might modulate NMDAR-dependent synaptic transmission by tuning NMDAR lateral dynamics through D1R–NMDAR physical interactions. To address this question and investigate the role of the D1R/NMDAR surface crosstalk in synaptic physiology, we here assessed the surface distribution and trafficking of D1 and NMDA receptors in rat hippocampal neurons using a combination of high-resolution single-nanoparticle tracking, bulk imaging, and electrophysiology.

Results

Distribution and Surface Dynamics of D1R in Hippocampal Neurons.

Although their surface organization and trafficking remain elusive, D1Rs have been previously detected in hippocampal and cortical pyramidal neurons (4–6). Using immunohistochemistry and electron microscopy, we first labeled these receptors in the hippocampus and confirmed their presence in several cellular compartments (Fig. 1A). It is notable that endogenous D1R expression levels in vitro were comparable to those observed in native brain structures (Fig. 1B). D1R labeling was observed in the vicinity of—but not within—postsynaptic densities (PSDs) (Fig. 1A). Because single-receptor detection is not yet operational in or ex vivo, we then used high-resolution single-particle detection in cultured hippocampal neurons to investigate the live surface distribution of D1R. To do so, we designed a receptor tagged on its N terminus with a cyan fluorescent protein (D1R-CFP), which does not affect ligand binding nor dopamine-mediated activation (19). No difference was observed when comparing the overall content of D1R in transfected (D1R-CFP) and nontransfected neurons (Fig. S1A), discarding a possible transfection-induced overload of surface D1R and validating the use of exogenously expressed D1R in the rest of study. As endogenous receptors, surface D1R-CFPs appeared to be expressed at the level of both the soma and dendrites, with a heterogeneous distribution in the latter compartment (Fig. S1 B and C). Indeed, D1Rs were found to form clusters along the dendritic tree (Fig. 1C), most of them (>90%) being detected within less than 1.5 µm from the closest glutamatergic PSD as attested with the coexpression of the GFP-labeled synaptic scaffold protein Homer 1c (Fig. 1C and Fig. S1 C and D). Their linear density was comparable to that of glutamatergic synapses (D1R surface cluster density, 4.8 ± 0.2 clusters per 10 µm of dendrite; Homer 1c cluster density, 5.1 ± 0.4 per 10 µm of dendrite) (20), suggesting that each glutamate synapse is closely related to a D1R surface cluster. When examined over a period of 30 min, cluster intensity and shape were stable (Fig. S1E), indicating that these structures are not labile. We also used the direct stochastic optical reconstruction microscopy (dSTORM) as a dual-color single molecule-based superresolution microscopy method to compare the nanoscale distribution of D1R and glutamate synapses, detected through the labeling of endogenous GluA2-AMPA receptors (GluA2-AMPARs) (21). This high-resolution (20-nm) imaging technique further confirmed that D1Rs form surface clusters in the vicinity of AMPAR-containing synapses (Fig. 1D). Finally, taking advantage of the high signal-to-noise ratio and the subsequent pointing accuracy (capacity to detect any movement with an accuracy of <25 nm) of single-quantum dot (QD) detection (Fig. 1E), we labeled surface D1R-CFP with an anti-GFP antibody coupled with a single QD to establish a high-resolution map of surface D1R–QD in live hippocampal neurons. Consistent with the electron microscopy, immunohistochemistry, and dSTORM data, single D1Rs were not enriched in the PSD of glutamatergic synapses but rather observed in the perisynaptic (320-nm annulus around the PSD) and extrasynaptic compartments (Fig. 1E). The fraction of single D1R–QD complex observed in the PSD was consistently threefold lower than in the perisynaptic area (Fig. 1F). In addition, single D1R–QD appeared to form perisynaptic aggregates (dotted black line in Fig. 1E), consistent with the presence of D1R surface clusters.

Fig. 1.

