Introduction

G protein-coupled receptors (GPCRs) are the largest and most investigated class of cell surface receptors transmitting the signal of numerous extracellular stimuli into the cell. Once a ligand binds to and activates a GPCR, heterotrimeric G proteins are activated. In turn, the G protein subunits dissociate to interact with effector proteins to initiate downstream signaling1,2,3. Prolonged receptor activation is determined by phosphorylation by one of several G protein-coupled receptor kinases (GRKs). The phosphorylation in turn promotes the binding of proteins of the β-arrestin family. The binding of β-arrestin sterically obstructs G protein coupling and triggers the desensitization and internalization of the GPCR4,5. Besides their function of GPCR desensitization, β-arrestins are also capable to act as adaptor initiating distinct signal transduction pathways6,7,8. Abnormal GPCR activities and consecutively deregulated signalling pathways also impact on the pathophysiology of various diseases, such as metabolic disorders, immune diseases, neurodegenerative diseases, and psychiatric disorders like schizophrenia and bipolar disorder9,10. As GPCRs have critical roles in the pathophysiology of many diseases, GPCRs are key drug targets. Hence, GPCRs are currently targeted by 33% of all marketed drugs, making them the largest druggable class of receptors11. Within the drug development process it is key to test the selectivity of a compound towards its destined target GPCR among various subfamilies12. To do this, the activities of multiple relevant GPCRs, i.e. the target GPCR as well as closely and distantly related GPCRs, must be monitored upon compound treatment.

For the analysis of GPCR activities and downstream signalling effects, a large number of tools were developed over the last decades utilizing several steps of the GPCR signalling cascade. Various biochemical and biophysical approaches based on the recruitment of G-proteins have been developed13,14,15,43. However, serotonin receptor HTR5A was activated by dopamine in U2OS cells only, possibly due to cell type specific effects44.

For the analysis of drugs, it is critical to assess its functional properties, as a drug may act as agonist, partial agonist, or antagonist. Therefore, monitoring these diverse drug effects in GPCR profiling assays requires a flexible setup that allows measuring agonistic and antagonistic effects in separate experimental paradigms that can be, however, performed and analysed in parallel. In our profiling approach, we applied a co-treatment paradigm using an agonist mix of ligands that keeps numbers of samples low when compared to applying the agonists separately. We challenged the GPCR profiler with the two antipsychotic drugs paliperidone and aripiprazole to assess the performance of our approach.

Paliperidone, the primary active metabolite of risperidone, is a classical atypical antipsychotic drug that acts predominantly as antagonist. The monitored inhibitory effects on HTR2A, DRD2, the α-adrenergic receptors and HRH1 are consistent with the literature39,45. Aripiprazole is reported to act functionally selective as agonist and antagonist depending on both the receptor and cell type42. Aripiprazole exerted agonistic effects on DRD2 and HTR2A, but antagonistic effects on HRH1, and these findings were validated in single luciferase assays. The activation of HTR2A is contrarily discussed in the literature. Commonly, an antagonism for this receptor is described46. However, a weak partial agonistic effect was reported for aripiprazole in GF62 cells42. It may be possible that aripiprazole acts functionally selective not only on DRD241 but also on HTR2A in a cell type-dependent manner. For aripiprazole, an additionally partial agonism on the serotonin receptor HTR1A is reported, which was, however, not detected in this multiplexed GPCR assay, possibly caused by a limited sensitivity of the assay in U2OS cells. A comparison of our findings using the GPCRprofiler versus public databases and literature shows consistent overlap (Supplementary Table S4).

