Divergent response of homologous ATP sites to stereospecific ligand fluorination for selectivity enhancement

Article abstract of DOI:10.1039/c7ob00129k

Acquiring a divergent response from homologous protein domains is essential for selective ligand-protein interactions. Stereospecific fluorination of (?)-balanol, an ATP mimic, uncovers a new source of selectivity from integrated chemical and conformational perturbation that differentiates homologous sites by the level of congruency in their response to local and remote fluorine effects.

Full text of DOI:10.1039/c7ob00129k

Organic &
Biomolecular Chemistry
View Article Online
View Journal
COMMUNICATION
Divergent response of homologous ATP sites to
stereospecific ligand fluorination for selectivity
enhancement†
Cite this: DOI: 10.1039/c7ob00129k
Received 7th November 2016,
Accepted 18th January 2017
Alpesh Ramanlal Patel,‡ Ari Hardianto, Shoba Ranganathan and Fei Liu*
DOI: 10.1039/c7ob00129k
www.rsc.org/obc
Acquiring a divergent response from homologous protein domains protein kinase A and B (PKA and PKB), belonging to the AGC
is essential for selective ligand–protein interactions. Stereospecific superfamily of protein kinases.10,11 Although a potent antagon-
fluorination of (−)-balanol, an ATP mimic, uncovers a new source ist, balanol presents limited selectivity within the family.12 The
of selectivity from integrated chemical and conformational pertur- seminal PKA/balanol binary crystal structure13 identified
bation that differentiates homologous sites by the level of con- balanol as a type I ATP mimic that occupies the full ATP site, in
gruency in their response to local and remote fluorine effects.
which the p-hydroxybenzamide moiety of balanol occupies the
adenine-binding subsite, the azepane moiety the ribose subsite,
and the tetrasubstituted benzophenone moiety the triphosphate
subsite (Scheme 1a). The aryl rings of the benzophenone moiety
induce protein movement in the flexible glycine-rich loop,
which allows the loop to interact extensively with the benzophe-
none moiety. Perturbation of benzophenone substitution fre-
quently leads to severe loss of activity,10,12,14,15 however the
azepane ring is more amenable to derivatisation.12
Selectivity in ligand–receptor interactions is a general require-
ment in chemical and biological systems, and both chemical
diversification and conformational regulation are important
avenues for specificity optimisation.1–4 However, selectivity in
highly homologous receptors or protein domains remains a
challenge. New insight into how conformational perturbation
could be fused with chemical functionalisation to create com-
pounded effects may open new avenues of selectivity optimi-
sation. Here we investigate the potential of stereospecific
ligand fluorination in instigating composite perturbation
effects that challenge both the ligand and protein to adapt
congruently. Fluorine is small with minimal steric interference
yet a rich source of physical/chemical perturbation and stereo-
electronic effects for shape control.5,6 We employ the natural
product balanol as a model system to demonstrate how new
selectivity potential in highly homologous ATP sites of protein
kinases can be unlocked by using fluorine through integrated
chemical and conformational control in a protein-dependent
manner. As protein kinases are validated drug targets but
require high ATP site selectivity, this selectivity enhancement
by perturbation congruency may introduce new design con-
cepts in kinase drug development.
We proceeded to constructing balanoids 1a–1e with stereo-
specific C–F bonds on azepane. We have investigated exten-
Balanol was originally discovered as a potent antagonist
(IC50 4–9 nM) for the ATP site of the protein kinase C (PKC)
isozymes7–9 and soon after for other related members, such as
Department of Chemistry & Biomolecular Sciences, Macquarie University, Sydney,
NSW 2109, Australia. E-mail: fei.liu@mq.edu.au
†Electronic supplementary information (ESI) available: Details of synthesis,
spectra, affinity measurements, and docking analysis. See DOI: 10.1039/
c7ob00129k
‡Current address: Faculty of Pharmacy, The University of Sydney, Sydney, NSW
2109, Australia.
