Research Article

Deconstructing behavioral neuropharmacology with cellular specificity

See allHide authors and affiliations

Science  07 Apr 2017:
Vol. 356, Issue 6333, eaaj2161
DOI: 10.1126/science.aaj2161

A tailored look at behavioral pharmacology

It is important to understand how animal behavior is mediated by molecular, cellular, and circuit components of the brain. However, it has been difficult to link the activity of specific molecules in defined cells to behavioral roles. Shields et al. developed an approach to deconstruct behavioral neuropharmacology with cellular specificity. The technique, termed DART (drugs acutely restricted by tethering), uses enzymatic capture to restrict standard drugs to the surface of genetically specified cells without prior modification of the native pharmacological target. The method provides cell-type specificity, endogenous-protein specificity, acute onset, and utility in behaving animals. This enables the activity of specific molecules in defined circuit elements to be causally linked to behavior.

Science, this issue p. eaaj2161

Structured Abstract

INTRODUCTION

Animal behavior is mediated by molecular, cellular, and circuit components of the brain. However, because many proteins are broadly expressed, it has been difficult to link the activity of specific proteins in defined cells to behavioral roles. This challenge has particular relevance to neuropsychiatric disorders, which have largely been understood in relation to clinically effective drugs that act on known molecular targets. Such knowledge has been difficult to extend to circuit-level insight because cell type specificity has not been possible with traditional pharmacology.

RATIONALE

Here, we combined the speed and molecular specificity of pharmacology with the cell type specificity of genetic tools. DART (drugs acutely restricted by tethering) is a technique that uses a bacterial enzyme called HaloTag to capture and tether drugs to the surface of defined cells. A key feature is that the method does not require prior modification or overexpression of the native pharmacological target because HaloTag is expressed as a separate protein. Drug capture proceeds rapidly over seconds to minutes, producing a factor of ~100 enrichment of the drug at the surface of HaloTag-expressing cells. The method provides a unique feature set: (i) Cell type specificity arises from expression of HaloTag under control of a cell type–specific promoter; (ii) molecular specificity is inherited from the drug that is tethered; and (iii) acute onset upon delivery of the DART ligand is similar to that of traditional pharmacology, thus averting compensatory phenomena.

RESULTS

We first developed a DART that antagonizes the α-amino-3-hydroxy-5-methylisoxazole-4-propionic acid receptor (AMPAR), a broadly expressed postsynaptic glutamate receptor. We validated the speed, cell type specificity, and molecular specificity of the method in cultured neuronal assays, in coronal slices of mouse dorsal striatum, and in behaving mice. We then applied the technique to a mouse model of Parkinson’s disease (PD). In PD, AMPARs are subject to dysregulated long-term potentiation (LTP) and long-term depression (LTD) in distinct cell types of the striatum, a brain region critical for movement. AMPAR antagonists have been studied extensively in animal models of PD and in human clinical trials for the disorder. However, it has been difficult to link AMPAR activity in defined cells to motor deficits. We found that motor deficits were causally attributed to AMPARs in indirect spiny projection neurons (iSPNs) and to excess phasic firing of tonically active interneurons (TANs) of the striatum. Together, iSPNs and TANs (i.e., D2 cells) drove akinesia, whereas movement execution deficits reflected the ratio of AMPARs in D2 versus D1 cells. Finally, we designed a muscarinic antagonist DART in one iteration, demonstrating applicability of the method to diverse targets.

CONCLUSION

Neuropsychiatric disorders have been examined with acute manipulations featuring either circuit or molecular specificity. DART combines these features, enabling interrogation of specific proteins in defined cells. The approach may provide a platform whereby the mechanism of action for widely prescribed drugs can be examined with cellular specificity in animal models of several disorders. Such studies could inform new translational strategies by advancing nonobvious drug combinations, or by providing a road map for the design of bivalent therapeutics based on the “message-address” concept of Schwyzer.

Deconstructing behavioral neuropharmacology.

Drugs that manipulate specific molecules in the brain (e.g., green but not purple postsynaptic receptor) have shaped our understanding of neuropathologies. The DART technique uses HaloTag (red) to capture and spatially restrict drugs to the surface of genetically defined cells. Behavioral effects of drugs can thus be deconstructed into the individual and combinatorial contributions produced by defined cell types.

Illustration: Julia Kuhl

Abstract

Behavior has molecular, cellular, and circuit determinants. However, because many proteins are broadly expressed, their acute manipulation within defined cells has been difficult. Here, we combined the speed and molecular specificity of pharmacology with the cell type specificity of genetic tools. DART (drugs acutely restricted by tethering) is a technique that rapidly localizes drugs to the surface of defined cells, without prior modification of the native target. We first developed an AMPAR antagonist DART, with validation in cultured neuronal assays, in slices of mouse dorsal striatum, and in behaving mice. In parkinsonian animals, motor deficits were causally attributed to AMPARs in indirect spiny projection neurons (iSPNs) and to excess phasic firing of tonically active interneurons (TANs). Together, iSPNs and TANs (i.e., D2 cells) drove akinesia, whereas movement execution deficits reflected the ratio of AMPARs in D2 versus D1 cells. Finally, we designed a muscarinic antagonist DART in one iteration, demonstrating applicability of the method to diverse targets.

Animal behavior is mediated by molecular, cellular, and circuit components of the brain (1). Nonetheless, it has been difficult to link the activity of specific proteins in defined cells to behavioral roles (2). This challenge has particular relevance to neuropsychiatric disorders such as Parkinson’s disease (PD), the leading movement disorder worldwide (3). PD is caused by the degeneration of dopaminergic cells, leading to a downstream circuit disruption (49). The α-amino-3-hydroxy-5-methylisoxazole-4-propionic acid receptor (AMPAR), a broadly expressed postsynaptic glutamate receptor, has been implicated in the disorder and is subject to dysregulated long-term potentiation (LTP) and long-term depression (LTD) in distinct cell types of the affected circuitry (1012). AMPAR antagonists have been studied extensively in animal models of PD (1319) and in human clinical trials for the disorder (2022). However, because these receptors are widely expressed, it has been difficult to causally link AMPAR activity in defined cells to motor deficits. More broadly, many neuropathologies have largely been understood in relation to clinically effective drugs that act on known molecular targets, but such knowledge has been difficult to extend to circuit-level insight (2).

