Enhanced A3 adenosine receptor selectivity of multivalent nucleoside-dendrimer conjugates

Background An approach to use multivalent dendrimer carriers for delivery of nucleoside signaling molecules to their cell surface G protein-coupled receptors (GPCRs) was recently introduced. Results A known adenosine receptor (AR) agonist was conjugated to polyamidoamine (PAMAM) dendrimer carriers for delivery of the intact covalent conjugate to on the cell surface. Depending on the linking moiety, multivalent conjugates of the N6-chain elongated functionalized congener ADAC (N6-[4-[[[4-[[[(2-aminoethyl)amino]carbonyl]methyl]anilino]carbonyl]methyl]phenyl]-adenosine) achieved unanticipated high selectivity in binding to the cytoprotective human A3 AR, a class A GPCR. The key to this selectivity of > 100-fold in both radioreceptor binding (Ki app = 2.4 nM) and functional assays (EC50 = 1.6 nM in inhibition of adenylate cyclase) was maintaining a free amino group (secondary) in an amide-linked chain. Attachment of neutral amide-linked chains or thiourea-containing chains preserved the moderate affinity and efficacy at the A1 AR subtype, but there was no selectivity for the A3 AR. Since residual amino groups on dendrimers are associated with cytotoxicity, the unreacted terminal positions of this A3 AR-selective G2.5 dendrimer were present as carboxylate groups, which had the further benefit of increasing water-solubility. The A3 AR selective G2.5 dendrimer was also visualized binding the membrane of cells expressing the A3 receptor but did not bind cells that did not express the receptor. Conclusion This is the first example showing that it is feasible to modulate and even enhance the pharmacological profile of a ligand of a GPCR based on conjugation to a nanocarrier and the precise structure of the linking group, which was designed to interact with distal extracellular regions of the 7 transmembrane-spanning receptor. This ligand tool can now be used in pharmacological models of tissue rescue from ischemia and to probe the existence of A3 AR dimers.


Background
Dendrimers bearing multiple ligands may have increased avidity to a receptor compared to the monovalent ligand, particularly if the ligand has a weak affinity for the receptor [1]. While this phenomenon has only been loosely demonstrated with PAMAM dendrimers, it is well established that multivalent oligo-and poly-saccharides, including PAMAM glycodendrimers, show some enhancement in binding compared to the monovalent saccharide, which is known as the cluster glycoside effect [2]. Dendrimer-ligand complexes have also been used as imaging agents [3] and for gene delivery [1]. Recently, we also attached CGS21680, an A 2A adenosine receptor (AR) agonist, to G3 PAMAM dendrimers, providing the first example of a GPCR ligand to be conjugated covalently to a dendrimer while retaining its biological activity [4].
The ARs are GPCRs that have a generally cytoprotective role and their ligands are of increasing therapeutic interest. The A 1 AR and A 3 AR inhibit adenylyl cylase through the coupling of the G i protein and are also involved in activating phospholipase C and potassium channels [5]. The A 1 AR is highly expressed in the brain, spinal cord, eye, and atria while intermediate expression is found in the liver, kidney, and adipose tissue [6]. The A 3 AR is upregulated in peripheral blood mononuclear cells of patients with rheumatoid arthritis as well as in several breast, colon and pancreatic carcinoma tissues [7], but more studies are needed to learn about the expression of this protein in normal patients. Preconditioning of cardiomyocytes with either A 1 or A 3 AR agonists protects against myocardial ischemia. This cardioprotection occurs through extracellular signal-regulated kinase (ERK) signaling and activation of the mitochondrial K + -ATP channels [5]. A 1 AR agonists also inhibit lipolysis [6] and may act as anti-epileptic agents [8], while A 3 AR agonists may protect against lung injury and cancer [9,10].
The AR ligands chosen for conjugation to both G2.5 or G3 PAMAM dendrimers in the present study are the A 1 AR agonist N 6 - [4-[[[4-[[[(2-aminoethyl)amino]carbonyl]methyl]anilino]-carbonyl]methyl]phenyl]adenosine (ADAC, 1) and related functionalized congeners ( Figure  1). Functionalized congeners are designed by adding a chain substituent to a pharmacophore in a strategic, permissive location so that conjugation to other large mole-Synthesis of novel functionalized congener monomers related to ADAC Figure 1 Synthesis of novel functionalized congener monomers related to ADAC. cules is possible [11]. Ideally, the linker is modified to enhance the interaction of the pharmacophore with the receptor. This approach has been used to study A 1 [12], A 2A [13], and A 3 ARs [11]. ADAC is a highly selective A 1 AR agonist at the rat ARs and also displays some selectivity towards the human A 1 AR and human A 3 AR in comparison to the human A 2A AR [5,14]. ADAC protects against neuronal damage and mortality after either acute or chronic administration prior to a ten-min bilateral cerebrovascular occlusion in gerbils. Significantly higher doses of other A 1 AR agonists are needed to produce an equivalent effect [15]. ADAC also provides neuronal protection when given up to twelve hours post-ischemia [16]. Each of the dendrimer nucleoside conjugates also contained a fluorescent moiety for in vitro and in vivo localization.

