Topographical Mapping of Isoform-Selectivity Determinants for J-Channel-Binding Inhibitors of Sphingosine Kinases 1 and 2

Article abstract of DOI:10.1021/acs.jmedchem.9b00162

Sphingosine kinase enzymes (SK1 and SK2) catalyze the conversion of sphingosine into sphingosine 1-phosphate and play a key role in lipid signaling and cellular responses. Mapping of isoform amino acid sequence differences for SK2 onto the recently available crystal structures of SK1 suggests that subtle structural differences exist in the foot of the lipid-binding "J-channel" in SK2, the structure of which has yet to be defined by structural biology techniques. We have probed these isoform differences with a ligand series derived from the potent SK1-selective inhibitor, PF-543. Here we show how it is possible, even with relatively conservative changes in compound structure, to systematically tune the activity profile of a ligand from ca. 100-fold SK1-selective inhibition, through equipotent SK1/SK2 inhibition, to reversed 100-fold SK2 selectivity, with retention of nanomolar potency.

Full text of DOI:10.1021/acs.jmedchem.9b00162

Subscriber access provided by UNIV OF NEW ENGLAND ARMIDALE
Article
Topographical mapping of isoform-selectivity determinants for
J-channel-binding inhibitors of sphingosine kinases 1 and 2
David Roger Adams, Salha Tawati, Giacomo Berrretta, Paula Lopez Rivas, Jessica
Baiget, Zhong Jiang, Aisha Alsfouk, Simon Paul Mackay, Nigel Pyne, and Susan Pyne
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.9b00162 • Publication Date (Web): 19 Mar 2019
Downloaded from http://pubs.acs.org on March 19, 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 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Topographical mapping of isoform-selectivity determinants for
J-channel-binding inhibitors of sphingosine kinases 1 and 2
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
‡
†
†
†
†
David R. Adams, Salha Tawati, Giacomo Berretta, Paula Lopez Rivas, Jessica Baiget, Zhong
‡
†
,†
†
*,†
Jiang, Aisha Alsfouk, Simon P. Mackay Nigel J. Pyne, and Susan Pyne,
†Strathclyde Institute of Pharmacy and Biomedical Sciences, University of Strathclyde, Glasgow
G4 0RE, UK
‡School of Engineering & Physical Sciences, Heriot-Watt University, Edinburgh EH14 4AS, UK
ABSTRACT
Sphingosine kinase enzymes (SK1 and SK2) catalyse the conversion of sphingosine into
sphingosine 1-phosphate and play a key role in lipid signaling and cellular responses. Mapping of
isoform amino acid sequence differences for SK2 onto the recently available crystal structures of
SK1 suggests that subtle structural differences exist in the foot of the lipid-binding ‘J-channel’ in
SK2, the structure of which has yet to be defined by structural biology techniques. We have probed
these isoform differences with a ligand series derived from the potent SK1-selective inhibitor, PF-
5
43. Here we show how it is possible, even with relatively conservative changes in compound
structure, to systematically tune the activity profile of a ligand from ca. 100-fold SK1-selective
inhibition, through equipotent SK1/SK2 inhibition, to reversed 100-fold SK2 selectivity, with
retention of nanomolar potency.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 82
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
INTRODUCTION
Sphingosine 1-phosphate (S1P) is a key signalling lipid derived from sphingosine (Sph) by the
action of sphingosine kinases (SK1 and SK2). S1P is transported to the extracellular environment,
where it functions as a ligand for a family of five S1P-specific G protein coupled receptors (S1P ),
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
-5
but it also acts on specific intracellular target proteins, such as histone deacetylases 1/2
(
HDAC1/2).1, 2 Irreversible cleavage of S1P is catalysed by S1P lyase, producing (E)-2-
3
hexadecenal and phosphoethanolamine, but S1P is also reversibly dephosphorylated by S1P
phosphatase to regenerate Sph,4-6 levels of which are additionally influenced by flux through the
7
ceramide (Cer) synthesis pathway. The effects of S1P on cell function favour proliferation,
migration, differentiation and survival and are generally opposed by Cer, which induces apoptosis,
senescence and growth arrest.8, 9 As a consequence the concept of the ‘sphingolipid rheostat’ was
proposed,10, 11 in which the inter-conversion of Cer via Sph to S1P controls cellular fate (recently
reviewed12, 13). Altered S1P signalling has been associated with numerous pathophysiologies,
including cancer, cardiovascular disease, neurodegenerative conditions, diabetes and
1
4
inflammatory disease. The possibility of manipulating the sphingolipid rheostat for therapeutic
purposes has therefore provided a rationale for exploring the development of SK inhibitors and
their potential to reduce S1P formation and signalling and to raise pro-apoptotic Cer levels.
The SK1 and SK2 enzymes are encoded by distinct genes and differ in their subcellular
localisation, biochemical properties and functions; there are three N-terminal variants of SK1 and
two N-terminal variants of SK2.15, 16 A number of classes of SK inhibitors have emerged in the
last two decades, including the recent discovery of an ATP-competitive SK inhibitor chemotype
that decreases cellular S1P levels, elevates Sph/Cer, induces apoptosis, and that inhibits cell
1
7
proliferation and colony formation. Most work to date, however, has focused on the development
ACS Paragon Plus Environment

