Design and evaluation of xanthine based adenosine receptor antagonists: Potential hypoxia targeted immunotherapies

Article abstract of DOI:10.1016/j.bmc.2013.09.043

Molecular modeling techniques were applied to the design, synthesis and optimization of a new series of xanthine based adenosine A_2A receptor antagonists. The optimized lead compound was converted to a PEG derivative and a functional in vitro bioassay used to confirm efficacy. Additionally, the PEGylated version showed enhanced aqueous solubility and was inert to photoisomerization, a known limitation of existing antagonists of this class.

Full text of DOI:10.1016/j.bmc.2013.09.043

Accepted Manuscript
Design and Evaluation of Xanthine Based Adenosine Receptor Antagonists:
Potential Hypoxia Targeted Immunotherapies
Rhiannon Thomas, Joslynn Lee, Vincent Chevalier, Sara Sadler, Kaisa
Selesniemi, Stephen Hatfield, Michail Sitkovsky, Mary Jo Ondrechen, Graham
B. Jones
PII:
S0968-0896(13)00823-7
DOI:
http://dx.doi.org/10.1016/j.bmc.2013.09.043
BMC 11127
Reference:
Bioorganic & Medicinal Chemistry
To appear in:
Received Date:
Revised Date:
Accepted Date:
31 July 2013
9 September 2013
17 September 2013
Please cite this article as: Thomas, R., Lee, J., Chevalier, V., Sadler, S., Selesniemi, K., Hatfield, S., Sitkovsky, M.,
Ondrechen, M.J., Jones, G.B., Design and Evaluation of Xanthine Based Adenosine Receptor Antagonists: Potential
Hypoxia Targeted Immunotherapies, Bioorganic & Medicinal Chemistry (2013), doi: http://dx.doi.org/10.1016/
j.bmc.2013.09.043
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Design and Evaluation of Xanthine Based Adenosine Receptor
Antagonists: Potential Hypoxia Targeted Immunotherapies
Rhiannon Thomas,1 Joslynn Lee, 1 Vincent Chevalier, 1 Sara Sadler,1 Kaisa Selesniemi, 2
Stephen Hatfield, 2 Michail Sitkovsky, 2 Mary Jo Ondrechen, 1 and Graham B. Jones1
*
1Bioorganic and Medicinal Chemistry Laboratories, Department of Chemistry and Chemical
Biology, Northeastern University, 360 Huntington Avenue, Boston, MA, 02115 Fax: 617-373-
8795 Email: gr.jones@neu.edu 2New England Tissue Protection Institute, Northeastern
University, 360 Huntington Avenue, Boston, MA, 02115
Abstract: Molecular modeling techniques were applied to the design, synthesis and optimization of a new series of
xanthine based adenosine A2A receptor antagonists. The optimized lead compound was converted to a PEG
derivative and a functional in vitro bioassay used to confirm efficacy. Additionally, the PEGylated version showed
enhanced aqueous solubility and was inert to photoisomerization, a known limitation of existing antagonists of this
class.
Introduction
The adenosine receptors are a class of GPCR’s divided into four subtypes, A1, A2A, A2B, and A3,
differentiated based on their pathways of signal transduction.1 Activation of A1R and A3R
induces coupling to Gi pertussis-toxin sensitive proteins causing downregulation of adenylate
cyclase while activation of the A2AR and the A2BR results in Gs protein coupling and subsequent
upregulation of adenylate cyclase.1 These mechanisms result in control of intracellular cAMP
levels. Additionally, adenosine receptors, when bound to other G proteins, play a role in the
control of ion channels via protein kinases and phospholipase C (PLC) activity.1a,1c Until
recently, homology models of the four AR subtypes (based on the X-ray structure of rhodopsin)

represented the principal means of probing receptor function and binding modes. However, in
2008 Jaakola reported the X-ray crystal structure of the human A2AR bound to the high-affinity
antagonist ZM241385 (Figure 1).2
Figure 1. Crystal structure of the human A2AR bound to ZM241385 antagonist
This report prompted considerable research to identify additional antagonists, as A2ARs are
implicated in several diseased states offering the potential to impact neurodegenerative disorders,
cardiac regulation, inflammation and cancer,1b, 3 the latter a focus of interest to these laboratories.
In normal tissues, pathogen exposure causes accumulation of extracellular adenosine and
triggers “adenosinergic” signaling, mediated by A2AR and A2BR, which in turn induces increased
production of cAMP.4 Increased levels of cAMP then function to inhibit overactive
immunosuppressive T cells from attacking healthy tissues.4 However, this response is also
demonstrated in the hypoxic tumor microenvironment, and results in misguided protection of

