Sequences corresponding to full-length CHT (NM_021815.4) and LV-AA mutation at residues 531–532 (Ribeiro et al., 2005; Ruggiero et al., 2012) were codon-optimized for expression in human cell lines and custom-synthesized by GeneArt (LifeTech). Constructs were designed with a N-terminal FLAG epitope (Cuddy et al., 2012) and cloned into the pLenti6.3/V5 DEST vector which also contained a C-terminal V5 epitope tag (LifeTech). Lentiviral particles were generated (ViraPower, Thermofisher Scientific) and used to transduce HEK-293 cells (Sigma Aldrich). Pools of stable transformants were selected with 8 μg/mL Blasticidin. In addition, SH-SY5Y cells (ATCC) were transduced with a CHT construct tagged with GFP at the C-terminal (Origene, PS100071) or a FAP tag at the N-terminus (Sharp Edge Laboratory). Both stable pools and single clones were expanded. HEK-293 cells were cultured in DMEM High glucose (Gibco # 21969-035) supplemented with 10% FBS and 4 mM Glutamine (Thermofisher Scientific). SH-SY5Y cells were cultured in DMEM:F12 with Glutamine (Gibco # 11320-033) supplemented with 15% FBS and 1x NEAA (Thermofisher Scientific).

The Cresset field based virtual screening tool, Blaze (formerly called FieldScreen), (Cheeseright et al., 2008, 2009) was utilized to search the full Pfizer compound screening collection to identify compounds similar to literature CHT positive allosteric modulator (PAM) MKC-351/coluracetam (Takashina et al., 2008a,b) or CHT negative allosteric modulator (NAM) ML-352 (Ennis et al., 2015). Similarity was assessed using 50% 3D electrostatic and hydrophobic properties (Cheeseright et al., 2006) and 50% shape (Grant et al., 1996). This field was then used as a template to virtually screen the Pfizer file for additional compounds with a similar field and potentially related biological activity. For each virtual screening run, the top 500 compounds from the Pfizer compound collection, based on Blaze score, were selected. The set of 1,000 compounds identified from the PAM and NAM virtual screening campaigns was further filtered based on compound availability and removal of chemically unattractive groups (Stepan et al., 2011) to generate a test-set set of 887 compounds.

The Pfizer Chemogenomic Library (CGL) contains 2,753 selective small molecules covering 1,043 distinct biological targets (Jones and Bunnage, 2017). The CGL was created to support phenotypic screening with the purpose of expediting target identification. A hit from the set suggests the annotated activities of that pharmacological agent may be involved in perturbing the observable phenotype. Multi-parameter optimization was used in the creation of the library to ensure appropriateness of molecules for cell-based screening (including assessments of permeability, solubility, cytotoxicity and selectivity). CGL compounds were selected on their potency against their primary annotated target at a concentration equal to or less than 500 nM where possible.

Uptake of [³H]Choline (American Radiochemicals; ART 0197; 1,000 Ci/vial; Specific Activity 60–90 Ci/mmol) was measured in HEK cells expressing the high affinity choline transporter (CHT), clone WT4. Cells were plated at 50,000 cells /well in 200 μL culture media on a PDL coated Cytostar-T 96 well plate (Perkin Elmer) 24 h prior to assay. On the day of assay, following two washes and a 30 min pre-incubation at 37°C in 50 μL Na(−) free buffer (0.54 mM KCl, 1.3 mM CaCl₂, 0.53 mM MgCl₂, 0.4 mM MgSO₄, 0.37 mM KH₂PO₄, 240 mM Sucrose, 4.4 mM Tris-phosphate, 5.5 mM Glycine, 5.6 mM D-Glucose; pH 6.5 with KOH), 25 μL 4X control or test compounds in Na(−) buffer were added to the cells. Hundred and zero percent controls were defined with 10 μM Staurosporine and DMSO respectively. Final DMSO concentration was 0.25%. Following a 30 min incubation at 37°C, 25 μL [³H]Choline diluted to 320 nM (80 nM final concentration) was added in Na(+) buffer (0.54 mM KCl, 1.3 mM CaCl₂, 0.53 mM MgCl₂, 0.4 mM MgSO₄, 0.37 mM KH₂PO₄, 138 mM NaCl, 0.28 mM Na₂HPO₄, 5.5 mM Glycine, 5.6 mM D-Glucose; pH 6.5 with KOH). Plates were sealed with TopSeals (Perkin Elmer) and incubated in darkness for 3 h at room temperature before reading using a Wallac Microbeta, 1 min/well. The assay window was typically 3-fold signal to background ratio. A number of datasets fell outside of assay acceptance criteria (defined as signal to background ratio >3, individual assay plate Z prime (also known as z-factor) value >0.3 and EC₅₀ of a standard compound within the expected range) and were therefore not included in the analysis.

