Characterization of cancer-related somatic mutations in the adenosine A2B receptor.

In cancer, G protein-coupled receptors (GPCRs) are involved in tumor progression and metastasis. In this study we particularly examined one GPCR, the adenosine A2B receptor. This receptor is activated by high concentrations of its endogenous ligand adenosine, which suppresses the immune response to fight tumor progression. A series of adenosine A2B receptor mutations were retrieved from the Cancer Genome Atlas harboring data from patient samples with different cancer types. The main goal of this work was to investigate the pharmacology of these mutant receptors using a 'single-GPCR-one-G protein' yeast assay technology. Concentration-growth curves were obtained with the full agonist NECA for the wild-type receptor and 15 mutants. Compared to wild-type receptor, the constitutive activity levels in mutant receptors F141 L4.61, Y202C5.58 and L310 P8.63 were high, while the potency and efficacy of NECA and BAY 60-6583 on Y202C5.58 was lower. A 33- and 26-fold higher constitutive activity on F141 L4.61 and L310 P8.63 was reduced to wild-type levels in response to the inverse agonist ZM241385. These constitutively active mutants may thus be tumor promoting. Mutant receptors F259 S6.60 and Y113 F34.53 showed a more than one log-unit decrease in potency. A complete loss of activation was observed in mutant receptors C29 R1.54, W130C4.50 and P249 L6.50. All mutations were characterized at the structural level, generating hypotheses of their roles on modulating the receptor conformational equilibrium. Taken together, this study is the first to investigate the nature of adenosine A2B receptor cancer mutations and may thus provide insights in mutant receptor function in cancer.


Introduction
G protein-coupled receptors (GPCRs) are a family of membranebound proteins that have seven-transmembrane (7TM) domains, connected by three intracellular (IL) and three extracellular (EL) loops, an extracellular amino terminus, and an intracellular carboxyl terminus (Vassilatis et al., 2003). GPCRs are responsive to a diverse set of physiological endogenous ligands including hormones and neurotransmitters. In total approximately 800 GPCRs are present in the human genome which can be subdivided in five main families, namely glutamate, rhodopsin, adhesion, frizzled/taste, and secretin, according to the GRAFS classification system (Fredriksson et al., 2003).
GPCRs have been relatively underappreciated in preclinical oncology, where primary focus over the last two decades has been on kinases due to their central role in the cell cycle. However, there is a growing body of evidence showing a more prominent role of GPCRs in all phases of cancer (Lappano and Maggiolini, 2012). In addition, recent work has shown that the function of GPCRs present in patient isolates is altered due to mutation, and indeed GPCRs are mutated in an estimated 20% of all cancers (Kan et al., 2010;Watson et al., 2013).
One particular class of rhodopsin-like GPCRs are the adenosine receptors (ARs), which respond to adenosine as their natural ligand (Fredholm, 2010). There are four adenosine receptor subtypes, namely A 1 , A 2A , A 2B , and A 3 . The adenosine A 1 and A 3 receptors couple to the G i protein and consequently inhibit adenylate cyclase and cAMP production. Conversely adenosine A 2A and A 2B receptors activate adenylate cyclase via coupling to the G s protein (Fredholm et al., 2001a(Fredholm et al., , 2010. While the role of the immune system in cancer defense is vital, the https://doi.org/10.1016/j.ejphar.2020.173126 Received 18 November 2019; Received in revised form 16 April 2020; Accepted 20 April 2020 underlying mechanisms are yet not very well understood (Candeias and Gaipl, 2015). However, studies in immune cells provide evidence for the involvement of adenosine and adenosine receptors in these mechanisms (Antonioli et al., 2014). Additionally, increased concentrations of adenosine are present in the tumor microenvironment. Hence adenosine may activate all subtypes of the adenosine receptors in cancer, including the low-affinity adenosine A 2B receptor (Gessi et al., 2011).
Adenosine A 2B receptors expressed in human microvascular cells are likely modulating expression of angiogenic factors like vascular endothelial growth factor, interleukin-8 and basic fibroblast growth factor (Feoktistov et al., 2002;Fredholm et al., 2001b). In HT29 colon (Merighi et al., 2007), U87MG glioblastoma  and A375 melanoma cancer cells (Merighi et al., 2001), increased interleukin-8 levels were observed after stimulation of adenosine A 2B receptor resulting in cell proliferation. Moreover, it has been suggested that adenosine A 2B receptors also regulate immunosuppression in the tumor microenvironment (Ryzhov et al., 2008;Sorrentino et al., 2015). Recently, constitutive activity of the adenosine A 2B receptor has been determined and proven to be involved in the promotion of cell proliferation in prostate cancer (Vecchio et al., 2016). Furthermore, the adenosine A 2B receptor is frequently overexpressed in oral squamous cell carcinoma and triple negative breast cancer cell lines, and promotes cancer progression (Kasama et al., 2015;Mittal et al., 2016). On the other hand, it was shown that this receptor inhibits ERK 1/2 phosphorylation, resulting in an anti-proliferative action in ER-negative MDA-MB-231 cells (Bieber et al., 2008). However, a more recent study provided evidence for enhanced MAPK signaling via activation of the A 2B receptor, promoting tumor progression in bladder urothelial carcinoma (Zhou et al., 2017). Taken all evidences together, it appears that activation of the adenosine A 2B receptor promotes tumor progression (Allard et al., 2016;Sepúlveda et al., 2016).
From a mechanistic point of view, it has been shown that mutations in this receptor can result in altered constitutive and agonist-induced receptor activity (Liu et al., 2015;Peeters et al., 2011aPeeters et al., , 2014Peeters et al., , 2012. Mutant receptors that show increased basal activity independent of an agonist, are referred to as constitutively active mutants (CAMs), while those with decreased basal activity are termed constitutively inactive mutants (CIMs) (Peeters et al., 2011b). Based on the two-state-receptor model (Leff, 1995), the equilibrium between the inactive (R) and active (R*) receptor conformation is shifted in these CAMs and CIMs.
In the present study, we selected 15 mutations in adenosine A 2B receptor from all cancer types using a bioinformatics approach. Subsequently these mutations were screened in an S.cerevisiae strain to study the effect of these mutations on receptor activation using different reference ligands (Fig. 1). We found that these mutations resulted in 3 CAMs and 5 less active mutants. Moreover, we found 4 mutants that behaved similar to the wild-type receptor (i.e. no effect mutants) and 3 loss-of-function mutants. Taken together, this study is the first to characterize cancer mutations on adenosine A 2B receptor at the molecular level and may thus provide insights in mutant receptor function in cancer.