This gap between molecular and circuit knowledge can largely be attributed to the lack of cell type specificity with traditional pharmacology (Fig. 1A, top row). Nonetheless, pharmacology outrivals other methods in several respects: (i) In comparison to genetically encoded actuators such as light-gated channels (2331) (Fig. 1A, middle row), which introduce signals from exogenous proteins, pharmacology can manipulate endogenous proteins. (ii) Pharmacology has acute onset upon application; this is a temporal advantage over genetically encoded toxins (32) or gene editing (33) (Fig. 1A, bottom row), in which compensatory gene expression can occur during the days needed for protein expression or turnover. Methods for acutely manipulating native proteins with cellular specificity have thus far met technical challenges, preventing utility in behaving animals: Prodrugs (i.e., inert drugs that can be enzymatically activated) must act intracellularly and have been difficult to design because of solubility versus cell permeability constraints (34, 35), and light-sensitive toxins have exhibited limited contrast in dark versus illuminated potency (36, 37) (fig. S1).

Fig. 1 DART pharmacology.

(A) Features of commonly used tools with proven utility in behaving animals. (B and C) Schematic of DART method. Standard drug (Rx) is acutely tethered to the surface of genetically specified cells with an enzymatic capture technology; see text for details. (D) Assay design. Hippocampal neurons are separately nucleofected and cocultured such that presynaptic neurons (white schematic cell) express ChR2, and postsynaptic neurons express GCaMP6s alone (green cell) or with HaloTagTM-2A-dTomato (yellow cell). Illumination induces glutamatergic release from presynaptic cells while simultaneously imaging postsynaptic GCaMP responses (movie S1). (E) Example readouts under conditions isolating AMPAR transmission (see materials and methods). ΔF/FO reflects AMPAR-mediated GCaMP activity in postsynaptic neurons. YM90K-DART has distinct effects on HT‒ (black) versus HT+ (red) cells. (F) Summary AMPAR data. Each symbol is the per-coverslip mean from HT‒ (black) and HT+ (red) cells from an independent coverslip; error bars are means ± SEM. Dashed line accounts for a small degree of rundown in the assay (fig. S2G); solid curves are binding-relation fits with Hill coefficient 1 and IC50 ± SEM (determined via nonlinear regression) as indicated. HT+ versus HT‒ comparisons via Wilcoxon rank sum test (***P < 0.0001; n = 53 coverslips). (G and H) Under conditions isolating NMDAR transmission (see materials and methods), there is no effect (n.s., not significant) of YM90K-DART up to 100 μM (n = 34 coverslips). The assay is sensitive to the standard NMDAR antagonist (10 μM CPP).

Method development

We sought to spatially restrict drugs to the surface of defined cells with HaloTag (38), an engineered bacterial enzyme designed to capture a small molecule called the HaloTag ligand (HTL). This reaction is covalent and highly efficient, and the components are otherwise inert (38). We envisioned fusing the HTL to a standard drug (Rx) via a long flexible linker (Fig. 1B). In particular, if the Rx portion of the compound had modest potency (50% inhibitory concentration IC50 > 1 μM), then low (100 nM) doses of the compound would have minimal pharmacological effects in free solution but would suffice for HTL capture, producing a high, spatially restricted Rx concentration near the HaloTag protein (Fig. 1B). This approach is related to a prior strategy that anchored drugs directly to engineered ion channels or receptors (2729, 31). However, overexpression of engineered receptors can have deleterious effects [for example, AMPAR overexpression can alter synaptic strength (39)], and it has not been technically feasible to express an engineered receptor from an endogenous gene locus with cell type specificity (31) (fig. S1D). Thus, we pursued an alternate approach in which HaloTag was expressed as a separate protein via fusion to a transmembrane (TM) domain—a configuration that would require no alteration or overexpression of endogenous receptors (Fig. 1B). In this scheme, cell type specificity arises from expression of HaloTagTM under a cell type–specific promoter. Molecular specificity is inherited from the Rx, which binds a specific endogenous protein [e.g., AMPAR but not NMDAR (N-methyl-d-aspartate receptor)]. Finally, acute onset upon delivery of the DART ligand is similar to that of traditional pharmacology—a unique threefold feature set (Fig. 1C).

We began with the AMPAR antagonist YM90K, an analog of DNQX with improved selectivity for AMPARs versus NMDARs. This drug is more potent than ideal for the method (IC50 ~ 80 nM) (40, 41); thus, we sought a site of linker conjugation that would attenuate potency (yielding IC50 > 1 μM). The crystal structure of an antagonist-bound AMPAR (42) and structure-activity relationships of YM90K analogs (40, 41) suggested that attachment to the 1-position of the quinoxaline ring would satisfy these requirements, likely allowing linker exit from the binding pocket while producing a potency attenuation due to steric bulk (fig. S2, A and B). Given uncertainty in optimal linker length, three variants were synthesized via polyethylene glycol (PEG) linker (43) conjugation to a 1-butyric acid variant of YM90K (40, 41) (fig. S2, C and D). The shortest (PEG12, length ~4.5 nm) was based on a minimum-distance structural analysis, and the longest (PEG36, ~14 nm) was chosen to avert tethered drug effects on neighboring cells [~20 nm between apposed membranes (44)]. Finally, HaloTag surface expression was optimized with endoplasmic reticulum (ER) export signals and codon optimization to maximize the density of drug that could be tethered (fig. S2, E and F).

Initial characterization was performed with an optical cocultured neuronal assay that quantified AMPAR transmission from “presynaptic” neurons expressing ChR2 [a genetically encoded opsin that depolarizes neurons in response to blue light (23)] to “postsynaptic” neurons expressing GCaMP6s [a genetically encoded green fluorescent indicator of neural activity (45)], either alone (HT−) or with HaloTagTM (HT+) (Fig. 1D; see materials and methods, below). Drug effects on HT− cells confirmed the known IC50 of 80 nM for YM90K and revealed a desired attenuation (IC50 of ~3 μM) for PEG-conjugated YM90K in free solution (fig. S3, A and B, black). Before drug addition, adjacent HT− and HT+ cells had similar activity (Fig. 1E, No Rx), indicating intact AMPAR function in HT+ neurons. The longest (PEG36) YM90K conjugate had distinct effects on HT− versus HT+ cells (movie S1): An HT− cell was minimally affected by the compound at 1 μM (Fig. 1E, black), whereas its HT+ neighbor exhibited pronounced AMPAR antagonism at 100 nM (Fig. 1E, red). Drug effects were reversed upon washout for the HT− cell but persisted on the HT+ cell, as expected for a covalently tethered drug (Fig. 1E, wash). These observations were consistent over many replicates, indicating a ~75-fold therapeutic window during drug incubation [i.e., IC50 for HT− cells corresponded to ~75 times the IC50 for HT+ cells (Fig. 1F)]. After washout, the method provided essentially perfect cell type specificity [i.e., nearly complete AMPAR antagonism on HT+ cells with no effects on HT− cells (Fig. 1F, wash)].

