Enzymatic Preparation of 2′-5′,3′-5′-Cyclic Dinucleotides, Their Binding Properties to Stimulator of Interferon Genes Adaptor Protein, and Structure/Activity Correlations

Article abstract of DOI:10.1021/acs.jmedchem.9b01062

Cyclic dinucleotides are second messengers in the cyclic GMP-AMP synthase (cGAS)-stimulator of interferon genes (STING) pathway, which plays an important role in recognizing tumor cells and viral or bacterial infections. They bind to the STING adaptor protein and trigger expression of cytokines via TANK binding kinase 1 (TBK1)/interferon regulatory factor 3 (IRF3) and inhibitor of nuclear factor-κB (IκB) kinase (IKK)/nuclear factor-κB (NFκB) signaling cascades. In this work, we describe an enzymatic preparation of 2′-5′,3′-5′-cyclic dinucleotides (2′3′CDNs) with use of cyclic GMP-AMP synthases (cGAS) from human, mouse, and chicken. We profile substrate specificity of these enzymes by employing a small library of nucleotide-5′-triphosphate (NTP) analogues and use them to prepare 33 2′3′CDNs. We also determine affinity of these CDNs to five different STING haplotypes in cell-based and biochemical assays and describe properties needed for their optimal activity toward all STING haplotypes. Next, we study their effect on cytokine and chemokine induction by human peripheral blood mononuclear cells (PBMCs) and evaluate their cytotoxic effect on monocytes. Additionally, we report X-ray crystal structures of two new CDNs bound to STING protein and discuss structure-activity relationship by using quantum and molecular mechanical (QM/MM) computational modeling.

Full text of DOI:10.1021/acs.jmedchem.9b01062

Subscriber access provided by - Access paid by the | UCSF Library
Article
Enzymatic Preparation of 2‘-5‘, 3‘-5‘Cyclic Dinucleotides, Their Binding
Properties to STING Adaptor Protein, and Structure/Activity Correlations
Barbora Novotná, Lenka Vaneková, Martin Zav#el, Miloš Bud#šínský, Milan Dejmek,
Miroslav Smola, Ondrej Gutten, Zahra Aliakbar Tehrani, Markéta Pimková Polidarová,
Andrea Brázdová, Radek Liboska, Ivan Št#pánek, Zdenek Vavrina, Tomáš Jandušík, Radim
Nencka, Lubomír Rulíšek, Evžen Bou#a, Ji#í Brynda, Ond#ej Páv, and Gabriel Birkus
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.9b01062 • Publication Date (Web): 12 Nov 2019
Downloaded from pubs.acs.org on November 13, 2019
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a service to the research community to expedite the dissemination
of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in
full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully
peer reviewed, but should not be considered the official version of record. They are citable by the
Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore,
the “Just Accepted” Web site may not include all articles that will be published in the journal. After
a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web
site and published as an ASAP article. Note that technical editing may introduce minor changes
to the manuscript text and/or graphics which could affect content, and all legal disclaimers and
ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or
consequences arising from the use of information contained in these “Just Accepted” manuscripts.
is published by the American Chemical Society. 1155 Sixteenth Street N.W.,
Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 46
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 46
1
2
3
4
5
6
7
8
9
Enzymatic Preparation of 2‘-5‘, 3‘-5‘Cyclic
Dinucleotides, Their Binding Properties to STING
Adaptor Protein, and Structure/Activity Correlations
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Barbora Novotná1,2, Lenka Vaneková1,2, Martin Zavřel1, Miloš Buděšínský1, Milan Dejmek1,
Miroslav Smola1, Ondrej Gutten1, Zahra Aliakbar Tehrani1, Markéta Pimková Polidarová1,2,
Andrea Brázdová1, Radek Liboska1, Ivan Štěpánek1, Zdeněk Vavřina1,2, Tomáš Jandušík1,3, Radim
Nencka1, Lubomír Rulíšek1, Evžen Bouřa1, Jiří Brynda1, Ondřej Páv1 and Gabriel Birkuš1,*
1 Institute of Organic Chemistry and Biochemistry of the Czech Academy of Sciences and Gilead
Sciences Research Centre at IOCB, Flemingovo nam. 2, Prague, 16610, Czech Republic
2 Faculty of Science, Charles University, Prague, Czech Republic
3 Faculty of Food and Biochemical Technology, University of Chemistry and Technology,
Prague, Czech Republic
1
ACS Paragon Plus Environment

Page 3 of 46
Journal of Medicinal Chemistry
1
2
3
4
ABSTRACT
5
6
7
8
9
Cyclic dinucleotides are second messengers in the cGAS-STING pathway which plays an
important role in recognizing tumor cells and viral or bacterial infections. They bind to the STING
adaptor protein and trigger expression of cytokines via TBK1/IRF3 and IKK/NFκB signaling
cascades. In this report we describe an enzymatic preparation of 2’-5’, 3’-5’ cyclic dinucleotides
(2’3’CDNs) with use of cyclic GMP-AMP synthases (cGAS) from human, mouse and chicken. We
profile substrate specificity of these enzymes by employing a small library of NTP analogues and
use them to prepare thirty-three 2’3’CDNs. We also determine affinity of these CDNs to five
different STING haplotypes in cell-based and biochemical assays, and describe properties needed
for their optimal activity toward all STING haplotypes. Next, we study their effect on cytokines
and chemokines induction by human peripheral blood mononuclear cells (PBMC) and evaluate
their cytotoxic effect on monocytes. Additionally, we report X-ray crystal structures of two new
CDNs bound to STING protein and discuss structure activity relationship by using quantum and
molecular mechanical (QM/MM) computational modeling.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
2
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 46
1
2
3
4
INTRODUCTION
5
6
7
8
Cyclic dinucleotides (CDNs) play an important role as second messengers in vertebrates and
prokaryotes.1–3 They are naturally synthesized from NTPs by cyclic dinucleotide synthases.
Prokaryotic and vertebrate CDNs differ in the type of phosphodiester linkages connecting two
nucleotide monophosphates together. While bacteria produce 3’3’CDNs (e.g. 3’3’c-diGMP,
3’3’c-diAMP, 3’3’cGAMP) with two 3’-5’ phosphodiester bonds,4–6 only one CDN with mixed
2’-5’ and 3’-5’ phosphodiester linkage (2’3’cGAMP) is found in mammalian cells.7
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
2’3’cGAMP is a product of cyclic-GMP-AMP (cGAMP) synthase which is evolutionary
conserved from fish to human.2,8,9 The enzyme belongs to a family of DNA sensors and detects
dsDNA released into cytosol during pathogen infection or disruption of host homeostasis.10 Binding
of dsDNA to cGAS occurs in a sequence-independent manner and induces a conformational change
ultimately allowing synthesis of 2’3’cGAMP from ATP and GTP.11 The 2’3’ and 3’3’ CDNs bind
to an adaptor protein called Stimulator of Interferon Genes (STING, also denoted as TMEM173,
MITA, ERIS) residing in endoplasmic reticulum.12 STING is a 379 amino acid long protein
consisting of the N-terminal transmembrane domain, the C-terminal ligand-binding domain and
C-terminal tail (CTT).11,13 Upon a CDN binding, STING homodimer transforms from an open to
closed conformation and forms oligomers.8 This conformation change leads to the recruitment of
TBK1 kinase, trans-phosphorylation of CTT of STING and subsequent recruitment and
phosphorylation of the transcription factor IRF3.14–17 IRF3 then forms homodimers that translocate
to the nucleus where they trigger expression of type I and III interferons. In parallel, a crosstalk
between TBK1 and IKK kinases leads to NFB activation and TNF and IL-6 expression.18
cGAS-STING pathway plays an essential role in host defense against invading pathogens and in
immune surveillance of tumor cells.12,19 Its importance from the perspective of drug discovery is
3
ACS Paragon Plus Environment