Data mining
Mutation data from The Cancer Genome Atlas (TCGA, version August 8th, 2015) was downloaded using Firehose (Broad Institute TCGA Genome Data Analysis Center, 2016; Weinstein et al., 2013). Subsequently the data was extracted, when available MutSig 2.0 data was used, in cases where this was not available MutSig 2CV was used (specifically Colon Adenocarcinoma, Acute Myeloid Leukemia, Ovarian Cerous Cystadenocarcinoma, Rectum Adenocarcinoma). Natural variance mutation data was downloaded from Uniprot in the form of 'Index of Protein Altering Variants' (version November 11th , 2015) (The 1000 Genomes Project Consortium, 2015).Sequence data was filtered for missense somatic mutations and the protein of interest (human A 2B receptor, Uniprot accession P29275) ("UniProt: the universal protein knowledgebase," 2017). The GPCRdb alignment tool was used to assign Ballesteros Weinstein numbers (Ballesteros and Weinstein, 1995;Isberg et al., 2016), indicated by superscripts after the corresponding residue number. Finally, 15 cancer-related mutations were identified of which one mutation (L310P) was a duplicate, i.e. present in samples from two separate patients both suffering from colon adenocarcinoma.

Materials
The MMY24 strain, the S. cerevisiae expression vectors, the pDT-PGK plasmid, and the wild-type human adenosine A 2B receptor in pDT-PGK plasmid were kindly provided by Dr. Simon Dowell from GSK (Stevenage, UK). The QuikChange II® Site-Directed Mutagenesis Kit was purchased from Agilent Technologies, which includes XL10-Gold ultracompetent cells (Amstelveen, the Netherlands). The QIAprep mini plasmid purification kit was purchased from QIAGEN (Amsterdam, the Netherlands). NECA (Fig. 1), 3-amino-[1,2,4]-triazole (3-AT), bovine serum albumin (BSA) and 3, 3′,5,5′-tetramethyl-benzidine (TMB) were purchased from Sigma-Aldrich (Zwijndrecht, the Netherlands). BAY 60-6583 ( Fig. 1) was synthesized in house. ZM241385 was purchased from Ascent Scientific (Bristol, United Kingdom). PSB603 ( Fig. 1) was purchased from TOCRIS Bioscience (Abingdon, United Kingdom). The Hybond-ECL membrane and the ECL Western blotting analysis system were purchased from GE Healthcare (Eindhoven, the Netherlands). The antibody directed against the C-terminal region of the human adenosine A 2B receptor was kindly provided by Dr. I. Feoktistov (Vanderbilt University, Nashville, TN, USA), the antibody directed against the extracellular region of the human adenosine A 2B receptor was purchased from Alpha Diagnostic International (San Antonio, USA) and goat antirabbit IgG was purchased from Jackson ImmunoResearch Laboratories (West Grove, PA, USA).

Generation of human adenosine A 2B receptor mutations
The mutations were generated as described previously by Liu et al. (2015). In short, DNA primers for the human adenosine A 2B receptor (uniprot: P29275) mutations were designed by the QuikChange Primer Design Program of Agilent Technologies (Santa Clara, CA, USA). These primers contained a single substitution resulting in a codon change for the desired amino acid substitution. Primers and their complements were synthesized (Eurogentec, Maastricht, the Netherlands) and used to generate mutation plasmids according to the QuikChange method from Agilent Technologies. The mutagenic PCR was performed in the presence of 50 ng of template DNA, 10 μM concentration of each primer, 1 μl of dNTP mix, 2.5 μl of 10 x reaction buffer and 2.5 U PfuUtra HF DNA polymerase. The number of mutagenic PCR cycles was set to 22 (PCR cycling conditions: 95°C for 30 s, 55°C for 1 min, and 68°C for 10 min). The methylated or hemimethylated non-mutated plasmid DNA was removed by 5 U Dpn I restriction enzyme incubating for 2 h at 37°C. The mutated DNA plasmids were transformed into XL-1 Blue supercompetent cells according to the manual of the QuikChange II® Site-Directed Mutagenesis Kit. After plasmid isolation by using a QIAprep mini plasmid purification kit, the mutations were verified by double-strain DNA sequencing (LGTC, Leiden University, the Netherlands).