Page 3 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
of Sph-competitive inhibitors. This includes the development of inhibitors selective for SK1 that
1
8
19
20
exhibit nanomolar potency, such as PF-543 (1), Genzyme-51 (2) and Amgen-23 (3), Chart 1.
2
1
PF-543 reduces sickling of red blood cells in vitro and in vivo and attenuates post-infarction
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
22
23
cardiac remodeling. We have also demonstrated that PF-543 protects against cardiomyocyte
apoptosis and reduces cardiac maladaptive hypertrophy in a mouse model of pulmonary
hypertension.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 82
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Chart 1. Lipid substrate (Sph), product (S1P) and established inhibitors for SK1 and SK2.
OH
X = OH
sphingosine (Sph)
SK1 / SK2
X
6
X = OPO32-
NH3
sphingosine 1-phosphate (S1P)
SK1-selective inhibitors
a
Ki (IC50) / nM
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
SK1
SK2
OH
O
O
S
O
d
b
b
1 PF-543
(28)
(58)
(20)
(>5,000)
N
O
OH
H
N
c
2
3
Genzyme-51
H
ND
N
HN
N
O
OH
HO
d
Amgen-23
N
(1,600)
S
Cl
N
N
H
Cl
SK2-selective inhibitors
N
H
N
4
5
6
ABC294640
K145
no inhibition 9,800
at 100 µM
Cl
O
O
NH2
N
S
no inhibition 6,400
at 10 µM
O
O
N
SLR080811
12,000
1,300
N
H
N
O
H2N
NH
7 ROMe
no inhibition 16,000
at 100 µM
O
H2N
OH
OH
8
9
Pfizer-27c
(
25)
(2.4)
90
N
O
N
S
N
O
VT-20dd
9,000
N
O
N
NH
NH2
N
N
N
H
Other inhibitors
S
OH
16,000
6,700
d
1
0 SKi
Cl
N
N
H
a
b
c
d
For references see text. This study. No data. Co-crystal structure available with SK1.
ACS Paragon Plus Environment

Page 5 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
24
Early prototype Sph-competitive SK2-selective inhibitors include ABC294640 (4), K145
2
5
26
27
(
5), SLR0808110 (6) and (R)-FTY720 methyl ether (ROMe; 7), Chart 1. Of these compounds,
ABC294640, which induces cell death via apoptotic and autophagic pathways,24,28 has been tested
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
in several disease models for cancer,24 rheumatoid arthritis29 and ulcerative colitis.30 The
compound has proceeded to phase I/II clinical studies targeting multiple oncology and
inflammatory indications. However, ABC294640 also indirectly targets dihydroceramide
desaturase and induces the proteasomal degradation of SK1 in prostate cancer cells and in
HEK293T cells.31-33 ROMe, an SK2-selective inhibitor (K 14 μM) with no activity against SK1 at
i
1
00 μM, inhibits DNA synthesis in breast cancer cells and induces autophagic death of T-cell acute
lymphoblastic leukemic cells.27, 34 Very recently two reports have disclosed examples of more
35
36
potent SK inhibitors, Pfizer-27c (8) and VT-20dd (9), with selectivity for SK2. Despite these
developments and a growing interest in the (patho)physiological roles of SK2, a clear structural
rationale for the development of high potency SK2-specific inhibitors has yet to emerge.
No SK2 crystal structures are available as yet to support structure-based design of SK2-
selective inhibitors, but amino acid sequence alignment suggests a high degree of homology to
SK1, the structure of which has recently been solved by X-ray crystallography.37,20,38 SK1 exhibits
a bidomain architecture comprising an ATP-binding N-terminal domain (NTD) and a C-terminal
domain (CTD) that hosts the Sph binding site, Figure 1. The CTD folds as a two-layer β-sandwich
with three extended inter-strand loops packing across one face to generate the Sph binding site as
a J-shaped cavity, termed the ‘J-channel’. One of the three ‘lipid binding loops’, LBL-1, consists
of two reverse-paired helices (α7 and α8) that may gate substrate entry and product egress by a
3
9
hinged motion. Residues in LBL-1 and in the second loop, LBL-2, define much of the contact
surface for the polar head group and central region of the lipid. The third loop, LBL-3, folds with
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 82
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
an additional helix (α9), running antiparallel along α8, and provides enclosure for the lipid tail. To
date, SK1 co-crystal structures have been acquired with three inhibitors bound in the J-channel,
3
7
20
namely with SKi (10; also known as SKI-II), with Amgen-23 (3), and most recently with PF-
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
38
5
43 (1). We have used the latter structure as a starting point to explore differences in the foot of
the J-channel of SK1 and SK2. Here we report our results from a ligand-based structure activity
relationship (SAR) study that highlights key determinants for switching between SK1-selective
and SK2-selective inhibition.
Figure 1. The structure of SK1 is shown with bound PF-543 (1; orange surface) (PDB: 4V24).38
PF-543 occupies the substrate Sph binding site, formed by the packing of three lipid binding loops
(LBL-1 to LBL-3) against a β -sandwich core. The nucleotide binding site is shown by
superimposition of Mg-ADP (green sphere/rose surface) from its separate co-crystal structure
(3VZD) with SK1.37
ACS Paragon Plus Environment

Page 7 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
RESULTS AND DISCUSSION
Design rationale for ligand selectivity switching. Of the 20 or so residues that contribute to the
J-channel in SK1, SK2 differs at only 3 points in the direct hydrophobic contact surface of the
binding site—namely in Ile174, Met272 and Phe288 of SK1, corresponding respectively to
Val304, Leu517 and Cys533 in SK2 (Figure 2, orange). Phe288 stoppers the toe of the SK1 J-
channel, and its substitution by Cys533 in SK2 may result in a longer J-channel and potentially
also confer greater surface plasticity in that region of SK2 due to loosened packing against adjacent
hydrophobic residues. Ile174 and Met272 contribute to the throat of the J-channel and, in the
SK1/PF-543 co-crystal structure, are seen to form part of the flanking surface on either side of the
inhibitor’s p-xylylene subunit; their replacement by Val and Leu may make this region of the J-
channel slightly wider in SK2 than in SK1. Ile174 additionally contributes surface on one edge of
the PF-543 terminal subunit, with contact distances of 3.6 and 3.4 Å between its Cδ methyl carbon
and the PF-543 phenyl ortho and meta carbons respectively. A leucine (Leu261; Figure 2, yellow),
that is packed against Ile174 in SK1, also contributes to the contact surface around the edge of the
PF-543 terminal phenyl and, although conserved in SK2 (as Leu506), may differ in its
conformational profile due to the replacement of the adjacent Ile174 in SK1 by Val304 in SK2.
Together, these considerations suggest that the J-channel toe might be expanded in SK2 on a
contact surface mapping to the para-position of the terminal phenyl of SK1-bound PF-543 (Figure
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
, green arrow).
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 82
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 2. J-channel surface detail (mesh) from the PF-543/SK1 co-crystal (PDB: 4V24),38
highlighting key isoform residue differences that likely impact on the J-channel structure of SK2.
SK1 residue numbering is shown; cognate SK2 residues, where different, are in parenthesis.
(Residue color notation: see text.)
In principle, an expanded surface in SK2 at the J-channel toe might contribute to reduced
affinity for PF-543 and hence, at least in part, to the established SK1-over-SK2 selectivity of this
inhibitor (Chart 1). However, SK2 also differs from SK1 in several other residues that, whilst not
contributing directly to the ligand contact surface, are likely to indirectly affect the J-channel
structure. Thus, a set of residues in LBL-3 of SK1 (Val294, Met298, Met312, Tyr314, Tyr318;
Figure 2, green) differs in SK2 (as Ile539, Ala543, Phe557, Leu559 and Gln563 respectively).
Inspection of the SK1/PF-543 co-crystal structure suggests that these differences may impact on
the conformation of a conserved leucine (Leu302 in SK1; Figure 2, magenta), which contributes
to the contact surface for the sulfonyl group of PF-543. We hypothesized that the cognate residue
(Leu547) in SK2 might be driven inwards to narrow the J-channel at that point. If so, surface
encroachment in the J-channel heel of SK2 might then interdict comfortable accommodation of
ACS Paragon Plus Environment