Results
This study was designed to probe the feasibility of modulating the potency and selectivity of nucleoside agonist ligands of ARs based on conjugation to a PAMAM nanocarrier.

Synthesis of ADAC-Related Functionalized Congeners and Dendrimer Conjugates
ADAC, an amine-derivatized nucleoside that potently binds to and activates the A 1 AR, was coupled covalently to the surface of polyamidoamine (PAMAM) dendrimers of generation 2.5 (G2.5). Two other linker moieties were applied for comparison: one containing a secondary amine and another containing an extended arylthiourea group, which was attached to a G3 PAMAM dendrimer as shown in Figure 1B. Two nucleoside intermediates related to ADAC, 4 and 7, which had chains that could be coupled to PAMAM dendrimers, were synthesized as shown in Figure 1. 3-(p-Aminophenyl)propanoic acid 2 was converted to 3-(p-isothiocyanatophenyl)propanoic acid 3 by addition of thiophosgene in aqueous medium. The isothiocyanate group of 3 was then conjugated to the terminal amino group of ADAC to form a thiourea linkage in 4, which had a terminal carboxyl group that could be coupled to the amino group of the G3 PAMAM dendrimer. To synthesize the diamino derivative 7, diethylenetriamine 6 was heated with methyl ester 5, which was similar to a previous method [17]. This product has a terminal primary amine group that was coupled to the G2.5 PAMAM dendrimer, with preference for its acylation over the secondary amine.
Another goal was to compare 2.5 PAMAM -conjugates of A 1 AR agonists with G3 PAMAM -conjugates of similar agonists. However, in order to attach AF488 to the carboxylic G2.5 dendrimer, it was necessary to synthesize a new Synthesis of compounds 12 and 13 -derivatives of G3 PAMAM dendrimer AF488 derivative having a terminal primary amine. Initial attempts were made to couple ethylenediamine to 10 using triethylamine in DMF or DMSO, but the AF488 did not appear stable under these conditions. However, a variation of the method using ethylenediamine in 0.1 M NaB 4 O 7 , pH 8.5, was successful. After HPLC purification and lyophilization, compound 14 was isolated in 93% yield.
The conjugates were purified using size exclusion chromatography and characterized using NMR and electrospray ionization (ESI) mass spectrometry (MS) (see Additional file 1). The parent G3 dendrimer matched its theoretical weight, but the parent G2.5 dendrimer appeared to be missing 2 propionate groups in the largest peak in the mass spectrum, as shown in Figure S1 (Additional file 1). Due to the excessive amount of sodium, the G2.5 spectrum was significantly more fragmented than the G3 spectrum. These spectra appear to be one of the first examples of using ESI MS rather than MALDI MS to obtain data on the PAMAM dendrimers.
After removal of the monomers by dialysis, NMR showed that approximately three and eight molecules of 4 were attached per dendrimer, on average, in derivatives 12 and 13, respectively. Interestingly, while the mass spectrum of 13 was very close to the theoretical mass, the mass spectrum of both compounds was very fragmented as shown in Figure S2 (Additional file 1). The largest peak of 12 appeared to differ from the theoretical mass by approximately 1.8%, possibly due to the molecule breaking down in the mass spectrometer. The majority of the amino groups on the dendrimer appeared to be acylated, which has previously been shown to significantly decrease toxicity [21].
Synthesis of compounds 16 and 17 -derivatives of G2.5 PAMAM dendrimer  NMR indicated that on average there were approximately three nucleoside ligand moieties attached to each dendrimer in purified derivatives 16 and 17. The mass spectrum of 16 was different from the theoretical mass by approximately 1 nucleoside moiety, possibly due to the compound decomposing in the mass spectrometer. The mass spectrum of 17 was too fragmented to be useful as shown in Figure S3 (Additional file 1). The largest peak in the spectra was smaller than 15, the dendrimer with only the AF488 moiety attached. However, smaller peaks in the spectrum were closer to the theoretical weight. Unlike the spectra for 13 and 14, there was significantly more sodium in these spectra, which may have caused the difficulties in obtaining these spectra. Also, the parent G2.5 dendrimer had more fragmentation than the parent G3 dendrimer. The difficulty in obtaining mass spectral data for dendrimers is a known phenomenon [21].