cancerous cells. It is important to note that the effectiveness of the immune response is
dependent on the densities of the receptor.5 This means that if the number of receptors has
decreased by half, twice as many agonist ligands would need to be present to activate the
receptor. Differences in receptor affinity can also be observed across the four subtypes.6 Agonist
binding varies widely between A1, A2A, and A3 with Kd values of 0.6, 20, and 6 nM, respectively.
Ohta et al7 reported that based on these mechanisms, ZM241385 significantly delays the rapid
growth of CL8-1 melanoma in mice with endogenously developed antitumor T cells. Based on
this promising result, we embarked on an effort to ascribe structure with function in a series of
synthetic antagonists. However, based on the limited circulating half-life of ZM241385 in vivo
(<60 min.) we sought a more robust platform and elected to investigate the xanthine class.
Caffeine, a prototypical xanthine owes its CNS activity to adenosine receptor antagonism and
both it and its structural analog theophylline, demonstrate antagonism in the micromolar range
for A2AR.1b The 1, 3, 7, and 8-postions of the xanthine scaffold have been
most widely exploited in pursuit of A2AR selective antagonists. Substitution at N1 is most
important for both A1R and A2AR binding. The A2AR tolerates smaller alkyl substituents, i.e.
methyl, allyl, propargyl, ethyl, propyl, and cyclopentyl, while the A1R binding pocket is able to
accommodate larger substituents at this position.8 Small alkyl chains are also most commonly
found at N3, including methyl, ethyl, propyl, and 3-hydroxypropyl, but larger substituents,
including phenoxy groups are also tolerated.9

Methyl substitution at N7 appears to impart the A2AR specific activity of caffeine and
paraxanthine analogs. 7-methylation was found to reduce affinity for both A1R and A2AR in the
case of 8-substituted xanthines, with one exception; 8-styrylxanthines. 7-methyl-8-
styrylxanthines demonstrated increased selectivity and either comparable or improved potency at
the A2AR.8 The first highly potent and selective A2AR antagonist, 54 nM with 529-fold
selectivity, CSC, bears a m-chlorostyryl group at C8. The discovery of CSC provided a thrust for
further exploration of the SAR around this class of compounds.1b Meta- and para- substitution of
the styryl phenyl group is generally well tolerated. Meta- halogens and meta/para- methoxy
substituents impart improved selectivity of analogs with small alkyl chains at the 1 and 3
positions of the xanthine core. Replacement of the phenyl group with thienyl (DMPTX), furyl,
pyridyl and pyrazolyl groups yield compounds inferior to the styryl derivatives.8 These studies
eventually resulted in the clinical candidate KW-6002 (Istradefylline), with meta- and para-
methoxy substituents, which was submitted to phase III clinical trials for the treatment of PD. As
KW-6002 displays a favorable t1/2 of ~2h (rat), we conducted preliminary experiments using a
tumor immunotherapy model which confirmed its potential.10-12
Unfortunately, KW-6002 and related 8-styrylxanthines are plagued with issues of vinyl
photoisomerization in dilute solution and have been reported to undergo [2+2] cycloaddition in
the solid form. The Z-isomers are significantly less potent and selective than the E-isomers, and
hence replacements for the vinylene linker have been sought. Significant loss of affinity is
observed when vinylene is replaced with an alkynyl, cyclopropyl, and 2-naphthyl linkers and

also when tricyclic constraint is indroduced.1b Azo derivatives retain selectivity, but not affinity
for A2AR, while the imine analogs are both more potent and selective, but are significantly less
stable due to their Schiff base structure.1b, 8
Based on the promising data supporting use of KW-6002 in immunotherapy models,4c we
proposed to use molecular modeling guided synthesis of a library of analogs and optimize for
chemical stability and target specificity. Specific goals were to:
1. produce a readily synthetically accessible library of compounds
2. control biodistribution by minimizing bbb penetration
3. reduce photo-isomerization and other degradative pathways
4. improve water solubility
According to the above criteria it was evident that considerable structural modifications would
be required. KW-6002 was designed for use in Parkinson’s disease, its lipophilicity
advantageous for localization in the brain striatum environment, as confirmed by subsequent
PET imaging studies with an [11C] analog.1b An appropriate strategy for our purposes appeared
to be to derivitize an optimized lead compound as a PEGylated bioconjugate, enhancing its
hydrophilicity and by virtue of molecular weight diminishing its blood-brain-barrier crossing
ability, based on consideration of Lipinski’s rules.
Results and Conclusions
Previous work in our laboratory has focused on the development of an efficient method for
synthesis of 8-substituted xanthines from 5,6- diaminouracils and carboxaldehydes using
bromodimethylsulfonium bromide.13 Our first synthesis candidate was an ortho substituted aryl
ester, which would be subsequently converted to the PEGylated derivative for evaluation.
Diaminouracil was coupled to methyl-2-formylbenzoate under BDMS conditions to afford

xanthine benzoate 3 after filtration. Subsequent xanthine methylation and saponification of the
benzoate yielded the key methylxanthine benzoic acid 5 for esterification (Scheme 1).
Scheme 1. Synthesis of ester-linked polyethylene glycol conjugates.
To demonstrate the accessibility of a range of PEGylated analogs, three PEG conjugates
of different lengths were synthesized using EDCI mediated coupling conditions viz. a 600, 1500
and 2000 MW (6a-c). In addition to the desired conjugates, amounts of bis-xanthine conjugates
were detected, which proved difficult to remove from the desired adducts. Remedy was found by
allowing a pre-incubation period of 2h with the xanthine and coupling agent prior to addition of
the PEG-OH. To extend synthetic versatility, a benzyl ester derivative 8 was also prepared
(Scheme 2), accompanied by quantities of the bis-conjugate 9. Unfortunately attempts to remove
all traces of the dimer chromatographically were unsuccessful.