D1Rs are clustered in the vicinity of glutamate synapses and dynamically explore the plasma membrane. (A) Electron micrographs demonstrating the subcellular localization of immunogold-labeled endogenous D1R protein (arrow) close to the plasma membrane (PM) in a dendritic spine (Sp) (Upper) and in the cytoplasm (Cy) (Lower). In Lower, the intracellular compartment of a dendrite was filled (light brown), and detected D1R was circled (dotted line). (B) Detection of endogenous D1R by Western blot in Cos7 cells, cultured hippocampal neurons (Hipp. cult.), striatum, hippocampus (Hipp.), and cortex (15 µg of total proteins for each condition). The molecular weight (MW) is indicated on the left. (C) Immunofluorescence detection of surface D1R-CFP (Alexa 568) (Left) in neurons expressing Homer 1c-GFP (Center). D1R-CFP clusters do not colocalize with Homer 1c staining (Right). (Scale bar, 10 µm.) (Lower) High magnifications of the staining in Upper. (Scale bar, 1 µm.) (D) Gallery of superresolution images acquired by dSTORM technique demonstrating the subdiffraction organization of D1R-CFP (Center) at GluA2-AMPAR–containing synapses (Left). (Right) AMPARs (red) are concentrated in the PSD and do not colocalize with D1R-CFPs (green) that are distributed perisynaptically. Spine shapes are drawn based on diffracted epifluorescence images. (Scale bars, 1 µm.) (E) Surface D1R-CFP was labeled with a single QD–antibody complex (D1R–QD), allowing high pointing accuracy detection. (Lower Right) A 30 × 30 pixel 3D projection of a single D1R–QD complex. Representative surface distribution of a single D1R–QD complex detection (500 frames, 20-Hz acquisition rate; each dot represents the detection of a single receptor during a frame), revealing a trapping zone of the single surface D1R–QD (dashed black line) in the vicinity of the PSD. (F) Percentage of single D1R–QD particles detected in the synaptic core (PSD) or the perisynaptic area (Perisyn., 320-nm annulus around PSD) (n = 51 neuronal fields; n_{total PSD} = 1,236 particles; n_{total peri.} = 5,223 particles). (G) Live detection of single D1R–QD complexes, allowing single-particle tracking. (Left) Representative trajectories (1,000 frames, 20-Hz acquisition rate) of surface D1R in a dendritic field of hippocampal neurons. (Scale bar, 4 µm.) (Right) Enlarged single D1R–QD trajectory from the dendritic field. (Scale bar, 1 µm.) Arrowhead, first point of the trajectory (t_0s); black dot, last point of the trajectory (t_50s). (H) Plot of the MSD versus time of total endogenous (black; n = 588 trajectories) and exogenous (green; n = 2,236 trajectories) surface D1R. The two curves are not linear (dotted line, free diffusion) but tend toward a negative curvature, characteristic of a confined behavior. (Inset) Representative trajectories (1,000 frames, 20-Hz acquisition rate) of surface endogenous D1R labeled with a single QD–antibody complex in a dendritic field of hippocampal neurons. (Scale bars, 5 µm.) (I) Representative trajectory (600-frame duration, 20-Hz acquisition rate) of a single surface D1R–QD in a dendritic fragment containing two Homer 1c-DsRed clusters (white arrows). D1R–QD diffuses between PSDs and is temporarily retained in the perisynaptic area. (Scale bar, 1 µm.) (J) Cumulative distributions of the instantaneous diffusion coefficients of total (black; n = 2,296 trajectories), synaptic (light gray; n = 421), or extrasynaptic (dark gray; n = 1,875) surface D1R–QD. The first point corresponds to the percentage of immobile receptors (diffusion coefficient < 0.005 µm²/s). The surface diffusion of D1R–QD inside synaptic areas was significantly lower than in extrasynaptic compartments (extrasynaptic, n = 1,875 trajectories; synaptic, n = 421 trajectories; *P < 0.05; Kolmogorov–Smirnov test on cumulative distributions: extrasynaptic vs. synaptic, ***P < 0.001).