Pharmacological Characterization of ADAC-Related Functionalized Congeners
The human AR binding affinity of these functionalized congeners was measured prior to attachment to the dendrimers (Table 1) [22]. Both 4 and 7, the new ADAC derivatives, had slightly lower affinity than ADAC itself at the A 1 AR, with K i app values of 30 nM and 43 nM, respectively. While 4 retained selectivity similar to ADAC towards the A 1 AR in comparison to A 2A AR, 7 was slightly less selective. Interestingly, 7 had a similar affinity for the A 3 AR as ADAC, while 4 had significantly lower affinity at this receptor. [ 35 S]GTPγS binding, a functional assay for G i protein activation [23], was completed in membranes expressing the A 1 AR ( Table 2). Compound 7, with an EC 50 value of 63 nM in activation of GTPγS binding via the A 1 AR, was 20-fold more potent than 4 and slightly more potent than ADAC. In an assay measuring the inhibition of the production of cAMP (Table 2), compound 7 was also the most potent monomer at the A 1 and A 3 ARs. Compound 7 was 3 -7 fold more potent in adenylyl cyclase assays at these two ARs than either ADAC or 4, which were nearly equipotent at both the A 1 and A 3 ARs. All compounds were shown to be full agonists at the A 1 and A 3 ARs in both assays.

Pharmacological Characterization of Nuceloside-Dendrimer (G3) Conjugates
In the radioligand binding studies, the G3 dendrimer-ligand conjugates 12 and 13 had a comparable affinity to the free monomer 4 at the A 1 AR, but maintained a lower degree of A 1 selectivity compared to the A 2A AR. However, both conjugates had a higher affinity at the A 3 AR than the free monomer. The control dendrimer, 11, which contained AF488 and multiple acetamide moieties but not the nucleoside ligand, showed no binding at the A 1 AR. At the A 2A and A 3 ARs, weak binding inhibition was evident at 10 μM, which might be a result of association of the radioligand with the dendrimer conjugate at high concentrations. This phenomenon was seen in the A 2A AR agonist-dendrimer conjugates as well [24]. The control dendrimer 11 also showed slight activity at 10 μM in the stimulation of [ 35 S]GTPγS binding. However, at 10 μM, 11 was unable to significantly inhibit cAMP production at the A 1 AR or the A 3 AR. In an assay measuring [ 35 S]GTYγS binding at the A 1 AR, the G3 dendrimer ligand conjugates 12 and 13 had EC 50 values that were at least 4 fold lower than the free monomer. Both of the dendrimer ligand conjugates 12 and 13 were almost 5 -10 fold more potent at the A 1 AR than the free monomer in an assay measuring inhibition of cAMP production. Therefore, conjugating the nucleoside 4 to the dendrimer improved the potency in activation of the A 1 AR even though the affinity was similar in the radioligand binding.

Pharmacological Characterization of Nuceloside-Dendrimer (G2.5) Conjugates
Radioligand binding was completed for each of the G2.5 dendrimer conjugates. Compound 17 showed a 2.4 nM affinity for the A 3 AR while compound 16 had a 14 nM affinity for this receptor. Interestingly, 16 displayed at least a 10-fold selectivity, and compound 17 displayed over a 100 fold selectivity for the A 3 AR in comparison to the A 1 and A 2A ARs (Figure 4). Compound 17 was also 100 fold selective for the A 3 AR in comparison to A 1 AR in assays of adenylate cylase inhibition with an EC 50 value of 1.6 nM at the A 3 AR ( Figure 5). However, in this assay, 16 was only 8 fold more potent at the A 3 AR than at the A 1 AR. In GTPγS studies, 16 was 15 fold less potent at the A 1 AR than in an assay measuring the suppression of cAMP production; however, 17 had similar potency at both A 1 AR functional assays, and both compounds were full agonists in both assays. In the GTPγS study, DPCPX, an A 1 antagonist, was able to fully inhibit the binding of [ 35 S]GTPγS when incubated with 17 ( Figure 6), showing that the binding is due to the specific interaction of 17 with the A 1 receptor. The control dendrimer 15 showed no binding or activity in either cAMP or GTPγS assays of A 1 AR activation. The stably transfected CHO A 1 and A 3 cells had B max values of 530 ± 210 fmol/mg protein and 253 ± 19 fmol/ mg protein, respectively, showing that there is similar receptor expression in both cell lines.