Scheme 2. Synthesis of PEG-diacid conjugate from the xanthine benzoate.
With quantities of PEG-ylated derivatives in hand, we initiated preliminary in vitro assays to
determine A2A affinity. However, though the derivatives were freely water soluble and
essentially stable in solution, affinity for A2A was considerably inferior to control (KW-6002),
and molecular modeling studies were initiated to guide identification of additional analogs for
synthesis. Initial docking studies were performed in Yet Another Scientific Artificial Reality
Application (YASARA) Structure which includes a derivative of AutoDock 4.014 and in
Schrodinger’s GLIDE115 to asses affinity of potential analogs for the active site of the human
A2AR (PDB file: 3EML)2 based on predicted theoretical binding energies (Figure 2). The first
important interaction of ZM241385 is the aromatic stacking of Phe168. The second is a
hydrogen bond interaction of Asn253 to N15 atom of ZM241385.

Figure 2. Scrutinized residues (in black) and water molecules (in red) in the active site of A2AR. A2AR from PDB ID
3EML bound to ZM241385 (in green) represents a conformation for the receptor in an inactive state.
A family of arylxanthine isomers derivitized as ortho-, meta-, and para- methyl esters
were scrutinized as surrogates for the PEG-ylated versions, as the PEG chain was not well
tolerated in AutoDock due to the number of rotatable bonds. As depicted in Figure 3, the o-
methyl ester derivative was not planar, and the distortion of the arene-xanthine bond resulted in a
collision with Phe168 and a positive theoretical binding energy. A similar distortion was
observed with the p-methyl ester arylxanthine. The highly positive predicted binding energy was
presumably due to poor orientation of the xanthine core. The m-methyl ester arylxanthine had a
significantly improved binding energy over the ortho- and para- isomers resulting from π–
stacking between the arene and Phe168, which induced planarity in the molecule. However the
predicted energy of this compound was not a highly negative value; therefore, it was concluded
that the arylxanthine series was not the optimal scaffold for conjugation and further explains the
poor outcome in the preliminary binding assays.

B.E.ortho = 11.76 kcal/mol
B.E.meta = -0.44 kcal/mol
B.E.para = 48.66 kcal/mol
Figure 3. Models of the o, m and p- methyl ester arylxanthines with corresponding theoretical binding energies.
The second extracellular loop (ECL2, residues Gln148 to Ser156) was missing from the
two structures 3EML due to weak experimental electron density in that region. The homology
modeling module in YASARA16 was used to build the missing loop in the 3EMLstructure. To
evaluate model quality, the structures were analyzed using SWISS-MODEL Workspace

PROCHECK (see methods).17 The docking studies revealed the ECL2 provides important
interactions to the ligands and affects ligand confirmation.
Closer examination of ZM241385 docking provides insight into an additional issue with
the arylxanthines. In the binding of ZM241385, Phe168 underwent π–stacking with the
triazolotriazine core of the molecule, and not with the phenyl moiety. This difference suggested
that the arylxanthines were positioned too deeply within the active site, and did not extend
through the pocket sufficiently to have a high affinity. Accordingly, analogs with high structural
similarity to KW-6002 were designed wherein the styryl moiety could provide appropriate
spacing. Two compounds of immediate interest were styrylxanthines that maintained one
methoxy group of KW-6002, at the meta- or para- position of the arene, and were conjugated to
PEG in the respective para- or meta- position (Figure 4).
Figure 4. Styrylxanthine analogs selected for additional docking studies. The portion of the compound shown in red
is analogous to KW-6002.

Figure 5. Molecular models of styryl xanthines n = 0, bound to the A2AR crystal structure. (Left) The para-PEG
derivative docked into the 3EML structure (light brown), GLIDE pose (orange) and AutoDock pose (pink) have
similar docking poses and (right) the meta-PEG derivative docked into the 3EML structure (light brown), GLIDE
(orange) and Autodock pose (pink) have similar results to a docking pose.
At this point, the tolerance of AutoDock 4.0 and Schrodinger’s GLIDE for rotatable
bonds was expanded so that more authentic modeling studies could be accomplished. This
exercise allowed for the addition of a diethylene glycol chain to the docked compounds and
analysis of the placement of the oligomer in the binding pocket (Figure 5). Both the meta- and
para- PEG derivatives demonstrated significantly improved binding energies over the
arylxanthine compounds, -9.00 kcal/mol and -9.67 kcal/mol respectively in AutoDock and Glide
Score -7.723 and -7.960, respectively. Critical π–stacking between the xanthine core and
Phe168 and hydrogen bonding between xanthine C2-carbonyl and Asn253 contribute to the high
affinity of these conjugates.2 Additionally, the incorporation of the vinyl linker extended the
molecule so that the arene was stabilized by hydrophobic contact with Met270, which is a
contact observed in the binding of the phenyl pharmacophore in ZM241385.2
Since neither analog displayed a significant advantage over the other in binding at the
xanthine core or the styryl pharmacophore, attention was directed to the conformation of the