To shed light on the cellular pathways that shape such a distribution, we tracked single D1R-CFP-QD at the surface of hippocampal neurons (Fig. 1G). The vast majority of D1R (94%) was found to be diffusive (6% immobile receptors, defined as presenting diffusion coefficients <0.005 µm²/s) with a relatively high instantaneous diffusion coefficient [median = 0.08 µm²/s, interquartile range (IQR) = 0.03–0.18 µm²/s, n = 2,296 trajectories]. Plotting the mean square displacement (MSD; surface explored) versus time of both endogenous D1R and D1R-CFP revealed similar diffusion properties and showed that D1Rs diffuse with a confined pattern reflecting the presence of clusters and/or regulatory mechanisms in neurons (Fig. 1H). Because endogenous and D1R-CFP dynamics were undistinguishable, we took advantage of the higher detection specificity of exogenously expressed D1R-CFP to study the fine regulation of D1R in dendritic compartments. It is notable that D1R-CFPs in neighboring glial cells (mostly astrocytes) were significantly more diffusive with virtually no immobile receptors (<0.4% immobile D1R; Fig. S2), confirming that regulation processes of D1R surface dynamics are restricted to neurons. Within the synaptic area (PSD + perisynaptic annulus), D1R surface diffusion was reduced, mostly because of a higher fraction of immobile D1R (21% compared with only 6% extrasynaptically) (Fig. 1 I and J), which is fully consistent with the observation of perisynaptic D1R clusters (Fig. 1 C and D). These aggregates locally impacted on receptor dynamics because the dwell time of D1R within perisynaptic clusters was, for instance, sevenfold higher than within the spine area (D1R clusters, 6.8 ± 1.2 s, n = 132 trajectories; spine area, 1.2 ± 0.7 s, n = 173 trajectories). Together, these results reveal that D1Rs actively explore the neuronal surface and form clusters in the vicinity of glutamate synapses in which they are temporarily trapped.

Direct Interaction Between GluN1 Subunit and D1R Shapes Their Surface Dynamics.

Because interactions with NMDAR have been reported to modulate D1R trafficking (16), we then tested whether they could participate in this perisynaptic D1R clustering process. Costaining of surface D1R and GluN1-NMDAR revealed that a fraction of surface GluN1-NMDAR clusters colocalize with neighboring D1R clusters (12.5 ± 4% colocalization, n = 12 neurons) (Fig. 2A). To further investigate whether interactions with NMDAR shape the surface distribution of D1R, we took advantage of the fact that the D1R/GluN1-NMDAR interaction can be disrupted, either physiologically through the activation of D1R or artificially with the application of a cell-permeant (TAT-conjugated) peptide (TAT-t₂) mimicking the C-terminal sequence of D1R required for the interaction with GluN1-NMDAR subunits (13). Strikingly, disrupting D1R–NMDAR interactions either through D1/5R agonist (SKF-38393, 10 µM, 15 min) or TAT-t₂ competing peptide (10 µM, 15 min) application significantly and rapidly decreased D1R cluster fluorescence intensity (Fig. 2 B and C), indicating that the direct interaction between D1R and GluN1 subunit regulates the size of D1R clusters. This reduction in the size of D1R clusters was not observed when disrupting another interaction between D1R and GluN2A-NMDAR subunits with a specific competing peptide (TAT-t₃, 10 µM for 15 min) (Fig. 2C), which indicates that the D1R–GluN1, but not D1R–GluN2A, interaction controls the clustering of D1R in spines. Based on these findings, one may predict that favoring the presence of perisynaptic NMDAR in the spine area would increase their potential binding to D1R and thereby modulate the size of D1R clusters. To address this question, we artificially increased the amount of perisynaptic NMDAR by blocking their synaptic retention with a biomimetic ligand (TAT-[N2A₁₅], 10 µM, 20 min) that prevents their C-terminal interaction with cytosolic PDZ domain-containing proteins (22). Strikingly, this treatment resulted in a rapid and significant increase in D1R cluster intensity (Fig. 2C). Together, these data reveal that D1Rs are retained in clusters in the close vicinity of glutamate synapses through interactions with GluN1-NMDAR, constituting a perisynaptic reserve pool of surface D1R and NMDAR that can possibly be mobilized on demand following dopamine release.

Fig. 2.