Page 9 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
the sulfonyl oxygens of PF-543 and also contribute to the observed SK1-selectivity of this
inhibitor.
Two further residue differences, Ala175 and Ala339 in SK1 (corresponding to Ser305 and
Thr584 in SK2), might also potentially be relevant to isoform discrimination by ligands targeting
the J-channel. These residues (Figure 2, cyan) lie proximal to the 2-(hydroxymethyl)pyrrolidine
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
‘head group’ of PF-543, and their variation in SK2 might affect a network of polar interactions
that engages the PF-543 head group or/and slightly reposition LBL-1 with respect to the β -
sandwich core, which could then impact on the surface of a notch in the J-channel that
3
8
accommodates the PF-543 methyl group (Figure 2). However, it is known that the hydroxyl
group of SK1-bound PF-543 precisely maps to the position of the Sph-3-OH and, given that Sph
serves as a substrate for both SK1 and SK2, it is likely that the ‘head group’ hydrogen bonding
features in SK2 should also support the binding of the PF-543 2-(hydroxymethyl)pyrrolidine
subunit. Thus, in seeking to identify determinants for SK2-selective inhibition, we focused
attention on a series of PF-543 analogues that retained the 2-(hydroxymethyl)pyrrolidine head
group but embodied alterations to the tail targeted to the enclosed J-channel foot, predicted by our
analysis to offer scope for isoform-discriminating interactions with ligands. Accordingly, we
prepared a concise compound test set to evaluate whether SK2-inhibitory activity might be
enhanced by alleviating steric demand in the sulfonyl linker and/or by deletion of the PF-543
methyl group. We also sought to test the hypothesis that optimisation of ligand fit to the J-channel
toe in SK2 might confer selectivity over SK1.
Synthesis of probe compound set. In order to assess the contribution of the methyl group
to the SK1 selectivity of PF-543, we prepared the normethyl analogue (18; Scheme 1) for
evaluation. The compound was prepared from m-methoxybenzyl bromide (11) by reaction with
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 82
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
sodium benzenesulfinate followed by boron tribromide mediated cleavage of the aryl methyl ether
in 12 to afford phenol 14. The latter was then alkylated with 4-(bromomethyl)benzaldehyde and
the resulting aldehyde (16) subjected to reductive amination with (R)-(–)-prolinol.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Scheme 1. Synthesis of probe compounds 18 and 19.a
OMe
Ph
OMe
Br
(a)
S
O
O
11
12
b)
(
HO
OH
Ph
Y
OH
(c)
S
O
O
1
3
1
1
4
5
Y = CH2
Y = O
(77% from 12)
(21% from 13)
(d)
CHO
Ph
Y
O
S
O
O
1
1
6
7
Y = CH2
Y = O
(81% from 14)
(88% from 15)
(e)
N
Ph
Y
O
S
O
O
OH
1
1
8
9
Y = CH2
Y = O
(71% from 16)
(33% from 17)
a
®
Reagents and conditions: (a) NaSO Ph, Aliquat 336, 85 °C (47%); (b) BBr , CH Cl , –78 °C
2
3
2
2
(
77%); (c) phenylsulfonyl chloride K CO / MeCN (21%); (d) 4-(bromomethyl)benzaldehyde,
2 3
K CO , MeCN, 60 °C; (e) (R)-2-(hydroxymethyl)pyrrolidine, NaBH(OAc) , DCE.
2
3
3
A collection of compounds was then prepared in order to evaluate the contribution of the
–
SO -CH – tail linker to inhibitory performance and selectivity. Analogues were designed to test
2 2
the impact of replacing the PF-543 sulfonyl group by methylene, but we also included an isosteric
sulfonate (19) to probe the impact of subtle changes in ligand bond angle preferences. Sulfonate
1
9 was prepared by initial sulfonylation of resorcinol (13) followed by alkylation of the resulting
ACS Paragon Plus Environment

Page 11 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
phenol (15) with 4-(bromomethyl)benzaldehyde and reductive amination with (R)-(–)-prolinol,
Scheme 1.
The ethylene-linked analogue (23, Scheme 2) was prepared to assess the impact of reducing
steric demand in the locus occupied by the PF-543 sulfonyl group. For reasons of synthetic
convenience, this and other sulfonyl deletion analogues were prepared for comparative evaluation
against 18 without the methyl group in place on the central arene ring. Synthesis of 23 was
undertaken from acetylene 20, prepared by ambient temperature Sonogashira coupling of 3-
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
4
0
iodophenol with phenylacetylene. The acetylene was hydrogenated to afford phenol 21, which
was alkylated with 4-(bromomethyl)benzaldehyde to afford 22. Reductive amination of the latter
with (R)-(–)-prolinol afforded 23.
Scheme 2. Synthesis of probe compound 23.a
Ph
OH
OH
(a)
2
0
21
(b)
CHO
O
22
(c)
N
O
OH
2
3
a
Reagents and conditions: (a) H , Pd/C, EtOH (95%); (b) 4-(bromomethyl)benzaldehyde, K CO ,
2
2
3
MeCN, 20 °C, 16 h (87%); (c) (R)-2-(hydroxymethyl)pyrrolidine, NaBH(OAc) , DCE (82%).
3
An ether-linked variant (49) of 23 was prepared according to Scheme 3. Thus, resorcinol
13) was alkylated with benzyl bromide and the resulting phenol (25) transformed into 49 by the
(
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 82
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
two-step alkylation/reductive amination procedure applied in Schemes 1 and 2 for compounds 18,
1
9 and 23. A subset of six halogenated analogues (50–55, Scheme 3) was then expanded to
facilitate systematic exploration of the space in the J-channel foot. Two further analogues were
included in the ether-linked test set to explore the impact of steric demand in the critical toe region
of the J-channel––namely, methyl ether 56 and nitrile 57, with bent and linear two-atom extensions
from the para-position of the tail arene. The sulfide counterpart (58) of the ether-linked parent (49)
was prepared in a similar manner starting from benzenethiol (24). The para-fluorophenyl and
para-chlorophenyl analogues (59 and 60) of this latter compound were made by a similar strategy.
Sulfides 58–60 were included to test the impact of incorporating a heteroatom in the linker with
greater steric demand and differing bond angle preferences, which might impact on positioning of
the ligand’s terminal para-site in the critical J-channel toe region.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
ACS Paragon Plus Environment