Fluorescent Detection of Dendrimer (G2.5) Conjugates Bound to A 3 AR Expressed in CHO Cells
10 μM of compounds 15 or 17 were incubated for 1 h with CHO cells that did or did not stably express the A 3 AR. After one wash with PBS, the cells were imaged at 100× magnification on a Zeiss AxioVision D1 Imager, and both light and fluorescent pictures were obtained. As shown in Figure 7, only the cells expressing the A 3 AR were bound by 17. Neither type of CHO cells were bound by 15, the control dendrimer with no ligand attached. 17 was cAMP inhibition curves for 7 and 17 Figure 5 cAMP inhibition curves for 7 and 17. After 30 min incubation with increasing concentrations of 7 or 17, forskolin was added to CHO cells expressing A 1 or A 3 ARs to increase adenylyl cylase. The inhibition of adenylyl cylase was measured using the Direct cAMP Enzyme Immunoassay. For a summary of EC 50 values obtained, see Table 2. The results shown are means ± S.E.M. of three independent experiments.  Radioligand binding curves for 17  Table 1. unable to bind CHO cells that did not express the A 3 AR. While background fluorescence was seen for both compounds 15 and 17 when incubated with the CHO cells, this fluorescence did not correspond to the location of the cells. The fluorescence bound to the surface of the A 3 ARexpressing cells was not evenly distributed, but rather showed a punctuated distribution, possibly due to receptor aggregation.