The D1R–NMDAR interaction bidirectionally regulates the surface distribution and dynamics of D1R and NMDAR. (A) Immunostaining of surface D1R-CFP (green) and GluN1 subunit (red) in hippocampal neurons. The yellow arrow shows overlay. (B) Immunostaining of surface D1R-CFP in control or after D1/5R agonist, TAT-t₂, or TAT-[N2A₁₅] application. (Scale bar, 250 nm.) (C) Normalized measures of D1R-CFP clusters intensity in control (n = 32 neuronal fields), D1/5R agonist-treated (n = 24 neuronal fields; *P < 0.05 compared with control), TAT-NS^t2–treated (non-sense of TAT-t₂, n = 19 neuronal fields), TAT-t₂–treated (n = 21 neuronal fields; **P < 0.01 compared with TAT-NS^t2), TAT-NS^t3–treated (non-sense of TAT-t₃, n = 11 neuronal fields), TAT-t₃–treated (n = 12 neuronal fields; P > 0.05 compared with TAT-NS^t3), TAT-[NS₁₅]–treated (n = 27 neuronal fields; P > 0.05), or TAT-[N2A₁₅]–treated (n = 21 neuronal fields, *P < 0.05 compared with TAT-[NS₁₅]) conditions. (D) Representative trajectories (1,000 frames, 20-Hz acquisition rate) of surface single D1R-CFP (Left) (green) (scale bar, 400 nm) and GluN1-NMDAR (Right) (blue) (scale bar, 300 nm) in the absence and presence of either D1/5R agonist (10 µM, 15 min) or TAT-t₂ (10 µM, 15 min). Bold dotted line, perisynaptic area; thin dotted line, PSD area. (E) Plot of the MSD of surface D1R-CFP (Upper) (green) and GluN1-NMDAR (Lower) (blue) versus time in presence of TAT-NS or TAT-t₂ peptides (10 µM, 15 min). The SEM is included for each data point (D1R: TAT-NS, n = 986 trajectories, and TAT-t₂, n = 1,326; GluN1-NMDAR: TAT-NS, n = 198, and TAT-t₂, n = 134). (F and G) Representative surface distributions of single D1R-CFP (green) (F) and GluN1-NMDAR (blue) (G) in the synaptic area (PSD + perisynaptic area) in control, D1/5R agonist, and TAT-t₂ conditions. Each dot represents the detection of a single receptor during a frame. Comparisons of the time spent in the synaptic area (dwell time) by single D1R-CFP (control, n = 173 trajectories; D1/5R agonist, n = 142, **P < 0.01; TAT-t₂, n = 752, *P < 0.05) (F) and GluN1-NMDAR (control, n = 189 trajectories; D1/5R agonist, n = 157, *P < 0.05; TAT-t₂, n = 134, **P < 0.01) (G) and the synaptic fraction of detected single D1R-CFP (control, n = 14 neuronal fields; D1/5R agonist, n = 19, **P < 0.01; D1/5R agonist in the presence of dynasore, n = 47, **P < 0.01; TAT-t₂, n = 15, ***P < 0.001) (F), D5R-CFP (n = 16, P > 0.05) (F), and GluN1-NMDAR (control, n = 11; D1/5R agonist, n = 15, *P < 0.05; TAT-t₂, n = 14, *P < 0.05) (G). Dyn., dynasore; D1/5 ago., D1/5 receptor agonist SKF-38393.

To identify the cellular pathways involved in this regulatory process, we then tracked surface D1Rs before and after disrupting their interaction with NMDAR, either by activating D1R with SKF-38393 (10 µM, 15 min) or through acute TAT-t₂–mediated disruption (10 µM, 15 min). Both treatments rapidly dispersed D1Rs by significantly and reversibly increasing both their lateral diffusion within the synaptic area and the surface they explored (Fig. 2 D and E and Fig. S3 A and B). Furthermore, D1/5R agonist and TAT-t₂ treatments decreased both the D1R–QD fraction (control, 9.3 ± 1.1%; D1/5R agonist, 4.1 ± 1%; P < 0.05) and dwell time (TAT-NS, 1.7 ± 0.1 s; TAT-t₂, 1.3 ± 0.09 s; P < 0.01) in the synaptic area (Fig. 2F), indicating that the acute disruption of D1R–NMDAR interactions laterally displaced D1R out of the synaptic area. The effects of SKF-38393 and TAT-t₂ treatments on D1R could theoretically be explained by an increased internalization of the receptors rather than a surface redistribution. However, the SKF-38393– and TAT-t₂–mediated increase in D1R diffusion already occurred after 1 min of exposure, and D1R instantaneously exhibited a less confined motion inconsistent with the reduced diffusion usually observed during receptor internalization (23). In addition, single puffs of SKF-38393 through a glass pipette similarly induced instantaneous, local, and reversible increases in D1R surface diffusion (Fig. S3 C and D). Finally, preventing receptor endocytosis by bath application of dynasore (80 µM) did not interfere with D1/5R agonist-mediated decrease in D1R synaptic fraction (Fig. 2F). Together, these data argue in favor of an agonist-induced surface redistribution rather than internalization process. To note, similar experiments performed on CFP-tagged D5R—which does not interact with NMDAR—showed no effect of TAT-t₂ on the synaptic content or surface diffusion (Fig. 2F and Fig. S4), consistent with a specific D1R–NMDAR interaction-based mechanism. Altogether, these experiments demonstrate that at the plasma membrane of hippocampal neurons, D1R/GluN1-NMDAR interactions strongly contribute to the regulation of D1R surface dynamics and distribution.