Page 13 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Scheme 3. Synthesis of probe compounds 49–60.a
m
p
o
A
OH
Y
OH
(a) or (b)
X
1
2
3
4
A = OH
A = SH
25 Y = O X = H
26 Y = O X = o-F
27 Y = O X = m-F
(55% from 13)
(53% from 13)
(58% from 13)
(38% from 13)
(54% from 13)
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
2
3
3
8
9
0
1
Y = O X = p-F
Y = O X = o-Cl
Y = O X = m-Cl (49% from 13)
Y = O X = p-Cl (48% from 13)
32 Y = O X = p-OMe (32% from 13)
3
3
3
4
Y = O X = p-CN (47% from 13)
Y = S X = H
(68% from 24)
(71% from 24)
(79% from 24)
(c) or (d)
35 Y = S X = p-F
3
6
Y = S X = p-Cl
CHO
Y
O
X
3
3
3
4
7
8
9
0
Y = O X = H
(88% from 25)
(62% from 26)
(64% from 27)
(77% from 28)
(71% from 29)
Y = O X = o-F
Y = O X = m-F
Y = O X = p-F
41 Y = O X = o-Cl
4
4
2
3
Y = O X = m-Cl (55% from 30)
Y = O X = p-Cl (96% from 31)
44 Y = O X = p-OMe (68% from 32)
Y = O X = p-CN (95% from 33)
4
5
46 Y = S X = H
47 Y = S X = p-F
(79% from 34)
(93% from 35)
(84% from 36)
(e)
4
8
Y = S X = p-Cl
N
Y
O
X
OH
4
9
Y = O X = H
(70% from 37)
(69% from 38)
(57% from 39)
(73% from 40)
(58% from 41)
50 Y = O X = o-F
5
5
5
5
1
2
3
4
Y = O X = m-F
Y = O X = p-F
Y = O X = o-Cl
Y = O X = m-Cl (50% from 42)
(86% from 43)
Y = O X = p-OMe (84% from 44)
Y = O X = p-CN (38% from 45)
55 Y = O X = p-Cl
5
5
6
7
58 Y = S X = H
(95% from 46)
(78% from 47)
(53% from 48)
5
6
9
0
Y = S X = p-F
Y = S X = p-Cl
aReagents and conditions: (a) 25–33: (substituted)benzyl bromide (or 4-chlorobenzyl chloride in
the case of 31), K CO , acetone, 65 °C, 5 h; (b) 34–36: (substituted)benzyl bromide (or 4-
2
3
chlorobenzyl chloride in the case of 36), K CO , MeCN, 20 °C, 16 h; (c) 37–45: NaH, DMF, then
2
3
4
-(bromomethyl)benzaldehyde, 16 h, 20 °C; (d) 46–48: 4-(bromomethyl)benzaldehyde, K CO ,
2 3
MeCN, 20 °C, 16 h; (e) (R)-2-(hydroxymethyl)pyrrolidine, NaBH(OAc) , THF (for 49–57) or
3
DCE (for 48–60).
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 82
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
SK-inhibitory performance and selectivity of probe compound set. Compounds were
evaluated for inhibition of purified SK1 and SK2 using a radiometric assay, as detailed in our
earlier work.41 In brief, the procedure uses [ γ - P]ATP and measures incorporation of the
radiolabelled phosphate into the S1P product. SK1 and SK2 were assayed, respectively, at 3 and
32
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
0 µM Sph substrate concentration. These concentrations correspond to the K values of SK142-44
m
2
7
and SK2 and, therefore, allow direct comparison of the sensitivity of the two enzymes at 50%
saturation with Sph. Assays were performed with ATP at 250 µM concentration, well above the
K value for the nucleotide, to ensure saturation with ATP. Activity data are summarised in Chart
m
2
.
ACS Paragon Plus Environment

