Fluorescent ligands for dopamine D2/D3 receptors

Fluorescent ligands are versatile tools for the study of G protein-coupled receptors. Depending on the fluorophore, they can be used for a range of different applications, including fluorescence microscopy and bioluminescence or fluorescence resonance energy transfer (BRET or FRET) assays. Starting from phenylpiperazines and indanylamines, privileged scaffolds for dopamine D2-like receptors, we developed dansyl-labeled fluorescent ligands that are well accommodated in the binding pockets of D2 and D3 receptors. These receptors are the target proteins for the therapy for several neurologic and psychiatric disorders, including Parkinson’s disease and schizophrenia. The dansyl-labeled ligands exhibit binding affinities up to 0.44 nM and 0.29 nM at D2R and D3R, respectively. When the dansyl label was exchanged for sterically more demanding xanthene or cyanine dyes, fluorescent ligands 10a-c retained excellent binding properties and, as expected from their indanylamine pharmacophore, acted as agonists at D2R. While the Cy3B-labeled ligand 10b was used to visualize D2R and D3R on the surface of living cells by total internal reflection microscopy, ligand 10a comprising a rhodamine label showed excellent properties in a NanoBRET binding assay at D3R.


Results
Ligand design and synthesis. For the development of fluorescent probes targeting the D 2 -like receptor family, we made use of four different privileged scaffolds that are frequently found in dopamine receptor ligands 30 . The N-propyl substituted 2-amino-dihydro-1H-indene (building block A, Fig. 1), is a well-known dopamineisostere and exhibits agonistic properties 31,32 . In contrast, 1,4-disubstituted phenyl-or pyrimidyl-piperazines (1,4-DAPs, building blocks B-D, Fig. 1) with different substituents attached to the aromatic core represent the main receptor recognition element of atypical antipsychotics such as aripiprazole and cariprazine 30 . While the 2-methoxy or 2,3-dichloro substituted phenylpiperazines are known to bind to all D 2 -like receptor subtypes, the pyrimidylpiperazine (building block D) selectively targets D 3 R 33 . Previous studies have revealed the importance of a second lipophilic moiety for high-affinity binding to dopamine receptors 30,34 . Thus, we envisioned connection of the primary pharmacophores to a triazole-benzylamine or triazole-benzoic acid moiety through a flexible four-carbon aliphatic linker. This linker size was chosen so that ligands should possess high affinity for both D 2 R and D 3 R, while a considerably shorter linker would have been beneficial for the D 4 R subtype 30 . For an initial set of fluorescent ligands, we focused on the incorporation of a dansyl label, a naphthalene-derived fluorophore emitting green light. The dansyl fluorophore has a relatively low molecular weight and is commercially available as reactive sulfonyl chloride (dansyl chloride). The dansyl dye is widely used to label proteins for fluorescence polarization measurements due to its favorable lifetime or as an acceptor for resonance energy transfer from protein tryptophan residues. It is also known for its large Stokes shift, resulting in a good signal-to-noise ratio 35 . Since we have previously found that the binding pockets of D 2 R and D 3 R can accommodate flexible linkers 14 but also sterically more demanding substituents in their extended binding pocket 36,37 , installation of the dansyl fluorophore was envisioned through sulfonamide-formation either directly with the primary benzylamine (ligands 4a-d) or through addition of a short 1,3-diaminopropane spacer to the benzoic acid precursors (ligands 8a,b). The latter strategy has already been successful for the development of Cy3B-labeled dopamine receptor ligands in the context of TIRF microscopy 14,17 .