Scheme 1 (a) General scheme of fluorinated balanoids as ATP site
probes. (b) Retrosynthesis of fluorinated balanoids.
This journal is © The Royal Society of Chemistry 2017
Org. Biomol. Chem.

View Article Online
Communication
Organic & Biomolecular Chemistry
Table 1 Binding affinities of
1
and 1a–1e against selected PKA/C
sively stereospecific fluorination by deoxyfluorination reactions
on substituted azepanes,16–19 and secured the synthesis and
structural elucidation methods for preparing these stereospeci-
fically fluorinated azepanes 4a–4e from a common precursor
5.16 These fluorinated azepane fragments were converted to
3a–3e after debenzylation and acylation, followed by esterifica-
tion to furnish the final fluoro-balanoids 1a–1e using reported
protocols.20,21 The natural product (−)-balanol (1) was also pre-
pared for comparison.21
isozymesa,b
Kd c (nM)
1
1a
1b
1c
1d
1e
The initial hypothesis was that a minimal level of fluorina-
tion on balanol (e.g. monofluorinated 1a–1c) should influence PKA
5.9 0.5 7.9 0.5 6.9 0.4 6.4 0.1 9.2 0.8 43
4
PKCδ 4.5
PKCε 0.73 0.06 19
PKCη 14 0.5
0
17
1
8
15 1.5 4.9 0.4
24 3.5 0.4 0.02 110 19 38 9.5
19 0.5 12
120 15 140 10 24 0.5 580 60 850 80
19 0.5 18 0.5
conservatively on the binding of this high affinity ligand. This
would then permit more reliable structural modelling for
38 3.5 16 2.5 13
0
2
fluorinated balanoids without drastic changes in the binding PKCθ 25
mode of the ligand. Multiply fluorinated balanoids 1d (C6-(R,S)
6
a Each shaded square indicates the ratio between the Kd of each bala-
noid for a particular enzyme over that of balanol. b Affinity gain or loss
under 1.5 fold is considered no change (within error) as the error
margin is between 0–41%. c Standard error of the mean (SEM) calcu-
lated from duplicate experiments.21
bisfluorination) and 1e (C5-(S) and C6-(R,S) trifluorination)
were prepared to examine the tolerance limit of these enzymes
to fluorination. Given the central position of the azepane ring
in the binding pocket, this balanol system offers a rare oppor-
tunity to examine how a protein could contend with shape-
controlling fluorination that not only modifies interactions
directly in the azepane-binding subsite but also indirectly in in enhancing both potency and selectivity for PKCε. This high
the other subsites by ligand conformation perturbation. Can positional sensitivity of PKCε towards single fluorine substi-
new binding interactions in the azepane subsite arise from the tution (i.e. ∼50–60 fold relative difference in affinity between
stereospecific C–F bond in a protein-dependent fashion? Can 1a/b and 1c) is either absent or subdued in the other enzymes.
the (sp3)C–F bond potentially confer “remote control” in For 1a and 1b, the relative selectivity ranking, even for this
p-hydroxybenzamide or the benzophenone binding subsite, small test panel of enzymes, is altered by single fluorination,
given the flexibility of both the ligand and the protein?
from PKCε ≫ PKCδ/PKA > PKCη/θ to PKA > PKCδ/ε/η ≫ PKCθ;
The binding affinities of balanol (1) and balanoids (1a–1e) however, the relative selectivity ratio (i.e. the ratio between the
to PKA and the PKC isozymes PKCδ, PKCε, PKCη, and PKCθ highest and lowest affinity case) did not improve for 1a and 1b.
were then measured using the KINOMEscan™ binding The PKC isozymes consistently exhibit higher sensitivity
assay.21,22 Isozymes PKCδ/ε/η/θ belong to the novel subclass of toward monofluorination at the C6 position than PKA,
the PKC family that shares a very high level of sequence homo- although the fluorine effect is not stereochemistry-dependent
logy in their catalytic domain.8 Without a crystal structure of at this position as all five enzymes showed comparable
balanol bound to any PKC isozyme available, the crystal struc- affinities for 1a and 1b.