Page 5 of 46
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
highlighted by the fact that two cyclic dinucleotides ADU-S100 and MK-1454 are currently
undergoing evaluation in Phase I clinical trials for the treatment of cancer.20,21 Moreover CDNs
could also find utility in the treatment of viral infections such as chronic hepatitis B, human
immunodeficiency virus (HIV ) infection or as adjuvants in vaccines.22,23
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Several synthetic routes for 2’3’ and 3’3’ CDN preparations have been established.24–26 However,
these are usually complex multistep syntheses. Meanwhile, an elegant idea of using dinucleotide
synthases for CDN preparation was pursued and few 3’3’ and 2’3’CDNs were thus enzymatically
prepared from the commercially available ATP and GTP analogues.7,27–30
The objective of this study is to identify the most suitable enzyme for synthesis of 2’3’ CDNs,
characterize binding properties of prepared CDNs toward STING adaptor protein and determine
structure activity relations between STING and CDNs. To achieve this goal, we profile substrate
specificity of chicken, mouse and human cGAS using a library of forty-one NTP analogues and
describe preparation of thirty-three 2’3’CDNs. We characterize affinity of prepared CDNs toward
five STING haplotypes in cell-based and biochemical assays. We also study their effect on
induction of cytokines and chemokines by employing human peripheral blood mononuclear cells
(PBMC) and evaluate their cytotoxic effect specifically on monocytes. Finally, we also discuss a
structure–activity relationship (SAR) between prepared molecules and their biological activity by
employing state-of-the-art methods of computational chemistry, such as QM/MM calculations.
4
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 46
1
2
3
4
RESULTS
5
6
Substrate specificity of human, mouse and chicken cGAS
7
8
9
Binding of DNA to cGAS triggers rearrangement of its active site and catalysis of 2’-5’
phosphodiester bond between 2’-OH of GTP and -phosphate of ATP.11,29 The formed
intermediate AMP-2’-GTP flips in the active site of the enzyme and 3’-5’ phosphodiester bond
between 3’-OH group of AMP and -phosphate of GTP is formed.11,29 Despite being responsible
for the synthesis of the same product, cGAS from different species have relatively low amino acid
homology. For example, human and mouse or human and chicken cGAS share 56% and 51%
amino acid identity, respectively, based on BLAST® alignment31 of amino acid sequences
Q8C6L5, Q8N884 (UNIPROT) and XP_419881.4 (NCBI). Because of low level homology of
cGAS from these three species, we decided to express them and profile their substrate specificity
with a goal to identify the most suitable enzyme for synthesis of CDNs. We performed a small-scale
synthesis of CDNs in such a way that one of the NTPs was always either ATP or GTP and the other
an NTP analogue 1-41 (Tables 1A-1B).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
For the arbitrarily chosen cut-off of >25% reaction conversion, twenty-eight CDNs could be
prepared using mouse cGAS (mcGAS) compared to sixteen and eleven CDNs that could be made
with human (hcGAS) and chicken (ccGAS) cGAS, respectively (Tables 1A-1B). Thus, mouse
cGAS shows the lowest substrate specificity and is the most suited for enzymatic synthesis of
CDNs. Nevertheless, human and chicken cGAS should not be disregarded. For example, human
cGAS more efficiently catalyzed the reaction of GTP with 2-aminoadenosine-5’-triphosphate (3)
or isoguanosine-5’-triphosphate (4), and the conversion of ATP with O6-methylguanosine-5’-
triphosphate (16) to the appropriate CDN was only possible with human cGAS (Table 1A).
5
ACS Paragon Plus Environment

Page 7 of 46
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Table 1A. Substrate specificity of human, mouse and chicken cGAS toward NTP analogues.
R1
NH
O
N
N
N
N
N
6
N
N
7
N
5
4
1
2
O
O
O
O
O
O
O
O
O
8
9
3
N
O
N
R4
O
P
HO
P
O
P
O
O
P
HO
P
O
P
O
O
P
HO
P
O
P
R2
N
N
NH2
O
O
O
R6
OH
OH
OH
OH
OH
OH
OH
OH
3'
R5
2'
R3
OH
OH
OH
OH
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
(1)
(2)
(3-31)
Reaction Conversion (%)a
GTP
ATP
NTP
1
R1
R2
R3
R4
R5
R6
hcGAS mcGAS ccGAS hcGAS mcGAS ccGAS
-
-
-
29
45
18
6
5
0
-
-
-
2
NH2
NH2
NH2
NH2
SH
NH2
OH
F
OH
OH
H
H
H
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
90
74
22
0
54
50
60
29
73
0
47
36
0
0
0
-
31
52
-
0
0
-
3
4
OH
H
H
H
H
H
H
H
H
H
H
H
H
OH
F
5
Cl
H
0
-
-
-
6
H
OH
H
78
0
19
0
-
-
-
7
Cl
H
-
-
-
8
NHCH3
SCH3
H
H
OH
OH
OH
OH
OH
OH
OH
OH
OH
H
67
100
90
91
95
82
9
-
-
-
9
H
67
48
-
-
-
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
NH2
NH2
NH2
OH
H
6
0
0
66
2
50
0
-
6
0
5
78
6
1
0
-
0
0
0
0
0
13
23
-
SH
Cl
91
0
100
26
0
71
0
OH
OH
0
0
OCH3
SCH3
NH2
NH2
NH2
NH2
NH2
NH2
NH2
NH2
NH2
NH2
OH
NH2
NH2
H
-
-
-
70
25
0
67
51
0
34
14
0
H
H
-
-
-
H
H
H
H
H
H
H
H
H
H
H
H
H
H
0
57
20
38
47
0
0
-
-
-
H
NH2
F
9
0
-
-
-
H
0
0
-
-
-
F
F
12
0
0
-
-
-
H
Cl
0
-
-
-
H
Br
0
0
0
-
-
-
H
I
0
0
0
-
-
-
H
OH
OH
OH
OH
OH
OCH3 OH
0
0
0
-
-
-
NH2
NH2
NH2
H
H
F
OH
OH
SH
SH
-
-
-
0
3
28
-
17
72
48
-
0
34
45
-
OH
-
-
-
OH
OH
OH
0
0
0
NH2
0
52
0
a 1mM ATP or GTP was reacted with 1mM NTP analogue in the presence of 5M human, mouse
or chicken full-length cGAS for 16 h at 37oC in 20l volume. Reaction conversions were
6
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 46
1
2
3
4
5
determined by HPLC using UV detection at 260 nm. Conversions are defined as follows; a ratio of
AUC of a CDN over the sum of AUCs of the CDN, NTPs and NDPs. - not determined.
6
7
8
9
As shown in Table 1A, N1-methylguanosine-5’-triphosphate (1) was a mcGAS substrate but N1-
methyladenosine-5'-triphosphate (2) was not. Nucleotide triphosphates with 2-amino (3), 2-
hydroxy (4), 2-fluoro (5 and 23), and 2-chloro (6) substitutions of adenosine were tolerated by
mcGAS as well as NTPs with 6-mercaptopurine (7), 6-N-methyladenine (9) and 6-
methylthiopurine (10) nucleobases; however, 2'-deoxy-6-chloropurine-5'-triphosphate (8) was not
a mcGAS substrate. NTPs with 2-aminopurine (11), 6-thioguanine (12), 2-amino-6-chloropurine
(13), xanthine (14), hypoxanthine (15) and 2-amino-6-methylthiopurine (17) nucleobases were
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Table 1B. Substrate specificity of human, mouse and chicken cGAS toward NTP analogues.
R1
O
O
X
N
N
N
N
N
6
NH2
7
5
4
NH2
1
2
O
O
O
O
O
O
Y
O
O
O
8
9
3
N
N
O
Z
R2
HO
P
O
P
O
P
O
P
HO
P
O
P
O
O
P
HO
P
O
P
O
NH2
O
O
O
OH
OH
OH
OH
OH
OH
OH
OH
OH
OH
(32-39)
OH
OH
OH
(40)
OH
OH
(41)
Reaction Conversion (%)a
GTP ATP
hcGAS hcGAS ccGAS hcGAS hcGAS ccGAS
NTP
32
33
34
35
36
37
38
39
40
41
R1
NH2
R2
H
X
CH
Y
Z
CH
CH
CH
N
N
N
N
N
N
N
N
C
94
96
83
-
-
-
OH
OH
NH2
NH2
H
CH
-
-
-
-
-
-
N-CH3
-
-
-
0
-
0
-
0
-
NH2
NH2
NH2
OH
N
N
N
N
C
29
0
58
0
24
0
H
C-N3
C-OH
C-OH
CH
-
-
-
H
37
-
0
0
-
-
-
NH2
NH2
-
-
0
24
-
0
62
-
0
41
-
OH
-
-
-
16
24
45
51
23
31
0
33
0
a 1mM ATP or GTP was reacted with 1mM NTP analogue in the presence of 5M human, mouse
or chicken full-length cGAS for 16 h at 37oC in 20l volume. Reaction conversions were
determined by HPLC using UV detection at 260 nm. Conversions are defined as follows; a ratio of
AUC of a CDN over the sum of AUCs of the CDN, NTPs and NDPs. - not determined.
7
ACS Paragon Plus Environment