Despite the advantages of the dansyl label, it is not ideally suited for all popular fluorescence-based technologies, for example bioluminescence resonance energy transfer. For instance, its fluorescence properties are highly sensitive to solvent polarity and the fluorescence excitation maximum (λ max ≈ 350 nm) 35 prevents usage in NanoBRET assays. In recent years, many organic fluorophores with emission maxima in the red spectral range and enhanced photochemical properties have been developed. Chemically, these fluorophores often belong to the xanthene or cyanine dye families, but differ in their substituents, leading to a variety of net charges and different degrees of hydrophilicity 44 . We selected three different fluorophores, the commercially available dyes Alexa488 and Cy3B and a recently described bis-trifluoroethyl substituted rhodamine 38 for the synthesis of our fluorescent dopamine receptor probes. Alexa488 and Cy3B, but also tetramethylrhodamine dyes (TAMRA), have already www.nature.com/scientificreports/ been shown to be suitable for characterizing ligand binding to GPCRs 11,22,29,45 , and especially Cy3B-labeled ligands have proven useful for fluorescence microscopy 14,17,29 . For the synthesis of these ligands, we focused on the N-propyl substituted 2-amino-dihydro-1H-indene scaffold. 5 -Carboxy-N,N -bis(2,2,2-trifluoroethyl)rhodamine (9) was prepared as described previously 38 and connected directly to the benzylamine 3a using TBTU as the coupling reagent to give the desired fluorescent ligand 10a in a yield of 28%. Alexa488 and Cy3B were purchased as tetrafluorophenyl (TFP) or N-hydroxysuccinimid (NHS) esters, respectively, and reacted with the 1,3-dipropylamino-substituted benzoic acid 7a under basic conditions in DMF, affording the fluorescent ligands 10b (Cy3B) and 10c (Alexa488).

Molecular docking.
To explore whether addition of the fluorescent moieties would be tolerated by the dopamine receptors and to explore their location upon binding of the probes to D 2 R and D 3 R, docking studies were performed with the representative ligands 8a and 10b, comprising a dansyl and a Cy3B fluorophore, respectively, and the crystal structures of D 2 R (complex with risperidone, PDB-ID: 6CM4 3 ) or D 3 R (complex with eticlopride, PDB-ID: 3PBL 5 ). As expected from the ligand design, typical receptor-ligand interactions in the orthosteric binding pocket 3,5 including the salt bridge formation with Asp 3.32 were observed for the indanylamine moiety of 8a and 10b in both receptors (Fig. 2, Supplementary Fig. S1). These results indicate that the presence of the fluorescent moiety does not hinder the pharmacophore from adopting its canonical binding pose. Thus, the addition of long and flexible aliphatic spacers, which were previously employed in fluorescent probes targeting D 2 R or D 3 R 14,17 , is not strictly required. According to the docking results, the fluorescent moieties of both ligands lay between the extracellular loops of the receptors, mainly interacting with hydrophobic amino acid sidechains. Between the two receptor subtypes, binding poses of the ligands only differed marginally (Fig. 2, Supplementary Fig. S1).  www.nature.com/scientificreports/ Binding affinity and functional activity. Affinities for the dopamine receptor subtypes D 1 -D 4 along with the related 5-HT 1A , 5-HT 2 and α 1 receptors were determined by radioligand binding for the synthesized fluorescent ligands and some of the central intermediates (Table 1, Supplementary Table S1). In general, ligand affinities were found to be mostly dependent on the pharmacophore. Ligands comprising an N-propyl-2-aminoindane or phenylpiperazine substructure showed K i values in the low nanomolar range for all D 2 -like receptors, with the highest affinities observed at the D 3 R subtype (0.26-13 nM), while the affinities for the D 1 R were substantially lower (180 -> 10,000 nM). The direct comparison of the dansyl-labeled