In silico and in vitro screening for potential anticancer candidates targeting GPR120

The G-protein coupled receptor GPR120 has recently been implicated as a novel target for colorectal cancer (CRC) and other cancer managements. In this study, a homology model of GPR120S (short isoform) was generated to identify potential anti-cancer compounds targeting the GPR120 receptor using a combined in silico docking-based virtual screening (DBVS), structure–activity relationships (SAR) and in vitro screening approach. SPECS database of synthetic chemical compounds (~350,000) was screened using the developed GPR120S model to identify molecules binding to the orthosteric binding pocket followed by an AutoDock SMINA rigidflexible docking protocol. The best 13 hit molecules were then tested in vitro to evaluate their cytotoxic activity against SW480 – human CRC cell line expressing GPR120. The test compound 1 (3-(4-methylphenyl)-2-[(2-oxo-2-phenylethyl)sulfanyl]5,6-dihydrospiro(benzo[h]quinazoline-5,1-cyclopentane)-4(3H)-one) showed ~ 90% inhibitory effects on cell growth with micromolar affinities (IC50 = 23.21–26.69 μM). Finally, SAR analysis of compound 1 led to the identification of a more active compound from the SPECS database showing better efficacy during cell-based cytotoxicity assay –5 (IC50 = 5.89–6.715 μM), while a significant reduction in cytotoxic effects of 5 was observed in GPR120-siRNA pre-treated SW480 cells. The GPR120S homology model generated, and SAR analysis conducted by this work discovered a potential chemical scaffold, dihydrospiro(benzo[h]quinazoline-5,1-cyclopentane)-4(3H)-one, which will aid future research on anti-cancer drug development for CRC management. GPR120 is a member of the Class A Rhodopsin-like G-protein coupled receptor (GPCR) family and the free-fatty acid receptors (FFAR) subfamily and exists in two homologous human isoforms (a short isoform at 361 amino acids and a long isoform at 377 amino acids).1,2 It is also known as Free Fatty Acid 4 receptor (FFA4) – as its endogenous ligands have been identified as poly-unsaturated long-chain free fatty acids (PUFFA).3–5 GPR120 over-expression has been reported in colorectal cancer (CRC) cell lines, at approximately 2–3 folds higher occurrence than for normal colorectal cells.6 In CRC, GPR120 overexpression has been correlated with enhanced protein expression of proangiogenic factors such as vascular endothelial growth factors (VEGF), cyclooxegenase-2, and interleukin-8, as well as enhanced chemotactic activity of tumour cells.6 This increased activity has been characterised as GPR120-induced activation of the PI3K/Akt-NF-κB signalling pathway.6 GPR120 has also been reported to augment chemoresistance in breast cancer treatment by cross-talk through the Akt/NF-κB pathway.7 An extensive review by Senatorov and Moniri8 outlines the regulating role of GPR120 in various human cancers such as breast cancer, osteocarcinoma, melanoma prostate cancer, lung cancer, pancreatic cancer, and CRC.8 Tumour cells are known to secrete pro-angiogenic factors9,10 and over-expression of GPR120 in CRC cell lines might be further stimulating the development of tumour angiogenesis. As well, GPR120 agonism is reported to result in lower rate of cell proliferation and cell migration and increased rate of apoptosis in various cancer cell lines and animal models.8 Most of the FDA approved anti-angiogenic drugs10 are * Corresponding author. E-mail address: gemma.kinsella@tudublin.ie (G.K. Kinsella). c ORCID ID: https://orcid.org/0000-0002-8641-1497. d ORCID ID: https://orcid.org/0000-0002-9320-9254. e ORCID ID: https://orcid.org/0000-0002-6329-5841.