Page 9 of 46
Journal of Medicinal Chemistry
1
2
3
4
5
good mcGAS substrates. As mentioned above mcGAS did not tolerate O6-methyl-guanosine-5'-
triphosphate (16).
6
7
8
9
It was also possible to prepare CDNs by employing 7-deazaadenosine-5'-triphosphate (32) and
7-deazaguanosine-5'-triphosphate (33), but not by the use of N7-methylguanosine-5'-triphosphate
(34) (Table 1B). 8-Azaadenosine-5'-triphosphate (35) was a substrate for mcGAS, but 8-azido-
adenosine-5'-triphosphate (36), 8-oxo-adenosine-5'-triphosphate (37) and 8-oxo-guanosine-5'-
triphosphate (38) were not. Interestingly, mcGAS could also catalyze synthesis of CDNs from
thienoguanosine-5'-triphosphate (39), ribavirin 5'-triphosphate (40) and 5-aminoimidazole-4-
carboxamide-1-β-D-ribofuranosyl-5’-triphosphate (AICAR-5'-triphosphate) (41).
Ara-adenosine-5’-triphosphate (18), 2’-deoxy-adenosine-5’-triphosphate (20), 2’-amino-2'-
deoxy- (21), and 2’-fluoro-2'-deoxy-adenosine-5’-triphosphates (22) were also good mcGAS
substrates; however, 2'-fluoro-2'-deoxy-ara-adenosine-5'-triphosphate (19), 2’-chloro- (24), 2’-
bromo- (25), and 2’-iodo-2'-deoxyadenosine-5’-triphosphates (26) were not accepted. Similarly,
3’-deoxy- (28) or 3’-fluoro- 3’-deoxy-guanosine-5’-triphosphate (29) were mcGAS substrates as
were guanosine-5'-O-(1-thiotriphosphate) (30) and adenosine-5'-O-(1-thiotriphosphate) (31)
(Table 1A).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Large scale synthesis of CDNs and profiling of their activity
Next, we initiated a larger scale synthesis of CDNs from NTP analogues. We attempted to
prepare about 2 mol of CDNs which was sufficient amount for their downstream biological and
physico-chemical characterization. Some of the CDNs from nucleotide-5’-triphosphates listed in
Tables 1A-1B could not be prepared in sufficient purity since nucleotide-5’-diphosphates or linear
intermediates often co-eluted with CDNs. In the end, we prepared thirty-three CDNs enzymatically
8
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 46
1
2
3
4
5
6
7
8
and in order to get more complete SAR, we also prepared four CDNs (CDN-29, 35-37) through
chemical synthesis (Table 2, 3, 4 and 6, for more details on synthesis of CDN-29, 35-37 see
Supplementary Information).
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
To profile the activity of CDNs we employed cell based reporter assays. These assays are based
on CDN/STING dependent expression of firefly luciferase from a reporter plasmid with four ISRE
sites placed upstream of the firefly luciferase reporter gene minimum promoter. The luciferase
expression is induced by IRF3 transcription factor as a result of STING signaling cascade
activation.32 We have developed the reporter assays for five STING haplotypes that are present in
humans33 and run them in two assay formats. In digitonin assay we performed permeabilization of
cellular membranes with a detergent (digitonin A). As a consequence, the uptake of negatively
charged CDNs into cells was not the limiting step in STING activation. In standard assay, the
detergent was absent and the activity of CDNs was therefore influenced by efficiency of their
cellular uptake. Finally, our biochemical assay (differential scanning fluorimetry, DSF) determined
difference between melting temperature of STING protein with and without ligand (ΔTm) and was
an indirect measure of CDNs binding affinity to STING protein. Tm values from the DSF assay
correlated better with log EC50 values from digitonin (Pearson r = -0.73, P (two-tailed) < 0.0001)
than with log EC50 values from standard assay (Pearson r = -0.56, P (two-tailed) 0.001) (SI Figure
S1). EC50 values were calculated from dose response curves of individual CDN treated cells in
digitonin or standard assay. Representative dose response curves are shown in Figure S2. To
compare results of presented cell-based and biochemical assays for all prepared CDNs see Table
2, 3, 4, and 6.
9
ACS Paragon Plus Environment