Transformation in a MMY24 S. Cerevisiae strain
The yeast strain was derived from the MMY11 strain and further adapted to communicate with mammalian GPCRs through a specific Gpa1p/G αi3 chimeric G protein. The genotype of the MMY24 is: MATa his3 leu2 trp1 ura3 can1 gpa1_::G_i3 far1::ura3 sst2_::ura3 Fus1::FUS1-HIS3 LEU2::FUS1-lacZ ste2_::G418R and the last five amino acid residues of C-terminus of Gpa1p/G αi3 chimera were replaced by the same length sequence from mammalian G αi3 protein (Dowell and Brown, 2009). The pheromone pathway was coupled to HIS3 reporter gene via FUS1 promotor, so that the level of receptor activation can be measured directly by the yeast growth rate on histidine-deficient medium. The plasmids containing the mutant adenosine A 2B receptors were transformed into an S. cerevisae strain according to the Lithium-Acetate procedure (Gietz and Schiestl, 2007).

Solid growth assay
To characterize the activation of various mutated receptors, concentration-growth curves were generated from a solid growth assay. This assay was performed with yeast cells from an overnight culture in selective YNB medium lacking uracil and leucine (YNB-UL) as this yeast strain can produce leucine and the plasmid also contains a gene encoding for uracil production. The yeast cells were diluted to 400,000 cells/ml (OD 600 ≈ 0.02) and droplets of 1.5 μl were spotted on selection agar plates, i.e. YNB agar medium lacking uracil, leucine and histidine (YNB-ULH). In addition, the agar on the plates contained 7 mM 3-AT and the adenosine A 2B receptor full agonist NECA in a concentration range from 10 −9 to 10 −4 M, or the non-ribose adenosine A 2B receptor agonist BAY 60-6583 (Eckle et al., 2007) in a concentration range from 10 −9 to 10 −5 M, or the adenosine A 2B receptor inverse agonist ZM241385 (Li et al., 2007) in a concentration range from 10 −9 to 10 −5 M. After 50 h incubation at 30°C, the plates were scanned and receptor activation-mediated yeast cell growth was quantified with ImageLab 5.2.1 software of a ChemiDoc MP Imaging System from Bio-Rad (Hercules, CA, USA). The level of yeast cell growth was calculated as the intensity of each spot after correction for local background on the plate.

Liquid growth assay and schild-plot analysis
To characterize the mutant adenosine A 2B receptors further, similar concentration-growth curves were obtained using liquid YNB-ULH medium and 96-well plates. Yeast cells were inoculated in YNB-UL and diluted to 4·10 6 cells/ml (OD 600 ≈ 0.2). To each well, 150 μl YNB-ULH medium containing 7 mM 3-AT, 2 μl various concentrations of ligands and 50 μl yeast cells were added. The 96-well plate was then incubated at 30°C for 35 h in a Genios plate reader, and was shaken every 10 min at 300 rpm for 1 min to keep the cells in suspension. The level of yeast cell growth was determined by the optical density at a wavelength of 595 nm.
For the wild-type adenosine A 2B receptor, mutant receptor F141L 4.61 , Y202C 5.58 and L310P 8.63 , concentration-growth curves of NECA were generated in the absence or presence of different concentrations of the selective adenosine A 2B receptor antagonist PSB603.

Whole cell extracts and immunoblotting
In order to determine the expression level of the different mutant adenosine A 2B receptors in the MMY24 yeast strain, an immunoblotting method was used as described previously (Liu et al., 2015). In short, whole cell protein extracts were prepared using trichloroacetic acid (TCA) from an overnight culture (1.2·10 8 yeast cells). After two washing steps with 20% TCA, yeast cells were broken by thoroughly vortexing in the presence of glass beads. Then, a semi-automated electrophoresis technique (PhastSystem™, Amersham Pharmacia Biotech) was used to separate 4 μl (≈24 μg) of the yeast cell extract on a 12.5% SDS/PAGE gel. Subsequently the proteins were blotted on a Hybond-ECL membrane, where a rabbit anti-human adenosine A 2B receptor primary antibody was used directed against the C-terminus of the human adenosine A 2B receptor. The remaining unbound antibody was washed off the membranes repeatedly with TBST (0.05% Tween 20 in Tris-buffered saline), and the HRP-conjugated goat anti-rabbit IgG (Jackson Im-munoResearch Laboratories) was added. Via the ECL Western blotting analysis (GE Healthcare, the Netherlands), the specific human adenosine A 2B receptor bands were found at 29 kDa and 48 kDa, whereas a nonspecific band was detected at approximately 45 kDa, which was used as loading control. Densitometric analysis was performed by the volume analysis tool in the ImageLab 5.2.1 software from Bio-Rad (Hercules, CA, USA). The ratio between the intensity of specific human adenosine A 2B receptor protein bands and nonspecific protein band was determined, where the yeast strain carrying wild-type hA 2B R was set to 100% and the yeast strain carrying the empty vector pDT-PGK was set to 0%.