Several methods are available to assess ligand actions on GPCRs. The most conventional approach to identify ligand receptor interactions are cell-free binding assays. These assays have the intrinsic disadvantage that compounds are frequently radioactively labelled, and that binding affinity data does not provide any information on the functional properties of ligands. Thus, it is neither clear whether ligands act as agonist or antagonist, nor which efficacies are exerted under physiological conditions16. These limitations are overcome in cell-based functional assays. These approaches address ligand-dependent receptor activation and signalling within a cellular context. Most of these approaches rely on receptor-mediated activation of G proteins, either by directly measuring of second messenger levels (e.g. cAMP and calcium)47,48,49 or by using reporter gene assays that respond to second messenger activities50. Monitoring multiple GPCR-ligand interactions within one experimental setup was, however, not feasible as it is known that different G proteins couple to various types of GPCRs. A sensitive and robust method to monitor receptor activation of GPCRs uses the dynamic recruitment of β-arrestin, which is a measurement independent of G protein activity27,28. In addition, a recent study supports this view as β-arrestin can be recruited to activated GPCRs in the absence of active G proteins51. Recently, a large collection of cell-based assays using β-arrestin recruitment was introduced to interrogate the druggable GPCRome in a parallel manner28. The assay uses a modified Tango assay27 for screening compound libraries against a multiplicity of orphan and non-orphan GPCRs. However, the approach is based on single compound per target assays. Therefore, when for example analysing target selectivity for a given drug on multiple GPCRs, screening of GPCR activities will lead to extensive consumption of materials at increasing costs.

The GPCR profiling assay we introduce provides the following benefits: (I) the GPCRprofiler enables the rapid and easy detection of multiple GPCR activities in response to several ligands within one experiment. The GPCRprofiler monitors GPCR activation at a time point that is defined by optimal performance of individual GPCR assays. As the readout is based on molecular barcodes, large amount of data can be obtained for one compound within one measurement. This stands in contrast to conventional single assays, wherein each receptor-ligand combination must be measured separately. Therefore, our approach may be of advantage when compounds are limited e.g. in early phases of drug discovery programs. (II) The presented multiplexed GPCR assay is highly flexible and scalable in terms of combinations of receptors and ligands using the mix & divide design. The GPCRprofiler provides sensitive and specific activity readouts using transient transfections, enabling the addition of new GPCR activity assays easily. However, stable cell lines expressing the receptor and/or adapter protein might enhance the robustness and sensitivity of the assay, especially for weak receptor activations caused by partial agonists that may require a close-to-endogenous expression of GPCRs. (III) The GPCRprofiler can be principally performed in all cell lines and primary cells that are amenable to transient transfections allowing to screen for cell type-dependent effects of compounds.

GPCR recruitment assays based on β-arrestin2, like split TEV-based GPCR activity assays, use tagged GPCRs, and it may be possible that modified receptors show altered signalling properties52. For split TEV, it has been shown that GPCR-NTEV-tevS-GV fusions do not substantially interfere with assay performance and signalling23. Split TEV-based GPCR recruitment assays are designed to monitor receptor activation only, but not downstream signalling effects.

However, the GPCRprofiler may be complemented with genetically encoded pathway reporter assays that use overexpressed native GPCRs and molecular barcodes as final readout25. Using this approach, GPCR target profiling can be combined with cellular pathway profiling to provide a holistic view of compound actions complementary to biochemical or transcriptomic drug profiling53.

In summary, we present a multiplexed GPCR activity assay that uses a β-arrestin recruitment assay and molecular barcoding as readout to assess multiple GPCR activities in parallel. We validated the GPCRprofiler using cognate ligands and challenged the assay using the antipsychotic drugs paliperidone and aripiprazole. As the GPCRprofiler can be flexibly expanded by additional targets, it therefore might represent a suitable platform technology for early drug discovery.