Whereas all length variants of the YM90K conjugate had similar HTL capture efficiency and Rx potency in free solution, the shortest had lower potency when tethered to HT+ cells (after washout; fig. S3C). This suggested that long PEG linkers are needed to span intermolecular HaloTag-AMPAR distances. We saw no evidence that the longest variant could affect adjacent HT− cells when tethered to an HT+ cell (fig. S3D), and a model of tethered diffusion indicated that drug concentration drops steeply over a length scale of ~2 nm (fig. S3E); hence, transcellular effects would be very unlikely to occur, even for intermembrane distances well below the ~20-nm estimate (44). Our longest PEG36 variant, termed YM90K-DART, was thus used for subsequent work.

To evaluate protein specificity, we modified the assay to isolate the NMDAR component of transmission (see materials and methods) on the basis that AMPAR antagonists can inhibit NMDARs at high doses (40). This assay was sensitive to the specific NMDAR antagonist CPP (Fig. 1, G and H), but even 100 μM YM90K-DART produced no evidence of NMDAR block for HT− or HT+ cells at the highest levels of HaloTagTM expression (Fig. 1, G and H, and fig. S3F). Further screening revealed no significant activity of 300 μM YM90K-DART against 30 molecular targets (fig. S3G).

Manipulation of endogenous AMPARs in slices of striatum and in behaving mice

The striatum is critical for mammalian motor control. It contains two major cell types that can be distinguished by expression of D1 versus D2 dopamine receptors (5). Most D1 cells are direct spiny projection neurons (dSPNs) that monosynaptically project to the substantia nigra reticulata (SNr), whereas D2 cells are mostly indirect spiny projection neurons (iSPNs) that connect to the SNr via intermediate nuclei. dSPNs and iSPNs are thought to have opposing roles in locomotion (7), are tightly intermingled, and together constitute >95% of striatal neurons. The remaining cells are interneurons, notably D1 fast-spiking and D2 tonically active cholinergic neurons (TANs) (7). Whereas striatal neurons express distinct dopamine receptors, they all express pharmacologically indistinguishable AMPARs.

We first tested DART in acute striatal slices from D1-cre or D2-cre mice injected with a cre-dependent HaloTagTM virus, thereby restricting expression to D1 or D2 cells (5, 7, 46). Illumination of ChR2-expressing corticostriatal fibers evoked AMPAR-mediated postsynaptic potentials (47) (Fig. 2A). AMPAR antagonism occurred with lower YM90K-DART doses for HT+ versus HT− cells, with effects essentially permanent on HT+ but reversible on HT− cells upon washout (Fig. 2B). Data from D1 versus D2 cells were not statistically different and were combined (Fig. 2C). Pooled data indicated a ~30-fold therapeutic window after 15 min of drug incubation (Fig. 2D, solid red versus black) and >300-fold after 15 min of washout (Fig. 2D, dashed red versus gray), thus validating the approach in brain tissue.

Fig. 2 Striatal slice validation.

(A) Acute striatal slices from mice with corticostriatal projections expressing ChR2 (blue) and SPNs expressing HaloTagTM-2A-dTomato (HT+) or a control virus (HT‒). (B) Example of optically evoked excitatory postsynaptic potentials (EPSPs) for HT‒ (black) and HT+ (red) SPNs at baseline, with 10 μM YM90K-DART, and after washout. Blue tick mark indicates optogenetic illumination. (C) Time course of drug effects. Thin traces are normalized peak EPSPs for individual cells, binned in 2-min intervals; error bars are means ± SEM across cells; curves are single-exponential fits. HT+ versus HT‒ comparisons are via Wilcoxon rank sum test (*P < 0.01, **P < 0.001). (D) Dose response. Symbols are individual cells; error bars are means ± SEM; curves are binding relations determined via nonlinear regression. Data underlying solid curves (circles) were collected after 15 min of drug incubation (Rx) for HT+ (red, IC50 = 0.6 ± 0.2 μM) and HT‒ (black, IC50 = 20 ± 5 μM). Data underlying dashed curves (squares) were collected after 15 min of washout (wash) for HT+ (light red, IC50 = 0.6 ± 0.2 μM) and HT‒ (gray, IC50 ~ 350 μM). HT+ versus HT‒ comparisons (*P < 0.01, **P < 0.001; black, Rx; gray, wash) are via Wilcoxon rank sum test.

Unilateral striatal manipulations are known to produce turning biases (7, 46). To test DART in behaving mice, we expressed HaloTagTM in D1 or D2 cells of the left dorsal striatum and implanted a unilateral cannula (Fig. 3A, top). Histology confirmed cell type–specific expression: Axonal projections to the SNr were labeled in D1-cre but not D2-cre animals (Fig. 3A). Mice were infused while awake and immediately monitored in an open-field arena (Fig. 3B, top). A strong turning bias in HT+ animals was seen after YM90K-DART infusion: D1-cre mice turned toward the manipulated hemisphere (left), whereas D2-cre mice turned in the opposite direction (right), indicating cell type–specific drug effects (movie S2). Turning manifested immediately after infusion, indicating rapid drug capture (Fig. 3B, red). Effects were constant over the first day, recovered by the next morning, and were qualitatively recapitulated by a second YM90K-DART infusion 1 week later (fig. S4C). This time course was consistent with dye-chase experiments using spectrally distinct cell-impermeant HTL dyes, indicating that HaloTagTM is replenished by continuous protein turnover (fig. S4B). No behavioral effects were observed when saline was infused into HT+ animals or when YM90K-DART was infused into HT− animals expressing a control virus (Fig. 3B, black and blue). Shown in Fig. 3, C and D, are behavioral summaries and a correlation of turning versus histological penetrance in each animal. These opponent responses to AMPAR antagonism in specified cells recapitulate recent results, in which transgenic (but not viral) optogenetic silencing produced similar behavioral effects (46).

Fig. 3 Validation in behaving mice.