Page 11 of 46
Journal of Medicinal Chemistry
1
2
3
4
Structure and conformation of CDNs in water solution.
5
6
1
31
19
All prepared CDNs (CDN 1 – 37) were characterized by H, P and F NMR spectra and
7
8
31
high-resolution MS. The NMR data are summarized in Tables S1a, S1b, S2, S3, S4, S5. P
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
chemical shifts clearly distinguished the presence of thiophosphate group ( 52-57 ppm in CDNs
19
33 – 37) from phosphate group ( +1.5 to -1.5 ppm in all other CDNs 1 - 32). F signals in the
range  -196 to -200 ppm proved fluorine substituent on furanose ring (CDNs 1 – 8) while signal
around  -50 showed CDNs with fluorine substituent in position 2 of nucleobase (CDNs 6 and 7).
Basic conformation features of CDNs can be derived from vicinal coupling constants and observed
homonuclear NOEs. Furanose rings of CDNs with ribose sugar units and a 2’-phosphate group
adopted C2’-endo conformation (characterized by large J(1’,2’) = 7 – 9 Hz and small J(3’,4’) = 0
- 1.5 Hz) while rings with 3’-phosphate group preferred C3’-endo conformation (manifested by
very small J(1’,2’) = 0 – 1.5 Hz and large J(3’,4’) = 7 - 9 Hz). Orientation of the nucleobase was
found as anti- throughout the whole series of our CDNs as indicated by characteristics observed
from NOE contacts: between H-8 and protons H-1’, H-2’ in 2’-phospate units and between H-8
and protons H-1’, H-2’, H-3’ in 3’-phosphate units. The structure modifications of the nucleobase
had only a very small effect on the conformation of furanose ring. For conformation analysis of
phosphate and thiophosphate linkages of CDNs there was not a complete set of proper NMR
parameters accessible. While torsion angles  and  were still related to vicinal heteronuclear
couplings J(P,H) observable in 1H NMR spectra and J(P,C) obtainable from 13C NMR spectra, for
torsion angles  and  such usable couplings were missing.
2’ and 3’ substituted 2’3’CDNs
10
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 46
1
2
3
4
5
6
7
8
9
We were able to prepare fifteen 2’3’CDNs with substitutions at 2’-position of adenosine or 3’-
position of guanosine monophosphate units (Table 2). Compared to the parent 2’3’cGAMP, all but
one CDNs (CDN-11) had less than five-fold shift in EC50 values in the digitonin reporter assay
using cells expressing WT, HAQ and AQ STING haplotypes (Table 2). The important difference
appeared in the case of REF STING haplotype that is present in about 13% of humans.33,34
2’3’CDNs with 2’-amino-2’-deoxyadenosine (CDN-9), ara-adenosine (CDN-10), and some with
deoxyribonucleotides (CDN-13, -14) were at least 10-fold less active against REF than WT STING
haplotype (Table 2). Similarly, fluoro- (CDN-6, CDN-11) or chloro- (CDN-15) substitutions at C2
position of adenine resulted in decreased activity against the REF STING.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Despite log EC50 values from digitonin and Tm values from DSF assay correlated well (Figure
S1), clear outliers could be found (e.g. CDN-1 and -2 vs 2’3’cGAMP). This is probably due to
differences in efflux and/or intra/extracellular stability of these CDNs that are captured in digitonin
cell-based assay but not in DSF assay.
We also determined activities of the prepared CDNs in the standard assay using WT STING
293T reporter cells (Tables 2, 3, 4, 6). Due to the dramatic effect of the cellular uptake on the
activity of CDNs, EC50 values for all prepared CDNs were more than 100-fold higher compared to
EC50 values from the digitonin assay (Tables 2, 3, 4, 6).
Throughout the study, we attempted to explain all experimentally observed binding affinities of
CDNs with STING by employing crystallography and computational modelling. For relative
comparison of binding affinity of CDNs to STING, we used Tm values from our biochemical DSF
assay. In order to highlight changes brought about by individual modifications of a ligand we
discuss Tm, defined as the difference between Tm of a ligand X and a reference ligand specified
11
ACS Paragon Plus Environment

Page 13 of 46
Journal of Medicinal Chemistry
1
2
3
4
5
6
in the text below. Positive values of Tm (X) indicate higher stability of protein-ligand complex
than in the case of a reference ligand.
7
8
9
Table 2. Activity of 2‘OH and 3‘OH substituted 2‘3‘CDNs in biochemical and cell-based assays.
O
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
N
NH
OH
P
O
O
N
N
NH2
O
R1
R2
2'
O
3'
R3
3'
2'
O
P
O
O
O
R4
N
N
3
9
2
1
4
5
OH
8
7
N
6
N
NH2
DSF
STANDARD
ASSAY
EC50 (µM) b
WT
DIGITONIN ASSAY
EC50 (µM) c
Tm (OC)a
WT
Compound
CDN-1
R1
F
R2
H
R3
OH
R4
H
WT HAQ REF AQ
Q
20.5
13.8
0.01 0.01
0.02 0.05
0.02 0.02
0.02 0.05
0.01 0.04
0.05 0.10
0.02 0.08
0.02 0.04
0.02 0.03
0.04 0.08
0.01 0.03 0.02
0.02 0.09 0.20
0.07 0.01 0.04
0.04 0.04 0.18
0.07 0.02 0.06
2.00 0.17 0.70
0.07 0.03 0.08
0.17 0.03 0.13
1.40 0.02 0.24
0.40 0.04 0.09
6.40 0.06 0.60
0.14 0.03 0.34
3.30 0.03 0.04
0.40 0.03 0.12
4.70 0.09 0.96
0.07 0.04 0.05
0.21 0.03 0.17
4.30 0.26 2.06
CDN-2
CDN-3
F
OH
F
H
H
H
H
H
H
H
H
F
F
H
H
H
H
F
20.3
16.2
17.4
16.6
10.7
14.7
14.4
9.4
13.3
16.7
10.7
15.1
>300
11.0
27.6
107.5
nd
CDN-4
H
CDN-5
H
F
CDN-6
H
F
CDN-7
F
OH
F
CDN-8
F
OH NH2
CDN-9
NH2
H
OH
H
H
F
CDN-10
CDN-11
CDN-12
CDN-13
CDN-14
CDN-15
2'3'-cGAMP
2'2'-cGAMP
3'3'-cGAMP
3'3'c-diAMP
3'3'c-diGMP
OH OH
OH OH
17.0
12.3
15.8
13.5
13.2
9.5
H
186.8
31.9
75.4
21.6
58.2
13.7
39.2
70.1
14. 9
184.2
0.20
0.1
H
H
H
H
H
OH
H
H
H
H
Cl
0.02 0.06
0.06 0.03
0.01 0.03
0.07 0.08
0.02 0.02
0.03 0.02
0.12 0.12
H
OH
H
H
OH
15.3
11.6
5.1
2.5
0.3
4.5
0.2
0.9
>45
0.2
1.1
7.1
2.4
8.40
>45
a Results of differential scanning fluorimetry assay performed with WT STING haplotype. Tm
values are the mean of two independent double determinations. b Results of standard assay in 293T
reporter cells expressing WT STING haplotype. c Results of digitonin assay in 293T reporter cells
expressing different STING haplotypes. EC50 values are the mean of two independent experiments
measured in triplicates with standard deviations <50% of EC50 values.
12
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 46
1
2
3
4
5
6
7
8
Several modifications of the sugar ring on both nucleotides have been examined. For the 3'
nucleotide, these modifications include substitution of the 2'-hydroxyl with 2'-fluorine (CDN-1),
hydrogen (i.e. deoxyribose, CDN-12), amine group (CDN-9), or changed stereochemistry of 2'
carbon (i.e. arabinose, CDN-10).
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Of these modifications, CDN-9 forms the least stable ligand-protein complex. The amine group
of the unbound ligand is expected to be protonated. However, it is improbable that the binding site
would favor presence of a positively charged substituent. Thus, binding of this ligand is expected
to incur a desolvation penalty, which in turn leads to decreased thermal stability of the protein-
ligand complex (ΔTm(CDN-9) = +9.4 ºC, i.e. ΔΔTm = -5.9 ºC with respect to 2’3’cGAMP (see
Table 2)).
According to QM/MM models, the arabinose modification (CDN-10) is accommodated by the
2' hydroxyl interaction with Arg238. This differs from 2’3’cGAMP-STING complex where this
amino acid is primarily in contact with the phosphate group.8 In the case of CDN-10 the phosphate
is displaced and interacts with Arg232 and the 2' hydroxyl via intramolecular hydrogen bonding.
Thus, the arabinose modification is well tolerated by WT STING which is reflected by ΔΔTm(CDN-
10) = +1.7 ºC with respect to 2’3’cGAMP .
In the crystal structure of 2’3’cGAMP with STING, the environment surrounding 2'-hydroxyl
includes both hydrophobic and hydrophilic residues and a nearby water molecule (Figure 1). The
amphipathic character of this region contributes to the ability of the protein to accommodate less
hydrophilic substitutions of 2'-hydroxyl, e.g. hydrogen (i.e. deoxyribose, ΔΔTm(CDN-12) = +
0.5 ºC or fluorine, ΔΔTm(CDN-1) = +5.2 ºC, both in respect to 2’3’cGAMP). On the other hand,
the environment around the 3'-hydroxyl of the 2'-nucleotide is significantly more hydrophilic.
Thus, these same modifications (i.e. hydrogen or fluorine substitution in the 3'-position of the
13
ACS Paragon Plus Environment