ture of PKA/balanol13 has served as a representative for the
For the multiply fluorinated balanoids 1d and 1e, the
AGC superfamily. These proteins thus form a small but highly affinities in most cases are retained in the low nM range,
homologous panel for testing if this stereospecific fluorination suggesting that fluorination, even in multiple positions, can
approach could solicit divergent responses from these be well tolerated. Isozymes PKCδ and PKCη, along with PKA,
isozymes.
appear to be particularly tolerant of fluorine substitution (loss
The observed Kd values of balanol 1 to the enzymes (4–25 under 10 fold). The trifluorinated balanoid 1e also switched
nM, Table 1) generally are in good agreement with earlier, bio- the isozyme preference from PKA to PKCη. However, PKCε and
logical activity-based measurements for balanol,7 except for PKCθ exhibited a much higher level of sensitivity toward 1d
PKCε that exhibited a higher binding affinity, a small variation and 1e with an affinity loss of 20–150 fold. Interestingly, with
that may be attributed to the biological activity measurements an additional C5(S)-fluorine, 1e again improves the binding
in earlier reports being performed with the substrate. affinity to PKCε by nearly 3 fold compared to that of 1d, a case
However, as expected the overall specificity of balanol for these analogous to that of the monofluorinated balanoid 1c in
enzymes is limited (i.e. under 100 fold).23
In general, the binding profile of the monofluorinated bala- affinity by nearly two fold.
which the C5(S)-fluorine substituent improves the binding
noids 1a–1c across the enzyme panel shows either no or mild
This affinity profile of the fluoro-balanoids opens the next
affinity loss under 10 fold, except for PKCε that exhibited a question on the likely structural basis of this divergent protein
substantial affinity reduction of 26–33 fold for 1a and 1b but a response that results in improved affinity and selectivity of 1c
slight gain for 1c. Single fluorine substitution at the C5-(S) for PKCε. With no X-ray structure currently available for PKCε,
position (1c) improves the affinity to 0.4 nM and also the rela- the binary structure of mouse PKA bound with balanol
tive selectivity ratio of the parent balanol 1 for PKCε, showing (1BX6)21 and another of human PKCη with naphthyridine
that fluorine perturbation at the C5-(S) position is productive (3TXO)24 were used to construct a structure homology model
Org. Biomol. Chem.
This journal is © The Royal Society of Chemistry 2017

View Article Online
Organic & Biomolecular Chemistry
Communication
similar to that of 1c (Fig. 1b). Also, the C6-fluorination of 1a,
in the bound conformation, does not satisfy the strong F/N+
gauche preference in the azepane ring.5 Our earlier confor-
mational analysis of 1a as a free ligand identified three major
azepane ring conformations in equal distribution.16 One of
these three conformations is similar to the bound azepane
ring conformation of 1a here, except that the puckering of the
protonated nitrogen is inverted in the free ligand that does
satisfy the F/N+ gauche preference. In the presence of the
protein, the nitrogen puckering of the ligand is flipped in
order to H-bond to Asp493 and Asp536, however at the cost of
violating the F/N+ gauche preference (Fig. 1b, Newman projec-
tion). This ligand conformational perturbation in 1a is absent
in 1c. For 1a, a mild interaction gain in the azepane-binding
region, combined with ligand conformational perturbation
and some contact loss in the benzophenone subsite, renders
an overall affinity loss of 26 fold for 1a. Thus, the differential
affinity profile of 1a and 1c for PKCε is due to the congruent
chemical and conformational control conferred specifically by
C5-(S), but not C6-(S), monofluorination.
The large difference between the affinities of 1c and 1d to
PKCε (275 fold) also prompted docking analysis of 1d in the
ATP site of PKCε.21 Comparison of the interactions of 1c and
1d in the ATP site of PKCε (Fig. 2a) shows that C6-bisfluorina-
tion of 1d causes conformational changes in the azepane ring
and is accompanied by rotation of the p-hydroxybenzamide
amide bond (Fig. 2a, red highlight) that is binding in the
opposite orientation compared to that of 1c (Fig. 2a, pink high-
Fig. 1 Docking analysis showing PKCε-ligand interactions in the
azepane-binding region for (a) 1 (grey), (b) 1a (green), and (c) 1c (pink).