(A) Top: Experimental design. D1- or D2-cre mice were unilaterally injected with a cre-dependent AAV2/1-DIO-HaloTagTM-2A-dTomato (HT+) or control (HT‒) virus, with cannula implanted in the left dorsal striatum. Bottom: Post-behavior histology shows striatal (Str) cannula placement, infusion volume (blue; 1 μl of fluorogold), and viral expression (red), which exhibited a strong projection to the SNr in D1-cre but not D2-cre mice. All images are displayed with identical contrast. (B) Top: Behavioral timeline. Mice were infused while awake (arrow; 1 μl; 10-min infusion, 5-min rest) and placed in an open-field arena for several 1-hour sessions (solid black boxes). The assay timeline was performed twice for each mouse, first with saline, then with 30 μM YM90K-DART (Rx). This YM90K-DART dose was one-tenth the amount needed to produce turning in HT‒ mice (fig. S4A). Bottom: Net number of 360° turns per 1-hour session (left minus right turns) for HT‒ (gray, blue) and HT+ (black, red) animals. One thin line per animal; thick lines and shading represent mean ± SEM across animals. For finer temporal analysis, rotations were analyzed in 15-min bins; data are means ± SEM. Top graph, D1-cre mice; bottom graph, D2-cre mice. (C) Day 1 rotation data (average of first three sessions). Each connected symbol pair represents one animal (saline, Rx); error bars denote means ± SEM; P values, Wilcoxon signed rank test. (D) Behavior data from (C) plotted as a function of penetrance of the manipulation, estimated from histology [e.g., (A); see materials and methods]; r2 and P values via regression slope analysis. See movie S2 for example open-field recordings; fig. S4, B and C, for HaloTagTM protein turnover analysis and repeated-dosing characterization; and fig. S4D for additional behavioral metrics.

Interrogating glutamatergic contributions to motor deficits in parkinsonian mice

Optogenetic manipulations of D1 versus D2 cell output have shaped theories of striatal function in mouse models of PD (7). PD also affects the many forms of input to these cells. In particular, dopaminergic inputs to striatal cells degenerate, leading to substantial remodeling of glutamatergic pathways (1012). Certain features of this remodeling are common to D1 and D2 cells, such as a reduction in the number of glutamatergic spines per cell (12). Other features are divergent: AMPARs in the remaining spines are subject to unopposed LTP (hyper AMPAR signaling) in D2 cells and LTD (hypo AMPAR signaling) in D1 cells (1012). It has been speculated that these divergent phenomena are “anti-homeostatic” (12); however, a direct test of the behavioral role of glutamatergic signaling onto D1 versus D2 cells has not previously been possible.

We used DART to interrogate glutamatergic inputs onto defined cells in the striatum of parkinsonian mice. HaloTagTM was expressed in the left dorsal striatum, with cannulas implanted in the left striatum (as before) and left medial forebrain bundle (MFB) (Fig. 4A). After surgery, mice exhibited normal locomotion, with an equal frequency of large-diameter left and right turns (Fig. 4D, bottom, dashed gray). Subsequent 6-OHDA infusion into the MFB (Fig. 4B) ablated dopaminergic afferents to the left striatum (Fig. 4C). Prototypical motor deficits of this PD model (48, 49) stabilized after ~7 days (fig. S5B). Akinesia manifested as a factor of 10 increase in 1- to 10-s periods of complete stillness (Fig. 4D, top, black). During motion, several movement execution deficits were apparent: Asymmetrical turning was pronounced with a loss of right turns and prevalence of pathologically narrow left turns (Fig. 4D, bottom, black), tortuosity increased ~50%, and bradykinesia was evident with velocity decreased to one-third of pre-lesion values (fig. S5).

Fig. 4 AMPAR contributions to parkinsonian motor deficits.

(A) Disease model. Cre-dependent HaloTagTM virus and cannula are placed in left striatum (Str), with additional cannula in left medial forebrain bundle (MFB). (B) Behavioral timeline. Solid boxes indicate 1-hour open-field sessions. Arrows indicate infusion of either 1 μl of 6-OHDA into MFB or 1 μl of 30 μM YM90K-DART into Str. Color codes for session type are defined below. (C) Histology in D2-cre mouse: 1 μl of fluorogold infused into Str (blue), HaloTagTM (red cells), and tyrosine hydroxylase staining (white) to confirm left striatal dopaminergic ablation. (D) D2-cell AMPAR antagonism is therapeutic across all behavioral metrics (movie S3 and fig. S5C). Top: Akinesia; connected symbols for a single animal [color key as in (B)]; error bars denote means ± SEM; P values, Wilcoxon signed rank test. Bottom: Frequency of 360° turns per hour, binned by turn diameter (bin size 1 cm); left turns positive, right turns negative. Curves and shading, mean ± SEM of 14 animals; P values, Wilcoxon signed rank test (see fig. S5C). (E) D1-cell AMPAR antagonism has no behavioral effects in PD animals (fig. S5D). (F) AMPAR antagonism of all cells (via cre-independent HaloTagTM virus) is therapeutic with regard to akinesia but has no effect on other behavioral metrics (fig. S5E). (G) Left: HaloTagTM-expressing TAN from a ChAT-cre mouse. Inset: YM90K-DART attenuates phasic but not tonic output of these cells. Right: AMPAR antagonism of TANs is therapeutic across all behavioral metrics except velocity (fig. S6B).

In these PD mice, acute AMPAR antagonism of D2 cells had pronounced therapeutic effects—akinesia, turning, velocity, and tortuosity were all significantly ameliorated (Fig. 4D, red, fig. S5C, and movie S3), and deficits returned after drug reversal (Fig. 4D, blue). This supports the hypothesis that unchecked LTP in D2 cells (1012) contributes to motor deficits and is thus antihomeostatic (12).

With regard to D1 cells, AMPAR antagonism had no behavioral consequence on any metric (Fig. 4E and fig. S5D). We confirmed with histology that the penetrance of the manipulation was not different in PD versus healthy mice (fig. S5A), in which the same manipulation produced leftward turns (Fig. 3). Thus, AMPAR antagonism of D1 cells did not augment leftward turns beyond that already seen in PD animals. This behavioral saturation may be a consequence of unopposed LTD in D1 cells (1012), such that further AMPAR antagonism has little behavioral consequence.

The results obtained thus far suggest that a nontargeted AMPAR antagonist, similar to that used in PD clinical trials (2022), may produce therapeutic effects dominated by D2 cells. To test this idea, we examined a third cohort of PD mice with HaloTagTM in all striatal cells. Here, AMPAR antagonism ameliorated akinesia (Fig. 4F, top), consistent with a dominant role for D2 cells. In contrast, beneficial effects on turning, velocity, and tortuosity were surprisingly absent in the same cohort (Fig. 4F, bottom, and fig. S5E), indicating that these movement execution deficits are sensitive to the ratio of AMPAR activity on D2 versus D1 cells. However, the interplay between these cells was not symmetric, which suggests a higher-order circuit interaction. One possibility, for example, is that isolated D2 cell antagonism may disinhibit D1 cells (50), thereby magnifying the impact of concurrent D1 cell antagonism. Such higher-order phenomena underscore the utility of DART, which enables defined cells to be individually and combinatorially interrogated.