Could this interaction also affect the surface distribution and dynamics of NMDAR? Remarkably, both SKF-38393 and TAT-t₂ applications transiently increased the surface diffusion and MSD of GluN1-NMDAR (Fig. 2 D and E and Fig. S3E). Consequently, these treatments rapidly enhanced the synaptic fraction and dwell time of GluN1-NMDAR (Fig. 2G). These results are consistent with a scenario in which releasing NMDARs from their interaction with perisynaptic D1R leads to their lateral dispersal and their trapping within the PSD by PDZ proteins. D1/5R-associated signaling has also been reported to affect the trafficking of other glutamate receptors. Indeed, D1/5R activation induces long-term changes in the surface expression of AMPARs through intracellular protein kinase signaling (24). To tackle this point, we tracked single GluA2-AMPAR-QDs and observed that their diffusion was not significantly different before and after SKF-38393 application (Fig. 3 A and B). Because AMPARs are highly retained in the core of the synapse (23), the use of single-QD tracking might prevent the observation of minor effects induced by D1R activation, as previously assessed (25). We thus used another single-molecule detection approach providing an efficient way to simultaneously probe a large number of single AMPARs in highly confined compartments, i.e., the universal point accumulation for imaging in nanoscale topography (uPAINT) approach. This generalizes the previously developed point accumulation for imaging in nanoscale topography (PAINT) method for dynamic imaging of arbitrary membrane biomolecules, providing a powerful way to record thousands of single-molecule trajectories that appear sequentially at the neuronal surface upon continuous labeling of an endogenous AMPAR subunit (26). In these conditions, surface GluA2-AMPAR diffused and explored both extrasynaptic and synaptic membrane compartments (Fig. 3C), as previously described (26). It is notable that a larger fraction of the trajectories was confined within synaptic areas compared with QD tracking experiments, yielding lower diffusion coefficient values. This could be attributed to the smaller size of ATTO647–antibody complexes used for uPAINT experiments that are thus more likely to be enriched within synapses (27). Application of SKF-38393 did not significantly alter the diffusion of surface GluA2-AMPAR, both after 5 min (diffusion coefficient median ± IQR: control, 1.2 × 10⁻³ ± IQR 1 × 10⁻⁵ to 9 × 10⁻³ µm²/s, n = 6,826; after SKF-38393, 9 × 10⁻⁴ ± IQR 1 × 10⁻⁵ to 4 × 10⁻³ µm²/s, n = 30,280; P > 0.05) and 15 min (diffusion coefficient median ± IQR: control, 1.2 × 10⁻³ ± IQR 6 × 10⁻⁵ to 48 µm²/s, n = 1,811; after SKF-38393, 1.6 × 10⁻³ ± IQR 8 × 10⁻⁵ to 0.53 µm²/s, n = 1,588; P > 0.05), consistent with the single–GluA2-AMPAR-QD data. Similarly, SKF-38393 treatment did not affect the mobility of GluA2-AMPAR in the postsynaptic area (Fig. 3D), with an unchanged ∼70% of immobile GluA2-AMPAR after SKF-38393 applications. Thus, SKF-38393 application did not alter the surface diffusion of GluA2-AMPAR (Fig. 3E). Altogether, these data indicate that D1R–NMDAR interactions specifically and bidirectionally regulate the surface diffusion and distribution of both receptors in the synaptic area and that activating D1R laterally redistributes D1R and NMDAR, thereby specifically increasing the synaptic trapping of NMDAR.

Fig. 3.