Yeast enzyme-linked immunosorbent assay (ELISA)
Yeast ELISA was adapted from a method reported previously (Guo et al., 2010). About 2·10 7 yeast cells were collected from an overnight culture in YNB-UL medium. The cells were blocked with 2% BSA in Tris-buffered saline (TBS), followed by 1 h incubation with polyclonal rabbit anti-human adenosine A 2B receptor antibody (1:2000) in TBST containing 0.1% BSA at room temperature. After washing three times with TBST, HRP-conjugated goat anti-rabbit IgG (1:5000) was added and incubated for 1 h at room temperature. After removing the secondary antibody and washing the cells with TBS, TMB was added and incubated for 15 min in the dark. The reaction was stopped with 1 M H 3 PO 4 , and absorbance was read at 450 nm using a Wallac EnVision 2104 Multilabel reader (PerkinElmer).

Homology modeling
Homology models were generated using the GPCR-ModSim webbased pipeline for modeling GPCRs (Esguerra et al., 2016;Rodríguez et al., 2012), available through http://open.gpcr-modsim.org/. Shortly, this webserver performs a multiple sequence alignment against a curated set of crystalized GPCRs, divided in three categories: inactive, active-like and fully-active. The best templates are suggested based on the overall homology with the target sequence and the user can select between single-template or multi-template homology modeling. In this study, the adenosine A 2A receptor was the only template used to model the adenosine A 2B receptor, using the structures deposited with PDB codes 3EML for the inactive model, 2YDV for the active-like and 5G53 for the fully-active structure. After manual adjustment of the alignment of the extracellular loop 2 (ECL2) region, which is recommended due to the high variability within this loop region, homology models were created and sorted by the DOPHR scoring function by the server, using Modeller as a background engine (Webb and Sali, 2014). The best model was finally selected among the top scored models based on visual inspection.

Data analysis
All experimental data were analyzed using GraphPad Prism 7.0 software (GraphPad Software Inc., San Diego, CA, USA). Solid and liquid growth assays were analyzed by non-linear regression using "log (agonist or inhibitor) vs. response (three parameters)" to obtain potency (EC 50 ), inhibitory potency (IC 50 ) and efficacy (E max ) values. For Schild-plot analysis, log[DR-1] was calculated by equation log[DR-1] = log[(A'/A)-1], where A′ is the EC 50 value obtained from the concentration-growth curves in the presence of antagonist, A is the EC 50 value obtained from the concentration-growth curves in the absence of antagonist. pA 2 values were generated by using linear regression. Statistical evaluation was performed by a two-tailed unpaired Student's t-test between pEC 50 or E max values of mutant receptors and values of wild-type receptor obtained from the same experiments. All values obtained are means of at least three individual experiments performed in duplicate.

Data mining and expression of mutants
In total 15 point mutations in the adenosine A 2B receptor were identified in cancer patient isolates by data mining the TCGA database on August 8th , 2015. Three mutations were located at an extracellular loop (EL), two at an intracellular loop (IL), six in the 7-transmembrane domain and four at the C-terminus of the adenosine A 2B receptor. Of note, three mutations were found for position L310 8.63, i.e. L310F, L310I and L310P, of which the latter was present in two separate patient isolates. In the natural variance set 16 point mutations were identified, namely A35V 1.60 , T37S 12.49 , C72Y 23.51 , A82T 3.29 , I126T 4.46 , L129I 4.49 , G135R 4.55 , C171S 45.50 , L172P 45.51 , R223W 6.24 , R228W 6.29 , K267E EL3 , R293W 7.56 , R295Q 8.48 , D296G 8.49 and R298H 8.51 . However, none of these were found to match any of the residue positions of the cancer-related mutations. All 15 cancer-related mutants were constructed and transformed into the MMY24 yeast strain, and showed overall receptor expression, as determined by Western blot analysis of whole cell lysates ( Fig. 2A and Table S1). Besides, all mutant receptors also showed cell surface expression, as determined by ELISA performed on intact yeast cells ( Fig. 2B and Table S1).

