Molecular and Pharmacological Characterization of the Interaction between Human Geranylgeranyltransferase Type I and Ras-Related Protein Rap1B

Geranylgeranyltransferase type-I (GGTase-I) represents an important drug target since it contributes to the function of many proteins that are involved in tumor development and metastasis. This led to the development of GGTase-I inhibitors as anti-cancer drugs blocking the protein function and membrane association of e.g., Rap subfamilies that are involved in cell differentiation and cell growth. In the present study, we developed a new NanoBiT assay to monitor the interaction of human GGTase-I and its substrate Rap1B. Different Rap1B prenylation-deficient mutants (C181G, C181S, and ΔCQLL) were designed and investigated for their interaction with GGTase-I. While the Rap1B mutants C181G and C181S still exhibited interaction with human GGTase-I, mutant ΔCQLL, lacking the entire CAAX motif (defined by a cysteine residue, two aliphatic residues, and the C-terminal residue), showed reduced interaction. Moreover, a specific, peptidomimetic and competitive CAAX inhibitor was able to block the interaction of Rap1B with GGTase-I. Furthermore, activation of both Gαs-coupled human adenosine receptors, A2A (A2AAR) and A2B (A2BAR), increased the interaction between GGTase-I and Rap1B, probably representing a way to modulate prenylation and function of Rap1B. Thus, A2AAR and A2BAR antagonists might be promising candidates for therapeutic intervention for different types of cancer that overexpress Rap1B. Finally, the NanoBiT assay provides a tool to investigate the pharmacology of GGTase-I inhibitors.


Introduction
Post-translational modifications are important processes to expand and enhance protein functions in response to complex external stimuli. One example is the prenylation of proteins that are involved in biological regulation in eukaryotic cells [1]. Geranylgeranyltransferase-I (GGTase-I) and farnesyltransferase (FTase) are prenyltransferases that catalyze the transfer of a 20-carbon or a 15-carbon isoprenoid from geranylgeranyl diphosphate (GGPP) or farnesyl diphosphate (FPP), respectively, to proteins with a C-terminal CAAX motif-the CAAX motif is defined by a cysteine residue (C), two aliphatic residues (AA) and the C-terminal residue (X, determines enzyme prevalence; in the case of GGTase-I X refers mostly to leucine) that contribute to substrate specificity [2,3]. Both enzymes are functional as cytosolic heterodimers formed by an αand a β-subunit. The α-subunits of FTase and GGTase-I are identical (hereafter referred to as FTase α), whereas the β-subunits are different and harbor the active site. In different studies it was demonstrated that truncated rat (∆N29) and yeast FTase α-subunits are able to form homodimers [4]. In contrast, investigations conducted with full-length human FTase α-subunits showed no indication of homodimer formation by human FTase α [5]. Geranylgeranyltransferase type II (GGTase-II or RabGGTase) is more specialized and modifies only G-proteins from the Rab subfamily. Additionally, a recently identified geranylgeranyltransferase type III (GGTase-III) consists of an orphan prenyltransferase α-subunit and the catalytic β-subunit of GGTase-II [6,7].
One important substrate of GGTase-I is the Ras-like small GTPase Rap1, which has two isoforms Rap1A and Rap1B that differ only in a few amino acids [8]. The Rap1 protein has different cellular functions depending on its isoform and subcellular localization. For example, high Rap1B expression has been detected in different squamous cell carcinoma cell lines [9]. Many studies have indicated that Rap1B is required for cell survival and migration in human cancer cell models [10] and is targeted by several miRNAs [11][12][13].
The fact that prenylation is usually-depending on the cell-type-required for the oncogenic activity of small GTPases led to the development of GGTase-I inhibitors as potential anti-cancer drugs [14][15][16][17][18][19]. Recently, FGTI-2734-a Ras C-terminal mimetic dual farnesyltransferase and geranylgeranyltransferase inhibitor-inhibited membrane localization of K-Ras in human pancreatic, lung and colon cancer cells. Moreover, the compound induced apoptosis and inhibited the growth of mutant K-Ras-dependent human tumors in mice [20]. In a phase 1 study with the peptidomimetic small molecule inhibitor GGTI-2418 in patients with advanced solid tumors, the compound was safe and tolerable at all tested dose levels but was rapidly eliminated and could not achieve optimal GGTase-I inhibition [21]. Furthermore, GGTase inhibitors are postulated to have therapeutic potential in many other diseases, including multiple sclerosis, arteriosclerosis, viral infection, as well as osteoporosis and were developed as antifungals and antiparasitics [17,22]. The principal mechanism of the competitive, peptidomimetic CAAX GGTase inhibitors is to block the interaction of GGTase with their substrates, leading to lack of prenylation and therefore to mislocalization. However, many of the structurally diverse GGTase inhibitors were never characterized in detail, and their exact mode of action is unknown. Additionally, half maximal inhibitory concentration (IC 50) data for the same compound are often in disagreement when using different cell lines and assay systems.
In the present study, we developed a new NanoBiT assay to characterize the interaction of human GGTase-I and its substrate Rap1B in real-time. This assay system uses a split luciferase system (NanoLuc) consisting of an optimized small (1.3 kDa peptide, SmBiT) and a large (18 kDa polypeptide, LgBiT) NanoLuc subunit. Moreover, compared to other methods that are suitable to measure protein-protein interactions, such as most of the bimolecular fluorescence complementation systems (BiFC) [23], the interaction is reversible, and temporally dynamic protein interactions can be monitored in real-time [24]. Additionally, the NanoBiT technology was already evaluated successfully for many different drug targets, such as G protein-coupled receptors (GPCRs) and enzymes [25][26][27][28][29][30][31].
Here, the assay was also used to investigate the blocking effect of the specific GGTase-I inhibitor N-[4-[2(R)-amino-3-mercaptopropyl]amino-2-(1-naphthalenyl)benzoyl]-L-leucine methyl ester trifluoroacetate salt (GGTI-298) and the effect of A 2A AR (adenosine A 2A receptor) and A 2B AR (adenosine A 2B receptor) signaling on human GGTase-I and Rap1B interaction. The A 2A AR and A 2B AR are G protein-coupled receptors that are co-expressed on many different cell types, e.g., on cells of the immune system. Both receptors play important roles in inflammation and have recently become major drug targets in immunooncology [32]. A 2B ARs are highly expressed on different cancer cells, promoting cancer cell proliferation and angiogenesis, and A 2B AR antagonists have been suggested for cancer therapy [33][34][35][36][37][38]. Moreover, it has been shown that A 2A AR and A 2B AR are able to formdependent on their expression levels in different cell lines-heterodimers with an altered pharmacology compared to monomers [39,40]. Both receptors activate adenylate cyclase via Gα s -proteins, and the A 2B AR can additionally couple to calcium mobilization, mainly via G q -proteins in different cell lines [41][42][43][44]. The A 2A ARs are activated by low nanomolar adenosine concentrations, whereas A 2B ARs require higher concentrations, which are only reached under hypoxic or inflammatory conditions. Ntantie et al. have identified a signaling pathway by which activation of the A 2B AR with the non-selective adenosine receptor (AR) agonist 5 -N-ethylcarboxamidoadenosine (NECA) causes protein kinase A (PKA) to phosphorylate the polybasic region (PBR) of newly synthesized Rap1B [45]. This inhibits the interaction of Rap1B with the chaperone protein SmgGDS-607 and suppresses Rap1B prenylation (Supplementary Figure S1). That in turn leads to reduced Rap1B trafficking to the plasma membrane and to increased cell-scattering and tumor metastasis [45]. Reduced Rap1B prenylation was detected in different cancer cell lines and tumors [46].
In the present study, we further investigated whether the activation of both Gα scoupled A 2A AR and A 2B AR have an effect on human GGTase-I and Rap1B interaction and thereby potentially modulate the prenylation of Rap1B. Additionally, different Rap1B mutants (C181S, C181G, ∆CQLL) were evaluated to characterize the human GGTase-I and Rap1B interaction on the molecular level.