Methods

Plasmids

All expression plasmids were generated by Gateway recombination cloning (Invitrogen). ORFs of GPCRs were PCR amplified from a mix of human heart, liver and brain cDNA library using proofreading polymerases (PfuUltra, Stratagene or Pwo, Roche). DRD4 ORF was purchased from Bio Basic Inc. as synthesized linear sequence. ADRB3 ORF shuttle clone was purchased from Source Bioscience (IOH29805). The entry vector for β-arrestin2delta383 was described before23. All ORFs were BP-recombined into pDONR plasmid and the yielded Entry vectors were sequence verified. Entry vectors were then LR-recombined into the split TEV destination vectors pTag4C_X-V2R-NTEV-tevS-GV or pcDNA3.1_X-CTEV, plasmids were described in detail before22,23, to obtain the final expression vectors of GPCR-V2R-NTEV-tevS-GV and β-arrestin2delta-CTEV fusion constructs. EXT reporter gene plasmids were generated by three-fragment multisite Gateway recombination (Invitrogen). EXT synthesis and entry vectors were described before25. To yield final reporter constructs Entry vectors encoding ten-time clustered GAL4-responsive cis-elements (UAS, attL1/attL4), a CMV minimal promoter (attL4/attL3) were LR-recombined with an EXT entry vector library (attL3/attL2) to obtain final 10xUAS-CMVmin-EXT reporter library. All constructs were sequence verified.

Compounds

Serotonin hydrochloride, Histamine dihydrochloride, Asenapine maleate were purchased from Tocris, Dopamine hydrochloride, Epinephrine hydrochloride, [Arg8]-Vasopressin acetate salt, Paliperidone, Somatostatin were obtained from Sigma-Aldrich, Aripiprazole was purchased from Toronto Research Chemical.

Cell culture

U2OS cells were maintained in McCoy’s 5A medium (Gibco) including GlutaMAX (Gibco), supplemented with 10% fetal bovine serum (FBS, Gibco) and 100 U/ml each of penicillin and streptomycin (Lonza). PC12-tetOFF cells were maintained in low glucose DMEM medium (1 g/L, Lonza), supplemented with 10% FBS, 5% horse serum (HS, Gibco), 2 mM GlutaMAX (Gibco) and 100 U/ml each of penicillin and streptomycin. PC12 cells were grown on poly-L-lysine (Sigma) coated surfaces for maintenance and experiments; U2OS cells were grown on poly-L-lysine surfaces only for experiments. All cells were grown at 37 °C and an atmosphere of 5% CO2.

Luciferase reporter assays

For luciferase assays cells were plated on 96-well white plates at 4 × 104 cells/well (PC12) one day before the experiment. The assays were performed with 6 wells as replicates per condition. Cells were transfected with split TEV and EXT reporter plasmids using lipofectamine 2000 (Invitrogen). Plasmids and lipofectamine 2000 transfection reagent (Invitrogen) were diluted in opti-MEM (Gibco) and incubated for 20 min at room temperature. Medium was removed from cells, the transfection-mix was added and the plates incubated for 2 h at 37 °C, followed by the adding of double-volume of culture medium to dilute transfection reagents. On the next day, culture medium was exchanged by serum-reduced assay medium (U2OS: McCoy’s 5 A supplemented with 0.5% dialyzed FBS (Gibco), 0.1 mM NEAA (Gibco); PC12: low glucose DMEM (1 g/L) supplemented with 1% dialyzed FBS and 0.1 mM NEAA) to induce starvation of the cells for 17–18 h. On day 3, medium was removed and cells were treated with compounds diluted in assay medium at different concentrations for 6 h at 37 °C. After 6 h of compound treatment the medium was removed and the cells were lysed with 30 µl passive lysis buffer (Promega). The reading of firefly luciferase was carried out with a Mithras LB 940 Microplate Reader (Berthold Technologies) and the software MicroWin2000. The data was exported from MicroWin2000 to Excel and the mean and standard error of the mean (s.e.m.) of the 6 replicates for the firefly readings were calculated.

Transfection and stimulation conditions in single EXT reporter assays

For single EXT reporter assays cells were plated on 6-well plates at 4 × 105 cells/well (PC12) one day before the experiment. The assays were performed in duplicates for each stimulation condition. Cells were transfected as described for luciferase assays. After 2 h of incubation, the transfection mix was removed from the cells and culture medium added and incubated for 20 h at 37 °C. Cells were then starved by medium change to serum-reduced assay medium. Afterwards, compounds at different concentrations were applied in duplicates for 6 h at 37 °C. After compound treatment cells were washed with PBS and lysed with 500 µl RLT buffer (Qiagen). Lysates were processed for sequencing as described below.