ligands 4a-c with their benzonitrile precursors 2a-c clearly demonstrates that the insertion of the naphthyl-derived fluorophore is well tolerated by D 2L R, D 2S R, D 3 R and D 4 R, as highly similar binding affinities were determined in the presence and absence of the fluorescent moiety. As expected, ligands 2d and 4d, containing a pyrimidylpiperazine as primary dopamine receptor recognition element, were found to be selective for the D 3 R subtype. The introduction of the dansyl moiety further increased the D 3 R-selectivity, as it had no major influence on the D 3 R affinity (9.5 vs. 11 nM), but decreased the affinity for the D 2L R and D 2S R subtypes by a factor of ~ 20. Since 4d is also selective for D 3 R over the investigated serotonin and α 1 receptors, it could potentially be used in fluorescence microscopy for D 3 R localization studies in tissues. Connection of the dansyl label and the lipophilic moiety through a short 1,3-diaminopropylene-spacer did not substantially improve binding affinities at the D 2 R and D 3 R, indicating The stick representation of the indanylamine pharmacophore in the orthosteric binding pocket shows an ionic interaction of the protonated nitrogen with D114 3.32 , while the aromatic ringsystem is accommodated in a hydrophobic pocket formed by V115 3.33 , C118 3.36 , T119 3.37 , S197 5.46 , F198 5.47 , F382 6.44 , W386 6.48 and F389 6.51 . The propyl substituent of the protonated amine points into a hydrophobic cleft formed by W386 6.48 , F390 6.52 and T412 7.39 . (c) Surface representation of the extracellular loop region. 10b is colored yellow, while EL1, EL2 and EL3 are colored in green, red and cyan, respectively. The polar sulfonate group of 10b is directed outwards towards the solvent. Table 1. Binding affinities of the test compounds. Binding affinities of the test compounds for human D 2L R, D 2S R, D 3 R, D 4 R were determined by radioligand competition. a Data represent mean ± SEM of (n) individual experiments, each performed in triplicates.
2d 360 ± 60 (4) 350 ± 30 (4) 9.5 ± 1.9 (4) 5,000 ± 300 (4) 7,600 ± 3,000 (4) 6,000 ± 1,400 (4) 11 ± 2 (4) > 10,000 (4) 13 ± 6 (4) 3.6 ± 1.6 (4) 1.9 ± 0.9 (4) 130 ± 10 (4) 10a (NMP160) 46 ± 7 (4) 21 ± 4 (4) 0.97 ± 0.23 (4) 56 ± 4 (4) 10b (NMP137) 10 ± 2 (5) 4.8 ± 1.0 (5) 0.90 ± 0.26 (5)  www.nature.com/scientificreports/ that this additional spacer is not strictly required for the small dansyl fluorophore. However, with a K i value of 0.44 nM, ligand 8a displayed the highest D 2S R affinity of the entire series. When the small dansyl fluorophore was exchanged for the sterically more demanding xanthene or cyanine dyes, binding affinities slightly decreased for the D 2S and D 2L receptors (4.8-46 nM). Similar to their dansyl analogs, fluorescent ligands 10a-c displayed the strongest binding affinity for the D 3 R subtype (0.76-0.97 nM). With the exception of the D 3 R-selective ligands 2d and 4d, the series of fluorescent ligands and precursors also showed affinity for 5-HT 1A R (13-260 nM) and α 1 -AR (1.1-150 nM). This is not unexpected, since the 2-methoxyphenylpiperazine is a common motif in α 1 -AR antagonists 46 . At 5-HT 2 R, observed affinities ranged from 220 to 2,200 nM, only. In order to evaluate the impact of the fluorophores on the functional activity of the ligands, we determined their capacity to elicit β-arrestin-2 recruitment to the D 2S R employing an assay based on enzyme fragment complementation (Pathhunter, DiscoverX). Since the presence of GRK2 is known to be important for a sensitive detection of ligand effects in HEK293 cells, GRK2 was coexpressed together with the D 2S R fused to the enzyme donor. When the cells were incubated with the fluorescent ligands 8a and 10a-c, substantial stimulation of β-arrestin-2 recruitment was observed, confirming the agonistic nature of the N-propyl-2-aminoidane dopamine-isostere ( Supplementary Fig. S2). In good agreement with its high binding