Discussion
Many drugs have already been delivered using dendrimers by bioreversible covalent conjugation, including methotrexate [25] and penicillin V [26]. Methotrexate, which was covalently attached via a hydrolyzable ester bond to a generation 5 (G5) PAMAM dendrimer that also contained folic acid, was significantly more toxic to folic acid receptor-expressing cancer cells than was the free ligand. The dendrimer with the ligand and folic acid attached and monomeric folic acid appeared to have a similar affinity for the folic acid receptor [25]. Our study differs from the bioreversible approach in that it describes covalent nucleoside-dendrimer conjugates that do not require cleavage in order to achieve a biological effect. It extends previous studies in which another receptor subtype, the A 2A AR, was targeted [4,24].
The previous studies of A 2A AR-directed dendrimers utilized exclusively amino-terminal dendrimers, such as G3 PAMAM dendrimers. Dendrimers with free amino groups at the periphery typically display a toxicity that is dependent on both size (i.e. generation) and concentration. This toxicity can be ameliorated by using neutral or anionic dendrimers, such as half-generation PAMAM dendrimers that have terminal carboxyl groups [21]. For instance, G3 PAMAM dendrimers caused significant haemolysis at 4 mg/ml in red blood cells, whereas G2.5 PAMAM did not cause haemolysis at 10 mg/ml [27]. In fact, only G7.5 PAMAM and higher generations caused significant haemolysis at 4 mg/ml. Therefore, we have included in this study dendrimers of half-generation, e.g. 2.5, which have terminal carboxylate groups and are therefore likely to be less toxic. When the G3 dendrimer was used in the present study, most of the amino groups were blocked to avoid cytotoxicity.
The differences between half-generation compared to integral generation PAMAM dendrimers in drug delivery have not been adequately studied. One study did attach methotrexate to G2.5 and G3 dendrimers using the terminal amine and carboxyl groups of the ligand, respectively, and the G2.5 conjugate had increased drug activity compared to either free methotrexate or the G3 conjugate [28]. However, the major reason given for the increased activity with the G2.5 derivative was that the methotrexate was released from the dendrimer due to prolonged interactions with proteases in the lysosome since the G2.5 derivative has an anionic charge. The paper concluded that it was probably necessary for the drug to be released from the dendrimer in order to retain its cytotoxic activity. However, GPCR ligands activate their receptors from outside of the cell, and it is unlikely that the ligand will need to be released from the dendrimer to retain activity. Therefore, there must be a different explanation for the improvement of selectivity and affinity of the G2.5 dendrimer-ligand conjugates.
The two functionalized congeners related to ADAC as well as ADAC itself were attached to the dendrimer through a terminal side chain attached to the same position on the nucleoside and which should not interfere with the binding of the adenosine moiety to the receptor. Each of the nucleoside monomers (compounds 1, 4, and 7) showed less than a 5 fold difference in affinity between the human A 1 and A 3 ARs, but 17 had an enhanced affinity and selectivity at A 3 AR in both radioligand binding and cAMP assays. This enhancement can be explained by the difference in the linking moieties, such that the G2.5 dendrimer-ligand conjugates 16 and 17 allow a significant increase in selectivity towards the A 3 AR compared to the A 1 AR. This selectivity was not seen with the G3 dendrimer-ligand conjugates, which were approximately equipotent at the A 1 and A 3 ARs in both radioligand binding and cAMP assays. However, the decrease in affinity at the A 1 AR and increase in affinity at A 3 AR was seen in our previous work using A 2A AR-directed G3 dendrimer-ligand conjugates. This work showed that dendrimer-nucleoside conjugates increased the selectivity at the A 2A AR com-  pared to the A 1 AR by decreasing the binding affinity at the A 1 receptor, and that all of the dendrimer ligand-conjugates were most potent at the A 3 AR [24]. Dendrimer-conjugates 12 and 13, which are approximately equipotent at A 1 and A 3 ARs could be useful for cardioprotection [29], while A 3 AR selective conjugates 16 and 17 could be useful in the treatment of rheumatoid arthritis [30].
Since residual amino groups on dendrimers are associated with cytotoxicity, the unreacted terminal positions of the A 3 AR-selective G2.5 dendrimer 17 were present as carbox-ylate groups, which had the further benefit of increasing water-solubility. Interestingly, not only does the A 3 AR selectivity of 17 improve upon conjugation to the dendrimer, but the affinity also slightly improves compared to the parent nucleoside, 7. While this could be due to the fact that the nucleoside concentration is higher since there are multiple ligands attached per dendrimer, it is unlikely that this is the sole cause of the phenomenon. There are only on average three ligands attached per dendrimer, so it is unlikely that all of them are in the correct geometry to bind multiple receptor proteins in the membrane simulta- neously. It is possible that the dendrimer-ligand conjugate would be blocking other receptor binding sites for the radioligand since it is much larger than the monomer. However, if this was the case, it would be expected that each of the dendrimer-ligand conjugates is more potent than the free nucleosides; instead, the conjugate 16 is approximately equipotent to 1. It could also be possible that due to the overexpression of the A 3 AR on the CHO cells, A 3 AR dimers are forming and the dendrimer conjugate is able to bridge the binding sites of both receptors. A 3 ARs are known to accumulate in membrane microdomains and may form A 3 AR homodimers [31]. If there is only one G protein associated with the GPCR dimer, the receptor that is not attached to the G protein could act as an anchor for the dendrimer ligand complex, allowing for a lowering of the EC 50 and K i app values. Previously molecular modeling work at the A 2A AR has shown that one dendrimer with multiple ligands could bridge an AR dimer [32]. However, further studies are necessary to elucidate the mechanism for the improvement of selectivity and potency of 17 in comparison to 7.

Light and fluorescent microscopy of CHO or CHO A 3 cells with compound 17
Compound 17 was also studied using fluorescent microscopy. Interestingly, 17, but not 15, the control dendrimer with no ligand attached, was able to bind CHO cells expressing the A 3 AR. Neither compound significantly bound to cells that were not expressing the A 3 AR. The fluorescence remained associated with the cells expressing the A 3 AR after washing. Compound 17 appeared to bind some areas of the membrane more strongly than other areas as shown by increase in fluorescent signal. This finding could provide evidence that the A 3 AR is condensed into patches on the cell membrane, possibly as dimers or oligomers. The uneven distribution of the fluorescent signal might indicate the existence of higher-order receptor oligomers, as it been recently demonstrated for the A 2A AR [33]. We did not use transmission electron microscopy to determine if the fluorescent dendrimer conjugate was internalized by the cells. Internalization of other GPCRs under similar conditions, i.e., incubation with agonist for 1 hr at 37°C, is established. These two issues might be responsible of the punctuate distribution of A 3 ARdependent binding of 17.
Compounds 12 and 13 are identical, except that 13 contains an additional five ADAC moieties per dendrimer. Interestingly, attaching additional ADAC moieties to G3 PAMAM appeared to cause a slight decrease in affinity at all three ARs, although the selectivity remained similar. Our previous results comparing increasing numbers of A 2A AR ligand attachments to dendrimers also failed to show a significant improvement in affinity by adding multiple ligands to the dendrimer [24]. However, in both of these studies, attaching the monomer to the dendrimer did not create a significant enhancement in affinity, unlike in our new G2.5 conjugates. Therefore, it will be interesting to determine if there is an enhancement in affinity when increasing numbers of ADAC moieties are attached to the G2.5 dendrimers.