Development of a NanoBiT Assay to Monitor the Interaction of Human GGTase-I and Rap1B
The NanoBiT system provides a new platform to measure protein-protein interactions in living cells and in real-time. Here, we developed a new NanoBiT assay to monitor the interaction of the human GGTase-type-I and its substrate Rap1B. We used human embryonic kidney cells (HEK293) for our experiments because it was previously shown by qRT-PCR, Western blots, and pharmacologically evaluated, functional cyclic adenosine monophosphate (cAMP) accumulation assays that HEK293 cells express endogenous A 2B ARs and A 2A ARs (Supplementary Figure S5) [41,44,[47][48][49]. The A 2A AR selective agonist CGS-21680 was able to increase cAMP accumulation in HEK293-A 2B cells with an ED 50 -value of 15.2 µM [44]. Moreover, the HEK293 cells utilized for this study were additionally stably transfected with human A 2B AR [43] (Supplementary Figure S6) and co-transfected with combinations of SmBiT and LgBiT plasmid constructs.
An overview of all NanoLuc fusion constructs used in this study is given in Supplementary Figure S2. The N-terminal orientation of the NanoBiT tags on Rap1B was chosen because the C-terminus of the Rap1B protein carries the CAAX motif, which is important for the recognition and interaction with the active site of the GGTase-I-β-subunit. Moreover, in previous studies it was shown that a HA-tag (hemagglutinin) at the N-terminus of Rap1B does not modify its subcellular localization or its ability to be phosphorylated by PKA in vivo [50]. Additionally, the SmBiT tag was used for the human Rap1B wild-type (WT) protein as well as for the Rap1B prenylation-deficient mutants C181G, C181S, and ∆CQLL, considering their relatively small size of~21 kDa. The LgBiT was attached to the N-terminus of the human GGTase-I-β-subunit. For the structurally similar yeast GGTase-I, it was demonstrated that an N-terminal GS dipeptide or a Flag-tag at the β-subunit does not influence the function of the enzyme [51,52]. Furthermore, LgBiT and SmBiT tags were attached to the N-terminus of FTase α-subunit for control experiments. For initial experiments, the SmBiT-FTase α was co-transfected with LgBiT-GGTase-I-β-subunit as a heteromeric positive control, and the combination of human LgBiT-FTase α and SmBiT-FTase α served as a negative control [5]. Additionally, a non-interacting fusion protein (HaloTag ® -SmBiT) was co-transfected in combination with LgBiT-GGTase-I-β-subunit as a second negative control. Moreover, LgBiT and SmBiT tags were attached to the C-terminus of A 2A AR, which were co-transfected for A 2A AR-homodimer control experiments [40,53].
A total of nine fusion proteins were generated and connected with a flexible glycine/ serine linker sequence. Transfected plasmid amounts of 50 ng/well or 100 ng/well were evaluated, and first experiments were conducted with cells transfected solely with SmBiT or LgBiT plasmids, respectively. As expected, none of the mono-transfected cells (50 ng/well) showed any significant luminescence signal over a time period of 60 min (Supplementary Figure S3A,B).