oligoethylene glycol moiety. Upon conjugate binding to the receptor, the polymeric carrier
should not disrupt the binding of the antagonist. Therefore, the PEG chain should lie outside of
the active site, as it does with the para- PEG conjugate. The docking of the meta- PEG
derivative shows the diethylene glycol folded back into the pocket, and although this did not
effect the conjugate’s theoretical binding energy, it is indicative of an interference that may
occur with a longer ethylene glycol chain (n ≥ 4). Unfortunately, docking of a PEG chain with n
≥ 4 was not possible with AutoDock 4.0 or Schrodinger, and would most likely have been
unreliable. Based on these observations, the para- PEG analog of KW-6002 was selected for
development.
Synthesis of the styrylxanthine began with Heck coupling of acrolein to iodobenzoate
under phosphine-free conditions (Scheme 3). This reaction proceeded with full conversion to the
desired cinnamaldehyde product 10, and neither aryl-aryl homocoupling nor iodide elimination
was observed. Compound 10 was then coupled to diaminouracil 2 in the presence BDMS and the
product 11 isolated in high purity after recrystallization from DMSO/water. Xanthine
methylation and subsequent ester saponification yielded the functionalized styrylxanthine 13 for
polymer conjugation.

Scheme 3. Synthesis of styrylxanthine-PEG conjugate.
The styrylxanthine compound was conjugated to PEG monomethyl ethers (PEG-Me) to
alleviate the bis-coupling issues observed with derivatization to the PEG-diols. Two different
length chains, octaethylene glycol-Me and PEG750-Me (14a,b) were selected with the ultimate
goal of comparing their physiochemical properties in vitro and in vivo. Condensation was
initially attempted under the conditions used for the arylxanthine compound, but the reactions
were low yielding, <20%. Several alternative procedures were investigated, including use of
Mitsonobu conditions, the Mukaiyama reagent, and alumina to little benefit.18 Ultimately,
modification of the EDCI coupling (additional 1.1 equiv. EDCI and 10 mol % DMAP with
overnight reflux) produced the octaethylene glycol-Me conjugate and the PEG750-Me conjugate
in 40% and 27% isolated yields, respectively. Noteworthy in the route to 13 is that
chromatography is not required at any stage prior to PEG coupling, allowing efficient scale up.

Given there is a correlation between the CNS penetration of a molecule and its
octanol/water partition coefficient,19,20 (log P) thus we elected to determine this for the PEG-
esters using a modified HPLC method, based on the EPA Product Properties Test Guidelines
(OPPTS 830.7570).21
Using multiple reference compounds an R2 = 0.993 was obtained and the experimental
values compared to theoretically derived values (ChemAxon).22
Experimental
log P (HPLC) log P (ChemAxon)
Theoretical
Compound
KW-6002
log k
0.162
2.42
2.975
2.20
1.83
KW-octaethylene glygol-Me
KW-PEG750-Me
0.145
0.143
2.802
2.775
Table 1. Calculated lipophilicities of KW-6002 and synthetic variants 14
A marginal decrease in log P, approximately 0.2 log P units, is observed for the PEG conjugates
versus KW-6002, and while a relatively mild impact, the nearly two-fold increase in the
hydrophilicity of the compounds bodes well for development, with KW-6002 partitioning in a
944:1 ratio between the lipophilic and hydrophilic layers, and the conjugates partitioning in a
633:1 (octaethylene glycol- Me) and a 596:1 ratio (PEG750-Me) respectively. We also assessed
the degree of photo-stability of the PEG analogs versus KW-6002 using an HPLC assay.
Whereas freshly prepared KW6002 underwent photo-isomerization to the Z isomer (less than
10% abundance in solution within 24h, and >50% abundance within 7 days), no isomerization of
the PEGylated variants was detected even after 2 weeks in solution, as shown in Figure 6. While
the precise mechanism of this protective function is unknown, it is tempting to hypothesize that
the steric constraints in the PEG has an impact, favoring the E isomer.

(A)
(B)
(C)
(D)
Figure 6. Photoisomerization data of PEG conjugates and KW-60002. A) KW-octaethylene glycol-Me sample after
2 weeks B) KW-PEG750-Me sample after 2 weeks C) KW-6002 after 8h D) KW-6002 after 2 weeks
Functional Bioassay
Bioassays of the PEG analog 14a was conducted using two functional models that assess
A2AR binding-dependent signaling through A2AR on the surface of T cells. In the first series of
experiments, the functional efficacy of this antagonist (‘KW-PEG’) was tested by determining
the extent of inhibition of A2AR-induced intracellular cAMP accumulation in the A2AR-
expressing lymphocytes. The activation of A2AR in these assays was accomplished by using the
classical agonist, CGS 21680. It is shown (Figure 7) that both KW-6002 and KW-PEG are able
to prevent CGS 21680-mediated signaling. Similar, or even stronger, antagonism was observed
by KW PEG treatment when compared to KW-6002, and is directly attributable to the PEG chain
as routine control experiments showed the methyl ester of 13 to have near identical activity to
KW-6002 itself.