The G-protein coupled receptor -GPR120 has recently been implicated as a novel target for colorectal cancer (CRC) and other cancer managements. In this study, a homology model of GPR120S (short isoform) was generated to identify potential anti-cancer compounds targeting the GPR120 receptor using a combined in silico docking-based virtual screening (DBVS), structure-activity relationships (SAR) and in vitro screening approach. SPECS database of synthetic chemical compounds (~350,000) was screened using the developed GPR120S model to identify molecules binding to the orthosteric binding pocket followed by an AutoDock SMINA rigidflexible docking protocol.
GPR120 is a member of the Class A Rhodopsin-like G-protein coupled receptor (GPCR) family and the free-fatty acid receptors (FFAR) subfamily and exists in two homologous human isoforms (a short isoform at 361 amino acids and a long isoform at 377 amino acids). 1,2 It is also known as Free Fatty Acid 4 receptor (FFA4) -as its endogenous ligands have been identified as poly-unsaturated long-chain free fatty acids (PUFFA). 3-5 GPR120 over-expression has been reported in colorectal cancer (CRC) cell lines, at approximately 2-3 folds higher occurrence than for normal colorectal cells. 6 In CRC, GPR120 overexpression has been correlated with enhanced protein expression of proangiogenic factors such as vascular endothelial growth factors (VEGF), cyclooxegenase-2, and interleukin-8, as well as enhanced chemotactic activity of tumour cells. 6 This increased activity has been characterised as GPR120-induced activation of the PI3K/Akt-NF-κB signalling pathway. 6 GPR120 has also been reported to augment chemoresistance in breast cancer treatment by cross-talk through the Akt/NF-κB pathway. 7 An extensive review by Senatorov and Moniri 8 outlines the regulating role of GPR120 in various human cancers such as breast cancer, osteocarcinoma, melanoma prostate cancer, lung cancer, pancreatic cancer, and CRC. 8 Tumour cells are known to secrete pro-angiogenic factors 9,10 and over-expression of GPR120 in CRC cell lines might be further stimulating the development of tumour angiogenesis. As well, GPR120 agonism is reported to result in lower rate of cell proliferation and cell migration and increased rate of apoptosis in various cancer cell lines and animal models. 8 Most of the FDA approved anti-angiogenic drugs 10 are prescribed in combination with cytotoxic agents for effective CRC treatment. As GPR120 is involved at two extreme ends of the cancer therapeutics spectrum and tumour angiogenesis, it is of significant interest to develop and characterise small molecular anticancer agents targeting GPR120. Research studies, so far, have resulted in development of few potent GPR120 agonists only, such as TUG891 (see Fig 1), and a negative allosteric modulator, AH7614. 11 Due to advances in protein engineering, the availability of GPCR crystallographic structures in the RCSB Protein Data Bank 12 has increased and allows the collection of structural data on GPCRs for early stage computer aided drug discovery (CADD). GPCRs crystal structure data can also be used as the reference templates for comparative modelling or homology modelling (HM), to predict the three dimensional (3D) structure of homologous GPCRs. Such in silico homology models of GPCRs have been successfully employed in virtual screening (VS) studies to identify lead molecules for developing therapeutic agents acting as agonists, inverse agonists, or antagonists. 13,14 . This methodology is termed structure-based virtual screening (SBVS) and more specifically docking-based virtual screening (DBVS) is the most applied in practice. 15,16 .
In recent years, in silico VS studies followed by in vitro experiments have been successfully utilised to discover potential leads for developing GPCR ligands. 17 These in silico VS provide a faster and economical alternative to the wet-lab high-throughput screenings. 18 In addition to VS, analysis of structure-activity relationships (SARs) can lead to the design of potent and stable compounds by making rational structural modifications 19,20 .
In this study, a GPR120 DBVS protocol was developed utilising molecular docking with an in-house homology model in iterative combination with an in vitro cytotoxicity screening assay and SAR analysis, resulting in the identification of novel compounds with anticancer potential and predicted to act through GPR120.

Methods
VS was performed on the Irish Centre for High-End Computing (ICHECwww.ichec.ie) cluster. ICHEC provided CPU core units for computational processing and data storage space. The visual analysis and homology model building were carried out on an in-house 8 node (Intel® Core™ i7-4790 CPU @ 3.60 GHz × 8) Linux cluster. The visual analysis and image rendering was performed using PyMol Open-source version 2.1.0 21 and Biovia DS Client visualiser 2019. 22

Homology modelling
The FASTA sequences of human GPR120 short isoform, termed GPR120S -(UniProt ID: Q5NUL3-2) was retrieved from the UniProt database (https://www.uniprot.org Uniprot 2016). The sequence was used for BLASTP analysis (https://blast.ncbi.nlm.nih.gov/Blast.cgi) run against the protein databank (PDB) to identify 3D determined protein structures according to the best multiple sequence alignment (MSA) score generated. The human Delta-like opioid receptor PDB-ID: 4N6H and human Orexin 2 receptor PDB-ID: 4S0V were used as template structures. Using the multiple sequence alignment of GPR120S and the templates, comparative 3D structural models were generated and validated using MODELLER (v9.14). 23 An in-house python pipeline (incorporating MODELLER) was developed using the KNIME platform (Konstanz Information Miner) 24 to automate the process from generation of 100 models using the MSA of selected templates to side-chain rotamer optimisation using SCWRL4 25 and evaluation of homology models by MODELLER's DOPE (Discrete Optimized Protein Energy) score function. 23 The models were ranked by the DOPE statistical potential scoring models. The best models with the lowest DOPE values were then assessed for stereochemical properties using the MolProbity webserver 26 and their phi-psi Ramachandran plot were generated by using the PROCHECK online tool, 27 validation process (unpublished work).