Figure 2.
Effect of the cell-permeable, competitive CAAX (defined by a cysteine residue, two aliphatic residues, and the C-terminal residue) peptidomimetic GGTase-I inhibitor GGTI-298 on the interaction of GGTase-I and Rap1B and on A2AAR-homodimers. HEK293 cells stably expressing the adenosine A2B receptor (A2BAR) were transiently co-transfected as described in Figure 1. From the Rap1B mutant, ΔCQLL 50 ng DNA/well was used. Additionally, for a control of the specificity of GGTI-298, HEK293 cells were transiently transfected with combinations of plasmids encoding LgBiT and SmBiT attached to the C-terminus of the human A2AAR (each 100 ng DNA/well). (A) Structure of the cell-permeable, competitive CAAX peptidomimetic GGTase-I inhibitor GGTI-298. (B) Twenty-four hours after transfection, the cells were pre-treated for 30 min with the selective GGTase-I inhibitor GGTI-298 in different concentrations (10 µM, 1 µM, 0.1 µM, 0.01 µM) or with 1% DMSO at 37 °C in OptiMEM medium. Then the NanoGlo live-cell substrate was added and the association was immediately monitored at 25 °C over a time period of 30 min with a luminescence reading taken every 30 s. (C) After a time period of ~ 400-700 s the maxima of complex formation were reached (dependent on which GTI-298 concentration was used, see (B), and the relative luminescence units (RLU) were plotted. Significant differences were observed between the untreated LgBiT-GGTase-I-β/FTase α/SmBiT-Rap1B WT complex and the GGTI-298 inhibitor (10 µM, 1 µM) treated Lg-BiT-GGTase-I-β/FTase α/SmBiT-Rap1B complex (* p < 0.05, ** p < 0.01). (D) After 24 h the A2AAR transfected HEK293 cells were pre-treated for 30 min with the selective GGTase-I inhibitor GGTI-298 in different concentrations (10 µM, 1 µM, 0.1 µM) or with 1% DMSO at 37 °C in OptiMEM medium. Then the NanoGlo live-cell substrate was added, and the association was immediately monitored at 25 °C over a time period of 50 min, with a luminescence reading taken every 30 s. (E) After a time period of ~ 1005 s, the maxima of A2AAR-homodimer formation were reached, and the relative luminescence units (RLU) were plotted. No significant differences were observed between the untreated A2A-LgBiT/A2A-SmBiT homodimer and the GGTI-298 inhibitor (10 µM, 1 µM, 0.1 µM) treated A2A-LgBiT/A2A-SmBiT homodimer (ns, not significant). We considered the possibility that the decrease in luminescence with the GGTI-298 inhibitor was not the result of a specific competition versus Rap1B but rather a result of inhibitory effects on the reconstituted NanoLuc luciferase or of cytotoxicity. To investigate this possibility, we performed the same GGTI-298 inhibitor treatment with the well described A 2A AR-homodimer [53]. As before, HEK293 cells were co-transfected with A 2A -LgBiT/A 2A -SmBiT and pre-treated for 30 min with different concentrations of GGTI-298 inhibitor (10 µM, 1 µM, 0.1 µM) or with 1% DMSO (Figure 2D,E). The GGTI-298 inhibitor treated A 2A AR-homodimer showed similar kinetic properties as the untreated (1% DMSO) A 2A AR-homodimer ( Figure 2D). No significant differences of the GGTI-298 inhibitor treated A 2A AR-homodimer (10 µM, 1 µM, 0.1 µM) compared to the untreated A 2A AR-homodimer were observed (ns, not significant) ( Figure 2E). Thus, we conclude that the observed GGTI-298 inhibitor effect was specific and blocked the interaction between GGTase-I and its substrate Rap1B.