affinity, the dansyl-labeled ligand 8a acted as a full agonist (E max 98 ± 2%) and displayed a potency (EC 50 51 ± 11 nM) that is only slightly lower compared to the reference agonist quinpirole (EC 50 20 ± 3 nM, E max 100 ± 1%). Incorporation of the larger trifluoroethyl-rhodamine, Cy3B or Alexa488 fluorophores led to a two-to eightfold reduction in ligand potency alongside with slightly reduced ligand efficacy for the fluorescent agonists 10a (EC 50 410 ± 70 nM, E max 89 ± 3%), 10b (EC 50 410 ± 60 nM, E max 79 ± 3%) and 10c (EC 50 100 ± 20 nM, E max 93 ± 3%), respectively. In agreement with our docking studies, these results indicate that D 2 -like receptors can accommodate the small dansyl-but also larger xanthene or cyanine-derived fluorescent ligands. However, not only the size but also the type of the fluorophore can have an impact on binding affinity, intrinsic activity and potency. This is illustrated in particular by the Cy3B-labeled ligand 10b, which has threefold higher affinity but fourfold lower potency for β-arrestin-2 recruitment at D 2S R compared to 10c that only differs from 10b in terms of the fluorophore. TIRF microscopy. Fluorescence microscopy, in particular TIRF microscopy, is a powerful method to study the expression, distribution and interactions of GPCRs in the cytoplasmic membrane with high spatial and temporal resolution [14][15][16][17]47,48 . We employed our previously developed protocol for the imaging of dopamine receptors 14 to verify that our newly developed fluorescent ligands are generally suitable for fluorescence imaging. For TIRF microscopy, we focused on the Cy3B-labeled ligand 10b, as the same fluorophore was previously used for the imaging of D 2 R and D 3 R 14,17 . Indeed, when CHO cells stably expressing D 2S R or D 3 R were labeled with 10b in 10 nM (D 2S R) or 1 nM (D 3 R) concentration, receptors were detected in the cytoplasmic membrane as discrete fluorescent spots (Fig. 3a,b). While D 3 R was found to be evenly distributed over the entire cell membrane, fluorescently labeled D 2S R showed a more inhomogeneous distribution with clusters of fluorescent puncta. Employing a set of different fluorescently labeled D 2 R ligands, we could previously show that these clusters correspond to internalized receptors 17 . It is not surprising that we could not observe such clusters for D 3 R, because D 3 R is known to hardly undergo agonist-mediated β-arrestin recruitment and internalization 49 . When cells were pretreated with the D 2/3 R antagonist spiperone, neither clustered fluorescent puncta nor labeling of the receptors at the cell surface were observed, indicating that non-specific binding and uptake of ligand 10b were negligible (Fig. 3c,d).
NanoBRET. To further take advantage of the newly developed ligands as fluorescent probes, we planned to establish a NanoBRET assay for the detection of ligand binding at dopaminergic receptors. In this assay, ligand binding is detected through RET between the bioluminescent Nluc 19 enzyme fused to the receptor N-terminus and the fluorescent ligand, which only occurs if the two molecules are in sufficient proximity to each other 11 . As a first step, we obtained absorbance and emission spectra of ligands 8a and 10a-c to identify most suitable candidates. As shown in Fig. 4, ligands 8a, 10a and 10c possess a maximum emission in the range of 540-550 nm, while the emission maximum of the Cy3B fluorophore in 10b is red-shifted (emission maximum at ~ 600 nm). Analysis of the absorption spectra confirmed the expected spectral properties of the employed fluorophores. While absorption occurred only in the UV range for the dansyl-labeled ligand 8a, compounds 10a-c showed significant absorption in the area of 450-550 nm (up to 600 nm for 10b). Thus, ligands 10a-c could serve as BRET acceptors in combination with the Nluc enzyme as BRET donor, which shows luminescence in the range of 400-550 nm (maximum at 460 nm) 19 .