Page 15 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Chart 2. SK1 and SK2 inhibitory activity for probe compound set derived from PF-543 (1).a
SK1
IC50/nM
SK2
IC50/nM
SI
N
O
S
2
8
(33%)*
O
O
OH
PF-543 (1)
Me
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
methyl
deletion
X
X
X
X
18
19
X = H
144
11
(16%)†
<0.048
0.012
S
O
O
LINKER
TYPE
sulfonylmethylene
O
S
O
X = H
902
O
sulfonate
23
49
X = H
26
35
0– .74
ethylene
X = H
14
41
9
170
39
16
55
0.88
0.75
0.063
3.5
0.26
0.11
101
5
0
X = o-F
X = m-F
X = p-F
X = o-Cl
X = m-Cl
X = p-Cl
51
143
49
151
412
41
p
5
2
O
53
54
m
o
o
44
5
5
5
5
6
7
4,130‡
X = p-OMe >3,000 119
X = p-CN
>25
>2.7
methyleneoxy
>3,000 1,120‡
X
p
5
59
60
8
X = H
X = p-F
X = p-Cl
137
26
17
31
5.3
45
76
S
765
m
2,360‡
methylenesulfanyl
*
% enzyme inhibition at 5 µM test concentration
% enzyme inhibition at 3 µM test concentration
†
‡
Estimated IC50 from partial concentration response curve (1 nM – 3 µM range)
a
SK1 activity was assayed using 3 μM Sph and 250 μM ATP. SK2 activity was assayed using 10
μM Sph and 250 μM ATP. Data are reported as IC values (or % enzyme inhibition) vs control
5
0
determined in triplicate assays. NE: no inhibitory effect at maximal test concentration of 3 µM. SI
denotes selectivity index as IC (SK1) / IC (SK2).
5
0
50
4
1
We have previously obtained an IC of 28 nM for inhibition of SK1 by PF-543 in our
5
0
18
assay, considerably higher than the K value of 3.6 nM reported by Schnute et al. , with the caveat
i
that the latter determination was performed under conditions where ATP was not saturating and
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 82
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
therefore subject to two-substrate kinetics. As reported, however, PF-543 was highly selective for
inhibition of SK1, giving only 33% inhibition of SK2 at 5 μM and 72% at 50 μM.
Our analysis (vide supra) suggested that the most likely explanation for the SK1-over-SK2
selectivity of PF-543 is due to surface encroachment in the SK2 J-channel heel targeted by the
inhibitor’s sulfonyl group (Figure 2, red arrow). However, at the outset of our studies we could not
eliminate the possibility that structural differences in the sub-pocket that accommodates the PF-
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
5
43 methyl group (Figure 2, ‘notch’) might also be a contributory factor. We therefore evaluated
the normethyl analogue (18) of PF-543. This compound showed a modest (5-fold) loss in SK1-
inhibitory potency compared to PF-543, consistent with a slight reduction in favourable surface
contact with that isoform. However, the gain in SK2 inhibition that might have been expected for
relief of an unfavourable surface compression against the methyl, was not observed. Indeed, only
slight inhibition of SK2 (16%) was detectable at the maximal test concentration of 3 µM in our
current studies, suggesting that the PF-543 methyl group is not a strong determinant of selectivity.
35
During the course of our studies Schnute et al. disclosed further details of their development of
PF-543, including comparative activity data for normethyl analogue 18 using a different assay
system. In this study the methyl group was also shown to contribute only marginally to the strong
SK1-selectivity of PF-543; its deletion led to a change in selectivity by less than a factor of 2.
Interestingly the sulfonate analogue (19) was 13-fold more potent against SK1 than its
counterpart (18) with the sulfonylmethylene linker. A corresponding increase in inhibitory potency
was also seen against SK2, with a quantifiable IC of 902 nM determined. Replacement of the
5
0
sulfonyl group in 18 with a second methylene in 23 led to a modest 5.5-fold improvement in
potency against SK1 but a pronounced enhancement in potency against SK2 so that the compound
was almost equipotent against both isoforms (IC 26 and 35 nM for SK1 and SK2 respectively).
5
0
ACS Paragon Plus Environment

Page 17 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Taken together, these results suggest that the heel of the SK1 J-channel might be a tight fit for the
sulfonylmethylene linker in PF-543 and 18 and supported our hypothesis that in SK2 the J-channel
heel may be tighter still, encroaching sufficiently on the sulfonyl oxygens so as to be a key
determinant for isoform discrimination. Thus, in SK2 higher compression is postulated along a
trajectory (Axis-1 in Figure 3) between the head group and sulfonyl oxygens of the ligand, with
the head group locked in place by a highly organised network of polar interactions involving
features that normally serve to position the substrate for phosphorylation. These ‘headlock’
features include a structural water and an aspartic acid on helix-α7 (Asp178 in SK1, conserved as
Asp308 in SK2) that have been shown to form three tight hydrogen bonds with charge stabilisation
to the hydroxypyrrolidinium subunit of PF-543 in its SK1 co-crystal structure.38
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
HEAD LOCK
(water)
HO
O
N
H
O
SK1
SK2
O
TOE
9
Cys533
-

X
I
Phe285
L
E
H
Leu547
Leu302
HEEL
Compression Axis-1
SK1 relaxed
SK2 accentuated
Compression Axis-2
SK1 accentuated
SK2 relaxed
Figure 3. Key compression axes postulated to influence ligand binding to the J-channels of SK1
and SK2.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 82
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
The methyleneoxy-linked analogue (49) also exhibited near equipotent inhibitory activity
against the two isoforms, but with a marginal increase in overall potency relative to the ethylene-
linked compound (23). Thus, 49 exhibited IC values of 14 and 16 nM against SK1 and SK2
5
0
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
respectively. However, replacement of the oxygen by a larger sulfur atom in compound 58 caused
a 10-fold loss in potency against SK1 with only a slight loss (< 2-fold) against SK2, thereby
resulting in a compound with reversed selectivity (albeit modest) in favour of SK2 (IC values of
50
1
37 and 26 nM against SK1 and SK2 respectively). This was an important finding because it
supported our second hypothesis, namely that the J-channel toe is likely expanded in SK2 or/and
exhibits greater plasticity. Thus, compression along an axis linking the large sulfur atom and para
position of the phenyl ring in 58 (Axis-2 in Figure 3) might be less severe in SK2 than in SK1.
38
Indeed, analysis of our PF-543/SK1 co-crystal structure (Figure 2) had suggested a remarkably
close fit for the phenylsulfonylmethylene tail unit of PF-543 in the SK1 J-channel, with the para-
CH packed against Phe288 and the methylene of the linker wedged against helix-α9. Conceivably,
alleviation of compression between these two points (and associated ligand strain) might
contribute to the enhanced inhibitory potency of sulfonate 19 over the sulfonylmethylene-linked
analogue (18).
We next sought to exploit the hypothesised tighter compression axis between toe and heel
in SK1 (Axis-2 in Figure 3) still further by installation of a substituent in the terminal phenyl ring
of 49 and 58, with a particular focus on the para-position to see whether compounds with even
greater selectivity for SK2 could be achieved. Given the potential differences in J-channel surface
allied to the replacement of Ile174 in SK1 by Val304 in SK2, however, we also wished to examine
the impact of substitution at the ortho- and meta-positions. For that reason, we evaluated fluorides
50–52 and chlorides 53–55 derived from methyleneoxy-linked parent 49. The standout feature
ACS Paragon Plus Environment