Ligand library preparation
The SPECS database (www.specs.net) of synthetic chemical compounds (~350,000) was retrieved and processed in Biovia Discovery Studio's Pipeline Pilot v9.1 from Dassault Systèmes 22 to generate stereoisomers and tautomers, and remove compounds with molecular weight (greater than) > 650 and (less than) < 250 Dalton. The filtered chemical database was energy minimised using MMFF94 forcefield (steepest descent) in the open-source OpenBabel software package (htt p://openbabel.org/) and saved in sdf format.

Virtual screening and molecular docking
Molecular docking analysis was performed between the prepared ligands and the homology model of GPR120S (unpublished work) which was pre-processed in MGLs' AutoDock tools (www.autodock.scripps. edu) 28 to add polar hydrogens, charges by Kollman charge and assign torsion angles. The rigid-flexible docking protocol was followed by using Autodock SMINA (scoring and minimization in AutoDock VINA). 29 The binding site grid box was visually defined by employing AutoDock tools' Grid setting feature, based on the site-specific mutation study by Hudson et al. 2014 30 to include residues -Arg99(TM2), Trp104 (ECL1), Phe115 (TM3), Trp207, Phe211 (TM5), Trp277 (TM6) and Phe304 (TM7), deemed essential for ligand binding. The grid size dimensions were 40 × 60 × 60, with the (61.822, 59.75, 46.597) point set as the centre coordinates of the pocket. TUG891, being a selective potent agonist of GPR120, 31 was used as the reference ligand for docking analysis.
where: C score is consensus docking score. V score is docking using Auto-Dock Vina's default scoring function. S score is docking score using SMI-NA's default scoring function. NN score is docking score using NNScore 2.0 function. DL score is docking score using DLScore function. The ligands with a C score lower than − 9 were further analysed manually using the PyMol Open-source version 2.1.0 21 to enlist molecules for phase I of the in vitro screening based on their diverse scaffold chemistry. For phase II of in vitro screening, the chemical scaffold of the most cytotoxic test compound (from phase I) was used for a substructure search from the SPECS database docked pool using Discovery Studio's Pipeline Pilot from Dassault Systèmes 2017 22 .
Cell culture and materials SW480, a human CRC cell line was cultured in RPMI1640 growth media supplemented with 10% (v/v) Fetal Bovine Serum and incubated at 37 • C ± 1 • C in a humidified atmosphere with 5% CO 2 . The cell culture media and reagents were procured from Sigma-Merck, unless otherwise mentioned. The test compounds for in vitro screening assays were procured from SPECS (www.specs.net). A stock solution of test compounds (20 mM) was prepared in 100% (w/v) DMSO and stored at − 20 • C.

In vitro screening by cell-based cytotoxicity assay
Alamar blue based cytotoxicity assays were performed to determine inhibitory effects of the test compounds. SW480 cells were seeded at 10,000, 5,000 and 2500 cells per well in 96-well plates for 24, 48 and/or 72 h drug treatment, respectively. Required dilutions of test compounds were freshly prepared in 0.5% (v/v) DMSO growth media for cytotoxicity assays. After the treatment period, drug concentrations were replaced with 6% (v/v) alamar blue dye solution and the cells were incubated for 3 h under cell incubation conditions. Finally, the fluorescence signal was read using 560 nm excitation and 590 nm emission filters with Varioskan LUX Multimode Microplate Reader from Ther-moFisher Scientific.
siRNA transfected cells were harvested after 24 h of transfection and used for cytotoxicity assays as explained above.

Statistical analysis
All data was presented as mean ± standard error of the mean (SEM). For the analysis and graphic representation of biological experimental IC 50 values were calculated (GraphPad Prism 6 Software, La Jolla CA) using dose-response curves for the compounds. For plotting the dose-response curve, a non-linear regression curve fitting method was used where the mean positive control (50% DMSO in growth media) was defined as 0% and the mean vehicle control (0.5% DMSO in growth media) was defined 100%. For significance analysis, two-way analysis of variance (ANOVA) followed by Tukey's multiple comparison test was performed. P < 0.05 was considered statistically significant.