Page 15 of 46
Journal of Medicinal Chemistry
1
2
3
4
5
6
2'-nucleotide) are less favorable than their 2’-position counterparts, with ΔΔTm(CDN-3) = +0.9 ºC
and ΔΔTm(CDN-14) = -1.8 ºC compared to 2’3’cGAMP (Table 2).
7
8
9
While the above explanations help to rationalize some of the observed differences, it is
remarkable that the 2’-fluorine (CDN-1) and arabinose (CDN-10) modifications actually lead to
higher ΔTm with respect to the native ligand. One can speculate that this improvement in STING
binding is not due to their specific interaction with the protein but an inherently lower energetic
cost for these ligands to adapt the STING bound conformation.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 1. Amphipathic pocket around the 2’ position of 3’ nucleotide.
The environment includes both hydrophobic (Tyr163 and Pro264) and hydrophilic residues
(Thr263) and a water molecule. Hydrophobicity visualized by BIOVIA Discovery Studio35.
Structure obtained from a QM/MM model based on PDB 6T1X. Only the relevant residues and the
ligand are shown in stick representation.
14
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 46
1
2
3
4
5
6
7
8
9
By combining QM/MM models and conformational sampling of these two ligands we can
estimate this energetic penalty (see methods). Indeed, it turns out that both CDN-1 and CDN-10
require 4.7 and 4.6 kcal.mol-1 less of (free) energy, respectively, than 2’3’cGAMP to change the
conformation from unbound state in solvent to that found in the protein. These estimates are
qualitative but serve to illustrate that the ligand may be optimized not only by its interactions with
the protein but also via minimizing its conformational strain.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
The 2'- and 3'-hydroxyl groups of the native ligand are more than 7 Å apart. Based on the crystal
structure of CDN-1 with WT STING, substituting 2’ hydroxyl with fluorine does not lead to any
significant change in the binding mode. We presume the same is the case for substitution with
hydrogen (CDN-12) and analogous substitutions of the 3’ hydroxyl (CDN-3 and CDN-14). Thus,
it is not surprising that combined modifications result in additive changes of ΔΔTm. Indeed,
measurement of ΔΔTm for disubstituted ligands CDN-2, CDN-4, CDN-5 and CDN-13 give values
that can also be estimated by simply summing up the ΔΔTm values of monosubstituted ligands. The
error of such estimate is within 1.1 ºC, which is within the 3 times standard deviation of DSF
measurement. Further examples of such additivity or lack thereof will be discussed below and may
be used as an argument for cooperative effects of substitutions in different parts of a ligand.
2’3’CDNs with substituted nucleobases
2’3’CDNs with guanosine subunit combined with 2-aminoadenosine (CDN-16), N6-
methyladenosine (CDN-17), inosine (CDN-18), 2-aminopurineriboside (CDN-21) or 7-
deazaadenosine (CDN-24) (Table 3), and 2’3’CDNs with adenosine subunit combined with N1-
methylguanosine (CDN-28), inosine (CDN-29), 7-deazaguanosine (CDN-30), and
thienoguanosine (CDN-31) (Table 4) retained comparable activities in WT, HAQ and AQ STING
digitonin reporter assays as 2’3’cGAMP. Compared to 2’3’cGAMP, 2’3’CDN with xanthosine-
15
ACS Paragon Plus Environment

