Chemistry Reference
In-Depth Information
Crystallographic analysis of the
21
-BACE complex showed that the binding mechanism
remained consistent. The nitrogens at positions N1 and C2 of the dihydroisocytosine bind
to the catalytic aspartates of BACE. The aryl region binds within S3 using interactions that
are essentially the same as for the isocytosines. The methoxy substituent binds deeply in
the S
3
pocket at a new site termed the S3 sub-pocket, where it may undergo a hydrogen
bonding interaction with the Ser229 hydroxyl.
11.9 Summary and Conclusions
When HTS and other conventional methods for lead generation failed to identify useful
starting points as BACE inhibitors, we adopted fragment-based approaches. A library of
approximately 5000 low molecular weight compounds was screened at high concentration
(5 mM) by NMR spectroscopy. Among the numerous hits obtained, many appeared to be
false positives and we eliminated all those that were not displaced from their binding site
by addition of the high-affinity peptidic ligand OM99-2. To measure binding affinities, we
employed a SPR-based assay because conventional enzymatic assays could not be carried
out under the high ligand concentrations needed to evaluate such weak binders. One of the
key components in evaluating the resulting hits was consideration of the ligand efficiency.
Rather than prioritize based on strong affinity, we prioritized based on high ligand efficiency.
As a consequence, many higher affinity hits were discounted if the overall efficiency was
low. We focused on hit
1
which had very weak affinity (2500 M) but high ligand efficiency
(0.32). CrystallizationwithBACE allowed us to evaluate directly themechanismof binding.
With this information, we methylated the N3 position andmodified the aryl region; together,
these changes led to the identification of
14
, which showed 400-fold improved binding.
Once a validated hit has been identified, there exists the obvious potential to find addi-
tional structurally related hits. Screening of compounds similar to
1
led to the identification
of dihydroisocytosine
15
. Despite its weak affinity (IC
50
>1000 M), we were able to use
crystallographic information to make improvements rapidly, leading to
17
(IC
50
79 M).
The SAR knowledge gained in this process was applied to the original isocytosine hit to
explore a related series of extended isocytosines. Structural knowledge, combined with the
evaluation of two focused libraries, allowed us to efficiently identify the high-affinity lead
compound
21
(IC
50
80 nM).
Overall, fragment-based methods succeeded in the identification of useful starting points
as BACE inhibitors where other methods failed. By screening for small compounds and
allowing for weak binders, the hits that emerge tend to represent the minimal binding units.
Among the key consequences are the following. First, by restricting to smaller compounds
and (consequently) a smaller volume of structure space, many fewer compounds need to
be screened. Whereas our high-throughput screens included >500 000 compounds and
delivered no useful hits, our fragment screen included only 5000 compounds and delivered
multiple useful hits.Acorollary to this is that by iteratively evolving fromaminimal binding
unit, there exists greater potential to identify truly novel chemical series. For typical HTS
(test concentration around 10 M), hit identification is dependent on already having the
appropriate compounds in the collection. This may be more true for some target classes and
less so for others according to the target types and chemical classes of historical experience