For the development of the assay, we focused on the D 3 R, since our fluorescent ligands showed the highest affinity for this receptor subtype. To this end, a membrane targeted Nluc enzyme 11 (secNluc, Promega) and D 3 R were fused in frame by polymerase chain reaction and cloned into pcDNA3.1 for mammalian expression. A second construct carrying an N-terminal HA export sequence and a FLAG-tag 50 in front of the Nluc was generated, which allowed the detection of cell surface expression by ELISA 17 . Upon transient transfection into HEK293T cells, both variants of the Nluc-D 3 R fusion protein were well expressed, as determined by radioligand saturation (B max 2,500-16,200 fmol·mg −1 protein for secNluc-D 3 R and 5,000-5,700 fmol·mg −1 protein for FLAG-Nluc-D 3 R, compared to 2,200-4,800 fmol·mg −1 protein for wild type receptors (wtD 3 R), respectively), and both constructs showed the expected Nluc emission spectra in the presence of the substrate furimazine ( Supplementary Fig. S3). A direct comparison of Flag-D 3 R and FLAG-Nluc-D 3 R in an ELISA with an antibody directed against the N-terminal FLAG-tag ( Supplementary Fig. S3) showed that the presence of the N-terminal enzyme even improved cell surface expression. On the other hand, the N-terminal Nluc had no influence on the receptor-ligand recognition properties, as binding affinities for the reference ligands haloperidol, cariprazine, aripiprazole and fluspirilene were found to be highly similar to those obtained with wtD 3 R (Supplementary Table S2 www.nature.com/scientificreports/ assay (IPOne, Cisbio) with the reference agonist quinpirole, even slightly better potencies were observed for the Nluc-D 3 R fusion constructs compared to unmodified receptors ( Supplementary Fig. S3, EC 50 ± SEM: 14 ± 4 nM for wtD 3 R, 3.7 ± 0.4 nM for secNluc-D 3 R and 4.6 ± 0.8 nM for FLAG-Nluc-D 3 R, respectively), which are likely a result of their higher expression level. To find out whether fluorescent ligands 10a-c are indeed suitable for BRET-based ligand binding assays, we performed saturation binding assays with live HEK293T cells expressing secNluc-D 3 R. For all three ligands, typical saturation hyperbolas were observed (Fig. 5). Application of 10 µM haloperidol efficiently prevented binding of the fluorescent ligands and demonstrated a low contribution of non-specific binding to the detected netBRET signal. Analysis of the K D values revealed identical affinities for the trifluoroethyl-rhodamine and Alexa488labeled ligands 10a and 10c (K D ± SEM 0.72 ± 0.07 nM for 10a, n = 5; 0.72 ± 0.08 nM, n = 4 for 10c, respectively), that were in good agreement with their K i values obtained by radioligand competition (Table 1). In contrast, the determined affinity of the Cy3B-derivative 10b was substantially lower (K D ± SEM 12.1 ± 1.7 nM, n = 4), which was surprising, given that its radioligand K i value (0.90 ± 0.26 nM) was similar to those of 10a and 10c. Almost identical results were obtained, when additional saturation experiments with 10a (K D ± SEM: 0.90 ± 0.14 nM, n = 5) and 10b (K D ± SEM: 11.3 ± 1.6 nM, n = 4) were performed with membranes from HEK293T cells expressing secNluc-D 3 R instead of whole live-cells (Supplementary Fig. S4). This indicates that the higher K D value of 10b observed by NanoBRET is not a result from general differences between membrane and whole-cell assays.
Association and dissociation experiments performed with fluorescent ligand 10a at room temperature confirmed its excellent properties and showed a concentration dependent association profile (Fig. 6a), with an association rate constant of 1.02 ± 0.30 × 10 7 min −1 × M −1 (K on ± SEM, n = 10, Supplementary Table S3). As expected, dissociation of 10a was independent from the employed concentration (Fig. 6b), resulting in a mean residence time of 19 ± 1 min (mean ± SEM, n = 3). This is about 1.5-fold faster than previously observed for the radioligand [ 3 H]spiperone at D 3 R 51 .