The full images and interaction comparison list for 1, 1a, and 1c are in
the ESI.†
of PKCε, due to the high sequence homology between these light). As a consequence, 1d suffers the loss of a backbone
proteins in the catalytic domain (40% for PKA/PKCε and 71% H-bonding contact deep in the adenine subsite to Val489. For
for PKCη/PKCε). Ligands 1, 1a and 1c were then each docked 1d in PKCε, the loss of binding interactions, particularly
into the PKCε model, with the calculated binding energies
showing an excellent linear correlation with the measured
affinity values.21
The comparison of the bound conformations of balanol 1
(Fig. 1a) and C5(S)–F balanoid 1c (Fig. 1c) identified new inter-
actions in the azepane-binding subsite of PKCε for 1c (Fig. 1
table, full comparison list of interactions in the ESI†). The
C5(S)–fluorine induces azepane conformational change in 1c,
which satisfies the small fluorine–oxygen gauche preference,25
facilitates H-bonding from the protonated nitrogen to the carb-
oxylate of Asp493, and adds direct H/X-bonding interactions
to Lys416.26 Also, in the benzophenone and p-hydroxybenza-
mide binding subsites, the key ligand–protein contacts in 1c,
in particular the benzophenone contact made with the protein
by moving the flexible glycine-rich loop, are preserved. For 1c
in PKCε, the chemical and conformational change, instigated
by stereospecific fluorination, is synergistic such that the gain
from new interactions in the azepane subsite is not
accompanied by loss from ligand conformation perturbation
or interactions from other subsites. Thus, C5(S)-fluorination
effectively uncovers a highly selective ligand–protein pairing
between 1c and PKCε with stronger affinity that is unavailable
for the other isozymes.
Fig. 2 (a) 1c and 1d in PKCε: interaction differences highlighted in
yellow in the p-hydroxybenzamide and azepane-binding subsites. The
amide group of the ligand is highlighted in pink for 1c and red for 1d; (b)
1c and 1d in PKA showing similar azepane binding conformations. The
As for 1a, the C6(S)–fluorine is too far away to interact with
full comparison list of interactions for 1c and 1d in PKA/PKCε is in the
Lys416, even though the azepane ring conformation in 1a is ESI.†
This journal is © The Royal Society of Chemistry 2017
Org. Biomol. Chem.

View Article Online
Communication
Organic & Biomolecular Chemistry
H-bonds, is found in all three binding subsites for the p-hydro-
xybenzamide, azepane and benzophenone moieties, showing
that PKCε in this case is unable to adapt to the “remote
control” effect of azepane fluorine perturbation.
Contrary to the case of PKCε, PKA as a receptor is insensi-
tive to the fluorination perturbation in either 1c or 1d
(Fig. 2b). Comparison of 1c and 1d in the ATP site of PKA
shows highly preserved interactions in the p-hydroxybenz-
6 V. Gouverneur and K. Seppelt, Chem. Rev., 2015, 115, the-
matic issue.
7 P. Kulanthaivel, Y. F. Hallock, C. Boros, S. M. Hamilton,
W. P. Janzen, L. M. Ballas, C. R. Loomis, J. B. Jiang and
B. Katz, J. Am. Chem. Soc., 1993, 115, 6452–6453.
8 A. C. Newton, J. Biol. Chem., 1995, 270, 28495–28498.
9 D. Mochly-Rosen, K. Das and K. V. Grimes, Nat. Rev. Drug
Discovery, 2012, 11, 937–957.
amide and benzophenone binding regions, along with com- 10 J. Setyawan, K. Koide, T. C. Diller, M. E. Bunnage,
parable binding conformations in the azepane ring.21 This is
consistent with the observed binding affinities of these two
S. S. Taylor, K. C. Nicolaou and L. L. Brunton, Mol.
Pharmacol., 1999, 56, 370–376.