Homology model of GPR120
The human GPR120S receptor model was developed using the MODELLER homology modelling tool. 23 The model which scored best (DOPE score − 41156.738) as per the MODELLER scoring function was chosen as the final model and subjected to further energy minimization.
The α-carbon RMSD (root mean squared deviation) between the templates 4N6H, 4S0V and the developed human GPR120S model is 1.386 Å and 0.652 Å, respectively. Poor rotamers and sidechain rotations, above the permitted range of 0.3%, were removed and replaced by SCWRL4 with the average sidechain rotamer from its backbone-dependent rotamer library based on kernel density estimates. The comparative analysis of the stereochemical parameters of the GPR120S model by Molprobity webserver showed high reliability of the generated models. The Ramachandran psi-phi evaluation showed that all helical amino acids are located in the region favouring a right-handed α-helix with no residues (0%) in a sterically disallowed region and 98-99% residues in the favoured regions. The GPR120S model was embedded in a lipid bilayer model and energy minimised (unpublished work).

Molecular docking of TUG891 to the GPR120S model
The docked pose of TUG891 -a potent GPR120 selective agonist, with the receptor binding pocket is shown in Fig 1. An arginine in TM2 (Arg99) is critical for interaction between the receptor and the -COOH group of its ligands by various studies. 30,31,35,36 Six other specific residues are essential for TUG891, and other GPR120 agonists binding and interaction which are: Trp104 (ECL1), Phe115 (TM3), Trp207, Phe211 (TM5), Trp277 (TM6) and Phe304 (TM7). 30 Interactions with these seven residues were selected as the main criterion for docked binding pose selection of TUG891. 30,31 Automated protein-ligand interaction analysis using the PLIP webserver 37 identified strong noncovalent interactions of residues forming contacts with TUG891 docked to the orthosteric binding pocket of GPR120S (Fig 1).
The selected docked pose for TUG891 (binding score − 9.875) had interactions with Arg99 in all the predicted poses. The carboxylic acid of TUG891 forms a salt bridge (hydrogen bond and attractive charge interaction) with the Arg99 side-chain at a distance of 4.32 Å and a strong T-type (perpendicular) π-π stacking interaction between Phe115 and the second aromatic ring structure of TUG891. Other equitable hydrophobic and van der Waals interactions were also observed to stabilize the docked TUG891 in the orthosteric binding pocket.

In silico virtual screening
For virtual screening, the SPECS database (www.specs.net) containing ~ 350,000 commercially available well-characterised and druglike molecules was screened against the GPR120S homology model. AutoDock SMINA was used as the molecular docking algorithm and the docked poses were rescored using an in-house consensus scoring function from Equation I (see methods). The best scoring poses docked into the receptor were manually evaluated by comparison with the docked pose of TUG891.
The docking-based virtual screening with a score cut off set at − 9 (w. r.t binding score of TUG891 of − 9.8) followed by a manual docking-pose analysis resulted in 13 compounds (Table 1) being selected for the first phase of in vitro screening through Alamar Blue cytotoxicity assay. As per the docking evaluation, these compounds are predicted to bind to the orthosteric binding pocket of GPR120S (see Fig 2) as well as having similar hydrophobic and / or electrostatic interactions with one or more of the residues reported essential for the pharmacological activity of the receptor.
The electrostatic potential surface maps shown in Fig 2 reveal that the GPR120S binding pocket is lined with negatively charged residues (red colour) which can interact with ligand molecules via hydrogen bond contacts, whereas the opening of the pocket is surrounded by neutral or weak potential residues. Theoretically ligands with a strong  positive charge should be able to enter the binding tunnel without facing any resistance from the neutral or weakly charged residues at the opening. 38 Once the ligand enters the pocket the strong attractive charges should stabilise the protein-ligand binding, playing an important role in the binding kinetics. Steered molecular dynamics (SMD) studies can be insightful to further probe the binding kinetics of ligands w.r.t the charged surfaces of binding tunnel. 38,39 In vitro cytotoxicity assay The selected 13 compounds from the in silico VS were evaluated for their potential anticancer activity by an alamar blue cytotoxicity assay (see methods) using SW480 cells expressing GPR120. 6 For initial screening, the SW480 cell-line was treated with three concentrations (100, 10 and 1 µM) of each test compound for 72 h. As illustrated in Fig   3a, most of the compounds (11 out of 13) displayed null to negligible (~30%) cytotoxicity against SW480 cells at the highest tested concentration of 100 µM. However, two of the test compounds showed significant (>90%) inhibitory effects on cell growth with micromolar affinities (at 100 µM), reported in Fig 3a. The two active compounds -AK-968/12713190 (compound 1) and AG-690/40104520 (compound 2), were further tested at a wider concentration panel to construct a dose-response curve and determine their IC 50 values (see Fig 3b, c). Both the test compounds were active in cell line measurements with modest inhibitory activity at different treatment times. The 24 h drug treatments indicated IC 50 value of 23.21-26.69 µM for 1 and 26.55-33.2 µM for 2.
As the Dose-response-time (DRT) can highlight the dose-response patterns over time in pharmacological studies, 40 these two test compounds were tested over a longer treatment times of 48 h and 72 h for Percent relative cell viability for all treatments were quantified and normalised to the maximal response induced by vehicle control. Data for (a) was analysed by twoway ANOVA and asterisk values denote significant differences between 100, 10 and 1 µM treatment for each compound (*P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001). their cytotoxicity activity (see Fig 3b, c). After 48 h, the IC 50 of both compounds drops to ~50-60 µM. Similar effects over time have been reported in another oncogenic study 41 suggesting that the cells might have developed acquired resistance to test compounds at lower concentration after 48 h of exposure. This acquired resistance might enable them to escape the cytostatic state and start cell proliferation which can be traced back to the augmented chemoresistance in breast cancer treatment through GPR120 overexpression. 7 Also, the increased metabolic activity of cancer cells can be related to anticancer drug metabolism responsible for the resistance to cytotoxic agents, 42,43 hence reducing the cytotoxicity of these two compounds over time. While the 72 h experiments showed a slight decrease of 10 µM in IC 50 values of both compounds, it should be noted that the drug concentrations were not replaced over the treatment time intervals. The decreased cell growth or increased cytotoxic effects of these two compounds at 72 h might be the result of a lack of nutrients and increased metabolic waste in the culture solution. 44,45