Taking advantage of the high-affinity binding and the excellent signal to noise ratio of 10a in the membranebased BRET assay, we sought to determine the affinity of the antipsychotics haloperidol, fluspirilene, cariprazine www.nature.com/scientificreports/ and aripirazole for a direct comparison with the data from radioligand binding experiments (Fig. 6c,d). Independent from the employed concentration of 10a (10 nM or 100 nM), obtained binding affinities of the four antipsychotics showed nearly the same rank order and were only slightly lower (Supplementary Table S4, Fig. 6d) than those obtained by classical radioligand binding (Supplementary Table S2, Fig. 6c). These results demonstrate that 10a can be successfully employed in membrane-based NanoBRET assays to determine the K i values of non-labeled ligands.

Discussion
Fluorescent ligands represent versatile tools for the investigation of diverse biological questions. Similar to radioligands, they can be detected in very low concentration and with high specificity. In the context of GPCR research, fluorescent ligands have been successfully employed to study receptor internalization 17 , receptorreceptor interactions within the cellular membrane [14][15][16]18 , and recently also ligand binding 11,[22][23][24]29,45,53 , by techniques like fluorescence microscopy, fluorescence polarization and resonance energy transfer. Starting from phenylpiperazine and indanylamine scaffolds, known dopamine-isosteres with antagonistic or agonistic properties, respectively, we have designed and synthesized a small library of fluorescent ligands for D 2 -like receptors. Our initial efforts were directed towards the synthesis of dansyl-labeled probe molecules, as this fluorophore is cost efficient and readily available as reactive sulfonyl chloride. The obtained fluorescent ligands exhibited binding affinities for D 2 R and D 3 R in the subnanomolar range and ligand 8a also acted as highly potent D 2 R agonist. Despite these favorable characteristics, the application of the ligands is hampered by the fluorescence properties of the dansyl dye, that are in principle amenable to fluorescence microscopy, but not optimally suited for modern RET applications. Encouraged by the binding properties of ligands 4a and 8a, we www.nature.com/scientificreports/  www.nature.com/scientificreports/ exchanged the dansyl fluorophore by cyanine or xanthene moieties, both of which have been frequently used in fluorescence microscopy or NanoBRET binding assays 11,14,17,22,29,45 . While this exchange had almost no influence for ligand recognition properties at D 3 R, the fluorophore slightly reduced binding affinity at D 2 R.The fluorescent labels also slightly affected the intrinsic activity of the ligands when tested in a β-arrestin-2 recruitment assay at D 2 R. It should be noted that the employed fluorophores not only differ in their molecular weight and steric demand, but also their overall lipophilicity and net charge, which may influence not only receptor recognition and activation, but also the tendency of non-specific binding 44 . Taking advantage of our newly developed fluorescent ligands, we sought to establish a fluorescent ligand binding assay with the D 3 R subtype serving as an example case. In principle, ligands 10a-c comprising a bistrifluoroethylrhodamine, a Cy3B or an Alexa488 fluorophores should be well suited for both, FRET and BRETbased technologies. In FRET-based ligand binding assays, the receptor of interest is labeled with a fluorophore, either through the binding of a fluorescently labeled antibody or a self-labeling tag, that can be linked to a small molecule fluorophore 12 . In contrast, NanoBRET binding assays 11 make use of a small and bright luciferase variant (Nluc) 19 , that is fused to the extracellular part of the investigated GPCR. In both cases, binding of the fluorescent ligand is detected based on RET between the receptor as light emitting donor and the fluorescent ligand serving as the acceptor 10 . Both RET-assays have been proven extremely useful for the characterization of ligand binding at GPCRs, especially if high affinity radioligands are not available or if binding kinetics 10,23,24 are central to the investigated research question. Typical saturation hyperbolas were obtained for all three ligands when we performed NanoBRET assays with living HEK293T cells expressing an Nluc-D 3 R fusion protein. For two of the ligands, 10a and 10c, observed binding affinities were almost identical to those derived from classical radioligand binding studies. This was also true, when NanoBRET experiments were performed with cell membranes instead of whole cells. For the Cy3B-labled ligand 10b, the NanoBRET K D was approximately tenfold higher. On the other hand, 10b showed and excellent behavior in TIRF microscopy studies with CHO cells expressing D 2 R and D 3 R. This illustrates that despite similar absorption and emission spectra, the choice of the fluorophores may be critical not only for ligand binding to the target but has to be tailored to the desired application of the fluorescent ligand.