Page 19 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
with this compound set was indeed a pronounced loss in SK1-inhibitory potency for the para-
fluoro and para-chloro analogues (52 and 55), which showed ca. 12-fold and 295-fold increases
in IC relative to 49 respectively. Some loss in potency against SK2 was also observed for these
5
0
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
compounds (≤ 3-fold relative to parent 49), but SK1 was preferentially penalised, leading to highly
selective, nanomolar-potent SK2 inhibition in the case of chloride 55 (see concentration response
curves in Figure 4). Curiously the para-fluoro analogue (52; SK2 IC 49 nM) showed equivalent
50
or even rather more loss in SK2 potency relative to parent 49 (SK2 IC 16 nM) than its chloride
5
0
counterpart (55; SK2 IC 41 nM). This outcome may reflect a requirement for some induced fit
5
0
to accommodate the halogen in SK2. Thus, the cost associated with side chain perturbation caused
by binding of the ligand might be offset by greater surface contact for the chloride over the fluoride.
SK1-selective inhibition
equipotent SK1/SK2 inhibition
SK2-selective inhibition
compression along ‘Axis-1’ (Figure 3)
compression along ‘Axis-2’ (Figure 3)
log [Compound 49] (M)
log [Compound 52] (M)
log [Compound 55] (M)
log [Compound 18] (M)
log [Compound 23] (M)
log [Compound 58] (M)
log [Compound 60] (M)
Figure 4. Selected SK1 and SK2 concentration response curves for PF-543-derived J-channel
probes, highlighting the reversal of behaviour from SK1-selective inhibition (left) to SK2-selective
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 82
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
inhibition (right) with ligand features that induce compression along Axis-1 and Axis-2 of Figure
3
respectively. (Assay conditions: see legend to Chart 2.)
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
In the case of the methylenesulfanyl-linked analogue (58), the effects of para-fluoro and
para-chloro substitution (59/60) broadly mirror those seen in the methyleneoxy-linked series
52/55), where inhibition of SK1 but not SK2 is significantly penalised (Chart 2). Although
(
differences in the size and bond angle preferences of the sulfide and ether linkages may subtly
modulate terminal halogen fit to the J-channel toe and ligand strain, in both cases the presence of
the larger chlorine atom afforded greatest isoform discrimination, with selectivity for inhibition of
SK2 over SK1 in the range of 75-fold to 100-fold for 60 and 55. Introduction of bulkier or/and
more polar para-substituents may be counterproductive for SK2 inhibitory performance however.
Thus, in the methyleneoxy-linked series introduction of a methoxy group (compound 56, SK2 IC50
119 nM) led to a 7-fold loss in SK2 inhibitory potency relative to parent 49 and a linear nitrile
extension (compound 57, SK2 IC 1.12 µM) caused a 70-fold loss in activity. Whilst this fall off
5
0
in inhibitory potency may reflect a limit to the capacity of the SK2 J-channel toe, changes in dipole
and lipophilicity of the terminal subunit may also be factor in the activity fall off. Thus, a nitrile
group is sometimes exploited as a bioisostere for the halogen in (hetero)aromatic chlorides,
conferring a significant reduction in lipophilicity and, therefore, increased lipophilic ligand
efficiency.45, 46
In contrast to substitution at the para-position, the SK1-inhibitory activity of 49 is more
tolerant of halogenation at the ortho- and meta-positions on the terminal arene (Chart 2,
compounds 50, 51, 53 and 54). Here only modest activity loss (ca. 3-fold) is observed (50/53/54),
whilst meta-fluorination even marginally enhances activity (51, SK1 IC 9 nM) relative to parent
5
0
ACS Paragon Plus Environment

Page 21 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
4
9 (SK1 IC 14 nM). Interestingly SK2 is less tolerant of substitutions at the ortho- and meta-
5
0
positions, particularly at the latter site, where fluorination (51, SK2 IC 143 nM) and chlorination
5
0
(54, SK2 IC 412 nM) produce respectively 9-fold and 26-fold activity loss versus parent 49 (SK2
5
0
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
IC 16 nM). These results suggest that the SK2 J-channel foot, whilst extended compared to SK1,
5
0
is tighter in width, at least between certain points.
Structural inferences for the SK2 J-channel. Crystal structures of SK1 have only
recently emerged (since 2013), whilst no structures are currently available for SK2. To facilitate
qualitative rationalization of the isoform selectivity profiles of the ligands, we constructed an
homology model for SK2 based on the crystal structure (4V24) of the large SK1 splice variant
(SK1b, sometimes also called SK1c) with bound PF-543 that we had recently determined at 1.8 Å
38
resolution. In brief, an initial model was generated using the default automated routines of the
SWISS-MODEL server47-50 and this was refined using the Small Molecule Drug Discovery
software suite from Schrödinger, LLC. In particular, LBL-2 was remodelled from the structure of
the corresponding loop in the Enterococcus faecalis diacylglycerol kinase family protein, as
defined in the crystal structure 4WER (DOI: 10.2210/pdb4wer/pdb). The rationale for this
adjustment was that the C-terminal end of LBL-2 in SK2 (524-RFDDGL-529) has sequence
features that are more closely related to prokaryotic diacylglycerol kinases than to SK1 (Figure S1
3
9
in Adams et al ).
Our activity data for the normethyl analogue (18) of PF-543 suggested that the central ring
methyl group in PF-543 is not a key SK2-discriminating feature and determinant of the high SK1-
inhibitory selectivity exhibited by PF-543. Nevertheless, we did find that energy minimisation of
our ligand probe set docked into the SK2 model, as illustrated for 60 in Figure 5A, tended to result
in contraction of the J-channel notch that hosts the PF-543 methyl. This notch is formed at the
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 82
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
interface between helices-α8 and -α9 of the two flap-like LBL-1 and LBL-3 loops respectively (cf.
Figure 1). As such, size and surface contouring of the notch may be sensitive to subtle changes in
the seating of the two helices as well as to accessible conformational volumes of local residues.
Indeed, inspection of an overlay of all protein chains from the currently available crystal structures
revealed significant differences in the notch surface even across different SK1 structures, thus
suggesting a degree of plasticity in this region of the J-channel (Figure 5B and 5C). A particularly
important residue in this regard is Leu299 in SK1, which is located on helix-α9 and conserved in
SK2 as Leu544 (marked in Figure 5). This residue exhibits some conformational variance across
SK1 crystal structures and would expand the notch region in the SK2 model illustrated in Figure
5 to accommodate the PF-543 methyl group with only modest movement.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 5. Comparison of modelled binding mode for SK2-selective inhibitor 60 with SK1 co-
crystal structures of PF-543 (1) and SKi (10). Panel (A): 60 (blue stick) docked into SK2 model
(red mesh), highlighting key contact residues (cyan stick) and expanded SK2 J-channel toe surface.
Panel (B): PF-543 (1; orange stick) is shown from its co-crystal structure (4V24) with SK1 (green
ACS Paragon Plus Environment

