D. J. Gustin et al. / Bioorg. Med. Chem. Lett. 23 (2013) 4608–4616
4609
alleles of SphK1 and one copy of SphK2 have been ablated,24 sug-
gesting that circulating levels of S1P are correlated to SphK activity.
The SphK1/SphK2 double knockout genotype results in embry-
onic death on approximately day 12 of gestation. SphK1/SphK2
deficient embryos have impaired blood vessel maturation owing
to defects in the surrounding endothelium.22 Similar observations
have been made with S1P1 receptor knockout mice, suggesting that
a significant portion of these effects are due to S1P signaling
through the S1P1 receptor.25 Taken together, these results suggest
that interfering with S1P production and signaling may find use in
the treatment of diseases characterized by aberrant angiogenesis.
Indeed, the S1P1 receptor antagonist FTY720 inhibited tumor
growth and angiogenesis in a murine tumor allograft model,26
and an anti-S1P antibody has been shown to inhibit neo-vascular-
ization in both tumor models and a model of age-related macular
degeneration.27,28
Our primary interest in the SphKs was due to their role in angi-
ogenesis. Observations with SphK deficient mice suggested that
complete inhibition of both isoforms of SphK may be required in
order to block SphK driven angiogenesis. Figure 1 shows the struc-
tures and inhibition constants for both SphKs for a number SphK
inhibitors including SK1-I,29 SKI-II,30 ABC 294640,31 and inhibitors
1–3.32–34 To date, few examples of selective SphK2 inhibitors or
dual SphK1/SphK2 inhibitors have been reported.29,35,36
Herein, we describe the structure-based design of a series of
SphK inhibitors that includes examples of both potent dual
SphK1/SphK2 inhibitors and SphK1-specific inhibitors. The series
is highly selective for SphKs, has favorable physicochemical and
pharmacokinetic properties and includes the only known SphK
inhibitors that are also potent against rodent SphK1.
After screening efforts failed to identify tractable leads, we em-
barked on a structure-based design approach to create novel SphK
inhibitors. In doing so, we started with a number of design consid-
erations. First, we decided to target the Sph binding site rather
than the ATP binding site. Calculations based on the X-ray co-crys-
tal structure (see Fig. 2) of SKI-II and ADP with human SphK1 were
used to evaluate the druggability of the two binding sites.37 The
calculations showed that while the pockets are similar in enclosure
and volume, the Sph binding site is over three times as hydropho-
bic as the ATP binding site. Based on binding site surface area; the
ratios of nonpolar to total surface areas are 0.7 and 0.2, respec-
tively. Calculating druggability scores based on published models
that combine these physicochemical properties into a predictive
score yielded a score of 1.9 for the Sph site and 1.0 for the ATP
site.38,39 In this model, scores greater than 1.3 indicate druggable
pockets. Secondly, we proposed that inhibitor designs should be
Sph-like (but not SphK substrates), rather than S1P-like to avoid
building in potentially confounding activity at S1P receptors. Fur-
ther inspiration and a more concrete means to accomplish our
goals were suggested by the X-ray co-crystal structures of SphK1
in complex with SKI-II and Sph.
Figure 2A shows an overlay of the X-ray co-crystal structures of
human SphK1 with bound SKI-II and ADP (2.3 Å) and human SphK1
complexed with Sph (2.0 Å).37 The structural data indicate that
both the SKI-II and the hydrophobic tail of Sph bind to SphK1 in
a hydrophobic pocket that is formed between the strands b10-
b12 of the b-sandwich in the C-terminal domain and helices
a7–
a9. The para-chlorophenyl portion of SKI-II is positioned in the
back of the pocket occupying largely the same location as carbons
15–18 of Sph. In this binding mode, the phenol of SKI-II is located
near Asp178. Removal or methylation of the phenol oxygen of SKI-
II resulted in a complete loss of SphK inhibition. This, in combina-
tion with the structural data, suggested that SKI-II interacts with
SphK in part through donation of a hydrogen bond to Asp178.
The crystallographic data also suggested that Sph interacted with
SphK through donation of a hydrogen bond to Asp178, in this case,
from the secondary hydroxyl on C(3) (Fig. 2A).
Moreover, this hydroxyl group on C(3) appeared to make water
mediated hydrogen bonding interactions with the side chain of
Ser168, the main chain carbonyl of Ala339 and the backbone amide
of Gly342. Carbons 1 and 2 of Sph extend past Asp178 into a rela-
tively polar pocket where the C(1) hydroxyl group appeared to do-
nate a hydrogen bond to Asp81. Interestingly, the C(2) amine of
Sph is located nearly equidistantly between aspartates 81 and 178.
We envisioned that it might be possible to generate SphK inhib-
itors meeting our criteria by simply merging the structures of Sph
and SKI-II. As illustrated in Figure 2B, our plan called for the syn-
thesis of chemical libraries of general formula 4. We initially in-
tended to evaluate the SAR around the para-chlorophenyl
subunit, while at the same time preparing libraries that presented
both hydrogen bond donors and acceptors with the aim of mimick-
ing the hydrogen bonding interactions observed in the crystal
structures.
Cl
OH
H
N
Cl
N
O
N
H
S
After preparing a number of libraries, compound 6 (Table 1) was
identified as a lead. In 6, the para-chlorophenyl substituent on the
thiazole of SKI-II was replaced with a 2-napthyl group that was
identified in an earlier library. Additionally, the phenol of SKI-II
was replaced with an ethylene linker with an (S)-pyrrolidin-2-
ylmethanol group at its terminus. As shown, 6 had an IC50 of
N
SKI-II
ABC 294640
Ki (µM), SphK1
0.5
---
---
9.8
SphK2
NH2
H
N
OH
Me
Me
0.99 and 2.5 lM against SphK1 and SphK2 respectively. X-ray crys-
OH
NHMe
O
OH
tallography showed 6 bound to SphK in the Sph binding site as ex-
pected. Furthermore, 6 was not a substrate for either SphK1 or
SphK2 and showed no binding to S1P receptors 1 or 3 at concentra-
tions below 25 lM. Given that 6 met our initial design criteria, we
investigated the SAR of 6 with the goal of optimizing it for SphK
1
SK1-I
10
>100
1.4
31
Ki (µM), SphK1
H2N
NH HCl
SphK2
O
O
N
inhibition and physico-chemical properties (Table 1).
Me
NH HCl
NH2
N
H
Comparison of 6 to its enantiomer (7) revealed a slight prefer-
ence for the (R)-stereoisomer, while methylation of the hydroxy-
methyl group resulted in a loss of potency (cf. compounds 6, 7
and 8). Comparison of 2-hydroxymethyl substituted azetidine, pyr-
rolidine, piperidine and azepane rings indicated that the piperidine
ring was optimal (cf. compounds 5, 7, 9 and 10). Relocation of the
hydroxyl group to either the 3 or 4 position of the piperidine ring
or extending the hydroxymethyl group by one carbon resulted in a
N
O
2
3
Ki (µM), SphK1
0.2
0.5
0.05
4.2
SphK2
Figure 1. Representative SphK inhibitors.