Characterization of mutant adenosine A 2B receptors on receptor activation
To characterize the effects of the cancer-related mutations on receptor activation, the pharmacology of these 15 mutant receptors was investigated by yeast-growth assays. Concentration-growth curves and results of the mutant receptors in response to the full agonist NECA are shown in Fig. 3 and Table 1.
In this system, the wild-type receptor showed a pEC 50 value of 6.79 ± 0.09 for NECA and a low level of basal activity, if any. Overall half of the mutant receptors showed similar constitutive activity, potency and efficacy as the wild-type receptors (Table 1). In the ECL and ICL, two mutants behaved significantly different from wild-type adenosine A 2B receptor in response to NECA, i.e. mutant receptor F259S 6.60 (Fig. 3A) and mutant receptor Y113F 34.53 (Fig. 3B) displayed a more than 1 log-unit decreased potency. The maximum receptor activation (E max ) of these two mutant receptors was also decreased to 57% for both (Table 1). At the 7-transmembrane domain of the adenosine A 2B receptor, mutants C29R 1.54 , W130C 4.50 and P249L 6.50 showed a complete loss of function (Fig. 3C). In contrast, increased constitutive activities were observed for mutant receptors F141L 4.61 and Y202C 5.58 , which were 34-and 48-fold higher than the wild-type receptor, respectively ( Fig. 3C and Table 1). In response to NECA, mutant receptor F141L 4.61 showed a significantly increased potency with a pEC 50 value of 7.31 ± 0.03 and an increased E max value of 122%, while mutant receptor Y202C 5.58 showed a significantly reduced pEC 50 value of 6.30 ± 0.09 and reduced E max value of 73% in comparison to the wildtype receptor (Fig. 3C and Table 1). At the C-terminus, mutant receptor L310P 8.63 showed a relatively high level of constitutive activity, i.e. 26fold higher than wild-type receptor, as well as an increased potency for NECA. Mutant receptor L310I 8.63 showed a significantly decreased pEC 50 value of 6.40 ± 0.01 compared to wild-type receptor ( Fig. 3D and Table 1), while the E max value did not decrease significantly. Mutant receptor V315G 8.68 showed similar pEC 50 and E max values to wildtype receptor (Fig. 3D and Table 1), although the cell surface expression level was much higher than wild-type receptor (Fig. 2B). Additional yeast ELISA experiments were subsequently performed on yeast colonies carrying either wild-type or mutant receptor V315G 8.68 , and colonies with similar expression levels were selected in the yeast liquid growth assay to further investigate receptor activation in response to NECA. Interestingly, similar pEC 50 and E max values were still obtained for mutant receptor V315G 8.68 and wild-type adenosine A 2B receptor (Table S2).
The mutant receptors F259S 6.60 , Y113F 34.53 , C29R 1.54 , W130C 4.50 , P249L 6.50 , F141L 4.61 , Y202C 5.58 and L310P 8.63 which all showed the more altered pharmacological response to NECA were also studied with BAY 60-6583, a non-ribose adenosine A 2B receptor agonist (Fig. 1). On the wild-type receptor, BAY 60-6583 showed a pEC 50 value of 7.49 ± 0.33. As for NECA, decreased potencies and E max values were also observed for BAY 60-6583 at F259S 6.60 and Y113F 34.53 in comparison to the wild-type receptor (Table 1). Mutant receptors C29R 1.54 , W130C 4.50 and P249L 6.50 , which did not show any activity in the presence of NECA, were not activated by BAY 60-6583 either ( Table 1). As opposed to NECA, the effects of BAY 60-6583 on mutant receptors F141L 4.61 , Y202C 5.58 and L310P 8.63 were similar when compared to the wild-type receptor (Table 1).

Inverse agonism of the CAMs
To study whether receptors bearing CAMs (F141L 4.61 , Y202C 5.58 and L310P 8.63 ) could still be inhibited, an inverse agonist of the A 2B receptor, ZM241385, was used in the yeast solid growth assay, and compared to the wild-type receptor. The concentration-growth inhibition curves are shown in Fig. 4. The wild-type adenosine A 2B receptor had low basal activity in this system, and ZM241385 did not further reduce this. All CAMs displayed a decreased activity upon increasing concentrations of ZM241385. The level of constitutive activity of F141L 4.61 and L310P 8.63 was fully suppressed, with pIC 50 values of 7.43 ± 0.17 and 7.49 ± 0.27 for ZM241385, respectively ( Fig. 4 and Table 1). For the mutant receptor with the highest constitutive activity Y202C 5.58 , ZM241385 caused a partial reduction to a residual activity of 19% and with a lower pIC 50 of 6.62 ± 0.23 than for the less active CAMs, i.e. F141L 4.61 and L310P 8.63 (Fig. 4).

Characterization of CAMs on ligand binding
To investigate whether the affinity of an antagonist had changed on the CAM receptors F141L 4.61 , Y202C 5.58 and L310P 8.63 , a Schild-plot analysis was performed (Fig. 5). Concentration-growth curves were generated of the agonist NECA in the absence or presence of increasing concentrations of the competitive adenosine A 2B receptor antagonist PSB603 (Fig. 1). For the wild-type and mutant receptors F141L 4.61 and L310P 8.63 , a rightward shift of the concentration-growth curves was observed with increasing concentrations of PSB603 ( Fig. 5A-C). This resulted in decreased apparent pEC 50 values for NECA, while the maximal growth levels for these mutant receptors were still reached. The pA 2 values, as determined from the Schild-plot, were 8.47 for wildtype, 8.07 for F141L 4.61 and 8.27 for L310P 8.63 , showing that the antagonist affinity for these two mutants was not significantly influenced by the mutations (P = 0.0776 and P = 0.3097, respectively). Of note, a Schild-plot could not be generated for mutant receptor Y202C 5.58 , as the presence of PSB603 (at the selected concentrations) did not cause a shift in agonist potency (Fig. 5D).