The surface dynamics of extrasynaptic and synaptic AMPA receptors is unaltered by the activation of D1/5R. (A) Representative trajectories of surface GluA2-AMPAR in control and D1/5R agonist (10 µM) conditions. Dotted black line, PSD area. (Scale bar, 100 nm.) (B) Cumulative distribution of the instantaneous diffusion coefficients of surface GluA2-AMPAR in control or D1/5R agonist condition (control, n = 92; D1/5R agonist, n = 159; P > 0.05). (C) Representative trajectories (20-Hz acquisition rate) of surface GluA2-AMPAR either before or after incubation with D1/5R agonist. Each colored trace represents an independent trajectory lasting for more than 0.5 s. (Upper) Homer 1c-GFP–labeled PSD. (Right) Focus on trajectories located inside Homer 1c-GFP–labeled PSD. (D) Cumulative frequency distributions of synaptic GluA2-AMPAR instantaneous diffusion coefficients before (n = 2,341 trajectories) and after incubation with D1/5R agonist (n = 1,987 trajectories). The first point of the distribution corresponds to the immobile fraction (before, 73% of immobile GluA2-AMPAR; after D1/5R ago., 72% of immobile GluA2-AMPAR). (E) Incubation with D1/5R agonist did not significantly change the diffusion coefficients of GluA2-AMPAR (before D1/5R ago., n = 1,811; after D1/5R ago., n = 1,588; P > 0.05).

Functional Impact of the Dynamic Surface Interplay Between NMDAR and D1R.

To investigate the functional consequence of these selective surface redistribution processes, we first confirmed through electrophysiological recordings on hippocampal slices at CA1 synapses that activating D1R significantly increased the amplitude of NMDA-mediated synaptic currents (before, 63 ± 7 pA; SKF-38393, 77 ± 8 pA; P < 0.05) without affecting AMPA-mediated currents (before, 141 ± 19 pA; SKF-38393, 137 ± 20 pA; P > 0.05) (Fig. 4A and Fig. S5). To directly estimate the content of synaptic NMDAR, we expressed in hippocampal neurons the GluN1 subunit fused to a Super Ecliptic pHluorin (SEP) at its extracellular N terminus (GluN1-SEP) to isolate surface GluN1-NMDAR (SI Materials and Methods). SEP is a pH-sensitive variant of GFP that only emits fluorescence at neutral pH (fluorescence is quenched at acidic pH such as in intracellular vesicles). We confirmed that application of TAT-t₂ peptide significantly increased GluN1-SEP–associated synaptic fluorescence (Fig. 4B). Thus, the changes in NMDAR surface diffusion consecutive to D1R activation or D1R/GluN1-NMDAR interaction disruption result in a synaptic enrichment in NMDAR that is likely to have a profound impact on the adaptation of glutamate synapses. To directly address this point, we evaluated the consequences of TAT-t₂–induced D1R/NMDAR surface redistribution on synaptic long-term potentiation. Using a combined application of glycine/picrotoxin (Gly/Pic) as a chemical long-term potentiation stimulus (cLTP) (27, 28), we observed that a prior TAT-t₂ application significantly amplified the Gly/Pic-elicited increase in synaptic GluA1-SEP (Fig. 4 C and D). Moreover, the percentage of cLTP-potentiated synapses (defined as >2 × SD of the baseline GluA1-SEP synaptic content) was shifted from 56% in the presence of TAT-NS to 64% after preincubation with TAT-t₂. Together, these data reveal that through rapid changes in surface dynamics and distribution, the interaction between D1R and GluN1-NMDAR regulates NMDAR-mediated synaptic transmission and thereby the plastic range of excitatory synapses.

Fig. 4.

D1R activation or D1R/GluN1-NMDAR interaction blockade increases synaptic NMDAR content and favors AMPAR synaptic long-term potentiation. (A) (Left) Excitatory postsynaptic current traces recorded at −70 mV and +40 mV from a representative hippocampal CA1 pyramidal cell, before and 10 min after exposure to D1/5R agonist. (Right) Relative change over time of the AMPA/NMDA ratio at CA1 synapses in the absence or presence of D1/5R agonist (n = 13, *P < 0.05 10 min after agonist) and in the absence or presence of vehicle (n = 7, P > 0.05). (B) Surface imaging of GluN1-SEP in neurons incubated with either TAT-NS or TAT-t₂ (10 µM). (Scale bar, 5 µm.) (Right) Average value of GluN1-SEP content in the synaptic area after TAT-NS or TAT-t₂ application (n = 8 neurons per group, **P < 0.01). (C) Dendritic fragment of a hippocampal neuron expressing Homer 1c-DsRed (Upper) and GluA1-SEP (Lower). SEP only fluoresces at neutral pH when receptors are inserted at the plasma membrane. Ten minutes after chemical LTP induction (cLTP), the GluA1-SEP fluorescence intensity increased in postsynaptic clusters. (Insets) High magnification of a synaptic GluA1-SEP cluster. (Scale bar, 2 µm.) (D) Comparison of the synaptic GluA1-SEP fluorescence intensity before and after cLTP with prior TAT-NS (n = 198 synapses, *P < 0.05) or TAT-t₂ (n = 215 synapses, *P < 0.05) (TAT-NS versus TAT-t₂; *P < 0.05) application. (E) Schematic model of the D1R–NMDAR surface interplay in hippocampal neurons. D1Rs are highly diffusive at the neuronal surface and are dynamically retained in clusters in the vicinity of glutamate synapses where they interact with NMDAR. Dopamine release disrupts this interaction and favors the lateral redistribution of both receptors: D1Rs freely explore extrasynaptic areas, whereas NMDARs laterally reach the PSD where they impact on the long-term plasticity of glutamate synapses.