We next asked whether therapeutic D2 cell effects were entirely mediated by iSPNs, which represent 95% of D2 cells, or whether the TAN population also contributes. TANs have been difficult to study because they produce two components of output (51): tonic firing, an intrinsic pacemaker activity independent of synaptic input, and phasic firing, which requires glutamatergic input (Fig. 4G, inset, black). Whereas hyperpolarization-based silencing methods attenuate both tonic and phasic output (52) (Fig. 4G, inset, gray), YM90K-DART selectively blocks the inputs that drive phasic firing, leaving tonic output intact (Fig. 4G, inset, red)—a manipulation not previously possible in a cell type–specific manner. We repeated our PD experiments in A2A-cre and ChAT-cre mice to target iSPNs and TANs, respectively (5, 7, 46). AMPAR antagonism of iSPNs recapitulated findings in D2-cre mice (fig. S6A), as expected. Strikingly, AMPAR antagonism of TANs also ameliorated nearly all metrics (except velocity; Fig. 4G and fig. S6B), despite the subtlety of the phasic-firing manipulation.

Extension to a metabotropic receptor

We next explored whether the method could extend to a structurally distinct target, such as the muscarinic acetylcholine receptor (mAchR). Unlike the open clamshell of the antagonist-bound AMPAR structure (fig. S2A), the antagonized mAchR has a very narrow path for tether exit (53) (Fig. 5A), which implies that success with this drug class would be a stringent test of generalizability. Docking of the prototypical antagonist, atropine, indicated that a primary hydroxyl was positioned near this opening (fig. S7A). Dehydration of this hydroxyl enabled reaction with a thiol-containing PEG6 spacer and conjugation via PEG36 linker to the HTL (fig. S7B). Efficacy of this atropine-DART was characterized with an optical assay of muscarinic activity using cocultured HT+ and HT− neurons (54) (see materials and methods). Drug effects on HT− cells confirmed the known IC50 of ~2 nM for atropine (fig. S7C), and a desired attenuation (IC50 ~4 μM) for atropine-DART in free solution (Fig. 5, B and C, black). HT− and HT+ cells had similar responses in the absence of drug (Fig. 5, B and C, No Rx), indicating intact mAchR function in HT+ neurons. Antagonism occurred with lower atropine-DART doses for HT+ versus HT− cells, providing a ~60-fold therapeutic window (Fig. 5, B and C, red versus black) with effects essentially permanent on HT+ but reversible on HT− cells upon washout (Fig. 5, B and C, blue versus gray). Atropine-DART had no effect on AMPAR or NMDAR signaling; conversely, YM90K-DART had no impact in mAchR assays, indicating preserved pharmacological specificity across these molecular classes (fig. S7, C to E). Thus, the method generalized to ionotropic (AMPAR) and metabotropic (mAchR) targets, despite their structural divergence.

Fig. 5 Extending DART to a metabotropic receptor.

(A) Structure-guided design of atropine-DART with tether positioned to exit through narrow opening in muscarinic acetylcholine receptor (mAchR; fig. S7A). (B) Cultured neuron assay of mAchR activity using Ca2+ elevations in response to 10 μM muscarine (54) (see materials and methods). Neurons were pre-incubated in 0 to 100 μM atropine-DART and assayed with drug present (black for HT‒, red for HT+) or after drug washout (gray for HT‒, blue for HT+). Waveforms and shading represent means ± SEM from 50 to 100 neurons. HT+ versus HT‒ comparisons for a given dose via Wilcoxon rank sum test (***P < 0.0001). (C) Muscarinic dose response. Colors and statistical significance are as above; error bars denote means ± SEM. Fits and shading represent IC50 ± SEM determined via nonlinear regression.

Discussion

Neuropsychiatric disorders have been examined with acute manipulations featuring either circuit or molecular specificity (2). By providing specificity at both of these levels, DART enables interrogation of specific proteins in defined cells. Here, we showed that PD motor deficits are causally linked to excess AMPAR signaling in iSPNs, in support of a model in which unchecked LTP in these cells is antihomeostatic (1012). Deficits were also attributed to AMPARs on TANs, implicating excessive phasic (rather than tonic) activity in the hypercholinergic features of the disorder (51, 52). Together, iSPNs and TANs (D2 cells) exerted dominant effects over akinesia, whereas deficits in movement execution were sensitive to the ratio of AMPAR activity in D2 versus D1 cells. These findings shed light on the circuit basis of glutamatergic contributions to PD and may inform clinical trials of AMPAR antagonists for the disorder (2022).

In this work, two DARTs were successfully designed on the first site of drug conjugation. In both cases, a crystal structure of the parent drug bound to its receptor was available (42, 53), facilitating selection of a conjugation site compatible with tether exit. Moreover, both drugs benefited from potency reductions, which minimized drug effects on HT− cells without sacrificing efficacy on HT+ cells. Many antagonists and potentiators can tolerate a similar potency reduction, and rational design will be facilitated by the crystallographic and medicinal literature.

Future implementations could incorporate enhanced features: Cell type specificity could be improved by enhanced HTL capture at lower doses (38). For systemic delivery, molecular Trojan-horse methodologies (55) may enable DART ligands to bypass the blood-brain barrier. For rapid reversal, inducible HaloTagTM degradation (5658), cleavable linkers (59), or photoswitchable drugs (60, 61) could be used. To target pharmacologically indistinguishable isoforms, subcellular HaloTag trafficking or cell type–specific HaloTag insertion into an endogenous gene locus could be explored. Finally, bio-orthogonal capture technologies (62) could allow delivery of two or more drugs to distinct cellular populations.

DART may provide a platform whereby the mechanism of action for widely prescribed drugs can be examined with cellular specificity in animal models of several disorders (2). Such studies could inform new translational strategies, by advancing nonobvious drug combinations (63, 64) or by providing a road map for the design of bivalent therapeutics based on the “message-address” concept of Schwyzer (65, 66). Such translational endeavors may benefit from the use of DART in animal models of disease to inform selection of pharmacological message, choice of cell type–specific address, and optimization of tether design (67).

Materials and methods

All experiments were conducted according to National Institutes of Health guidelines for animal research and were approved by the Janelia Research Campus Animal Care & Use and Biosafety Committees.

Mice

Mice were housed in a 12-hour reverse light cycle with food and water ad libitum. Behavioral experiments used Drd1A-Cre (GENSAT EY262), Drd2-Cre (GENSAT ER43), A2A-Cre (GETSAT KG139), and ChAT-Cre (Jackson Stock 006410) lines. For slice experiments, Drd1A-Cre and Drd2-Cre lines were crossed with Thy1-ChR2-Line18 animals (Jackson Stock 007612). Mice were singly housed after surgery.

