.
Angewandte
Communications
itors and the protein, which are shown in Figures S3–S6 in the
Supporting Information.
geometry, which enables 8 and 9 to maintain the charge-
assisted H-bond interaction to Asp81, although the indole
moiety that previously accomplished this task is replaced by
the phenyl/benzyl group with the amide NH in thiophene 2-
position (Figure 4d). Even though the aryl- and benzyl-
substituted nitrogens are located at virtually identical posi-
tions, the increase in affinity by about threefold from 8 to 9
might be explained by the different pKa values of the aniline
NH compared to the benzyl NH. Thus, the latter, which is
being most likely protonated to a larger extent, is capable of
forming a more efficient charge-assisted H-bond to the water
molecule that mediates the H-bond to Asp33 and Gly221
(Figure 4e).
In our inhibitor series also less or inactive compounds
were synthesized which are essential for SAR interpretations
as well. However, in the present case the interpretation of
these data is almost impossible, as the assignment to one of
the binding mode clusters remains speculative.
The current example highlights the complexity of binding
events and their strong dependence on seemingly minor
effects of scaffold decoration and modifications. That 1, 7, 8,
and 9 do not adopt the same binding geometry as observed in
the green cluster can be rationalized retrospectively, but
without the crystal structure determinations the adopted
binding modes would have been difficult to predict, and, if so,
would have hardly been believed or attracted sustained
attention without our experimental evidence.
Such an unexpected flipping of binding modes after
modification of the scaffold substitutions has been detected
sporadically, however, mostly by serendipity.[6] As the com-
monly accepted hypothesis that similar ligands bind in
a similar fashion[7] apparently does not hold true on a com-
prehensive basis, the question arises, how frequently a
(re-)validation of an assumed binding mode is required. Are
there any easily available indicative hints to estimate whether
a crystallographic (re-)validation of the assumed binding
mode used as working hypothesis is required throughout the
lead exploration campaign?
In addition to the affinity determination via a fluores-
cence-based assay (Ki values), we performed a thermal shift
assay (TSA) to investigate the extent to which the studied
compounds stabilize the protein, especially with respect to the
different binding modes. TSAs are commonly used for hit
identification and affinity ranking of inhibitors.[8,9]
The results for our inhibitor series (Table 1) clearly
suggest that the observed shifts in the melting temperature
do not generally correlate with the Ki values. However, within
one binding mode cluster, the DTm values are ranked
correctly, with 5 being an exception: This might be attributed
to the loss of the charge-assisted H-bond to Asp81 due to the
absence of the indole moiety, while the other members of this
cluster establish this interaction. This observation is in
agreement with the concept that TSA measurements require
similar binding enthalpies for correct affinity ranking.[9]
However, how do TSA data perform for ligands of identical
chemotype that fall in the same affinity range, but exhibit
different binding geometries?
The binding mode of 1 apparently relies on the presence
of the ester moiety, whereas the amide derivative (3) adopts
an altered binding mode. Interestingly, the intramolecular H-
bond observed in 1 (Figure 2a) breaks up in favor of the
establishment of an H-bond to Thr222 Og through the ethyl-
substituted amide nitrogen. How does the mere replacement
of the ester by an amide translate into such a predominant
change in binding geometry? In all the EP crystal structures
determined in our study, the interaction pattern between
Thr222 Og and the Asp219 carboxylate (2.7 ꢀ) is conserved .
Likely, this Thr side chain donates an H-bond to the
carboxylate of Asp219 keeping the proton of Thr222 Og
oriented towards the aspartate (Figure 4b). This observation
supports the conclusion that Thr222 Og itself can only be
involved in favorable polar interactions if addressed by
appropriate H-bond donors on the ligand site. This is
impossible for the ester 1 as its oxygen can only serve as an
H-bond acceptor; hence the ester prefers to saturate its polar
contact inventory by an intramolecular H-bond. However,
when the ester is changed to an amide (3), the introduced NH
function is able to address Thr222 Og as a donor rather than
maintaining the intramolecular H-bond, while the concom-
itantly nonsaturated glycine amide NH rotates by ꢀ 1008 to
establish an H-bond to the Asp219 carboxylate (Figure 4a).
Although the overall SAR interpretation is inconclusive
in the absence of structural data, the SAR within the clusters
reveals a consistent picture, which will be discussed only
exemplarily. In the green cluster, 3 bearing an ethyl group at
P2 has the lowest affinity (4.0 mm); this most likely results from
the reduced hydrophobic interactions, compared to those of
the inhibitors equipped with a benzyl group at this position.
Furthermore, 3 and 6 lack the intramolecular p–p interaction
between the two phenyl rings observed in 4 and 5, which leads
to a loss of preorganization corresponding to the bound
conformation. In 5, due to the exchange of the indole moiety
for the benzyl group, the H-bond to Asp81 and the weak polar
contact to Ser83 as observed in 3, 4, and 6 cannot be
established (Figure 4c). In 4, the representative with the
highest affinity (460 nm) of this cluster, the four varied
substituents seem to be optimally chosen among all inves-
tigated variations, as the indole moiety is able to form an H-
bond to Asp81 whereas the benzyl moiety interacts via p–p
stacking to Tyr226. Inhibitor 8 (purple cluster) differs from 5
(green cluster) only by the position of the secondary amino
nitrogen in the P1 substituent of the inhibitor. Based on the X-
ray structure of 5, a lower affinity caused by a less efficient
interaction of the relocated nitrogen to the catalytic aspar-
tates would have been expected for 8. Interestingly though, its
affinity is nearly unchanged. The X-ray structure of 8 provides
the explanation for this initially surprising observation: The
switch in binding geometry between 5 and 8 seems to be
mainly triggered by the loss of the H-bond to Asp35 that
would be expected for an unchanged geometry of 5. The
thiophene moiety formerly located in the S1’ pocket (green
cluster) flips by about 1808 in order to populate now the S2
pocket. Likewise, inhibitor 9, in which the P1 substituent is
extended by one carbon atom, shows this altered binding
Only two of our four binding mode clusters qualify for
such a comparison, as the singletons 1 and 7 are slightly off-
4
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2015, 54, 1 – 6
These are not the final page numbers!