Discussion

Using high-resolution single-molecule imaging, we here provide evidence that D1Rs are highly dynamic at the surface of hippocampal neurons and form clusters in the vicinity of glutamatergic synapses, providing a perisynaptic reservoir of NMDAR (Fig. 4E). By disrupting D1R–NMDAR interactions, dopaminergic inputs laterally redistribute both receptors and enrich synapses in NMDAR, thereby extending their plastic range (Fig. 4C). A common feature shared by all surface receptors is that they appear to be on a dynamic equilibrium between diffusive and anchored states. NMDAR and D1R appear to stand at both ends of this range because most NMDARs are anchored in basal condition (∼70–80% immobile), whereas the vast majority of D1Rs are diffusive (∼5% immobile). We here show that a direct surface interaction between NMDAR and D1R contributes to the dynamic distribution of NMDAR in the vicinity of glutamate synapses. Although the kinetics of this molecular interaction remain unknown, the rapid effect of competing peptides (TAT-t₂) argues in favor of a rather dynamic process. Surface D1Rs are highly diffusive in hippocampal neurons, suggesting a limited number of anchoring mechanism(s). Scaffold proteins, such as PSD-95, likely modulate the surface content of D1R (29, 30), although they do not directly control their synaptic anchoring (31), which is rather regulated by direct interaction with NMDAR in the hippocampus. The activation of D1R by dopamine might affect its surface trafficking through the unbinding of D1R and NMDAR and/or the downstream activation of protein kinases, although whether these two mechanisms are linked still needs to be clarified. Based on the timing of these effects, these studies and our current data support a model in which D1R activation first (seconds to first minutes range) induces a lateral rearrangement of synaptic NMDAR to prime synapses for plasticity and second (tens of minutes range) regulates the surface distribution of NMDAR to support plasticity/metaplasticity processes. On the reverse, one may wonder about the impact of glutamate release on D1R receptor diffusion. In striatal neurons, it has previously been shown that D1Rs traffic at the neuronal surface (16, 30), and this trafficking is affected by NMDAR activation (16). Consistently, we found that activating NMDAR rapidly reduced the surface diffusion of D1R and increased the synaptic dwell time of D1R (Fig. S6). Clearly, further investigations are needed to fully uncover the role of ambient extracellular neurotransmitters on receptor dynamics, particularly in integrative areas such as dendritic spines. The molecular crosstalk of surface receptors should thus be envisioned as an agonist-sensing and integrating system regulating neuronal and synaptic responsiveness in function of the local ambient level of extracellular neurotransmitters.

Materials and Methods

The complete description can be found in SI Materials and Methods. Briefly, cultures of hippocampal neurons and glial cells were prepared from E18 Sprague–Dawley rats and used from days 10–17 in vitro. The dSTORM is described in SI Materials and Methods. For single-nanoparticle tracking, QD 655 coupled to goat anti-rabbit F(ab′)₂ or anti-mouse IgG (Invitrogen) was incubated (1:10,000, 10 min) onto neurons previously exposed for 10 min to either mouse monoclonal anti-GFP (1 µg; Invitrogen), rabbit polyclonal anti-D1R (1 µg; Lifespan Biosciences), mouse monoclonal anti-GluA2:00 AMPAR subunit (1 µg; Millipore), or rabbit polyclonal anti–GluN1-NMDAR subunit (1 µg; Alomone Laboratories) antibodies. For the specific experiments, neurons were incubated in presence of TAT peptides together with the QDs for 10 min at 37 °C in culture medium. The instantaneous diffusion coefficient, D, was calculated for each trajectory from linear fits of the first four points of the MSD versus time function using MSD(t) = <r²>(t) = 4Dt. The full methods are provided in SI Materials and Methods.