All LC/MS analyses were performed on a Sciex 6,500 triple quadrupole tandem mass spectrometer in positive electrospray ionization (ESI) mode. Other instrumentation consisted of Shimadzu LC-20AD pumps and an Apricot Design Dual Arm Autosampler (ADDA). Liquid chromatography was performed on a Waters Atlantis HILIC column (10 × 2.1 mm, 3μ). Mobile phase A consisted of water with 0.1% formic acid and mobile phase B consisted of water containing 0.1% formic acid and acetronitrile containing 0.1% formic acid (50:50). The flow rate was 0.6 mL/min and the gradient was as follows; hold at 100% B for 5 s and switch to 100% A for 10 s and return to 100% B for 5 s. Data were acquired with Analyst version 1.6. Quantitative analysis was performed in the multiple reaction monitoring (MRM) mode using MultiQuant software version 2.1. The MRM transitions monitored were m/z 113/60 m/z for choline-D9 (deuterated choline chloride-(trimethyl-d9), Sigma Aldrich) and 108/60 m/z for the internal standard, choline-D4 (deuterated choline chloride-1,1,2,2-D4, Sigma Aldrich).

A single cell clone of HEK293 overexpressing CHT (CHT-WT4) cells was grown in 96 or 384 well format to 50% confluency. Media was removed and cells were washed once with HBSS (+/+). 1 μg/mL Anti-FLAG antibody (F3165, Sigma) was diluted in HBSS (+/+) and added to the cellular monolayer for 30 mins. After incubation, cells were washed three times with HBSS (+/+) followed by incubation with 1:5,000 AF488 labeled secondary antibody (LifeTech) diluted in HBSS (+/+) for 30 min. After three washes, cells were fixed with 4% paraformaldehyde in HBSS (+/+) for 15 min. Nuclei were counterstained with the nuclear dye Hoechst. All images were captured and analyzed on an epifluorescent microsope or Cellomics platform.

Historically, animal-derived materials such as synaptosomes and primary culture have been used to study CHT function. We sought to move away from these approaches in accordance with UK NC3R guidelines. We generated and characterized a range of stable recombinant cell lines in both HEK-293 and SH-SY5Y backgrounds as neuronal-like cell lines in order to assess CHT transporter function for drug discovery. In the HEK-293 background, stable pools with either wild-type (WT) or mutant (LV-AA) CHT constructs were expanded and used to generate genetically homogenous clonal lines using single cell flow cytometry. Pools were used for initial radiometric assay development which facilitated subsequent selection of the CHT-WT4 line over other clones. This was done on the basis of greater assay window for inhibition and activation of [³H Choline] uptake (using 1 μM HC-3 and 10 μM STS respectively; Supplementary Figure 1) and data reproducibility. Interrogation of CHT-WT4 revealed overexpression of the codon-optimized CHT transcript but not endogenous SLC5A7 transcript as measured by qPCR but no differential expression of other choline transporters (SLC6A12/BGT1, SLC44A1-4/CLT1-4) (Supplementary Figure 2A). Similarly, there is no differential expression of other molecules required for acetylcholine synthesis (CHAT), transport (SLC18A3/vAChT) or hydrolysis (ACHE); nor in muscarinic and nicotinic acetylcholine receptor subunits (Supplementary Figures 2B,C). Parallel experiments were performed on the SH-SY5Y stable cell lines which overexpressed the codon-optimized CHT-GFP fusion transcript in addition to some basal SLC5A7 expression in parental cells in addition to ACHE, SLC18A3, and CHAT expression (Supplementary Figures 2D–F). These cell lines were the basis for assessing a range of assays to identify and characterize molecules that increased CHT-mediated transport (Table 1).