Interaction of GGTase-I with Prenylation-Deficient Rap1B Mutants C181G, C181S, and ∆CQLL
Having established that the new NanoBiT assay is suitable to measure the interaction of human GGTase-I and Rap1B, we were interested in studying the interaction of human GGTase-I with prenylation-deficient Rap1B mutants. Rap1B contains a C-terminal hypervariable region, which consists of a polybasic region (PBR) and the CAAX motif. The CAAX motif contains the cysteine that is the site of prenylation and is important for the coordination of the Zn 2+ ion in the active site of the enzyme [56]. Crystal structures are known of several complexes formed by the rat GGTase-I and different substrates [3,7], but no models or crystal structures are published for human GGTase-I and Rap1B complexes. We therefore further evaluated the effect of prenylation-deficient Rap1B mutants on the direct binding of human GGTase-I, which has not yet been characterized. For that, the cysteine in the CAAX motif was replaced by serine or glycine to generate the non-prenylateable Rap1B mutants C181S and C181G. Moreover, a complete CAAX motif deletion mutant, ∆CQLL, was generated.

Effect of A2BAR Agonists and Antagonists on the Interaction of GGTase-I and Rap1B WT
Earlier studies showed that the treatment of cells with compounds that elevate intracellular cAMP (cyclic adenosine monophosphate) levels results in PKA-dependent Rap1B phosphorylation [50,57]. For example, forskolin treatment resulted in an increase of Rap1B phosphorylation and in an increase in the bound GTP/GDP ratio compared to the basal ratio [50]. In following studies, Wilson et al. showed that the activation of Gαs-coupled β-adrenergic receptors phosphorylates Rap1B and inhibits its prenylation. This led to disturbed membrane localization and promoted the metastatic phenotype in breast cancer cells [58]. The same group showed that activation of A2BAR with the non-selective AR agonist NECA phosphorylates newly synthesized Rap1B and inhibits

Effect of A 2B AR Agonists and Antagonists on the Interaction of GGTase-I and Rap1B WT
Earlier studies showed that the treatment of cells with compounds that elevate intracellular cAMP (cyclic adenosine monophosphate) levels results in PKA-dependent Rap1B phosphorylation [50,57]. For example, forskolin treatment resulted in an increase of Rap1B phosphorylation and in an increase in the bound GTP/GDP ratio compared to the basal ratio [50]. In following studies, Wilson et al. showed that the activation of Gα s -coupled β-adrenergic receptors phosphorylates Rap1B and inhibits its prenylation. This led to disturbed membrane localization and promoted the metastatic phenotype in breast cancer cells [58]. The same group showed that activation of A 2B AR with the non-selective AR agonist NECA phosphorylates newly synthesized Rap1B and inhibits its interaction with the chaperone protein SmgGDS-607, leading to a decreased Rap1B prenylation and signaling at the cell membrane. This in turn led to reduced cell-cell contact and promoted tumor metastasis [45,46,59]. However, the direct binding of Rap1B to human GGTase-I was never characterized in detail nor was the pharmacological effect of specific and non-specific A 2A AR and A 2B AR agonists and antagonists on this protein-protein interaction.