Page 23 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
mesh) and the SK2 model surface (red mesh) overlaid from (A); SK1 contact residues are shown
(light orange stick) and the constrained SK1 J-channel toe surface is evident. Panel (C): As (B),
but showing co-crystal structure (3VZD) of SKi (10, magenta stick) with SK1 (blue mesh and pink
stick key contact residues) superimposed onto the SK2 model (red mesh) from panel (A); the heel
region adjacent to SK1 Leu302 is consistently less compressed in SK1 structures (relative to the
SK2 model), whilst the notch surface exhibits some plasticity. Panel (D): overlay of (A)–(C)
minus the J-channel surface; the chlorine atom of 60 is accommodated in SK2 between Cys533
and Leu506, with rotation of the latter possible whilst the cognate SK1 residue (Leu261) is
constrained by the packing of adjacent residues.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Although the notch region exhibits significant plasticity, inspection of the SK1 J-channel
across different crystal structures revealed that the protein surface is quite tightly defined
(irrespective of ligand occupancy) both around the J-channel heel that contacts the PF-543 sulfonyl
group and around the toe that interfaces with the end of the PF-543 terminal phenyl subunit. Our
data indicate that sulfonyl replacement by methylene in compounds 23 and 49 enhances SK1-
inhibitory potency compared to the sulfonylmethylene-linked parent (18), which suggests that the
sulfonylmethylene linker is already a little oversized for an optimal fit of the tail of the ligand to
the SK1 J-channel. Thus, only a modest additional contraction in the SK2 J-channel at this point
could generate a strongly unfavourable compressive effect for compounds such as PF-543 and its
normethyl analogue (18) to account for their SK1 selectivity. The SK2 model supported the
proposal that the SK2 J-channel may be contracted in the heel region, in particular due to inwards
movement of Leu547 (corresponding to Leu302 in SK1), Figure 6A. However, the model also
suggested that the basis for such movement may be a consequence, at least in part, of differences
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 82
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
in the LBL-2/LBL-3 interfaces of the isoforms and not merely due to local residue differences
immediately surrounding the leucine, as we had first conceived (Figure 2). In particular, Leu547
is located on helix-α9 in LBL-3, and the C-terminal end of this helix engages the hairpin-like tip
of LBL-2 (Figure 6a). The model suggested that residue differences in LBL-2 might promote an
inward movement at the C-terminus of helix-α9, thereby exerting leverage on the J-channel heel
through Leu547. As the key heel compression contact is against the Cβ atom of Leu547, rather
than the end of the side chain, the leverage cannot be alleviated by side chain rotation. However,
compounds with reduced steric demand in the heel (23 and 49) can be accommodated in both SK1
and SK2 (illustrated for 49 in Figure 6B and 6C).
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 6. The J-channel heel is a key determinant of isoform discrimination by SK inhibitors.
Panel (A): LBL-2 and LBL-3 are shown from the co-crystal structure (4V24) of PF-543 (orange
stick) with SK1 (green ribbon); the corresponding elements from our SK2 homology model (red
ribbon) are overlaid, illustrating inward movement in LBL-3 leading to heel compression in SK2
by Leu547. Panel (B): Potent unselective inhibitor 49 (blue stick) is shown modelled into the SK2
ACS Paragon Plus Environment

Page 25 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
J-channel (red mesh); the J-channel heel surface is driven significantly inwards relative to SK1 so
that the presence of a sulfonyl group in the ligand (49  19, 56-fold activity loss) is poorly
tolerated by SK2. Panel (C): Inhibitor 49 (magenta stick) is again shown, here modelled into the
SK1 J-channel (green mesh); the SK1 J-channel heel surface exhibits less compression and
accommodates a sterically more demanding sulfonyl containing linker more readily (49  19, no
activity loss; 23  18, 5-fold activity loss). Panel (D): overlay of (B) and (C) minus the J-channel
surface.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
The model also suggested that the strong SK2 selectivity of the para-halo derivatives of
49/58 is indeed likely due to J-channel toe surface expansion, but that this may not be solely due
to the single amino acid substitution of Phe288 in SK1 by Cys533 in SK2 (Figure 2). Indeed, in
the absence of halogen substituents on the ligand, the energy minimised SK2 model complexes
showed little apparent expansion in the toe surface compared to SK1, as illustrated for 49 docked
to SK2 and SK1 (Figure 6B and 6C respectively). The key to accommodation of the para-halo
atom appears to be perturbation in the conformation of an adjacent leucine (Leu506 in SK2,
conserved in SK1 as Leu261) that packs against the cysteine. The combination of Cys533 and
Leu506 generates the toe surface in SK2 (Figure 5), with movement of the latter predicted to allow
expansion to accommodate the chlorine of 55/60. The leucine additionally packs against two other
non-conserved residues––Val304 (Ile174 in SK1) and Ile508 (Leu263 in SK1)––and our model
suggests that loosened packing with these residues and the cysteine may confer greater plasticity
in the toe of the SK2 J-channel than in SK1. Thus, ligand binding in SK2 may involve an element
of induced fit in the J-channel toe, potentially with an associated energy cost that is more or less
offset depending on the effectiveness of surface contact between the ligand and the expanded toe.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 82
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
In contrast, the toe of the SK1 J-channel is likely to be more rigid and introduction of para-
substitution in the ligand is predicted to generate an unfavourable compression between the J-
channel toe and helix-α9, as illustrated in Figure 7 for chloride 60. Despite the enhanced plasticity
and capacity indicated for the SK2 J-channel toe, our results nevertheless indicate significant
limitations on the size of the para-substituent that can be accommodated in SK2. Thus, the axis of
the para-substituent is predicted to orientate directly at the Cβ atom of Ile508 presented from the
SK2 CTD β-sandwich core, Figure 5A. In the case of chloride 60, a distance of ca. 4.3 Å is
predicted between Ile508 Cβ and the chlorine atom, and this spatial constraint likely contributes
to the sharp drop in inhibitory potency against SK2 seen with the introduction of a para-cyano
group in 57.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 7. The J-channel toe is a second key determinant for isoform discrimination by SK
inhibitors. A comparison of modelled binding modes for the SK2-selective inhibitor (60) is shown
here, with the SK2-docked inhibitor (cyan stick) superimposed onto the SK1-docked inhibitor
(magenta stick). With reduced plasticity in the SK1 J-channel toe (pink mesh), the sulfur atom of
6
0 is predicted to be tightly wedged against the backbone of helix-α9 (pink ribbon). This generates
ACS Paragon Plus Environment