compounds that are within 1.5 fold for PKA. For 1c, more 11 J. M. Arencibia, D. Pastor-Flores, A. F. Bauer, J. O. Schulze
H-bonding interactions are made to the azepane-binding
and R. M. Biondi, Biochim. Biophys. Acta, 2013, 1834, 1302–
1321.
subsite of PKCε than to that of PKA.21 Effectively, the
fluorine perturbation from the azepane ring in 1c and 1d is 12 P. Akamine, Madhusudan, L. L. Brunton, H. D. Ou,
discernable by PKCε but not PKA, due to the ability of the
PKCε pocket to specifically “solicit” a productive binding con-
J. M. Canaves, N. H. Xuong and S. S. Taylor, Biochemistry,
2004, 43, 85–96.
formation from 1c but not 1d. Stereocontrolled fluorination 13 N. Narayana, T. C. Diller, K. Koide, M. E. Bunnage,
here is shown as the origin of integrated chemical and confor-
mational control that engenders recognition cooperativity
between the ligand and protein for specificity enhancement
K. C. Nicolaou, L. L. Brunton, N. H. Xuong, L. F. Ten Eyck
and S. S. Taylor, Biochemistry, 1999, 38, 2367–
2376.
(i.e. 275 fold diffidence in Kd for 1c and 1d in PKCε vs. 1.43 14 K. Koide, M. E. Bunnage, L. Gomez Paloma, J. R. Kanter,
fold diffidence in Kd for 1c and 1d in PKA) in a protein-
S. S. Taylor, L. L. Brunton and K. C. Nicolaou, Chem. Biol.,
dependent manner.
1995, 2, 601–608.
While the perturbation investigation here utilized a small 15 Y. S. Lai, J. S. Mendoza, G. E. Jagdmann Jr.,
number of fluorine atoms, the divergent response from these
highly homologous active sites toward stereospecific fluorina-
tion on the ligand is clearly evident. This is the first demon-
D. S. Menaldino, C. K. Biggers, J. M. Heerding,
J. W. Wilson, S. E. Hall, J. B. Jiang, W. P. Janzen and
L. M. Ballas, J. Med. Chem., 1997, 40, 226–235.
stration of shape-controlling fluorination that extracts 16 A. R. Patel, G. Ball, L. Hunter and F. Liu, Org. Biomol.
additional selectivity potential in highly homologous kinase Chem., 2013, 11, 3781–3785.
ATP sites, which complements the existing avenues of inhibi- 17 A. R. Patel and F. Liu, Tetrahedron, 2013, 69, 744–752.
tor discovery by high-throughput-screening, structure-guided 18 A. R. Patel, L. Hunter, M. M. Bhadbhade and F. Liu,
protein–ligand
engineering
or
fragment-based
lead
Eur. J. Org. Chem., 2014, 2014, 2584–2593.
design.27–30 Given the large number of members in this family 19 A. R. Patel and F. Liu, Aust. J. Chem., 2015, 68, 50.
and also other ATP-binding proteins, this new strategy will 20 A. Furstner and O. R. Thiel, J. Org. Chem., 2000, 65, 1738–
open additional avenues for addressing the selectivity issues
1742.
not just for kinases but other ligand–protein interactions in 21 Details in the ESI.†
general.23,31,32 Furthermore, complex natural product ligands 22 M. A. Fabian, W. H. Biggs 3rd, D. K. Treiber,
that are amenable to fragment-based development,33 such as
balanol, can incorporate this fluorine-moderated fragment
approach by incrementally challenging the adaptability of a
protein target to further enhance selectivity.