Discussion
GGTase-I represents an important drug target due to its involvement in different kinds of pathological processes [2,14,15]. In particular, GGTase-I inhibitors have attracted interest regarding their application in different types of cancer [18,22,55,63,64]. In this study, we developed a NanoBiT assay to measure the interaction of Rap1B and human GGTase-I, which is expected to become a pharmacological tool for the evaluation of novel GGTI inhibitors. The assay depends upon the expression of proteins tagged with the LgBiT and SmBiT reporters. Despite observing a robust luminescence signal for LgBiT-GGTase-I-β/SmBiT-FTase α and LgBiT-GGTase-I/FTase α/SmBiT-Rap1B, we were not able to quantify the expression of all tagged proteins by immunoblotting. However, as shown in the Western blots, all DNA constructs were clearly expressed in HEK293 cells. Especially, the negative control LgBiT-FTase α/SmBiT-FTase α was clearly expressed, although it elicited only a moderate luminescence signal, suggesting that the assay is specific and not a result of the overexpressed NanoBiT tags associating with each other. Secondly, the LgBiT-GGTase-I-β co-transfected with HaloTag-SmBiT ® , which was not expected to interact, showed only a low formation of functional NanoLuc. Moreover, the luminescence signal of LgBiT-GGTase-I-β/FTase α/SmBiT-Rap1B was reduced by the cell-permeable, competitive, and selective peptidomimetic CAAX motif inhibitor GGTI-298. GGTI-298 was proven to inhibit the processing of the structural homolog geranylgeranylated Rap1A with an IC 50 -value of 3 µM [52,53]. Toxic effects on HEK293 cells or inhibition of the functional NanoLuc can also be excluded, since GGTI-298 had no effect on the formation of A 2A AR-homodimers. Furthermore, the LgBiT and SmBiT subunits only weakly associate with each other, and their intrinsic affinity (K D = 190 µM) is outside of the ranges typical for protein interactions [23]. Thus, we conclude that the assay offers a new pharmacological tool to perform close to real-time measurements and to monitor reversible protein-protein interactions. In this format, it depends on transient transfection and could also be used for different cancer cell lines. However, it would be useful to establish stable cell lines with constant expression ratios for better comparability of IC 50 -values for tested GGTase-I inhibitors.
Despite the number of solved crystal structures, many aspects of enzyme substrate specificity and allosteric modulation of enzyme activity still remain unclear. So far, there is no crystal structure of the human GGTase-I in complex with its substrate Rap1B available, and understanding the key features of substrate specificity will contribute to optimization of anti-cancer drugs. In order to obtain further molecular insights into the interaction between human GGTase-I and its substrate Rap1B, we substituted the cysteine residue (C) in the CQLL motif for serine (S) or glycine (G). The prenylation-deficient mutants Rap1B GQLL and SQLL were obtained. Moreover, a complete CAAX motif deletion mutant, ∆CQLL, was generated. HEK293 cells stably expressing the A 2B AR were transiently co-transfected either with LgBiT-GGTase-I-β-subunit/FTase α-subunit/SmBiT-Rap1B WT or with LgBiT-GGTase-I-β-subunit/FTase α-subunit in combination with the SmBiT-Rap1B prenylationdeficient mutants C181G, C181S, and ∆CQLL. Interestingly, both Rap1B mutants C181G and C181S were still able to fully interact with GGTase-I, whereas the mutant missing the CAAX motif showed a decreased interaction with the enzyme.
In general, the prenylation reaction begins when GGTase-I binds its isoprenoid substrate geranylgeranylpyrophosphate (GGPP) and forms a binary enzyme diphosphate complex [56]. GGPP is a key intermediate in the isoprenoid biosynthesis pathway, and its concentration, e.g., in different human pancreatic cancer cell lines, varies from 1.96 nmol/10 6 cells to 9.96 nmol/10 6 cells [65]. After isoprenoid binding, the CAAX protein binds and its cysteine residue coordinates to a Zn 2+ ion (as a thiolate) in the active site of the enzyme, which is necessary for its catalytic activity [3,7,56,66]. Additionally, the Zn 2+ ion is coordinated by three conserved residues, D269β, C271β, and H321β, in GGTase-I [66]. Further, the cysteine thiolate forms a covalent thioether linkage to the isoprenoid at the C-1 position. Upon binding of new GGPP, the prenylated protein product is released from the enzyme, which is the rate-limiting step [56]. Regarding the enzyme's mechanism, our designed prenylation-deficient Rap1B mutants C181G and C181S were not able to form the bonded Zn 2+ -to-cysteine-thiolate interaction with GGTase-I, but we were still able to measure an interaction. It appears that the remaining three residues in the CAAX motif QLL (glutamine, leucine, leucine) are sufficient to stabilize the interaction with the enzyme. Modeling studies revealed that polar or charged a 1 residues in the Ca 1 a 2 X motif could form direct or water-mediated hydrogen bonds with rat GGTase-I and thereby enhance binding affinity [7]. The polar glutamine (Q) in the a 1 position of Rap1B might react in a similar way with human GGTase-I. Moreover, in rat GGTase-I the a 2 (isoleucine) side chain makes extensive hydrophobic contacts with Phe53β and Leu320β, as well as the fourth isoprene unit [3]. The same could be the case for the a 2 leucine residue in Rap1B. The X residue in the Ca 1 a 2 X motif is the primary determinant specifying whether the peptide is a substrate for FTase or for GGTase-I. GGTase-I harbors only one binding pocket for X residues, which is shaped to accommodate hydrophobic residues and discriminates against polar or charged residues [7]. It is obvious that the X leucine residue of Rap1B fits inside the hydrophobic binding pocket and contributes to the binding affinity to GGTase-I. In Western blots, our generated Rap1B ∆CQLL mutant exhibited a~5-fold lower expression level compared to the Rap1B WT protein and the mutants C181G and C181S. Additionally, the interaction with GGTase-I was more than 20-fold reduced compared to Rap WT interaction with GGTase-I, but the generated luminescence of ∆CQLL/GGTase-I was still higher (800-1000 RLU, Figures 2C and 3B) when compared to the clearly expressed negative control LgBiT-FTase/SmBiT-FTase alpha (200-300 RLU, Figure 1B,D).
Thus, the diminished interaction of ∆CQLL mutant with GGTase-I might not only be a result of lacking expression but could also be a result of a decreased binding affinity to the enzyme. This point would be supported by the presented observation that all CAAX motif residues are involved in the binding affinity to the enzyme. On the other hand, we could also not exclude that other amino acid residues of Rap1B are involved in the binding to GGTase-I.
In order to obtain further insights into the role of A 2A AR and A 2B AR signaling on the GGTase-I and Rap1B interaction, we investigated whether selected specific and non-specific A 2A AR and A 2B AR agonists and antagonists exhibit a pharmacological effect. Our results demonstrated that the non-selective agonists adenosine (100 µM), the metabolically stable adenosine analog derivative NECA (10 µM), and the selective A 2A AR agonist CGS-21680 (10 µM) showed a significant increase on human GGTase-I and Rap1B interaction. Additionally, forskolin (10 µM), which directly activates adenylate cyclase, showed similar effects when compared to adenosine (100 µM) and NECA (10 µM). In cAMP experiments at HEK293-A 2B cells, NECA (10 µM) showed a higher cAMP effect when compared to adenosine (100 µM) and forskolin (10 µM) (Supplementary Figure S5), but cAMP assays might not correlate with a much further downstream signal like GGTase-I/Rap1B interaction. In contrast, the selective A 2B AR partial agonist BAY60-6583 (10 µM) was not able to increase the interaction in the utilized HEK-A 2B cell line. The selective A 2B AR antagonist PSB-603 (1 µM) was able to reduce the interaction stimulated by 10 µM of NECA, but the difference was not significant. It is likely that the residual response elicited by 10 µM NECA is due to an endogenous A 2A AR population in HEK293 cells [41,49] since the specific A 2A AR ago-nist CGS-21680 (10 µM) also produced an increase in luminescence. The selective A 2B AR antagonist PSB-603 (1 µM) or the selective A 2A AR antagonist MSX-2 (1 µM) alone had no significant effect on the interaction, suggesting a relatively low endogenous adenosine concentration in HEK293 cells. The A 2A AR specific antagonist MSX-2 (1 µM) showed only a moderate but not significant decrease in the interaction elicited by the non-selective agonist NECA (10 µM) or the specific adenosine A 2A AR agonist CGS-21680 (10 µM), suggesting a relatively low A 2A AR population in HEK293 cells, which was previously shown by qRT-PCR [41].
Our results and previous studies demonstrate that A 2A AR forms homodimers or higher order oligomers, which may cause allosteric modulation and cooperativity [53,67]. For example, binding of the first ligand induces a conformational change in the A 2A -D 2 (dopamine D 2 receptor) heterotetramer and will then reduce the affinity of the second ligand [68]. Another explanation would involve the presence of different A 2A AR receptor conformations [69], which probably bind agonists and antagonists with different affinity. Based on these results, we speculate that regarding the observed pharmacology, different portions of A 2B AR/A 2A AR monomer/homodimer/oligomer populations might be involved in the increased interaction of GGTase-I with Rap1B. Alternatively, the relatively high concentration of 10 µM CGS-21680 might additionally activate A 2B ARs in HEK293 cells, which could explain that the specific A 2A AR specific antagonist MSX-2 (1 µM) could not fully reduce the luminescence signal elicited by CGS-21680.
Regarding the prenylation and the function of Rap1B, we can assume that an increase in the interaction of human GGTase-I with Rap1B might at least modulate the prenylation and localization of Rap1B. In earlier studies, Ntantie et al. demonstrated that A 2B AR activation with NECA increased total Rap1B protein but delayed its prenylation and localization in the plasma membrane [45,46]. Additionally, it was shown that NECA induced phosphorylation of S179 and S180 in the PBR region of Rap1B and inhibited the binding to the chaperone protein Smg-GDS-607, which was suggested to bind nonprenylated small GTPases and to assist in their prenylation [46]. On the other hand, Garcia-Torres et al. reported a dual role for Smg-GDS-607 activating and inhibiting farnesylation of small GTPases [70]. For example, they discovered that SmgGDS-607 increased the rate of farnesylation of H-Ras by enhancing product release from FTase [70]. One explanation for the increased GGTaseI/Rap1B formation induced by A 2A AR/A 2B AR signaling in the present study might be that the decreased binding of Smg-GDS-607 to Rap1B led to delayed product release of prenylated Rap1B. This could result in increasing complex formation and thereby a lower accumulation of prenylated Rap1B at the cell membrane, which is consistent with the studies of Ntantie et al. [45].
Finally, we can conclude that both A 2A AR and A 2B AR activation contributes to an increased interaction of GGTase-I with Rap1B. A greater understanding of the molecular mechanism of cancer development will lead to the identification of novel therapeutics and better treatments. Regarding the model that increased Rap1B geranylgeranylation and thereby increased cell adhesion decreases metastasis, A 2A AR and A 2B AR antagonists are promising candidates for therapeutic intervention in different types of cancer that overexpress Rap1B.