Page 27 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
an unfavourable compression axis (black arrow, corresponding to ‘Axis-2’ of Figure 3) that
preferentially penalizes binding to SK1. In SK2 the greater plasticity of the J-channel toe (cyan
mesh) is predicted to allow accommodation of 60 with less compression.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
During the course of our work, Childress et al36 described a series of pyrrolidine
carboximidamide SK inhibitors with a spectrum of SK1- and SK2-inhibitory selectivity. The
highest SK2-selectivity was seen with compound 9, which is notable for its extended tail structure.
With the guanidinium head group engaging key acidic residues in the mouth of the J-channel, the
aminothiazole subunit in 9 would be expected to overlay the middle arene subunit of our ligands,
whilst the oxadiazole ring of 9 would lie in plane and partially map onto the p-xylylene subunit
(Figure 8B/C). However, an intriguing feature with such a binding mode is that the rod like
biphenyl tail of 9 could only be accommodated if there is a significant expansion in the toe of the
J-channel, and this on the other side of Cys533 to that occupied by the chlorine in 55/60. There are
two key residue substitutions that may make this possible for SK2 but not readily achievable for
SK1: one is the replacement of SK1 Phe288 by Cys533 in SK2; the second is the exchange of SK1
Met312 for Phe557 in SK2. In the absence of compound 9 the latter residue, which is surface
exposed on LBL-3, is predicted to fold inwards towards the adjacent cysteine (in partial surrogacy
for the Phe288 residue of SK1). Rotation of these two side chains alone (Cys533/Phe557) was
sufficient to open an extended pocket in our model and accommodate the terminal phenyl group
of 9 with a reasonably relaxed biphenyl torsion (37°) and favourable packing interactions against
Cys533, Phe557 and other aromatic residues (Tyr, His). This postulated binding mode for 9, in
effect, amounts to tunnelling between LBL-2 and LBL-3 in SK2. In SK1 the more rigid J-channel
toe is stoppered by Phe288, for which a rotation analogous to that required of Cys533 in SK2 is
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 82
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
sterically precluded. The suggested binding mode for ligand 9 would also avoid encroachment on
the Leu547 heel compression point in SK2 that we propose biases inhibitor selectivity towards
SK1. Binding of compound 9 to SK1 would require much more extensive perturbation to the
protein structure and packing of the LBL-3 return leg.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 8. Plasticity in the J-channel toe of SK2 may be exploited to develop SK2-selective
inhibitors. Panel (A): Compound 60 (blue stick) docked into the SK2 model (red mesh),
highlighting key contact residues (cyan stick) and two predicted trajectories for possible toe
expansion (arrows). Small groups may be accommodated between Cys533 and Leu506
(Trajectory-1, hollow arrow). This is the predicted expansion for binding of 55/60, where the fit
of the Cl atom is likely made possible largely by rotation in Leu506. Larger groups may be
accommodated between Cys533 and Phe557 (Trajectory-2, filled arrow) upon rotation of these
side chains. Panel (B): VT-20dd (9, green stick) docked into the SK2 model (grey mesh),
highlighting key contact residues (orange stick). Here the ligand is predicted to exploit the second
toe expansion trajectory. Panel (C): Overlay of (A) and (B) showing the SK2 J-channel surface
ACS Paragon Plus Environment

Page 29 of 82
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
for docked 60 (red mesh) and illustrating the two key residue rotations in Cys533 and Phe557
required for Trajectory-2 expansion (cyan stick  orange stick positions). Panel (D): ABC294640
(4, orange stick) docked into SK2 (grey mesh). In this model, the Trajectory-1 expansion is made
possible chiefly by rotation of Cys533 rather than Leu506, but the space generated by the
expansion is underused by the Cl atom.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Although our analysis suggests that outward rotation of Phe557 might allow tunnelling of
substantially extended ligand tail groups in SK2 (Figure 8C), the preferred orientation of the
Phe557 side chain (in the absence of such ligand extensions) is predicted to lie inwards towards
Cys533 to maximise surface contact with adjacent hydrophobic residues. In this tightly packed
inward orientation, the end of the Phe557 side chain is predicted to come into close contact with
the meta-position of the terminal phenyl ring of our PF-543-derived ligands series. This might
contribute to the poor tolerance by SK2 of meta-halogenation, as seen with 51 and 54 (vide supra).
In SK1 the cognate residue (Met312, Figure 2) is more flexible and exhibits less reach compared
to Phe557 in SK2, which is consistent with the greater tolerance towards meta-halogenation in 51
and 54. A suitable tail extension to enable tunnelling in SK2 (‘Trajectory-2’ in Figure 8) may
require a large group to provoke the necessary rotation in Phe557 and compensate for the loss of
packing interactions between this residue and adjacent side chains. Whilst potentially interesting
for future investigation, our objective here was to achieve selectivity control with conservative
changes in the parent ligand structure, thereby keeping molecular weight within druglike bounds.
We also docked ABC294640 (4) into our SK2 model for comparison (Figure 8D). In this
case we found that the pyridyl group closely overlays the p-xylylene subunit of our ligands whilst
the adamantyl group sits well in the J-channel bend––partly overlaying the middle arene of our
ACS Paragon Plus Environment