Page 17 of 46
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
MP (CDN-19), 6-thio-IMP (CDN-20), 6-methylthio-GMP (22), 8-aza-AMP (CDN-25) and
AICAR-MP (CDN-26 and -32) had substantially diminished activity against all STING haplotypes
(Table 3 and 4). Moreover, with exemption of CDN-29, all prepared 2’3’CDNs had significantly
impaired activity toward REF STING haplotype in digitonin assays (Table 3 and 4). Similar to 2’
and 3’ substituted 2’3’CDNs, CDN-16 to 32 were at least two order of magnitude less potent in
standard than digitonin assay (Table 3 and 4).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Table 3. Effects of nucleobase modifications at 3’-nucleotide on CDN activity in biochemical and
cell-based assays.
O
N
O
N
N
N
N
N
NH
NH
OH
P
OH
P
O
O
O
O
NH2
NH2
O
O
O
O
OH
2'
OH
3'
2'
3'
OH
OH
O
P
O
P
O
O
O
O
O
O
H2N
R2
N
N
N
3
9
4
5
2
1
OH
OH
8
Z
7
H2N
N
6
N
X
O
CDN-26
R1
DSF
Standard Assay
EC50 (µM) b
WT
Digitonin Assay
EC50 (µM) c
WT HAQ REF AQ
Tm (OC) a
WT
10.0
9.0
Compound
CDN-16
CDN-17
CDN-18
CDN-19
CDN-20
CDN-21
CDN-22
CDN-23
CDN-24
CDN-25
CDN-26
R1
NH2
NHCH3
OH
OH
SH
R2
NH2
H
H
OH
H
NH2
NH2
H
H
H
X
Z
Q
N
N
N
N
N
N
N
N
CH
CH
CH
CH
CH
CH
CH
CH
35.9
63.7
148.8
>300
182.1
53.9
>300
13.3
66.4
>300
0.02
0.02
0.05
1.80
0.30
0.02
0.30
0.01
0.01
0.23
0.32
0.02
0.05 1.10 0.04 0.31
0.06 0.70 0.06 1.20
0.30 2.10 0.14 2.20
1.70 20.9 2.00 32.8
0.20 1.70 0.15 8.30
0.04 1.60 0.04 0.70
0.48 27.6 0.24 8.50
0.06 2.30 0.06 1.23
0.02 0.28 0.01 0.16
0.31 34.2 0.21 5.72
8.8
2.4
7.3
9.2
3.5
8.3
11.5
3.9
1.7
H
SCH3
SCH3
NH2
NH2
CH CH
N
N
>300
13.7
0.21
>45 0.21 45.3
2'3'-cGAMP
15.3
0.02 0.07 0.04 0.05
a Results of differential scanning fluorimetry assay performed with WT STING haplotype. Tm
values are the mean of two independent double determinations. b Results of standard assay in 293T
reporter cells expressing WT STING haplotype. c Results of digitonin assay in 293T reporter cells
expressing different STING haplotypes. EC50 values are the mean of two independent experiments
measured in triplicates with standard deviations <50% of EC50 values.
16
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 46
1
2
3
4
5
6
7
8
9
CDNs are anchored to STING protein through Arg238 and Tyr167 which stack via π - π and
cation - π interactions from both sides of nucleobases (Figure 2). The functional groups at C6
position on CDNs are without direct contacts with the protein. Instead, their interactions with
STING are mediated through a network of water molecules to backbone groups of residues 238 to
241, which are located in the lid region (Figure 3) of STING protein. This water network might be
crucial for understanding the effect of base modification on the activity of CDNs.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Table 4. Effects of nucleobase modifications at 2’-nucleotide on CDN activity in biochemical and
cell-based assays.
O
O
O
R2
X
6
N
7
NH
OH
P
1
2
5
OH
P
N
N
NH2
OH
P
S
8
4
9
3
N
O
O
O
O
N
O
O
R1
N
NH2
O
NH2
O
O
O
OH
2'
O
3'
OH
2'
3'
O
OH
2'
3'
2'
3'
2'
3'
2'
3'
OH
OH
O
P
O
P
OH
O
P
O
O
O
O
O
O
O
O
O
N
N
N
N
N
N
OH
OH
OH
N
N
N
N
N
N
NH2
CDN-31
CDN-32
NH2
NH2
DSF
Standard Assay
EC50 (µM) b
WT
Digitonin Assay
EC50 (µM) c
Tm (OC) a
WT
Compound R1 R2
CDN-27 OH
X
NH
WT
0.13
HAQ
0.19
REF
5.80
AQ
0.18
Q
H
7.6
>300
0.72
0.78
0.02
1.10
3.00
47.20
0.05
CDN-28 NH2 CH3 NH
9.0
16.1
6.6
40.0
29.9
0.02
0.02
0.04
0.11
0.83
0.02
0.04
0.03
0.10
0.25
0.40
0.02
0.90
0.04
3.50
6.00
>45
0.07
0.04
0.01
0.06
0.05
0.42
0.04
CDN-29
H
H
H
NH
CH
CDN-30 NH2
CDN-31
40.4
2.8
240.6
>300
13.7
CDN-32
0.6
2'3'-cGAMP
15.3
a Results of differential scanning fluorimetry assay performed with WT STING haplotype. Tm
values are the mean of two independent double determinations. b Results of standard assay in 293T
reporter cells expressing WT STING haplotype. c Results of digitonin assay in 293T reporter cells
expressing different STING haplotypes. EC50 values are the mean of two independent experiments
measured in triplicates with standard deviations <50% of EC50 values.
17
ACS Paragon Plus Environment

Page 19 of 46
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 2: Interactions of purine bases.
Purine bases of a CDN stack with Tyr167 and Arg238 residues. Detail from the structure of CDN-1
with human WT STING (PDB: 6S27). The π -π interaction shown in dashed-pink line and cation-
π interaction shown in dashed-orange line, respectively.
Information about the positions of water molecules is difficult to interpret due to symmetry
considerations. A 2’3’ cyclic dinucleotide allows for two equivalent binding modes to a symmetric
STING homodimer. Even if the water molecules are resolved in a crystal structure, the averaging
of their positions poses a challenge for a fundamentally asymmetric ligand interpretation. In the
following, we rely on the ability of QM/MM calculations to provide detailed structural information
concerning water positions. Moreover, most of the ligands presented herein contain modification
of the adenine base (with respect to the native 2’3’cGAMP ligand) and hence, the discussion of
water environment is focused on this region.
18
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 46
1
2
3
4
5
6
7
8
9
Figure 3 shows the proposed network of water molecules. This configuration not only provides
an ideal number of interactions for all water molecules in the network, but also accommodates the
asymmetry of the ligand, i.e. hydrogen bond acceptor at position 6 of guanine and hydrogen bond
donor at position 6 of adenine. This configuration agrees well with water molecules resolved in a
crystal structure of STING in complex with CDN-24. There is one less water molecule in the
QM/MM model in the vicinity of position 2 of 3’ nucleotide due to a different orientation of Ile235
side-chain in the crystal structure and in the QM/MM model.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 3. Proposed local water network near positions 6 of 2’3cGAMP.
Water-mediated hydrogen bond network between the purine bases of 2’3’cGAMP and the lid
region backbone groups of residues 238 (green), 239 (yellow), 240 (purple), and 241 (pink) located
in β3/ β4 strands of human WT STING’s lid domain. Four water molecules together with position
6 substituents of purine bases form an asymmetrical chain of hydrogen donors and acceptors.
Structure was obtained from a QM/MM model based on PDB 6T1X. Only the relevant residues
and the ligand are shown in stick representation.
19
ACS Paragon Plus Environment