Expression Vectors and Molecular Cloning
The human Rap1B WT and mutants C181G, C181S, ∆CQLL, the human GGTase-I-βsubunit, and the human FTase α-subunit coding regions were subcloned into the N-terminal Flexi vectors pFN217k LgBiT CMV-Hyg Flexi, pFN218k SmBiT CMV-Blast Flexi containing SgfI and PmeI sites. The following primers were used: The following PCR program was applied: 30 s at 98 • C and 30 cycles consisting of 10 s at 98 • C, 30 s at 60-63 • C, and 2 min at 72 • C, followed by a final elongation step of 2 min at 72 • C. The PCR products were purified and digested with SgfI and PmeI. The N-terminal Flexi vectors pFN217k LgBiT CMV-Hyg Flexi, pFN218k SmBiT CMV-Blast Flexi were cut with SgfI and PmeI and, after ligation, N-terminal fusion proteins were created, allowing translational readthrough of the SgfI side, which encodes the peptide sequence Ala-Ile-Ala. The correct assembly of the genes was verified by sequencing (Seqlab, Göttingen, Germany) using the following primers: The cDNA of the A 2A AR was subcloned with SgfI and PmeI without a stop codon into the C-terminal Flexi vectors pFC219k LgBiT and pFC220k SmBiT, which were cut with SgfI and EcoICRI. The following primers were designed: The PCR was conducted as described above. C-terminal fusion proteins were created by fusing the blunt-cut PmeI end of the protein coding region with the blunt end gener-ated by EcoICRI. The correct assembly of the genes was verified by sequencing (Seqlab, Göttingen, Germany) using the following primers: The following fusion proteins were received and tested in different plasmid amounts in the NanoBiT assay: SmBiT-Rap1B WT, SmBiT-Rap1B C181G, SmBiT-Rap1B C181S, SmBiT-Rap1B ∆CQLL, LgBiT-GGTase-I-β, LgBiT-FTase α, SmBiT-FTase α, A 2A -LgBiT, A 2A -SmBiT. For the expression of the untagged human FTase α-subunit, the cDNA was subsequently subcloned into the vector pcDNA3.1 (−). The coding region of human FTase α-subunit was amplified via PCR using Phusion-High-Fidelity-DNA-Polymerase (NEB, Ipswitch, USA). The following primers were designed: The PCR program was conducted as follows: 3 min at 98 • C and 30 cycles consisting of 30 s at 98 • C, 45 s at 58 • C, and 2 min at 72 • C, followed by a final elongation step of 10 min at 72 • C. The PCR product was purified and digested with EcoRI and BamHI and finally cloned into the vector pcDNA3.1 (−). The plasmids were transformed into competent Escherichia coli DH5α and single clones were isolated and sequenced (Seqlab, Göttingen, Germany).