Recombinant adeno-associated viral (rAAV) vectors

The following viral vectors were used: (i) rAAV2/1-CAG-DIO-HaloTagTM-2A-dTomato-WPRE; (ii) rAAV2/1-CAG-HaloTagTM-2A-dTomato-WPRE; (iii) rAAV2/1-CAG-DIO-tdTomato-WPRE (5 × 1012 GC/ml, Janelia virus core). CAG, promoter containing cytomegalovirus (enhancer, promoter, first exon), chicken β-actin (first intron), and rabbit β-globin (splice acceptor); DIO, double-floxed inverted orientation; HaloTagTM, described below; 2A, the self-cleavable p2A peptide; dTomato, red fluorescent protein; WPRE, woodchuck hepatitis posttranscriptional regulatory element.

Viral injections and cannula placement

Adult mice (10 to 20 weeks old) were stereotaxically injected with 0.8 to 1.2 μl of rAAV targeting the left dorsal striatum, distributed over 16 sites (4 tracks × 4 depths per track, 50 to 75 nl per site) to achieve uniform expression. Track coordinates: (0.5 ± 0.4 mm anterior to bregma) × (1.8 ± 0.4 mm left of midline); depths: 2.1, 2.5, 2.9, and 3.3 mm below dura. Mice were immediately implanted with plastic (peek) cannulas, which were favored over stainless steel to eliminate the hypothetical possibility that nitroaromatic compounds (such as YM90K) could be reduced by iron-containing surfaces. The striatal cannula (Plastics One; C315G/PK length 2.5 mm) was centered at the site of viral injection (0.5 mm anterior to bregma; 1.8 mm left of midline). For generation of parkinsonian mice, we used custom dual cannulas [Plastics One, C235G-1.8/PK length 2.5 mm (striatum) and 4.75 mm (MFB)], with identical striatal cannula placement, and MFB cannula 1.2 mm posterior to bregma; 1.2 mm left of midline. Mice were fitted with a plastic head bar adhered to the skull via UV-glue, enabling head fixation to facilitate drug infusions in awake animals. Mice were given 2 to 3 weeks to allow for viral expression, acclimation to head fixation, and recovery prior to behavioral assays.

Infusions, open-field assay, and 6-OHDA lesion

Mouse behavioral assays were performed with the experimenter blinded to viral payload in each animal. Drugs were infused in awake head-fixed animals via automated syringe pump (Harvard Apparatus PHD2000), Hamilton syringe (5 μl, model 75 RN SYR), and internal cannula (Plastics One, C315I) threaded into guide cannula (with 0.5-mm projection). For MFB infusions, 6-OHDA (3.6 mg/ml) was freshly dissolved in sterile degassed deionized water containing 0.1% ascorbic acid. For striatal infusions, 0 or 30 μM YM90K-DART was freshly dissolved in sterile ACSF consisting of 150 mM NaCl, 4 mM KCl, 2 mM MgCl2, 2 mM CaCl2, 10 mM HEPES, 10 mM glucose, pH 7.4. Of note, YM90K-DART is very water-soluble (to the high mM range) and thus does not require the use of solubilizers such as DMSO. In all cases, 1 μl was slowly infused (0.1 μl/min; 10 min total) followed by 5-min rest. Dummy cannulas were then replaced, and mice were immediately placed in a dark open-field arena (27 cm × 27 cm interior) and monitored via infrared video tracking (Noldus). Open-field monitoring was repeated for several 1-hour sessions (Figs. 3 and 4), and mice were returned to their home cage in the dark for intervening time intervals. After 6-OHDA infusion, mice were monitored daily and received daily diet supplementation and subcutaneous saline injections (1 to 3 ml, given after behavioral sessions). With these precautions, survival rate after 6-OHDA infusion was 100%. No stimulants (e.g., amphetamines or apomorphine) were used in this study for either healthy or PD animals; thus, all open-field recordings reflect spontaneous locomotion.

Behavioral analysis

Open-field videos were analyzed in a fully automated manner with a combination of commercial and custom-written scripts, as follows:

(i) Rotations: The positions of the nose, tail, and center of mass of each mouse were tracked using Noldus Ethovision XT 10 software, and analyzed offline in custom MATLAB scripts. Changes in orientation of the vector pointing from tail to nose were analyzed to identify vector rotations of at least 360° with no more than 25% cumulative rotation in the opposite direction (e.g., a 400° rotation to the right could contain no more than 100° cumulative hesitations to the left). Rotations were normalized (e.g., 400°/360° = 1.11 turns). False rotations attributed to nose/tail assignment flips were eliminated by excluding frames in which the nose-tail distance decreased to less than half of the median value. Left and right rotations were separately tallied, and the number of net turns (e.g., Fig. 3C) is the total of left minus right turns per hour. For analysis of rotations binned according to turn diameter (e.g., Fig. 4D, bottom), a sliding-window analysis was performed over each 360° portion of a turn; diameter was defined as the maximum Euclidean distance between all center-of-mass positions. Separate histograms were tallied for left and right turns (Fig. 4D, bottom).

(ii) Akinesia: Periods of immobility (continuous frames in which the average pixel change of the entire video image was less than 1%) were detected using Noldus Ethovision XT 10 (activity analysis). This definition of immobility was strict, such that any movement of the head, limbs or tail would be excluded (7). Cumulative akinesia was analyzed in custom MATLAB scripts to identify intervals of immobility lasting at least 1 s, but no more than 10 s (to exclude sleep).

(iii) Tortuosity: The center of mass of each mouse was tracked using Noldus Ethovision XT 10 software and analyzed offline in custom MATLAB scripts. Trajectories over a 1-hour session were segmented into intervals wherein the net Euclidian displacement (Dnet, linear distance between first and last point in an interval) reached a threshold value (Dnet,thresh). The integrated path length (Dpath), and tortuosity [= 1 – (Dnet/Dpath)] were calculated for each interval. Tortuosity over a session was averaged over all such intervals. Because DpathDnet, tortuosity would be 0 for straight trajectories and would approach 1 for infinitely curvy trajectories. Tortuosity monotonically increased as a function of Dnet,thresh; thus, we repeated the analysis for Dnet,thresh = 1, 2, 3, …, 12 cm and report the mean value.