Transfection and stimulation conditions in multiplexed EXT reporter assays

For EXT assays transfections were performed in-solution using lipofectamine 2000. Cells were resuspended in culture medium without antibiotics at a density of 1.5 × 106 cells/ml (U2OS) or 3 × 106 cells/ml (PC12). Split TEV and EXT-reporter plasmids and lipofectamine 2000 were mixed and incubated in serum-reduced opti-MEM for 20 min at room temperature. Transfection-mixes were added to cell suspensions and incubated for 2 h at 37 °C without shaking. Suspensions were centrifuged for 5 min at 900 rpm and resuspended in culture medium to remove transfection reagents. All transfection samples were combined within one pool and plated on multiple wells of 6-well plates and incubated for 24 h at 37 °C. Afterwards, cells were starved for 17–18 h and then treated with compounds for 6 h in duplicates. After compound treatment cells were washed with PBS and 500 µl RLT lysis buffer was applied for cell lysis. Lysates were processed for sequencing as described below.

Sequencing library preparation and next-generation-sequencing of EXT assays

For sequencing single and multiplexed EXT assays RNA was purified from cell lysates over RNeasy columns (Qiagen) including on-column DNase treatment according to the manufacturers’ protocol (RNeasy Mini Kit, Qiagen) and eluted in 50 µl H2O. RNA was additional in solution treated with Turbo DNase (Ambion) for 30 min at 37 °C and repurified over an RNeasy column. First-strand cDNA synthesis was performed with 1 µg total RNA using superscript III reverse transcriptase (Invitrogen) and 120 pmol of a random nanomer primer. For later data normalization, additional EXT reporter plasmids were added as a mix of twelve external calibrator EXTs (eCals) to each sample. Single eCals were added at semi-logarithmic concentrations covering three orders of magnitude, with three EXTs representing 3,333 molecules, three EXTs 1,054 molecules, three EXTs representing 333 molecules, and three EXTs representing 105 molecules. EXT sequences were PCR-amplified from cDNA samples using ‘decoding’ primer Dec1 and Dec2. For next-generation-sequencing Ion Torrent specific sequencing adapters were attached 5′ and 3′ to the amplified EXT sequences within a second round of PCR. On order to differentiate different samples 13mer barcodes were introduced between 5′ Ion adapter and target sequence. All processed samples of an experiment were combined within one pool and purified using PCR clean-up kit (Macherey-Nagel). The final samples were quantified using the dsDNA HS Assay Kit with the Qubit Fluorometer (Thermo Fisher Scientific) and adjusted to 23 pM. EXT samples were prepared and sequenced with the Ion Torrent Personal Genome Machine (PGM) system according to the manufacturer protocols for 200 bp template preparation and sequencing (Thermo Fisher Scientific).

Data analysis of EXT reporter assays

Raw sequencing data analyses were done by using BLAST algorithm and in-house developed R scripts. First, on the basis of the sample barcodes the data were split into the original samples encoding the specific treatment conditions and the reads of the assay EXTs, eCals and iCals were counted to obtain total read numbers of the individual samples. Samples with total reads below 20% of the average were excluded from the experiment to avoid variability caused by low read samples. Total read distributions between sample codes were balanced by normalization of the individual read numbers per EXT and sample to the highest total read number to enable the comparison of EXT reads between samples (Supplementary Fig. S4a). The eCals are also used to compensate an overamplification of highly expressed EXTs within the library amplification PCRs and the suppression of effects of lower expressed EXTs. These effects can be detected in samples with low read numbers of the eCals. To balance these effects, all reads of a sample were calibrated to a normalization factor calculated from the eCal reads of the corresponding sample (Supplementary Fig. S4b,c). Following to normalization and calibration steps, the biological sample replicates and the three EXTs assigned to the GPCRs were consolidated, and the resulting samples were standardized to a reference sample. In GPCR activation EXT assays, all samples that contain a stimulus were standardized to the unstimulated control sample to calculate fold changes (FC) of receptor activation.

$$FC=\frac{stimulated\,condition\,mean}{unstimulated\,mean}$$