Structural mapping and analysis of mutations
The mutations investigated in this study were mapped on a homology model of the adenosine A 2B receptor to provide structural hypotheses for the observed pharmacological effect (NEM, LFM, CAM and LAM) of the different mutations. As illustrated in Fig. 6A, the studied mutations are distributed over the whole receptor. The NEMs are positioned in the ECL, ICLs and C-terminus regions, while all three LFMs are exclusively located in the transmembrane region. These mutations are either part of highly conserved areas of the receptor (W130C 4.50 and P249L 6.50 , Fig. 6C) and/or introduce a drastic change in the properties of the amino acid (C29R 1.54 , W130C 4.50 ). The two LAMs are located in either the ECL (F259S 6.60 , Fig. 6C) or ICL (Y113F 34.53 , Fig. 6D) region, where F259 6.60 is located on the top of the helix 6, while Y113 34.53 is part of the TDY triad (see Discussion). Finally, the three CAMs are located in the transmembrane region or in helix 8. Mutant receptor L310 P 8.63 has been identified in two individual patients. Additionally, mutations of this residue to Phe (F) and Ile (I) were found, where mutation to F did not alter receptor functionality. The F141L 4.61 mutation is positioned at the membrane-helix interface on the top of TM4 and at the start of ECL2. Noteworthy are the opposing pharmacological effects observed between F259S 6.60 and F141L 4.61 (i.e. LAM and CAM, respectively), although they are both situated at a structurally similar location. Y202 5.58 is part of an activation switch with Y290 7.53 in the NPxxY motif, moving outward into the membrane in the agonist-bound receptor as observed in the adenosine A 2A receptor crystal structures (Fig. 6B).

Discussion
Numerous GPCR mutations are known to alter the pharmacological function of the receptor by affecting constitutive activity, ligand The lower panel shows the bar graph obtained from the densitometric analysis. Expression levels were determined between the density of specific bands and the density of the non-specific band that is present in all lanes. Wild-type receptor was set at 100% and the empty vector pDT-PGK was set at 0%. The combined bar graph is shown in mean ± S.E.M. from at least three individual experiments. B) Cell surface expression levels of wild-type and mutant adenosine A 2B receptors, determined in an enzymelinked immunosorbent assay. Wild-type receptor was set at 100% and the empty vector pDT-PGK was set at 0%. The bar graph is shown in mean ± S.E.M. of three individual experiments performed in duplicate.
binding, GPCR-G protein interaction, and/or cell surface expression, resulting in a wide range of disease phenotypes (Stoy and Gurevich, 2015). Moreover, a variety of mutations within GPCRs have been linked to different types of cancer (Bar-Shavit et al., 2016;O'Hayre et al., 2013;Sodhi et al., 2004), but the pharmacological effects of these cancer-related mutations are not yet fully understood. Previously, several studies performed on the adenosine A 2B receptor demonstrated some residues to be essential in receptor activation (Beukers et al., 2004;Isberg et al., 2016;Liu et al., 2015;Peeters et al., 2011aPeeters et al., , 2014Peeters et al., , 2012. Hence, in this study 15 single-site point mutations of adenosine A 2B receptor identified from The Cancer Genome Atlas (TCGA) (Broad Institute TCGA Genome Data Analysis Center, 2016) were examined in the S.cerevisiae system to improve our understanding of the mechanism of receptor activation in relation to cancer progression. As none of these mutations were found in the natural variance mutation dataset (The 1000 Genomes Project Consortium, 2015), they could be cancer specific, yet follow up research is warranted.

Table 1
Characterization of adenosine A 2B receptor mutations identified in cancer patient samples in yeast solid growth assays. Mutations are shown in the numbering of the adenosine A 2B receptor amino acid sequence as well as according to the Ballesteros and Weinstein GPCR numbering system. pEC 50 and E max values are shown as mean ± S.D. from at least three individual experiments, each performed in triplicate. The percentage maximum effect (% E max ) and the fold constitutive activity values were calculated by the mean values generated from the concentration-growth curves, compared to the wild-type receptor. Dose-growth curves of BAY 60-6583 were only generated for mutants that responded significantly different to NECA in comparison to the wild-type receptor. 1.4 6.89 ± 0.19 94 ± 4 --NEM a P < 0.05, b P < 0.01, c P < 0.001, d P < 0.0001 compared to the wild-type receptor in response to NECA; e P < 0.05, f P < 0.001, g P < 0.0001 compared to the wild-type receptor in response to BAY 60-6583, determined by a two-tailed unpaired Student's t-test. ND: not detectable, -: not measured, CA: constitutive activity; CAM: constitutively active mutant, LAM: less active mutant, LFM: loss of function mutant, NEM: no effect mutant.