Conclusion
In conclusion, it is feasible to modulate and even enhance the pharmacological profile of a ligand of a GPCR based on conjugation to a nanocarrier and the precise structure of the linking group, which was designed to interact with distal extracellular regions of a 7 transmembrane-spanning receptor. We have demonstrated the feasibility of potent and selective activation of specific subtypes of ARs using multivalent conjugates and the ability to modulate the selectivity based on the linkage between the pharmacophore and the polymeric carrier. Both G2.5 and G3 PAMAM dendrimers can be successfully used in covalent dendrimer-ligand conjugates directed to GPCRs. High selectivity in binding at the A 3 AR in comparison to the monomeric nucleosides could be achieved, depending on the nature of the linker moiety, i.e., a secondary amine linkage resulted in greater than 100-fold A 3 AR selectivity. The selective macromolecular agonist 17 can now be used in pharmacological models of tissue rescue from ischemia and as a fluorescent ligand tool to characterize the A 3 AR in situ and to probe the existence of A 3 AR dimers. Further studies will be completed using higher generation dendrimers and with new covalently-bound AR ligands. Other GPCRs may also be amenable to this approach to the design of multivalent ligands.

Chromatography and spectroscopy
To prepare a column for size exclusion chromatography (SEC), 100 g of Bio-Beads ® SX-1 beads were suspended in 1 L of DMF. After 24 h to allow for equilibration and expansion, the beads were added to the column as described previously [21]. High Performance Liquid Chromatography (HPLC) purification was performed using an Agilent 1100 Series HPLC (Santa Clara, CA) equipped with a Phenomenex Luna 5μ C18(2) 100A analytical column (250 × 10 mm; Torrance, CA). Peaks were detected by UV absorption using a diode array detector. Proton nuclear magnetic resonance spectra (NMR) were recorded on a Bruker DRX-600 spectrometer after being optimized for each sample using DMSO-d 6 as a solvent unless otherwise noted. To determine the number of ligands attached to each dendrimer, the integration of NMR resonances of the ligand was compared to the integration of signal from one of the sets of carbon-protons on the interior of the dendrimer as described previously [4]. Electrospray ionization mass spectra (ESI MS) were taken using a Waters LCT Premier mass spectrometer. Matrix Assisted Laser Desorption/Ionization Time-of-Flight (MALDI-TOF) spectra were obtained with a Waters Micro mass spectrometer using Waters MassPREP Direct Ionization on Silica Desorption/ionization (DIOS) target plates.
The ESI MS data for the dendrimer complexes was obtained using a Waters LCT Premier TOF mass spectrometer. The mass spectrometer was operated in positive ion W mode with a resolution of 10000 measured at half peak height. The capillary voltage was 2500 volts, the cone voltage was 40 volts, and the desolvation gas was dried nitrogen at 250°C and a flow of 300 L/h. The sample was dissolved in a 1:1 solution of water:acetonitrile containing 0.2% formic acid and injected directly into the eluting stream flowing at 200 μL/min and consisting of 20:80 water:acetonitrile and 0.2% formic acid. The relevant spectra were summed using the MassLynx software, and the summed spectrum was deconvoluted with the Max-EntI program.