Figure 7: Cyclic-AMP levels in lymphocytes after incubation with vehicle (VEH), 1uM CGS (CGS), 1uM CGS
plus 0.5-10um KW (KW), and 1uM CGS plus KW-PEG (KW-PEG) is shown. The intracellular cAMP levels were
determined 15 minutes following stimulation using quantitative cAMP ELISA and are expressed as
pmols/million cells. Data shown represent mean ± SEM of triplicate samples
In addition, we used a biological assay of cytokine secretion that is considered to be
sensitive in the determination of functional effects of A2AR signaling. In these assays,
immediately following TCR activation by CD3 ligand, T cells from splenocytes were incubated
with CGS 21680 since it was shown to prevent IFN-gamma secretion due to increases in A2AR-
induced immunosuppressive intracellular cAMP. It is shown (Figure 8) that the activated T cells
produced IFN-gamma, but were significantly inhibited by engagement of immunosuppressive
cAMP-elevating A2AR signaling. Importantly, repeating the same assay in the presence of
increasing quantities of KW-6002 and the PEG-ylated version restored IFN-gamma secretion.

The activity of the PEGylated version was nearly identical to KW-6002, noteworthy given that
substantial structural modification has been made. This result also vindicated the molecular
modeling strategy used in the design (Figure 5).
Figure 8: The INF-gamma production by lymphocytes after activation with 0.1 ug/ml mAB-CD3 and when
treated with vehicle (VEH), 1-100 nm CGS (CGS), 1-100nm CGS plus 0.5 um KW (KW), and 1-100nm CGS plus
-0.5 um KW-PEG (KW-PEG) is shown. The IFN-γ levels were determined in the supernatant one day following
stimulation using quantitative ELISA and are expressed as pg/ml. Data shown represent mean ± SEM of
triplicate samples.
These findings are noteworthy, and confirm that efficient chemical synthesis guided by
computational analysis can be used for the identification of functional xanthine analogs with
desirable physical and chemical properties. With the functional in vitro assays complete we are
now engaged in the design of in vivo tumor regression studies using hypoxic models, the details
of which will be reported in due course.
Conclusions
A computationally directed approach to the molecular design of improved adenosine A2A
receptor antagonists based on KW-6002 has been employed leading to identification of a
promising lead compound. Synthesis of the styrylxanthine analog and its PEGylated derivatives

was accomplished using an efficient method which allows facile modification of molecular
weight and thus lipophilicity and water solubility. The lead compound did not undergo
photoisomerization even after extended periods in solution. Preliminary functional bioassay
suggest the compound has a comparable antagonist profile to KW-6002, rendering them suitable
for studies as cancer immunotherapy agents, which will now be pursued.
EXPERIMENTAL PROCEDURES
General Synthetic Methods
1H NMR and 13C NMR spectra were obtained on either a Varian Mercury 400 (400 MHz)
or a Varian Inova 500 (500 MHz) spectrometer and are reported in ppm downfield relative to the
1
residual solvent peak. H NMR data is reported followed by multiplicity, coupling constant, and
number of protons. High resolution mass spectrometry and MALDI were conducted at the mass
spectrometry facility at the University of Illinois, Urbana-Champaign. HPLC was conducted on
an Agilent Technologies 1120 Compact LC equipped with a 100-tray autosampler, heated
column compartment, 100 μL variable volume injection loop, and single pump with solvent
selection valve. An Agilent ZORBAX Eclipse XDB-C18 column (4.6 x 150 mm, SN:
USRK025565) was used. Signals were detected using a variable wavelength UV detector (Serial
No.: DE80700574, Part No.: G4290A) and a standard UV flow cell (Part No.: G1314-600086).
High-resolution TEM analyses were performed on JEOL JEM-1010 and the images were
captured using a Xr41B 4-megapixel, bottom-mount camera purchased from Advanced
Microscopy Techniques. Analytical TLC was carried out using silica gel 60 F254 precoated
plates (Scientific Adsorbents, Inc.) and visualized using a 254 nm/ 366 nm UV lamp,
phosphomolybdic acid or ninhydrin stains, or an iodine chamber. Preparative TLC was carried

out using silica gel GF plates (Analtech, Inc.). Flash chromatography was performed using
SiliaFlash P60 (230-40 mesh) silica gel (Silicycle).

Synthetic Procedures
2-(1,3-diethyl-2,6-dioxo-2,3,6,7-tetrahydro-1H-purin-8-yl)benzoate (3):
BDMS (0.101 g, 0.451 mmol, 0.9 eq) was added to a stirring solution of 1,3-diethyl-5,6-
aminouracil 2 (0.100 g, 0.504 mmol, 1.0 eq) and methyl 2-formylbenzoate (0.070 mL, 0.504
mmol, 1.0 eq) in acetonitrile (10 mL). The reaction was stirred at room temperature overnight.
The reaction mixture was concentrated in vacuo and the resulting residue was purified via silica
gel chromatography (9:1 hexanes: ethyl acetate to 1:1 hexanes: ethyl acetate) to afford the
1
product as a white crystalline solid (0.099 g, 57%). mp = 198 – 201oC; H NMR (500 MHz,
DMSO-d6): δ 13.86 (brs, 1H), 7.86 (d, J = 7.0 Hz, 1H), 7.71 (t, J = 8.0 Hz, 1H), 7.67 (d, J = 8.0
Hz, 1H), 7.61 (t, J = 8.0 Hz, 1H), 4.03 (q, J = 7.0 Hz, 2H), 3.95 (q, J = 7.5 Hz, 2H), 1.24 (t, J =
7.0 Hz, 3H), 1.15 (t, J = 7.5 Hz, 3H); 13C NMR (100 MHz, DMSO-d6): δ 168.8, 154.6, 152.6,
152.4, 150.8, 148.1, 132.9, 131.8, 130.0, 129.6, 108.5, 52.9, 38.7, 36.5, 13.9, 13.8; HRMS (ESI)
m/z C17H18N4O4 (M+H)+ calcd 343.1403, obsd 343.1406.