Immunohistochemistry and analysis of histological penetrance

Mice were intracranially infused while awake with 1 μl of a fluorogold solution via striatal cannula. After 1 to 3 hours, mice were deeply anesthetized with isofluorane, then fixed via transcardial perfusion of PBS followed by 4% paraformaldehyde (PFA). The brains were dissected out of the skull, postfixed in 4% PFA at 4°C, then washed with PBS. The brains were sectioned (Leica, VT1200S) at 50 μm and mounted onto glass slides. Slides were air-dried, rehydrated with PBS, and then blocked in 5% normal goat serum (Jackson), 2% BSA (Sigma), and 0.3% Triton X-100 (Sigma) for 2 to 3 hours followed by incubation with rabbit anti-tyrosine hydroxylase (PelFreez, P40101-150; diluted 1:200 in half-block: 2.5% normal goat serum, 1% BSA, 0.15% Triton) for 2 days at 4°C. Slides were washed in PBS, 0.1% Tween, incubated with goat anti-rabbit AlexaFluor647 (Invitrogen, A21244; diluted 1:400 in half-block) for 2 to 3 hours at room temperature, and washed as before. Slides were coverslipped with Vectashield Hardset with DAPI (Vector Labs) and imaged on a Pannoramic 250 Flash II slide scanner (3DHISTECH). Analysis was performed in custom MATLAB scripts. For each coronal section, the striatum was manually segmented in both hemispheres. Only the dorsal halves of annotated regions were included in the analysis, and background fluorescence (median of right hemisphere) was subtracted. Histological penetrance in the left hemisphere (Fig. 3D) was calculated via a pixel-wise summation [Σ√(red × blue)] of the geometric mean of HaloTag expression (red) and infusion concentration (blue) over 25 coronal sections (cannula center ±12 sections; 50 μm per section). Note that this metric is ~0 for pixels lacking either blue or red, and thus reflects the integrated brain volume containing both HaloTagTM expression and drug infusion.

Slice electrophysiology

D1- or D2-cre mice crossed with Thy1-ChR2-Line18 animals (47) received striatal rAAV2/1-CAG-DIO-HaloTagTM-2A-dTomato-WPRE (HT+) or rAAV2/1-CAG-DIO-tdTomato-WPRE (HT−) viral injections at p21-24. Acute coronal slices of the dorsal striatum were prepared from p40-64 mice and incubated for ~30 min at 35°C in an ACSF solution containing 119 mM NaCl, 25 mM NaHCO3, 3 mM KCl, 1.25 mM NaH2PO4, 1 mM MgCl2, 1.3 mM CaCl2, 3 mM pyruvate, 1 mM ascorbate, 15 to 20 mM glucose (300 to 305 mOsm) saturated with 95% O2 / 5% CO2. Slices were subsequently kept at room temperature until experiments, typically within 4 hours of slice preparation. For electrophysiological recordings, slices were transferred to a recording chamber and constantly perfused at 4 to 5 ml/min with ACSF heated to 31° to 33°C. Whole-cell current-clamp recordings were performed using patch pipettes (resistance: 3 to 5 MΩ) pulled from Schott Glass #8250 (WPI Inc.) and filled with an internal solution containing 130 mM K-gluconate, 10 mM KCl, 4 mM NaCl, 4 mM Mg2-ATP, 0.3 mM Tris-GTP, 14 mM Tris-phosphocreatine, 10 mM HEPES, and 0.05 mM Alexa-488 (290 mOsm, pH 7.3 with KOH). Cells were selected according to their dTomato expression, visualized using a two-photon microscope, with verification of dTomato and Alexa-488 colocalization. Wide-field optogenetic stimulation was performed with a blue LED passed through the epifluorescence port of the microscope [Zeiss examiner Z1 with a 63× objective; light flash duration 1 to 2 ms; intensity adjusted to yield ~5 to 10 mV baseline excitatory postsynaptic potentials (EPSPs)]. Stimuli were given once every 20 s. Synchronization of illumination and electrophysiology recordings, as well as subsequent data analysis, were performed with custom MATLAB scripts (Fig. 2).

Cultured neuronal assays of synaptic transmission

Hippocampal neurons from postnatal day 0 to 1 Sprague-Dawley rat pups were isolated and dissociated with papain and gentle trituration as described (45). Cells were nucleofected (LONZA, V4SP-3096 or V4SP-3960), with 1.5 × 106 viable cells per shuttle well, along with the following constructs: (i) pre: SYN-hChR2(H134R)-dsfGFP-WPRE (1 μg/shuttle); (ii) HT− post: SYN-GCaMP6s-WPRE (0.3 μg); (iii) HT+ post: SYN-GCaMP6s-WPRE (0.3 μg) + CAG-HaloTagTM-2A-dTomato-WPRE (1 μg). SYN, synapsin promoter; CAG, promoter containing cytomegalovirus (enhancer, promoter, first exon), chicken β-actin (first intron), and rabbit β-globin (splice acceptor). hChR2(H134), mammalian codon-optimized channelrhodopsin (23); dsfGFP, chromophore-deleted (dark) superfolder green fluorescent protein; GCaMP6s, genetically encoded calcium indicator (45); HaloTagTM, described below; 2A, the self-cleavable p2A peptide; dTomato, red fluorescent protein; WPRE, woodchuck hepatitis posttranscriptional regulatory element. After a 10-min recovery period, nucleofected cells were mixed together, and plated on glass coverslips (Deckglaser) coated with high molecular weight poly-d-lysine (Sigma). Cultures were maintained in NbActiv4 (BrainBits), at 37°C in a 5% CO2 incubator. Half of the media was changed at 4 days, and once per week thereafter. Experiments were performed 16 to 17 days after plating.

For AMPAR synaptic transmission assays (Fig. 1, E and F), coverslips were transferred to 24-well glass-bottom plates containing 1 ml of a resting solution: 150 mM NaCl, 4 mM KCl, 2 mM MgCl2, 2 mM CaCl2, 10 mM HEPES, 10 mM glucose, pH 7.4, with 10 μM gabazine (to block GABAA inhibitory transmission) and 10 μM CPP (to block NMDAR transmission). Plates were transferred to an Olympus IX81 inverted wide-field microscope with a 10× air objective, mercury arc lamp illumination, and fluorescein isothiocyanate (FITC) (excitation: 460 to 500 nm; dichroic: 505 nm long-pass; emission: 510 to 560 nm) and TxRed (excitation: 533 to 588 nm; dichroic: 595 nm long-pass; emission: 608 to 683 nm) filter sets. One field of view was selected for each coverslip (12 coverslips per batch), and baseline dTomato images were obtained to indicate the degree of HaloTagTM expression in each cell. Cells with dTomato ≤ 5 fluorescence units were designated HT− and cells with dTomato ≥ 100 units designated HT+; a detailed analysis of intermediate levels of dTomato expression appears in fig. S3C. To assay AMPAR signaling, each coverslip was subjected to 16 frames of 3-Hz FITC illumination (exposure 50 ms), yielding a single trial (e.g., Fig. 1E, light gray traces). This was repeated six times for each coverslip, with 2 min between repetitions (e.g., Fig. 1E, heavy black and red traces are the mean of six repetitions). Experimental drugs were added by manual pipetting, and drug washout was achieved by exchanging two-thirds of the media 10 times. Analysis was performed in custom MATLAB scripts. GFP fluorescence was background subtracted (mean of cell-free region), and responses quantified as change in fluorescence divided by baseline fluorescence (i.e., ΔF/FO). Steady-state binding relations (Fig. 1F) report the time-averaged response for each dose (mean of 16 frames × 6 reps), normalized to the No-Rx condition for each cell.