CHT is sodium (Na⁺) and chloride (Cl⁻) dependent and requires a negative membrane potential; hence, alterations in membrane potential can be used as a measure of CHT-mediated transport. Indeed, Blakely and colleagues reported the discovery of ML-352, a novel non-competitive inhibitor and STS as a positive modulator of CHT using a membrane depolarized assay (Ruggiero et al., 2012; Ennis et al., 2015). Building on these observations, we evaluated the SURFE²R™ platform (Nanion Technologies), as a solid supported membrane (SSM) based electrophysiology platform to assess electrogenic transporter activity (Bazzone et al., 2013; Barthmes et al., 2016). Membrane fragments containing the protein of interest are immobilized on a gold electrode several millimeters in diameter to facilitate detection of transporter currents. Synchronous activation of the transport proteins is triggered by rapid application of a substrate-containing buffer. During electrogenic transporter action, charge accumulates in the membrane fragments which directly correlate with the measurable current on the gold electrode and thus enables flexible, robust real time measurement of transporter activity. Membrane preparations are essentially “cell free” preparations minimizing trafficking impacts such that any effect seen is likely to be a result of direct action of compound or other treatment on the transporter. However, it should be noted that the comparatively low proportion of CHT proteins at the plasma membrane may be a limiting factor for assays made using the SURFE²R™. In the first instance, we demonstrated that the Na⁺ and pH dependence of CHT mediated choline transport could be recapitulated using membrane preparations from CHT-WT4 (Figure 1A). Furthermore, we were able to saturate the assay with an apparent K_a of 25 μM (Figure 1B) and noted the measured signal was stable over time (Supplementary Figure 3). We observed inhibition of current with HC-3 (Figure 1C) and ML-352 (Figure 1D) with IC₅₀ estimates of 20.4 and 70.1 nM respectively (Table 2). However, we were not able to detect any effect of staurosporine (STS) or MKC-231.

SSM based approaches have utility as an orthogonal approach but may be currently limited by throughput; additionally we were also unable to observe positive effects of STS on CHT-related currents (see Section Discussion). A preponderance of literature precedent for measuring high affinity choline uptake (HACU) utilizes a radiometric assay to measure [³H]Choline transport in preparations. In order to measure CHT-mediated transport activity, we established a radiometric assay based on a measure of proximity-induced scintillation in recombinant cell lines (Figure 2A) which had increased throughput compared to traditional approaches. We observed an increase in [³H]Choline uptake in CHT-WT4 cells compared to the parental HEK293 background (Figure 2B) together with a concentration-dependent increase in [³H]Choline uptake (Supplementary Figure 4A). In addition, we recapitulated literature observations of sodium dependency (Figure 2C), consistent with known Na+ dependent choline co-transport activity as well as recapitulating pH dependency observations (Figure 2D) (Okuda et al., 2000). A number of parameters were assessed in assay development including cell seeding density, time of assay post-seeding, plate coating, buffer composition, read time and choline concentration. We noted that choline starvation and use of a sodium gradient and manipulating pH (decreasing pH increases STS response) improved assay window. We also observed the STS mediated enhancement of [³H]Choline was stable over time and there was reduced variability which led to the selection of a 3 h time point for screening purposes (Supplementary Figure 4B). However, a complete saturation curve could not be generated as the 3H-[choline] contained ethanol as a diluent which was toxic to the cells at higher concentrations; a limitation of this assay format. We proceeded to assess reported tools in this assay which was capable of reading out on both inhibitory and stimulatory activities (Table 2). Inhibition of CHT mediated transport was seen with the classic inhibitor hemicholinium-3 (HC-3) with an IC₅₀ estimate of 116 nM (Figure 2E) in addition to the recently described negative allosteric modulator ML-352 with an IC₅₀ estimate of 549 nM (Figure 2F) which is in broad agreement with literature observations (Okuda et al., 2000; Ennis et al., 2015). We further recapitulated the reported stimulatory effect of staurosporine (STS) on CHT-mediated transport (Ruggiero et al., 2012) with an EC₅₀ estimate of 1.7 μM in CHT-WT4 cells (Figure 2G) but no apparent effect in CHT-LVAA pools (data not shown). We were unable to satisfactorily demonstrate potentiation of CHT transport with STS in SH-SY5Y cell backgrounds due to toxicity. We were also unable to see any impact of positive allosteric modulator MKC-231 (Takashina et al., 2008b) on CHT mediated transport (Figure 2H).

We selected a set of 887 Pfizer compounds selected via the Cresset field-based virtual screening technology, (Cheeseright et al., 2007) using seed molecules CHT PAM MKC-351/coluracetam (Takashina et al., 2008b) and CHT NAM ML-352 (Ennis et al., 2015) (Supplementary Figure 4). This library was initially screened in the CHT radiometric assay in a 96 well format at 1 and 10 μM compound concentrations. Actives identified in the 1 and 10 μM single point screens were then assessed in a 10-point dose response format (Table 2). Compounds 1, 2, and 3 were confirmed as positive allosteric CHT modulators with EC₅₀ estimates of 0.3, 0.4, and 4.5 μM respectively. Compounds 4 and 5, with EC₅₀ estimates of 1.5 and 0.7 μM respectively, were characterized as negative allosteric modulators of CHT function. The Chemogenomic Library (CGL) was also screened in the CHT radiometric assay in a 96 well format at 10 and 1 μM concentrations. From this screening campaign, a number of positive CHT modulator hits (compounds 6-9, Table 2) were identified. The CHT radiometric assay identified CHT modulators from both the field-based virtual screening campaign and the Chemogenomic Library (CGL) screen. However, as a consequence of the small assay window, inter-assay variation was high and not all datasets fell within the assay acceptance criteria as defined in the Materials and Methods section. This, coupled with our inability to satisfactorily perform saturation analyses, prompted us to assess alternative screening platforms.