Methyl 2-(1,3-diethyl-7-methyl-2,6-dioxo-2,3,6,7-tetrahydro-1H-purin-8-yl)benzoate (4):
To a solution of methyl 2-(1,3-diethyl-2,6-dioxo-2,3,6,7-tetrahydro-1H-purin-8-yl)benzoate 3
(1.13 g, 3.29 mmol, 1.0 eq) in DMF (25 mL) was added K2CO3 (1.52 g, 11.0 mmol, 3.0 eq)
followed by iodomethane (0.686 mL, 11.0 mmol, 3.0 eq). The reaction mixture was heated to
60°C overnight. The solution was diluted with water (100 mL) and extracted with chloroform (2
x 100 mL). The combined organic extracts were washed with water (2 x 100 mL) and brine (100
mL), dried over MgSO4, and concentrated in vacuo to afford a yellow oil. The residue was
purified by silica gel chromatography (9:1 hexanes: ethyl acetate to 1:1 hexanes: ethyl acetate,
pre-neutralized with 19:1 hexanes: triethylamine) to afford the product as a yellow solid (0.699
g, 60%). mp = 270oC (decomposition); 1H NMR (500 MHz, CDCl3): δ 8.05 (d, J = 7.5Hz, 1H),
7.62 (t, J = 7.5 Hz, 1H), 7.55 (t, J = 7.5 Hz, 1H), 7.41 (d, J = 7.0 Hz, 1H), 4.09 (q, J = 7.0 Hz,
2H), 4.02 (q, J = 7.0 Hz, 2H), 1.26 (t, J = 7.0 Hz, 3H), 1.19 (t, J = 7.0 Hz, 3H); 13C NMR (125
MHz, CDCl3): δ 166.2, 155.5, 151.7, 151.0, 147.8, 132.9, 131.7, 131.0, 130.8, 130.1, 108.2,
38.7, 36.6, 32.9, 13.7, 13.6; HRMS (ESI) m/z C18H20N4O4 (M+H)+ calcd 357.1551, obsd
357.1563.

2-(1,3-diethyl-7-methyl-2,6-dioxo-2,3,6,7-tetrahydro-1H-purin-8-yl)benzoic acid (5):
A 1M aqueous LiOH solution (5.27 mL, 6.0 eq) was added to a solution of methyl 2-(1,3-
diethyl-7-methyl-2,6-dioxo-2,3,6,7-tetrahydro-1H-purin-8-yl)benzoate 4 (0.313 g, 0.878 mmol,
1.0 eq) in THF (20 mL) and the resulting solution was stirred at 50°C for 3 hours. The reaction
mixture was diluted with water (50 mL) and extracted with ethyl acetate (2 x 50 mL). The
combined organics were discarded, and the aqueous layer was acidified to pH ~ 2 with 5% v/v
aqueous HCl and then extracted with ethyl acetate (3 x 50 mL). The combined organic extracts
were dried over MgSO4 and concentrated in vacuo to afford the product as a white crystalline
solid (0.189 g, 65%). mp = 193 - 195oC; 1H NMR (500 MHz, CDCl3): δ 11.35 (brs, 1H), 8.06 (d,
J = 7.5Hz, 1H), 7.60 (t, J = 7.5 Hz, 1H), 7.54 (t, J = 7.5 Hz, 1H), 7.38 (d, J = 7.0 Hz, 1H), 4.06
(q, J = 7.0 Hz, 2H), 4.01 (q, J = 7.0 Hz, 2H), 1.21 (t, J = 7.0 Hz, 3H), 1.17 (t, J = 7.0 Hz, 3H); 13C
NMR (125 MHz, CDCl3): δ 168.1, 155.6, 152.1, 151.0, 147.5, 133.1, 131.8, 131.6, 131.0, 129.6,
108.3, 39.0, 36.9, 33.0, 13.6, 13.5; HRMS (ESI) m/z C17H18N4O4 (M+H)+ calcd 343.1409, obsd
343.1406.