In summary, we have developed a set of high affinity fluorescent ligands for D 2 R and D 3 R receptors, which are the main targets in the treatment of severe neurological and psychiatric diseases. Depending on the employed fluorophore, the ligands can be used in high-resolution TIRF microscopy or to study ligand binding by resonance energy transfer. The established D 3 R-NanoBRET assay can be equally performed in whole cells and membranes and can be used for ligand binding screenings and characterization of novel ligands for D 3 R in the future. GP I: Synthesis of terminal alkynes 39,40 . To a mixture of a secondary amine, K 2 CO 3 (2 eq) and KI (1 eq) in CH 3 CN was added 6-chlorohex-1-yne at room temperature and the reaction mixture was heated to reflux overnight. After addition of H 2 O, the aqueous phase was extracted with CH 2 Cl 2 . The combined organic layers were dried over Na 2 SO 4 and evaporated under reduced pressure to give the crude product. 41 . To a mixture of corresponding alkynes and aromatic azides in a solvent system of tert.-BuOH-H 2 O-CH 2 Cl 2 (1:1:1) was added CuSO 4 ·5H 2 O (5 mol %) and sodium ascorbate (10 mol %) and the suspension was stirred at room temperature. After the completion of the reaction, the suspension was diluted by the addition of H 2 O, extracted with CH 2 Cl 2 , and the combined organic layers were dried over Na 2 SO 4 . After evaporation, the crude residue was purified by silica-gel column chromatography with CH 2 Cl 2 followed by 95:5 CH 2 Cl 2 -MeOH.

GP II: Cu(I)-catalyzed 1,3-cycloaddition
GP III: N-dansylation 54 . To a mixture of the crude primary amine and triethylamine (1:1) in CH 2 Cl 2 was added dansyl chloride at 0 °C. The reaction mixture was stirred overnight at room temperature and the product was extracted with CH 2 Cl 2 . The combined organic layers were dried over Na 2 SO 4 and concentrated in vacuo to obtain the crude product. Target compounds were isolated by silica-gel column chromatography with CH 2 Cl 2 followed by 98:2 CH 2 Cl 2 -MeOH.
GP IV: Amide coupling. To a solution of the benzoic acid derivatives and DIPEA in CH 2 Cl 2 at 0 °C was added TBTU in anhydrous DMF and the mixture was stirred for 30 min before a solution of tert.-butyl 3-aminopropylcarbamate in CH 2 Cl 2 was added. After stirring for 2-3 h, saturated NaHCO 3 solution was added and the mixture was extracted with CH 2 Cl 2 . The combined organic layers were dried over Na 2 SO 4 and concentrated in vacuo to obtain the crude product. The compounds were isolated by silica-gel column chromatography with 95:5 CH 2 Cl 2 -MeOH.
Molecular docking. Docking of 8a and 10b was performed analogously to previously described protocols 55 .