Page 21 of 46
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
We speculate that this ideal setup is sensitive to modification of nearby functional groups.
Changing a substituent at the position 6 of adenine to an electron acceptor, such as oxygen (CDN-
18) or sulfur (CDN-20), leads to ΔΔTm ≈ -7 ºC with respect to 2’3’cGAMP. Bulkier substituents
like 6-methyl-amino (CDN-17) and 6-methylthio (CDN-23) are also disruptive, probably due to
the decreasing number of water molecules around these ligands. Regardless of the nature of the
methylated substituent, this disruption also leads to ΔΔTm ≈ -7 ºC. On the other hand, substitution
of NH2 group with hydrogen is accommodated without incurring a penalty to ΔTm (compare
CDN-16 and CDN-21).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Shifting the focus to substitutions on position 2, it seems that any substituent other than the native
hydrogen incurs a penalty of a ΔΔTm ≈ -5 ºC relative to an unsubstituted CDN, as evidenced by a
number of cases, including amine substitutions (compare ΔTm values of CDN-16 and 2’3’cGAMP;
CDN-22 and CDN-23; CDN-1 and CDN-8), oxygen (compare ΔTm values of CDN-18 and CDN-
19), fluorine (compare ΔTm values of CDN-10 and CDN-11, CDN-1 and CDN-7, CDN-5 and
CDN-6), and chlorine (CDN-15 and CDN-12).
Note the approximate penalties mentioned in the previous paragraphs, i.e. ΔΔTm ≈ -7 ºC for an
electron acceptor or methylated substituent in position 6 and ΔΔTm ≈ -5 ºC for any substituent at
position 2 of the 3' nucleotide. These two very simple rules are sufficient to estimate measured ΔTm
of all 14 ligands with substituted purine base of 3' nucleotide in our set for which Tm span from
+2.4 to +14.9 ºC to within 1.3 ºC (with mean absolute deviation of 0.7 °C), which is well within
the experimental error. The above QSAR analysis is summed up in Table 5.
20
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 46
1
2
3
4
Table 5. QSAR of ligand modifications.
5
6
7
8
9
Position
Native substituent
Modified substituent
DSFa
ΔΔTm (OC)
6 (3’nucleotide)
6 (3’nucleotide)
2 (3’nucleotide)
2’ (3’nucleotide)
2’ (3’nucleotide)
3’ (2’nucleotide)
3’ (2’nucleotide)
NH2
NH2
H
OH
OH
OH
OH
Hydrogen-bond acceptor
-7
-7
-5
+5
+0.5
+1
-2
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
methylated
any
F
H
F
H
a Tm, defined as the difference between Tm of a ligand X and reference ligand.
Thiophosphate 2’3’CDNs
We were able to prepare enzymatically two thiophosphate 2’3’CDNs. Mono-thiophosphate
2’3’CDN (CDN-33) is about 3-7-fold less active against all STING haplotypes than 2’3’cGAMP.
On the other hand, bis-thiophosphate CDN (CDN-34) retains similar activity toward all STING
haplotypes as the parent 2’3’cGAMP (Table 6). Since the enzymatic cyclization afforded only one
diastereomer out of four possible, we prepared bis-thiophosphate versions of 2’3’cGAMP through
chemical synthesis, which is detailed in Supplementary Methods. As shown in Table 6, none of the
prepared compounds had activity profile similar to the enzymatically prepared CDN-34. Moreover,
1H-NMR data of CDN-34 differed from those for the chemically prepared diastereomers CDN-35,
-36 and -37, indicating that CDN-34 is the diastereomer which was not obtained during chemical
synthesis of CDN-35, -36 and -37.
21
ACS Paragon Plus Environment

Page 23 of 46
Journal of Medicinal Chemistry
1
2
3
4
Table 6. The activity of thiophosphate 2’3’CDNs in biochemical and cell-based assays.
5
6
O
7
8
9
N
NH
R2
P
O
O
N
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
N
NH2
O
O
OH
2'
3'
3'
2'
OH
O
P
O
O
O
N
N
R1
N
N
NH2
DSF
Standard Assay
EC50 (µM) b
WT
Digitonin Assay
EC50 (µM) c
HAQ
0.11
Tm (oC) a
WT
Compound
CDN-33
R1
OH
R2
WT
0.06
REF
0.6
AQ
0.12
Q
0.14
SH
SH
SH
SH
SH
13.8
nd
CDN-34
CDN-35
SH
SH
SH
SH
14.6
13.6
15.1
14.8
15.3
70.7
7.0
0.03
0.09
0.02
0.03
0.02
0.02
0.13
0.02
0.03
0.02
0.01
0.7
0.03
0.14
0.03
0.03
0.04
0.03
0.19
0.07
0.08
0.05
CDN-36
120.6
1.0
0.2
CDN-37
0.15
0.07
2'3'-cGAMP
13.7
a Results of differential scanning fluorimetry assay performed with WT STING haplotype. Tm
values are the mean of two independent double determinations. b Results of standard assay in 293T
reporter cells expressing WT STING haplotype. c Results of digitonin assay in 293T reporter cells
expressing different STING haplotypes. EC50 values are the mean of two independent experiments
measured in triplicates with standard deviations <50% of EC50 values.
QM/MM modelling of CDN-34 and its diastereoisomers (CDN-35, CDN-36, and CDN-37)
reveals very few differences. The native ligand (phosphate) and R configuration of the
thiophosphate exhibit identical binding modes. In case of S-configuration of the thiophosphate (i.e.
P=S bond pointing in direction of Arg238), the interaction is shared with both Arg232 and Arg238.
Hence, we speculate that (S,S) configuration is the one that would be most affected in REF variant
of the protein (i.e. R232H allelic form). However, this difference alone is not sufficient for
explaining all the observed differences in activities.
22
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 46
1
2
3
4
The structures of WT STING with CDN-1 and -24
5
6
7
8
By employing our optimized protocol36, we were able to obtain crystal structures of human WT
STING in complex with CDN-1 and -24 with final resolutions of 3Å (I/σ = 2) and 2.2Å (I/σ = 2),
respectively. The crystal of STING in complex with CDN-1 belonged to the tetragonal P41212 and
in complex with CDN-24 to the orthorhombic P21212 space group (Tables S6 and S7). The
alignment of the structure of the WT STING in complex with CDN-1 or CDN-24 with structure of
STING in complex with 2’3’cGAMP (PDB entry 4KSY) revealed slight differences in protein
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
A
B
C
Figure 4. Structure of human WT STING in complex with CDN-1 (6S27) and its comparison to
the structure of STING in complex with its natural ligand 2’3’cGAMP (4KSY).
A) Structure of human WT STING in complex with CDN-1 (6S27). The Fo-Fc map is contoured
at I/σ = 3. B) Alignment of 6S27 and 4KSY (the RMSD over C-alpha = 0.5690). C) Comparison
of 2’3’-cGAMP and CDN-1 bound to the STING. Coloring: Nitrogen – blue, Oxygen – red,
Fluorine – bright green, Phosphate – orange, Carbon – yellow (CDN-1) or dark green
(2’3’cGAMP).
23
ACS Paragon Plus Environment

Page 25 of 46
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
conformation (the RMSD over C-alpha 0.57 and 0.78, respectively) and the protein structures can
be therefore considered virtually identical (Figure 4B and 5B). The binding mode and conformation
of the CDN-1 and CDN-24 correspond well to the binding mode and conformation of 2’3’-cGAMP
(Figure 4C and 5C).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
The water network chain in the vicinity of position 6 of purine bases coincides with those
used in QM/MM calculation, which a posteriori confirms the validity of our previous structural
considerations on water networks. However, compared to QM/MM model, both X-ray structures
show an extra water molecule near position 1 due to different orientation of Ile235 side-chain. Also,
both crystal structures lack one water molecule near position 3 (Figure 3).
Figure 5. Structure of human WT STING in complex with CDN-24 (6S26) and its comparison to
the structure of STING in complex with its natural ligand 2’3’cGAMP (4KSY).
A) Structure of human WT STING in complex with CDN-24 (6S26). The Fo-Fc map is
contoured at I/σ = 3. B) Alignment of 6S26 and 4KSY (the RMSD over C-alpha = 0.7833).
24
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 46
1
2
3
4
5
C) Comparison of 2’3’-cGAMP and CDN-24 bound to the STING. Coloring: Nitrogen – blue,
Oxygen – red, Phosphate – orange, Carbon – light blue (CDN-24) or dark green (2’3’cGAMP).
6
7
8
9
Effect of 2’3’CDNs on cytokine induction and monocytes
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Activation of cGAS-STING pathway results in the induction of interferons and TNF secretion.37
Consistently, PBMC treated with CDNs prepared within this study induced high level of IFN,
IFN and TNF secretion into cultivation medium (Table 7, Table S8). Since CDNs were reported
to induce monocytic cytotoxicity in A2a Gαs Protein-Coupled Receptor dependent manner,37 we
tested the effect of CDNs 1 - 37 on viability of monocytes in PBMC using a single concentration
of 12.5µM. Similar to the published data, the majority of the prepared compounds caused profound
monocytic cytotoxicity without affecting CD3+ T cell population (Table 7, Table S8).
Table 7. Effect of CDNs on monocyte viability and cytokine induction in PBMC assay.
Compound
Monocyte
viability a
Cytokines
(fold of 2’3’cGAMP induced cytokines)b
INF 
1.3
5.1
4
TNF 
0.6
4.3
0.8
4.4
4.2
2.3
3.2
0.7
3.7
2.6
3.4
4.1
0.6
1.9
4.3
4.3
4.1
0.2
IFN 
0.9
0.5
0.6
0.3
0.3
0
CDN-1
CDN-2
CDN-3
CDN-4
CDN-5
CDN-6
CDN-7
CDN-8
CDN-9
CDN-10
CDN-11
CDN-12
CDN-13
CDN-14
CDN-15
CDN-16
CDN-17
CDN-18
1.7%
0.4%
0.7%
0.7%
0.5%
0.2%
0.4%
2.5%
3.0%
5.2%
3.0%
2.3%
1.6%
3.3%
0.1%
1.5%
0.8%
17%
6.2
5.6
0.1
4.6
3.1
2.7
3.2
3.2
3.3
1.6
5
0.4
0.8
0.4
1
2.4
0.5
1.3
0.6
0.1
0.7
0.5
0.8
0.8
3.6
3.3
0.5
25
ACS Paragon Plus Environment