As highlighted in Section Radiometric assessment of CHT mediated transport function was not suitable for screening, given the challenges of our radiometric assay for high throughput screening, alternative methods for measuring choline uptake suitable for screening compound libraries were sought. We elected to focus on measuring transport of a stably labeled form of choline (deuterated choline chloride-(trimethyl-d9); D9-choline) coupled with mass spectrometric based quantification (Koc et al., 2002; Shariatgorji et al., 2014; Iwamoto et al., 2016). Using this method, we observed a clear increase in D9-choline uptake in HEK293 CHT-WT4 recombinant cells compared to the parental background (Figure 3A). Critically, we were also able to saturate transporter function using this method which led us to determine a K_m of 7.0 μM (Figure 3B), consistent with literature reports (Okuda and Haga, 2000; Okuda et al., 2000). Further validation of the mass spec assay using our tool compounds revealed the capacity of the assay to report on both inhibitory and potentiation activities. Pleasingly, we observed a good agreement with obtained IC₅₀ estimates for HC-3 (5.6 nM) and ML-352 (41.9 nM); and EC₅₀ estimates of STS (507 nM) with those obtained in the radiometric assay and in the published literature (Figures 3C–E and Table 2). Again, we saw no apparent effect of MKC-231 (Figure 3F), consistent with lack of activity in all other assay formats. We successfully miniaturized the assay to a 384 well format with consistent z prime >0.5 (Supplementary Figure 6). As a proof of concept study, we screened the Chemogenomic Library (CGL) at 1 and 10 μM with DMSO as the negative control and 10 μM STS as the positive control. A number of hits with either inhibitory or stimulatory activities were identified, and these hits confirmed with 10 point EC₅₀ dose response follow-up. As highlighted in Table 2, compounds 6-9, flagged as CHT PAMs in the single point radiometric assays, were identified as positive modulators in the mass spectrometry assay. These four compounds, all kinase inhibitors, were selected for further assessment in orthogonal assay formats (Tables 1, 2, Section Live cell staining reveals effect on transporter localization with compound treatment).

CHT function has been proposed to be tightly regulated through localization with a minority of transporters present on the presynaptic membrane (<10% of the total population; Ferguson and Blakely, 2004; Ribeiro et al., 2006, 2007; Black and Rylett, 2012). Thus, understanding compound effects on CHT density and activation state in the synaptic membrane may aid interpretation of the effects on cholinergic neurotransmission. Conventional methods such as cell surface biotinylation present challenges in terms of the requirement of large quantities of cells and low throughput. The predicted topology of CHT indicates 13 transmembrane domains, with an extracellular N-terminus and intracellular C-terminal tail (Okuda et al., 2012). Our CHT overexpression construct contains a FLAG tag on the N-terminus which we sought to utilize with antibody detection in non-permeabilized cells. However, we found that fixation of cells with PFA or methanol gave rise to some permeabilization and labeling of the internal CHT pool as indicated by immunostaining for the V5 epitope tag at the intracellular C-terminus (Supplementary Figure 7A). We therefore moved to live cell immunolabeling to reduce artifacts. We used the CHT-LVAA pool vs. CHT-WT4 cells to assess the effectiveness of this approach in the first instance as this mutant protein has been reported to exhibit increased cell surface localization through a decrease in endocytosis (Ferguson et al., 2003; Ribeiro et al., 2005, 2007; Ruggiero et al., 2012; Supplementary Figure 7B). We proceeded to investigate the effect of compound treatment in more detail. Cells were incubated with compounds at 1 or 10 μM for 1 h before immunolabeling and visualization. Interestingly, we observed an increase in cell surface localization with HC-3, ML-352 and STS treatments whereas again, MKC-231 had no apparent effect (Figure 4). We further tested compounds 6-9, initially identified by screening in the radiometric and mass spectrometry assays at 1 and 10 μM, and noted that compounds 8 and 9 both increased cell surface localization of CHT but compounds 6 and 7 appeared to have no effects based on qualitative observations (Figure 4).

All Pfizer-based authors (PC, EJA, CCJ, MP, RT, JSJ, RIS, SES, CLB) were full time employees of Pfizer Ltd., at the time of the study. MB was a full time employee of Nanion Technologies at the time of the study. SEJ generated data as part of her project as an industrial trainee at Pfizer Ltd. YM was a full time employee of Kissei Pharmaceutical Co. Ltd., and participating in a secondment at Pfizer Ltd., at the time of the study.