C. E. Atteridge, M. D. Azimioara, M. G. Benedetti,
T. A. Carter, P. Ciceri, P. T. Edeen, M. Floyd, J. M. Ford,
M. Galvin, J. L. Gerlach, R. M. Grotzfeld, S. Herrgard,
D. E. Insko, M. A. Insko, A. G. Lai, J. M. Lelias, S. A. Mehta,
Z. V. Milanov, A. M. Velasco, L. M. Wodicka, H. K. Patel,
P. P. Zarrinkar and D. J. Lockhart, Nat. Biotechnol., 2005,
23, 329–336.
Notes and references
23 L. N. Johnson, Q. Rev. Biophys., 2009, 42, 1–40.
1 E. Persch, O. Dumele and F. Diederich, Angew. Chem., Int. 24 M. J. van Eis, J. P. Evenou, P. Floersheim, C. Gaul,
Ed., 2015, 54, 3290–3327.
S. W. Cowan-Jacob, L. Monovich, G. Rummel, W. Schuler,
W. Stark, A. Strauss, A. von Matt, E. Vangrevelinghe,
J. Wagner and N. Soldermann, Bioorg. Med. Chem. Lett.,
2011, 21, 7367–7372.
2 J. A. Berrocal, F. Di Meo, M. Garcia-Iglesias, R. P. Gosens,
E. W. Meijer, M. Linares and A. R. Palmans, Chem.
Commun., 2016, 52, 10870–10873.
3 Q. Wang, P. Zhang, L. Hoffman, S. Tripathi, D. Homouz, 25 C. R. Briggs, M. J. Allen, D. O’Hagan, D. J. Tozer,
Y. Liu, M. N. Waxham and M. S. Cheung, Proc. Natl. Acad.
Sci. U. S. A., 2013, 110, 20545–20550.
A. M. Slawin, A. E. Goeta and J. A. Howard, Org. Biomol.
Chem., 2004, 2, 732–740.
4 S. J. Teague, Nat. Rev. Drug Discovery, 2003, 2, 527–541.
5 D. O’Hagan, Chem. Soc. Rev., 2008, 37, 308–319.
26 P. Auffinger, F. A. Hays, E. Westhof and P. S. Ho, Proc. Natl.
Acad. Sci. U. S. A., 2004, 101, 16789–16794.
Org. Biomol. Chem.
This journal is © The Royal Society of Chemistry 2017

View Article Online
Organic & Biomolecular Chemistry
Communication
27 T. Anastassiadis, S. W. Deacon, K. Devarajan, H. Ma and 31 M. W. Karaman, S. Herrgard, D. K. Treiber, P. Gallant,
J. R. Peterson, Nat. Biotechnol., 2011, 29, 1039–1045.
28 H. Park, Y. Shin, H. Choe and S. Hong, J. Am. Chem. Soc.,
2015, 137, 337–348.
29 B. S. Lauber, L. A. Hardegger, A. K. Asraful, B. A. Lund,
O. Dumele, M. Harder, B. Kuhn, R. A. Engh and
F. Diederich, Chem. – Eur. J., 2016, 22, 211–221.
C. E. Atteridge, B. T. Campbell, K. W. Chan, P. Ciceri,
M. I. Davis, P. T. Edeen, R. Faraoni, M. Floyd, J. P. Hunt,
D. J. Lockhart, Z. V. Milanov, M. J. Morrison, G. Pallares,
H. K. Patel, S. Pritchard, L. M. Wodicka and P. P. Zarrinkar,
Nat. Biotechnol., 2008, 26, 127–132.
32 Z. Zhao, L. Xie, L. Xie and P. E. Bourne, J. Med. Chem.,
2016, 59, 4326–4341.
30 P. Brear, C. De Fusco, K. Hadje Georgiou, N. J. Francis-
Newton, C. J. Stubbs, H. F. Sore, A. R. Venkitaraman, C. Abell, 33 B. Over, S. Wetzel, C. Grutter, Y. Nakai, S. Renner, D. Rauh
D. R. Spring and M. Hyvönen, Chem. Sci., 2016, 7, 6839–6845.
and H. Waldmann, Nat. Chem., 2013, 5, 21–28.
This journal is © The Royal Society of Chemistry 2017
Org. Biomol. Chem.