Less active mutations
Mutant receptors F259S 6.60 and Y113F 34.53 were identified from colon adenocarcinoma and lung adenocarcinoma located at opposite sides of the receptor, ECL3 and ICL2. However, both showed decreased potency and efficacy in the case of ribose and non-ribose agonists ( Fig. 1 and Table 1, LAMs). These data are consistent with a previous study on the CC-chemokine receptor 5 (CCR5) receptor, showing that mutation from phenylalanine to alanine at position 6.60 resulted in a decreased affinity of the agonist gp120 (Garcia-Perez et al., 2011).
Residue Y113 34.53 is completely conserved among adenosine receptors and several other class A GPCRs. Structures of the adenosine A 1 and A 2A receptor show that this residue is part of a conserved triad (T 2.39 -D 3.49 -Y 34.53 ). We and others previously hypothesized a regulatory role of this motif in receptor activation (Isberg et al., 2016) by mediating the strength of the D102-R103 ionic lock (Jespers et al., 2017;Rodrí;guez et al., 2011), preventing an outward movement of R 3.50 necessary for G protein binding (Fig. 6D). The Y113F 34.53 mutation increases the electrostatic attraction between D102 3.49 and Fig. 4. Concentration-inhibition curves of the adenosine A 2B receptor inverse agonist ZM241385 at the wild-type adenosine A 2B receptor and CAMs, F141L 4.61 , Y202C 5.58 and L310P 8.63 . The maximum activation level of wild-type adenosine A 2B receptor was set at 100%, the background of the selection plate was set at 0%. Combined graphs are shown from at least three individual experiments, each performed in triplicate. R103 3.50 , reducing the mobility of this residue. Notably, at the same position in the β 1 -adrenergic receptor, mutation Y149A 34.53 leads to a large decrease in thermal stability of the antagonist bound state of the receptor (Warne et al., 2008).

Loss of function mutants
Mutant receptors C29R 1.54 , W130C 4.50 and P249L 6.50 , identified from stomach adenocarcinoma, liver hepatocellular carcinoma, and colon adenocarcinoma, respectively, showed a complete loss of activation ( Fig. 3C and Table 1). Importantly, none of these mutant receptors showed severely reduced expression levels compared to the wild-type receptor ( Fig. 2A and B, Table S1), indicating that the loss of activation is not due to the loss of expression in our model system. At residue C1.54, a drastic change from cysteine to arginine resulted in a LFM in our study. At the same position (1.54) in CCR2 and CCR5, the conservative mutation CCR2-V62I 1.54 did not affect receptor expression or ligand binding (Mariani et al., 1999), yet the inverse mutation on CCR5-I52V 1.54 showed a decreased affinity for CCL5 (Saita et al., 2006).
The tryptophan at position 130 4.50 and proline at position 249 6.50 are highly conserved among all class A GPCRs with a presence of 96% and 100% (Deupi et al., 2007). Mutant W4.50C investigated on CXCR4 and melanocortin-4 receptor (MC4R) showed abolished ligand binding and cAMP response (Boulais et al., 2013;Fan and Tao, 2009), indicating that the introduction of a drastic change in the amino acid properties at highly conserved positions can dramatically change receptor functionality. A rigid body motion of TM6 related to TM3 is known to be facilitated through the presence of the conserved proline in TM6 (6.50) (Palczewski et al., 2000). Pro 6.50 is also known as a rotamer toggle switch, playing a role in the structural rearrangement of class A GPCRs on transition from the inactive to the active state (Fig. 6C) (Visiers et al., 2002). Hence the observed loss of function for the mutations at positions 130 4.50 and 249 6.50 has some precedent.