SAR/similarity search and in vitro screening of SAR compounds
The top-scoring docked poses of compounds 1 and 2 (see Fig 4) predicted that the two molecules interact with a number of residues reported significant for protein-ligand binding by Hudson et al 2014 30 and which also interacted with the selected docked pose of TUG891 (see Fig 1) such as Ile280, Ile284, Val307. Compound 1 consists of a benzoquinazoline ring structure as the chemical scaffold with smaller benzylmethyl and benzyl substituents. The phenylalanine residue at TM3 (Phe115) shows strong π-π interactions with the main scaffold as well as π-sulfur interactions with the sulfanyl linker (see Fig 4). While compound 2 consists of a 9-fluorenone as the chemical scaffold with  symmetric dimeric naphthalene groups at both ends linked by an aminosulfonyl. The presence of dimeric naphthyl substituents showing strong hydrophobic π-π stacked interactions at one end and simple π-Sigma interactions at the other suggesting strong binding interactions in the binding pocket. The chemical scaffold of 2 is 9-fluorenone (PubChem CID: 10241) which is actively used in preparation of antimalarial drugs, functional polymers, and dyes. 46 Of the two selected compounds, 1 was prioritized based on the strong docking predictions with the GPR120S model, its novelty with respect to the literature and the micromolar cytotoxic activity in CRC cell line, for SAR studies. To expand the SAR profile, the generic chemical structure, dihydrospiro(benzo[h]quinazoline-5,1 ′ -cyclopentane)-4(3H)-one, in combination with sulfanyl acetone tail was used as query for substructure search against the pre-processed SPECS database using DS pipeline pilot ( Table 2).
The substructure search resulted in 16 compounds from the prescreened SPECS database. This set of compounds explored R-groups in position R 1 and R 2 of the scaffold (see Table 2) in combination with in silico ADME profiling using SwissADME (http://www.swissadme.ch/) 47 to procure the selective compounds for in vitro screening. The analogues were selected exploring the alkyl to aryl substitutions at the R 2 position and simple halobenzene susbstitutions at the R 1 position connected by a sulfanyl acetone linker keeping dihydrospiro(benzo[h]quinazoline-5,1 ′cyclopentane)-4(3H)-one scaffold structure intact. Based on the docking analysis, manual screening and the availability of compounds at SPECS, seven compounds were tested for in vitro cytotoxicity assay in SW480 cell line as discussed above (see Fig 5a).
Based on the SAR study, the IC 50 values of the new hits enable an initial identification of the essential pharmacophore features required in the dihydrospiro(benzo[h]quinazoline-5,1 ′ -cyclopentane)-4(3H)-one scaffold. The presence of an aromatic halogen at the sulfanyl acetone tail exhibited an increase in potency (3, 5, 7). However, the substitution of chloride in comparison to bromide seems to be more effective for pharmacological activity. The drop in potency of 8 w.r.t the parent compound (1) and 5 can be related to larger atomic size of bromine atom compared to chlorine which can result in decreased solubility and hence lowering the bioavailability of the compounds. Although not predicted in molecular docking which showed similar binding scores − 10.22, − 10.706 and − 10.704 (compounds 1, 5 and 8 respectively), the greater size of bromine might be responsible for steric clashes with neighbouring residues in the binding pocket which can be further analysed by future MD studies. Incorporation of fluoroaromatics at the sulfanyl acetone tail may increase the solubility and hence bioavailability of these SAR analogues as the fluorine atom is smaller in size and also exhibit slight electronegativity which can increase the electrostatic bonding affinity of the compounds.