Cell Culture
HEK293 stably expressing the human A 2B AR were maintained in DMEM supplemented with 4.5 g/L D-glucose, 10% heat inactivated FCS, 100 U/mL of penicillin, 100 µg/mL streptomycin, and 800 µg/mL G418. The cells were cultured in an incubator with an atmosphere containing 5% CO 2 at 37 • C and were directly transfected in white 96-well plate format.

Transient Transfection of HEK293 Cells Stably Expressing the Human A 2B AR
Fifty thousand cells/well were seeded, resulting in~80% confluency at the time of transfection. Two hours before the transfection, the cell culture medium was changed to 100 µL of OptiMEM medium. For duplicates, 9 µL of OptiMEM medium without supplements was mixed to 1 µL of Lipofectamine 3000 reagent. Fifty to two hundred nanograms of plasmid constructs were mixed with OptiMEM medium and 1 µL of P3000 reagent to reach a final volume of 10 µL. The lipofectamine suspension was added and mixed to the DNA solution, incubated for 15 min at rt, and 10 µL of the transfection mix was finally added to the wells. After 20 h the cell culture medium was changed to 100 µL of fresh OptiMEM medium and further incubated for 4 h at 37 • C before performing the experiments.

NanoBiT Assays
For a first orientation screen, cells were transfected with 50 ng/well of each LgBiT and SmBiT plasmid alone or in combination with the investigated partner protein. The combinations LgBiT-FTase α/SmBiT-FTase α and LgBiT-GGTase-I-β/HaloTag-SmBiT ® were used as negative controls and the combination LgBiT-GGTase-I-β/SmBiT-FTase α as a positive control. The NanoGlo live-cell reagent containing the cell-permeable furimazine substrate was dissolved 1:50 in NanoGlo dilution buffer or in OptiMEM medium, and 25 µL of the solution was added to each well. The luminescence was immediately monitored at 25 • C over a time period of up to 60 min with a luminescence reading taken every 30 s using a Tecan microplate reader (Tecan group Ltd., Männedorf, Switzerland).

Treatment with Adenosine A 2A -and A 2B -Receptor Agonists and Antagonists
Cells were seeded and transfected as described above. After 24 h, the cells were pretreated for 30 min with the selective A 2A AR antagonist MSX-2 (1 µM), the selective A 2B AR antagonist PSB-603 (1 µM), or vehicle (1% DMSO) at 37 • C in OptiMEM medium. Then the cells were stimulated for 15 min with the non-selective adenosine receptor agonist NECA (10 µM), the selective adenosine A 2A receptor agonist CGS-21680 (10 µM), the selective adenosine A 2B receptor partial agonist BAY-60-6583 (10 µM), or with the endogenous adenosine receptor agonist adenosine (100 µM). The final DMSO concentration did not exceed 2%. Then 25 µL of the NanoGlo live-cell substrate was added and the association was immediately monitored at 25 • C over a time period of 42 min, with a luminescence reading taken every 30 s.