Page 27 of 46
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
CDN-19
CDN-20
CDN-21
CDN-22
CDN-23
CDN-24
CDN-25
CDN-26
CDN-27
CDN-28
CDN-29
CDN-30
CDN-31
CDN-32
CDN-33
CDN-34
CDN-35
CDN-36
CDN-37
2’2’cGAMP
3’3’cGAMP
3’3’c-diGMP
2’3’cGAMP
45%
10%
0
0
0
1.8
2.4
1.1
1.7
1.2
2.9
0
0.7
4.1
3.2
4.2
3.3
3.8
0
0.7
0.5
0
0.6%
4.8%
0.4%
0.7%
2.3%
83%
0.2
0.3
0.1
0
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
8.3%
2.0%
1.0%
0.9%
0.9%
87%
3
0.6
0.5
0.4
3.5
1.6
0
0.7
0.7
1.9
0.3
2.4
0
3.3
1.2
5.9
4
0
9.4%
2.1%
1.8%
0.3%
0.8%
2.8%
17%
2.2
2.4
1
1.6
1.3
0.2
0.4
0.5
0.2
0.3
0
1.2
2
0.6
1.2
0.1
0.6
1.4
0
3.9
0.2
1.8
1
32%
0
1.1%
1
1
1
a Viability of monocytes in PBMC culture treated with 12.5µM CDN for 16 h. Viability of monocytes in
untreated control equals 100%. Values are the mean of three independent determinations from one PBMC
donor. b Levels of INF , INF  and TNF secreted by PBMC treated with 12.5µM CDN for 16 h relative
to levels secreted by PBMC treated with 12.5µM 2’3’cGAMP. Values are the mean of three independent
determinations from the same PBMC donor as in monocyte cytotoxicity assay; Amount of cytokines
induced by 2‘3’cGAMP treatment: Interferon γ: 5570 pg/ml , TNFα: 5552 pg/ml, Interferon α: 648 pg/ml.
26
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 46
1
2
3
4
DISCUSSION AND CONCLUSIONS
5
6
7
8
In this study, we show that 2’3’CDNs can easily be prepared from NTP analogues by employing
vertebrate cyclic dinucleotide synthase cGAS. We profiled substrate specificity of human, mouse
and chicken cGAS and we developed a general protocol for enzymatic synthesis of CDNs. With
few exceptions (e.g., O6-methyl-guanosine-triphosphate (16) which was an exclusive substrate of
human cGAS), the mouse enzyme turned out to be the most promiscuous and was used for
preparation of thirty-four CDNs. Principally, it should be possible to prepare more CDNs from our
small library of NTPs. Majority of CDNs in this study contained either AMP as the 3' nucleotide
or GMP as 2' nucleotide, and more CDNs where both nucleotides are analogues such as CDN-2,
-4, -5 can be envisioned.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
During our efforts to identify novel STING agonists,38 we were guided rather by the digitonin
than standard assay. As mentioned above, activity of CDNs is influenced by the efficiency of their
uptake into cells in the latter assay. Consequently, EC50 values for the same CDN are at least
100-fold higher in the standard than digitonin assay (Table 2, 3, 4 and 6). The uptake issue can be
ultimately solved by synthesis of lipophilic prodrugs of CDNs as presented in recent patent
literature.39
An ideal CDN for clinical use should have similar activity toward all STING haplotypes; thus,
eliminating the need for genotyping of patients. CDNs with adenine 3' nucleotide and guanine 2'
nucleotide containing 2’-F or 3’-F substitutions or both (CDN-1, -2, and -3), 2’-F and 3’-deoxy
groups (CDN-4) or 2’-deoxy and 3’-F substitutions (CDN-5) showed good activity across all
STING haplotypes. In contrast to the general notion, not all 2’3’CDNs are good REF STING
agonists.33,34,40 CDNs with ara-adenosine, 2’-amino-2’-deoxyadenosine as 3' nucleotide and some
of CDNs with 2’deoxyribonucleotides had diminished potency against REF STING. Substitutions
27
ACS Paragon Plus Environment

Page 29 of 46
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
of nucleobases almost always resulted in impaired activation of REF STING haplotype with the
only exception being CDN-29 where 2' nucleotide guanosine was replaced for inosine.
All prepared CDNs stimulated human PBMC to secret IFN, IFN and TNF, consistently with
their ability to activate cGAS-STING pathway. In agreement with previously published data37,
CDNs decreased viability of monocytes in PBMC cultures without affecting T-lymphocytes (Table
7, S8). However, we cannot exclude that some other factors such as PBMC purification, cell density
and/or cultivation conditions could exacerbate the negative effect of CDNs on monocytes.
Monocytic cytotoxicity seemed to be somewhat proportional to the levels of induced cytokines
(Table 7, S8). Unfortunately, no CDN inducing high levels of cytokines and low monocytic
cytotoxicity was identified (Table 7, S8).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
We used DSF measurements to rationalize interactions of CDN ligands with human WT STING.
Crystal structures and docking calculations highlight large binding modes similarities of examined
ligands. This suggests the effects of modifications on ΔTm values will be dictated primarily by local
environment. This notion is further supported by the additivity of ΔΔTm values of ligands with
multiple modifications. As a result, the inferred structure-activity relationships show good
agreement with measured ΔTm values in all cases.
Crucial element for understanding these relationships is the network of water molecules, which
was obtained by combining information obtained from X-ray crystallography and computational
modelling. Furthermore, estimates of differences in conformational strain of various ligands
highlight that changes to the ligand affect not only its interactions with the protein but the ease of
adapting the necessary conformation as well.
28
ACS Paragon Plus Environment