Representative procedure for arylxanthine PEG conjugation:
2-(1,3-diethyl-7-methyl-2,6-dioxo-2,3,6,7-tetrahydro-1H-purin-8-yl)benzyl ester
polyethylene glycol 600 (6a):
EDCI•HCl (0.013 g, 0.065 mmol, 1.1 eq), and DMAP (0.001 g, 0.006 mmol, 0.1 eq) was added
to a solution of 2-(1,3-diethyl-7-methyl-2,6-dioxo-2,3,6,7-tetrahydro-1H-purin-8-yl)benzoic acid
5 (0.020 g, 0.059 mmol, 1.0 eq), in dichloromethane. After 2 hours, Poly (ethylene glycol) 600
(0.039 g, 0.065 mmol, 1.1 eq) was added to the solution. The reaction mixture was stirred
overnight at room temperature. The reaction mixture was concentrated in vacuo and the residue
was purified using preparative silica gel TLC (9:1 dichloromethane: methanol) to afford the title
1
compound as a clear, colorless oil (0.008 g, 14%). H NMR (400 MHz, CDCl3): δ 8.18 (d, J=
7.0, 1H), 7.70 (t, J = 7.2, 1.0 Hz, 1H), 7.64 (t, J = 7.2 Hz, 1H), 7.47 (d, J = 7.2, 1H), 4.32 (m,
2H), 4.18 (q, J = 7.0 Hz, 2H), 4.10 (q, J = 7.0 Hz, 2H), 3.78 – 3.57 (m, 50H), 1.32 (t, J = 7.0 Hz,
3H), 1.28 (t, J = 7.0 Hz, 3H); MALDI C17H18N4O4-PEG: calcd 915.47, obsd 915.34.

2-(1,3-diethyl-7-methyl-2,6-dioxo-2,3,6,7-tetrahydro-1H-purin-8-yl)benzyl ester PEG-1500
(6b):
The title compound was isolated following the general procedure (0.049 g, 46% as a clear,
colorless oil). 1H NMR (400 MHz, CDCl3): δ 8.19 (dd, J= 7.0, 1.0 Hz, 1H), 7.70 (dt, J = 7.0, 1.0
Hz, 1H), 7.64 (dt, J = 7.0, 1.0 Hz, 1H), 7.48 (dd, J = 7.2, 1.0 Hz, 1H), 4.32 (m, 2H), 4.18 (q, J =
7.0 Hz, 2H), 4.11 (q, J = 7.0 Hz, 2H), 3.71 (s, 3H), 3.70 – 3.58 (m, 134H), 1.34 (t, J = 7.0 Hz,
3H), 1.28 (t, J = 7.0 Hz, 3H); MALDI C17H18N4O4-PEG: calcd 1840.02, obsd 1840.33.

2-(1,3-diethyl-7-methyl-2,6-dioxo-2,3,6,7-tetrahydro-1H-purin-8-yl)benzyl ester PEG-2000
(6c):
The title compound was isolated following the general procedure (0.070 g, 51% as a clear,
colorless oil). 1H NMR (500 MHz, CDCl3): δ 8.18 (dd, J= 7.0, 1.0 Hz, 1H), 7.70 (dt, J = 7.5, 1.0
Hz, 1H), 7.67 (dt, J = 7.5, 1.0 Hz, 1H), 7.49 (dd, J = 8.0, 1.5 Hz, 1H), 4.32 (m, 2H), 4.17 (q, J =
7.0 Hz, 2H), 4.11 (q, J = 7.0 Hz, 2H), 3.80 – 3.38 (m, 178H), 1.35 (t, J = 7.0 Hz, 3H), 1.28 (t, J =
7.0 Hz, 3H); MALDI C17H18N4O4-PEG: calcd 2324.31, obsd 2325.39.

1,3-diethyl-8-(2-(hydroxymethyl)phenyl)-7-methyl-1H-purine-2,6(3H,7H)-dione (7):
DiBAlH (1M in hexanes, 1.67 mL, 1.67 mmol, 3.0 eq) was slowly added to a solution of methyl
2-(1,3-diethyl-7-methyl-2,6-dioxo-2,3,6,7-tetrahydro-1H-purin-8-yl)benzoate 4 (0.199 g, 0.558
mmol, 1.0 eq) in dichloromethane (20 mL) at -78 oC. The reaction mixture was stirred at -78 oC
for 4 hours then warmed to room temperature overnight. The reaction was quenched with
methanol (1 mL), diluted with 1% v/v aqueous HCl (50 mL) and extracted with ethyl acetate (3 x
25 mL). The combined organic extracts were washed with brine (50 mL), dried over MgSO4, and
concentrated in vacuo to afford an orange solid. The solid was purified via silica gel
chromatography (9:1 hexanes: ethyl acetate to 3:7 hexanes: ethyl acetate) to afford the product as
a white oily solid (0.112 g, 61%). 1H NMR (500 MHz, CDCl3): δ 7.50 (d, J = 7.5 Hz, 1H), 7.45
(t, J = 7.5 Hz, 1H), 7.38 (m, 2H), 4.91 (t, J = 7.0 Hz, 2H), 4.39 (d, J = 7.0 Hz, 2H), 4.07 (q, J =
7.5 Hz, 2H), 4.02 (q, J = 7.0 Hz, 2H), 3.91 (s, 3H), 1.27 (t, J = 7.5 Hz, 3H), 1.19 (t, J = 7.0 Hz,
13
3H); C NMR (125 MHz, CDCl3): δ 155.4, 151.1, 150.7, 147.0, 142.1, 131.2, 131.2, 130.1,
128.0, 127.5, 108.8, 64.1, 38.9, 36.7, 34.1, 13.6, 13.5; HRMS (ESI) m/z C17H20N4O3 (M+H)+
calcd 329.1605, obsd 329.1614.