Ligands were geometry optimized by means of Gaussian16 56 at the B3LYP/6-31 (d,p) level of theory (attributing a formal charge of + 1) and subsequently docked into the crystal structures of the D 2 R (PDB-ID: 6CM4) and the D 3 R (PDB-ID: 3PBL) using AutoDock Vina 57 . We applied a search space of 40 Å × 40 Å × 40 Å due to the large Cy3B moiety of 10b and to ensure a complete coverage of the binding pocket. Based on the scoring function and experimental data, four ligand-receptor complexes were selected, which were subsequently submitted to energy Scientific Reports | (2020) 10:21842 | https://doi.org/10.1038/s41598-020-78827-9 www.nature.com/scientificreports/ minimization using the PMEMD module of the AMBER 18 program package 58 . The all-atom force field ff14SB and the general AMBER force field (GAFF) were used for the receptors and ligands, respectively. Parameters for 8a and 10b were assigned using antechamber, and charges were calculated using Gaussian16 at the HF/6-31 (d,p) level of theory and the RESP procedure according to the literature 59

Measurement of absorption and emission spectra.
Absorption spectra were recorded on a CLARI-Ostar (BMG Labtech, Ortenberg, Germany) microplate reader. 1 mM solutions of the ligands were prepared in DMSO and measured either directly (8a) or after dilution to 10 µM in PBS (10a-c). Emission spectra were collected employing an excitation wavelength of 335/10, 470/10 or 520/10 nm, respectively. Spectra were background corrected and normalized to the maximum absorbance/emission of each sample.
cDNA constructs. For the generation of Nluc-D 3 R fusion constructs in pcDNA3.1, sequences of the Nluc enzyme 19 (pNLF1-N or pNLF1-secN, Promega) and D 3 R (DRD3, cdna.org) were amplified by polymerase chain reaction and fused in frame with a 4 AA linker (GSSG) by Gibson Assembly 62 (New England Biolabs). To achieve surface expression, the fusion protein was either N-terminally tagged with an HA-signal sequence and a FLAGtag 50 or the secretory version of the enzyme 19 (pNLF1-secN) was used. Sequence integrity was verified by DNA sequencing (Eurofins Genomics).

Radioligand saturation binding. HEK293T cells were grown to a confluence of 70-80% and transfected
with the Nluc-D 3 R plasmids using polyethyleneimine (PEI) as the transfection reagent (PEI/DNA ratio 3:1). After incubation in DMEM/F12 with 10% FBS at 37 °C, 5% CO 2 for 48 h, cells were harvested and cell membranes were prepared as described previously 34 . The protein concentration was determined using the method of Lowry and bovine serum albumin as standard 63  www.nature.com/scientificreports/ red with 5% FBS from a 1 mM DMSO-stock) were added. Non-specific binding was determined in the presence of 10 µM haloperidol. After 1 h incubation with the ligands at 37 °C, furimazine (Promega, final dilution 1:500 to 1:2,500) was added, followed by a 5 min incubation at room temperature in the dark. BRET was measured as the ratio of acceptor fluorescence and donor luminescence employing a CLARIOstar microplate reader equipped with 475/30 nm and 535/30 or 580/30 emission filters, respectively. Total, non-specific and specific binding were analyzed employing the algorithms for one-site saturation binding implemented in PRISM6.0 or 8.0.
For association kinetics, secNluc-D 3 R membranes were added to various concentrations of 10a (1-100 nM) in the presence or absence of 10 μM haloperidol and furimazine (1:2,500). For dissociation, membranes were preincubated with 10a for 1 h, before furimazine was added and dissociation was initiated with 10 μM haloperidol. Kinetic assays were monitored at room temperature and minimal possible cycle time of the CLARIOstar was used for each measurement. Data was fitted to one phase association and dissociation equations in PRISM6.0 to determine the kinetic constants. In competition binding experiments, membranes were incubated with serial dilutions of unlabeled ligands and the fluorescent ligand 10a (10 or 100 nM) for 90 min at 37 °C. Addition of furimazine (1:5,000) and BRET measurements were carried out as described above. Data analysis was performed using the one site-fit K i equation in PRISM6.0 to determine the inhibition constants (K i ) of the unlabeled ligands.

Data availability
The data that support the findings of this study are available within the Supplementary Information files and/or from the corresponding authors upon request.