NMDAR assays (Fig. 1, G and H) were identical to above, except solutions contained 150 mM NaCl, 4 mM KCl, 0.8 mM MgCl2, 2 mM CaCl2, 10 mM HEPES, 10 mM glucose, pH 7.4, with 10 μM gabazine (to block GABAA transmission), 10 μM NBQX (to block AMPAR transmission), and 50 μM glycine (co-agonist of NMDAR).

Cultured neuronal mAchR assay

Hippocampal neurons were dissociated and separately nucleofected with (i) HT−: CAG-SNAPfTM-WPRE (1 μg); or (ii) HT+: CAG-HaloTagTM-2A-dTomato-WPRE (1 μg). After a 10-min recovery period, nucleofected cells were mixed together, plated on poly-d-lysine–coated glass coverslips, and maintained in NbActiv4 (BrainBits), at 37°C in a 5% CO2 incubator. Experiments were performed 16 to 17 days after plating. Data was pooled from six independent cultures, with batch effects minimized by testing all drug conditions (Fig. 5, B and C) in parallel, as follows. Coverslips were pre-incubated for 30 min at 37°C in NbActiv media containing 1 μM SNAP-Surface Alexa Fluor 647 (to label HT− cells; New England Biolabs) and 2.5 μM Fluo2-a.m.-LeakRes (a green Ca2+ indicator; TefLabs; final 0.1% DMSO; 0.01% pluronic F-127). Coverslips were transferred to 24-well glass-bottom plates containing 0.5 ml of a resting solution: 150 mM NaCl, 2 mM MgCl2, 2 mM CaCl2, 10 mM HEPES, 10 mM glucose, pH 7.4, with 10 μM gabazine (to block GABAA inhibitory transmission), 10 μM NBQX (to block AMPAR transmission), 10 μM CPP (to block NMDAR transmission), and 1 μM tetrodotoxin (to block action potentials). Under these conditions, Ca2+ responses to muscarine are known to reflect the loading state of intracellular Ca2+ stores; thus, we used 7 mM KCl and 5 μM BayK to amplify Ca2+ responses to muscarine, as described (54). For each coverslip, one field of view was selected and imaged on an inverted IX81 microscope in multichannel time-lapse imaging mode: FITC, TxRed, and CY5 (excitation: 590 to 650 nm; dichroic: 660 nm long-pass; emission: 663 to 738 nm) channels were imaged once every 6 s. After 30 s of baseline imaging, 0.5 ml of a stimulation solution (= resting solution + 20 μM muscarine chloride) was added to produce a final concentration of 10 μM muscarine, and time-lapse imaging was continued for 2 min. In cases involving atropine-DART, coverslips were pre-incubated for approximately 20 min in resting solution that additionally contained 0 to 100 μM atropine-DART, and stimulated with drug present (Fig. 5, B and C, black/red for HT−/+) or following drug washout via five coverslip transfers into a resting solution lacking atropine-DART (Fig. 5, B and C, gray/blue for HT−/+). Analysis was performed in custom MATLAB scripts. Each channel was background-subtracted (mean of cell-free region). Alexa647-positive cells were designated HT− and cells with dTomato ≥ 50 units designated HT+. Calcium responses were quantified as change in FITC fluorescence divided by baseline fluorescence. Dose response relations (Fig. 5C) report the time-averaged response for 20 s after stimulation.

Genetic constructs

The final coding sequence for HaloTagTM contained the following elements (fig. S2E): SSnlg-HA-mHT-ΔECD.TM.CTnlg-ERXL, as follows: SSnlg, the signal peptide (residues 1 to 49) of mouse neuroligin-1; HA, the hemagglutinin epitope tag; mHT, mammalian codon-optimized (DNA2.0) variant of HaloTag7 (38); ΔECD.TM.CTnlg, the esterase-truncated 71-residue extracellular domain, the 19-residue predicted transmembrane domain, and the 127-residue C terminus of mouse neuroligin-1 (68); ERXL, the peptide sequence KSRITSEGEYIPLDQIDINVGGSGFCYENEV, a fusion of the trafficking and ER export signals from Kir2.1 (69). These elements were concatenated via standard PCR, restriction digest, ligation, and sequence verification.

Chemical synthesis

YM90K-DART and atropine-DART ligands were synthesized via two-step (fig. S2D) and four-step (fig. S7B) procedures, respectively, as detailed in the supplementary materials.

Statistics

Reported values are means ± SEM. Statistical tests were performed in MATLAB via Wilcoxon rank sum test for unpaired data, Wilcoxon signed rank test for paired data, regression slope analysis for behavior versus penetrance analysis, or nonlinear regression (MATLAB curve fitting toolbox) for estimates of IC50 ± SEM; P values are numerically indicated in each figure or legend. No statistical methods were used to predetermine sample size. Mouse behavioral analysis was fully automated, and assays were performed with the experimenter blinded to viral payload. HaloTag positive/negative injections were performed in littermate controls whenever possible.

Supplementary Materials

www.sciencemag.org/content/356/6333/eaaj2161/suppl/DC1

Figs. S1 to S7

Movies S1 to S3

Synthesis Methods

References (7080)

References and Notes

  1. Acknowledgments: We thank J. Magee, A. Lee, G. Murphy, D. Stern, M. Zlatic, and A. Hantman for discussions and feedback on the manuscript; C. Ruiz, S. DiLisio, K. Morris, M. Rose, A. Zeladonis, and J. Cox for surgical procedures, mouse care, and breeding; K. Ritola and K. McGowan for viral preps; S. Vaidya, D. Hunt, W. Sun, and W. Legant for pilot experiments; R. Egnor for input on behavioral protocols; J. Kuhl for artwork; and V. Custard for administrative support. Radioligand assays (fig. S3G) were generously provided by the National Institute of Mental Health’s Psychoactive Drug Screening Program, contract HHSN-271-2008-025C (NIMH PDSP), which is directed by B. L. Roth at the University of North Carolina at Chapel Hill and project officer J. Driscol at NIMH. This research was funded by the Howard Hughes Medical Institute. All data and reagents are available upon request.
View Abstract

Navigate This Article