1,3-diethyl-8-(2-(hydroxymethyl)phenyl)-7-methyl-1H-purine-2,6(3H,7H)-dione
polyethylene glycol carboxylic acid (8):
1,3-diethyl-8-(2-(hydroxymethyl)phenyl)-7-methyl-1H-purine-2,6(3H,7H)-dione 7 (0.020 g,
0.061 mmol, 1.0 eq) was added to a solution of PEG diacid (0.048 g, 0.070 mmol, 1.1 eq),
EDCI•HCl (0.013 g, 0.070 mmol, 1.1 eq), and DMAP (0.09 g, 0.070 mmol, 1.1 eq) in
dichloromethane (4 mL) and the reaction stirred at room temperature overnight. The reaction
mixture was concentrated in vacuo and purified via preparative silica gel TLC (19:1
1
dichloromethane: methanol) to afford the product as a light yellow oil (0.063 g, 14%). H NMR
(400 MHz, CDCl3): δ 7.58 (d, J = 8.0 Hz, 1H), 7.53 (dt, J = 7.2, 1.2 Hz, 1H), 7.46 (dt, J = 7.2,
1.6 Hz, 1H), 7.35 (dd, J = 7.6, 1.2 Hz), 5.14 (s, 2H), 4.17 (q, J = 7.2 Hz, 2H), 4.10 (q, J = 7.2 Hz,
2H), 3.68 (t, J = 6.8 Hz, 2H), 3.68 – 3.54 (m, 52H), 2.54 (t, J = 6.8 Hz, 2H), 1.34 (t, J = 7.2 Hz,
3H), 1.27 (t, J = 7.6 Hz, 3H). MALDI C17H20N4O3-PEG: calcd 1001.51, obsd 1001.66 and
1311.56.

(E)-methyl 2-methoxy-4-(3-oxoprop-1-en-1-yl)benzoate (10)23:
A solution of 4-iodo-2-methoxybenzoate (10.5 g, 36.0 mmol, 1.0 eq), tetrabutylammonium
bromide (14.3 g, 44.2 mmol, 1.2 eq), and K2CO3 (14.9 g, 108 mmol, 3.0 eq) in DMF (100 mL)
was degassed for 1 hour. Palladium(II) acetate (0.404 g, 5 mol %) was added followed by 2-
propenal (5.60 mL, 83.8 mmol, 2.3 eq) and the reaction stirred overnight at 60°C. The reaction
mixture was taken up in chloroform (300 mL) and washed with water (3 x 200 mL) and brine (2
x 200 mL). The organic extract was dried over MgSO4 and concentrated in vacuo. The resulting
residue was passed through a short plug of silica gel (1:1 hexanes: ethyl acetate) to afford the
1
title compound as an orange crystalline solid (3.07 g, 39%). mp = 109 - 114oC; H NMR (500
MHz, CDCl3): δ 9.61 (d, J = 7.5 Hz, 1H), 7.97 (d, J = 2.5 Hz, 1H), 7.65 (dd, J = 9.0, 2.5 Hz, 1H),
7.37 (d, J = 16.0 Hz, 1H), 6.99 (d, J = 9.0 Hz, 1H), 6.59 (d, J = 16.0, 7.5 Hz, 1H), 3.91 (s, 3H),
13
3.87 (s, 3H); C NMR (125 MHz, CDCl3): δ 193.6, 166.0, 161.4, 151.5, 133.6, 1332.6, 127.7,
126.5, 120.8, 112.8, 56.5, 52.5; HRMS (ESI) m/z C12H12O4 (M+Na)+ calcd 243.0633, obsd
243.0644.

(E)-methyl 4-(2-(1,3-diethyl-2,6-dioxo-2,3,6,7-tetrahydro-1H-purin-8-yl)vinyl)-2-
methoxybenzoate (11):
BDMS (7.32 g, 33.0 mmol, 1.5 eq) was added to a solution of 1,3-diethyl-5,6-aminouracil 2
(4.36 g, 22.0 mmol, 1.0 eq) and (E)-methyl 2-methoxy-4-(3-oxoprop-1-en-1-yl)benzoate 10
(6.22 g, 28.0 mmol, 1.3 eq) in acetonitrile (150 mL). The reaction mixture was stirred at room
temperature for 48 hours, filtered and the yellow precipitate collected. Pure product, a light
yellow crystalline solid, was obtained via recrystallization from DMSO/water (1.94 g, 22%). mp
= 285 - 287oC; 1H NMR (500 MHz, DMSO-d6): δ 13.70 (brs, 1H), 7.70 – 7.66 (m, 2H), 7.42 (s,
1H), 7.26 – 7.21 (m, 2H), 4.05 (q, J = 7.5, 2H), 3.94 (q, J = 7.0 Hz, 2H), 3.89 (s, 3H), 3.79 (s,
3H), 1.26 (t, J = 7.0 Hz, 3H), 1.14 (t, J = 7.0 Hz, 3H); 13C NMR (125 MHz, DMSO-d6): δ 165.8,
158.7, 153.7, 150.2, 149.0, 147.9, 140.6, 134.0, 131.4, 119.8, 118.6, 118.2, 111.1, 107.7, 72.3,
56.0, 51.9, 38.1, 35.8, 13.2; HRMS (ESI) m/z C20H22N4O5 (M+H)+ calcd 399.1668, obsd
399.1678.