Constitutively active mutants
Mutant receptors F141L 4.61 , Y202C 5.58 and L310P 8.63 , identified from skin cutaneous melanoma, liver hepatocellular carcinoma, and colon adenocarcinoma, respectively, showed increased constitutive activities compared to the wild-type receptor ( Fig. 3 and Table 1). These CAMs are located at TM4, TM5, and helix 8. It is known that TM4, TM5, ECL2, and helix 8 of the receptor are involved in receptor activation (Liu et al., 2015;Peeters et al., 2011b). Upon activation, a coupling between movements of ECL2 and TM5 has been observed as well as a rearrangement in the H-bond networks connecting ECL2 with the extracellular ends of TM4, TM5 and TM6 (Ahuja et al., 2009). Among the CAMs in this study, constitutive activities of F141L 4.61 and L310P 8.63 were reduced to wild-type level by inverse agonist ZM241385 with pIC 50 values of 7.43 and 7.49 (Fig. 4). These values are comparable to reference pK i values for wild-type adenosine A 2B receptor in the cAMP assay (de Zwart et al., 1999;Ongini et al., 1999). However, ZM241385 reduced the high constitutive activity of Y202C 5.58 roughly by half (Fig. 4), demonstrating that adenosine A 2B receptor is locked in an active conformation by mutation Y202C 5.58 , but not by F141L 4.61 or L310P 8.63 . Concordantly, PSB603 did not significantly inhibit the Y202C 5.58 mutant in a Schild-plot analysis at concentrations that inhibited the other two mutants (Fig. 5). Such "locked-in" receptor mutants have been described before, e.g. for the adenosine A 1 receptor (de Ligt et al., 2005). An altered potency value of NECA was observed on these CAMs compared to wild-type, while this difference was absent with BAY 60-6583 (Table 1). Similar results have been reported on adenosine A 2A receptor (Lane et al., 2012). The Color code is green for constitutively active mutants; black for loss-of-function mutants; red for less active mutants; and blue for no-effect-mutants. Panels (B, C and D) zoom in on selected residues mapped on the respective conformational models (inactive: orange, generated with 3eml; active-like: blue, generated with 2ydv; active: green, generated with 5g53). (B) The CAM Y202C 5.58 is located on part of an activation switch, which moves outwards to the membrane in the active-like structure and again inwards in the active structure, simultaneously F206 5.62 moves out in the active-like structure but remains in this position in the active structure. (C) The loss-of-function mutant P249L 6.50 disturbs a hinge region in the outward movement of TM6 observed upon receptor activation. (D) The less active mutant Y113F 34.53 is proposed to prevent the outward movement of R103 3.50 observed upon receptor activation. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) difference in receptor activation among wild-type and mutant receptors in response to CGS21680 (a ribose agonist) was not seen upon the activation mediated by LUF5834 (a non-ribose agonist). F141L 4.61 , located at the membrane-helix interface on the top of TM4 and at the start of ECL2 (Fig. 6A), has been reported to increase affinity and potency for NECA and BAY 60-6583 in a random mutagenesis study (Peeters et al., 2012). Similar results were observed here ( Fig. 3C and Table 1), while no significant changes on affinity of PSB603 were observed (Fig. 5E). As residue F141 4.61 is not located at the binding pocket of either NECA or BAY 60-6583 (Sherbiny et al., 2009) but still able to affect agonist activation, it is therefore likely that this mutation plays a role in the stabilization of ECL2 (Peeters et al., 2012) and also participates in the entry conformation of the agonist binding pocket. Interestingly, opposite pharmacological effects were observed between F259S 6.60 and F141L 4.61 , both situated at a structurally similar position (Fig. 3A, C, 6A and Table 1).
Three mutations were identified at "hotspot" L310 8.63 . L310P showed increased constitutive activity and potency on NECA, while L310F and L310I did not dramatically alter receptor functionality. Interestingly, the introduction of a proline mutation has a high impact as it may introduce a kink in helix 8, and thus potentially affect G protein coupling.
Residue Y5.58 is completely conserved among adenosine receptors and 88% conserved among class A GPCRs. Mutant receptor Y202C 5.58 showed the highest constitutive activity and reduced potency and efficacy of NECA compared to the wild-type receptor ( Fig. 3C and Table 1), indicating that maximal G protein coupling and signaling were decreased but consistently present. The residues Y 7.53 (part of NPxxY motif) and Y 5.58 were previously proposed as a possible activation switch for adenosine receptors, based on conformational changes observed in the agonist bound crystal structure (Xu et al., 2011) and their high conservation in class A GPCRs (Jespers et al., 2017). Comparing the inactive and active structures of adenosine A 2A receptor (PDB: 3EML, 2YDV and 5G53), we noticed the side chain of Y197 5.58 stretched into the membrane in the active structures (Lebon et al., 2015;Liu et al., 2012;Xu et al., 2011). As a consequence, TM5 and TM6 moved closer together enabling access of the G protein (Fig. 6B). Upon G protein binding, Y197 5.58 moves back into the receptor interior, filling up a space previously occupied by L235 6.37 and I238 6.40 . Additionally, the constitutive activities of the mutated receptors Y202C and Y202S (Peeters et al., 2012) are high, but not higher than the maximum observed effect of agonist bound receptors, further providing evidence that this residue is key for controlled modulation of the receptor.

Conclusion
In conclusion, 15 cancer-related somatic mutations on the adenosine A 2B receptor were retrieved from TCGA and characterized in a robust yeast system. We identified mutations that dramatically change receptor activation and function. Mutations in the adenosine A 2B receptor showing altered function in the yeast system may also be associated with cell proliferation and migration in cancer cell lines, and involved in cancer progression. Further studies in mammalian and/or cancer cell lines are warranted starting from the results in the present study to investigate mutation-mediated receptor activation and inactivation in a pathological setting. Since adenosine is an anti-inflammatory stimulus in the tumor microenvironment (Gessi et al., 2011), both wild-type and mutant adenosine receptors may play an important, yet largely undefined role in cancer progression, which eventually may be modulated with medicinal products.
CRediT authorship contribution statement

Acknowledgments
Xuesong Wang thanks the China Scholarship Council (CSC) for her PhD scholarship. Gerard van Westen thanks the Dutch Research Council, Domain Applied and Engineering Sciences (NWO AES), for financial support (Veni #14410). Hugo Gutiérrez-de-Terán and Willem Jespers acknowledge financial support from the Swedish research council (VR) and the Swedish National Infrastructure for computing (SNIC).

Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.org/10.1016/j.ejphar.2020.173126.