Substitution of the benzo-methyl at the R 2 position in the parent structure and 8 by a smaller methyl group (3) or an aromatic sixmembered group (5) resulted in a significant increase in potency of the analogues. The substitution of a non-aromatic cyclic group (4) resulted in reduced activity of the parent compound (1). The addition of a methoxy group to this aromatic ring at R 2 position in 7 resulted in the second most active compound of the SAR profiling. The quinazoline ring linked to five or six membered aromatic ring structure at R 2 position by a single C-C bond length seems to be the optimum as when the linker length in 5 is increase (-C-C 2 H 4 -C-), the cytotoxic activity of 6 registered a drastic decrease from ~ 6 µM to ~ 80 µM. The total inactivity of 9 may confirm the above inferences as it contains bromo-aromatic group at R 1 position and non-aromatic cyclic ring with a longer linker at the R 2 position.
SAR analysis is a state-of-the-art and precise method to explore bioactive analogues of an active compound which may improve binding affinity to a protein target. 19,20 Of the present SAR study, analogues (5 and 7) exhibited the best cytotoxic activity. When both these compounds were tested at 5 µM (below experimental IC 50 ) in GPR120-silenced SW480 cells, this cytotoxic effect of 5 was significantly suppressed in GPR120-siRNA transfected SW480 cells (Fig 5b) while 7 showed ~10% higher cytotoxicity levels in siRNA transfected cells. The comparative study between GPR120-silenced and control experiments suggested that 5 exhibited cytotoxic effects through GPR120 binding while cytotoxic activity of 7 might be through multiple targets including GPR120. However, further in silico and in vitro validation is required to confirm their anti-cancer potential targeting GPR120. As literature suggests that high levels of GPR120 expression in CRC cell lines increases the cell proliferation rate and reduces apoptosis, 6 it can be hypothesized that compound 5 and 7 inhibit GPR120 and hence increase the apoptosis rate. A comparative study of these test compounds against a competitive antagonist would be useful, but as mentioned earlier no GPR120 antagonists are available to date. AH7614 (4-Methyl-N-9H-xanthen-9-ylbenzenesulfonamide) was first reported as a GPR120 selective antagonist by GlaxoSmithKline in 2014 48 but its mechanism of antagonism was not known. Later collaborative research by the Ulven and Milligan labs in 2017 49 reported that AH7614 was a negative allosteric modulator of GPR120.
In summary, with the identification of the two most active analogues siRNA-mediated silencing of GPR120 in SW480 cells -Compound 5 and 7 were screened against GPR120-siRNA and scrambled-siRNA (control) transfected SW480 cells. Results from five replicates are expressed as the mean ± standard error, ****p < 0.0001 as indicated (ANOVA, Tukey's multiple comparisons test). Percent relative cell viability for all treatments were quantified and normalised to the maximal response induced by vehicle control.
from the SAR profiling of parent compound 1 (AK-968/12713190) with a dihydrospiro(benzo[h]quinazoline-5,1 ′ -cyclopentane)-4(3H)-one scaffold, our study has successfully identified a novel scaffold for developing potential therapeutics for CRC management. The main scaffold can be further enhanced focusing on the substitution and addition of key structural groups as mentioned in the SAR study. Future studies will confirm the role of GPR120 in the cytotoxic activity exhibited by these compounds.

Ethics approval
Not applicable.

Consent to participate
Not applicable.

Consent for publication
The authors hereby give consent for the publication of this manuscript.