2-(6-Amino-3-iminio-4,5-disulfonato-3H-xanthen-9-yl)-5-(2aminoethylcarbamoyl)benzoate
G3 PAMAM -23 Ac -AF488 (11) -Method 2 0.5 mL of G3 PAMAM methanol stock solution (18.8 mM, 9.4 μmol) was added to a flask, and the methanol was evaporated. The remaining polymer was dissolved in 0.5 mL of DMSO-d 6 . Acetic anhydride (20.4 μL, 216 μmol, 23 eq) was diluted in 0.5 mL of DMSO-d 6 , and this solution was added dropwise to the solution of G3 PAMAM with stirring. After 18 h, an NMR spectrum of the reaction mixture showed approximately 23 acetamide groups per den-drimer, as expected, to give 9. 234 μL of this solution (2.2 μmol, 9.4 mM) was removed and diluted to 500 μL with DMSO-d 6 . AF488-TFP (10) (2 mg, 2.3 μmol, 1.05 eq) was dissolved in 300 μL of 0.1 M MES, pH 5 and added to the mixture under nitrogen atmosphere. EDC (42 mg, 220 μmol) was dissolved in 300 μL of 0.1 M MES, pH 5 and added to the reaction mixture. After 48 h, the solution was vacuum filtered to remove a small amount orange precipitate that formed. The NMR spectrum was consistent with the assigned structure, but the signals resulting from AF488 could not be properly integrated due to the large G3 PAMAM peaks. The molecular weight of the compound was unable to be determined using either ESI or MALDI-TOF MS. Therefore, it was assumed that approximately one Alexa-Fluor 488 moiety was attached per G3 PAMAM based on previous data [4]. Finally, a mixture of triethylamine (28 μL, 202 μmol) and PyBOP (26 mg, 50 μmol) dissolved in 1.5 mL of DMSO was added. After 48 hr, the product was purified by SEC using DMF as the eluent. The fractions containing product which had the Alexa-488 moiety were dried and dissolved in DMSO-d 6 for NMR. The first and last fractions containing the product were excluded to provide a more homogenous sample. The remaining fractions were combined and dried to give 8.68 mg of product, which contained on average 8 moieties of 4 per dendrimer (0.694 μmol, 46% yield based on μmol of dendrimer

G2.5 PAMAM -AF488 (15)
This procedure was adapted from a carbodiimide coupling described previously [19,20] 120H). The mass spectrum of this compound was too fragmented to determine a molecular weight.

Cell Culture and Membrane Preparation
CHO (Chinese hamster ovary) cells stably expressing the recombinant human ARs were cultured in Dulbecco's modified Eagle medium (DMEM) and F12 (1:1) supplemented with 10% fetal bovine serum, 100 units/mL penicillin, 100 μg/mL streptomycin, and 2 μmol/mL glutamine. After harvesting, cells were homogenized and suspended. Cells were then centrifuged at 500 g for 10 min, and the pellet was resuspended in 50 mM Tris-HCl buffer (pH 7.5) containing 10 mM MgCl 2 . The suspension was homogenized and was then recentrifuged at 20 000 g for 20 min at 4°C. The resultant pellets were resuspended in Tris buffer, incubated with adenosine deaminase for 30 min at 37°C, and the suspension was stored at -80°C until the binding experiments. The protein concentration was measured using the BCA Protein Assay Kit from Pierce [20].

Radioligand Membrane Binding Studies
Radioligand binding assays were performed for A 1 and A 2A ARs, following the procedure described previously [22]. Each tube in the binding assay contained 100 μL of membrane suspension (20 μg of protein), 50 μL of agonist radioligand, and 50 μL of increasing concentrations of the test ligands in Tris-HCl buffer (50 mM, pH 7.5) containing 10 mM MgCl 2 . The concentration of the dendrimerligand complexes are measured by the concentration of the dendrimer, not the ligand. Therefore, all K i values are measured as K i app values. Nonspecific binding was determined using a final concentration of 10 μM 5'-N-ethylcarboxamidoadenosine diluted with the buffer. The mixtures were incubated at 25°C for 60 min. Binding reactions were terminated by filtration through Whatman GF/B filters under a reduced pressure using a MT-24 cell harvester (Brandell, Gaithersburg, MD). Filters were washed three times with 5 mL of 50 mM ice-cold Tris-HCl buffer (pH 7.5). The radioactive agonists [ 3 H]2-chloro-N 6cyclopentyladenosine and [ 3 H]2-(4-(2-carboxyethyl)phenylethylamino)-5'-N-ethylcarboxamido-adenosine were used for the A 1 and A 2A assays, respectively. All of the filters were washed 3 times with Tris-HCl, pH 7.5. Filters for