Preparation of Cytosolic Extract and Whole Cell Lysate of Transfected HEK-A 2B Cells
The transfected cells were harvested and disrupted by repeated aspiration at 4 • C through a 500 µL and afterwards a 10 µL pipette tip in 250 µL of RIPA-buffer containing 150 mM NaCl, 1% NP-40, 0.1% SDS, 0.5% sodium deoxycholate, 25 mM TRIS, pH 7.4, and protease inhibitor. The cytosolic extract was obtained by centrifugation at 20,000× g for 2 h at 4 • C. For the A 2B AR Western blots, the whole cell lysate was used, or it was purified by centrifugation at 1000× g for 3 min. The protein concentration of the supernatant fraction was determined, and different protein amounts (20-80 µg) were used for the Western blots.

Bradford Protein Determination
The protein concentration of the cytosolic extract was determined by the Bradford assay. A bovine serum albumin (BSA) standard in the range of 1 to 15 µg/mL was diluted in H 2 O to a final volume of 600 µL. Additionally, 1 µL of the samples were diluted in 599 µL of H 2 O. Then, 400 µL of Bradford reagent (Bio-Rad, Berkeley, USA) was added to the different standard concentrations, and the diluted samples and the absorbance were measured at 595 nm in a standard photometer.
After semidry blotting for 30 min and 25 V, the nitrocellulose membrane was blocked in blocking solution for 90 min at rt. Then, the membrane was incubated with the specific first antibodies, which were diluted in LI-Cor antibody solution, 0.1% Tween 20, for 60 min at rt. Afterward, the membrane was washed 4 times for 10 min with Tris-buffered saline (TBS), 0.1% Tween 20, and then incubated with the second infrared antibodies for 60 min at rt and shaking in the dark. Finally, the membrane was washed 3 times for 10 min with TBS, 0.1% Tween 20, and 5 min with TBS, dried in the dark, and visualized on an Odyssey imager (LI-Cor, Lincoln, USA). ImageJ was used for quantification of Western blot signals. At least 3 independent experiments were performed in single values for quantification, and data were presented as mean ± SEM.

cAMP Accumulation Experiments
The assays were conducted at the University of Bonn, Pharmaceutical Institute, Pharmaceutical and Medicinal Chemistry. cAMP accumulation experiments at HEK-A 2B cells were performed according to De Filippo et al. [72] with the following modifications. HEK-A 2B cells were detached from a confluent 175 cm 2 flask and centrifuged at 200× g at 4 • C for 5 min. Then the supernatant was removed, and the cell pellet was resuspended in Hanks' balanced salt solution (HBSS) buffer, pH 7.4 with 1 U/mL adenosine deaminase (ADA). When adenosine was used, ADA was omitted. Two hundred microliters of the cell solution containing~200,000 cells were transferred into 24-well plates. After an incubation time of 3 h at 37 • C, 5% CO2, 25 µL of the phosphodiesterase inhibitor Ro20-1724 (40 µM) dissolved in 100% HBSS buffer was added to each well. After an incubation time of 10 min, 12.5 µL of the antagonists were added and incubated for 30 min at 37 • C. Then, 12.5 µL of the agonists were added and incubated for 15 min at 37 • C. The final DMSO concentration did not exceed 1.4%. cAMP accumulation was stopped by the addition of 250 µL of hot lysis buffer (90 • C, 8 mM EDTA, 0.02% Triton X-100, pH 7.3). The 24-well plates were put on ice, and each well was subsequently homogenized. Fifty microliters of each cell lysate were transferred in 2.5 mL tubes. Radioligand competition binding experiments with [ 3 H]cAMP to determine the cAMP amount in the lysates were conducted according to De Filippo et al. [72].

Statistical Analysis
For all statistically analyzed studies, at least 3 independent experiments were performed, each in duplicates unless otherwise stated. Data were expressed as mean ± SEM and differences with p < 0.05 were considered significant. One-way ANOVA followed by Dunnett's post hoc test was used to evaluate the differences among three or more groups. In order to determine the differences between two groups, an unpaired parametric Welch's t-test was employed. GraphPad prism 8.0 was used for data analysis and statistics. Inkscape was employed for graphical representations.

Conclusions
The new NanoBiT assay might become a pharmacological tool for the evaluation of novel GGTase-I inhibitors. Moreover, both A 2A AR and A 2B AR activation contribute to an increased interaction of GGTase-I with Rap1B. In both cases, it has to be considered that prenylation of Rap1B is not always required for its oncogenic activity, and additionally, Rap1B prenylation and function appear to be cell-type dependent. Department of Molecular Medicine, School of Medicine, Cleveland Clinic. The adenosine A 2Aand A 2B -receptor agonists and antagonists, the plasmid pcDNA3.1(−), the cDNA of A 2A AR, the A 2B AR antibody for the Western blot, and the HEK293 cell line stably expressing the A 2B AR were kindly received from Christa E. Müller, Pharmaceutical and Medicinal Chemistry, University of Bonn. Moreover, we thank Christa E. Müller for the possibility to perform the cAMP accumulation experiment in her research group. The specific monoclonal LgBiT/NanoLuc antibody was a gift by Promega (Madison, WI, USA).