Last year Practical Fragments highlighted substrate activity screening (SAS)
as a means for identifying enzyme inhibitors. The idea is to make libraries
consisting of potential substrates, modified to reveal interactions: for
example, amides that release a fluorescent reporter group when cleaved by a
protease. These libraries are then screened against a protein of interest, and
any new substrates identified can be transformed into inhibitors by replacing
the cleavable bond with some sort of warhead. At the end of that post, we asked
why more people weren’t using SAS. In a new paper in ChemBioChem, Pieter Van der Veken and colleagues at the University of Antwerp
The researchers were interested in the
oncology target urokinase (uPA), a trypsin-like protease. They built potential substrates
from an amino-methylcoumarin designed to fluoresce when cleaved. They used this
to assemble a library of 137 molecules, each consisting of the amino-methylcoumarin
coupled to a variable fragment. Of these, about 50 contained positively charged
moieties likely to interact with the large S1 pocket of uPA, which has a
predilection for cationic species. (The rest were diverse molecules.) The
library was then screened against the enzyme to look for substrates, but the
researchers ran into several difficulties.
First, since all the substrates are poor,
the researchers had to use quite a bit of enzyme (about 2.5 micrograms per
well) to get a good signal. Second, for the same reason, they had to run the
assay for a long time (6 hours). Third, and somewhat unexpectedly, it turns out
that SAS is susceptible to an interesting artifact: low levels of contaminating
enzymes can cleave substrates, giving false positives. Indeed, the researchers
found that commercial uPA isolated from human urine produced a number of hits
that did not repeat with recombinant (and presumably purer) enzyme and could
not be competed by addition of a potent uPA inhibitor.
Despite these challenges, the researchers
identified 11 hits. However, notably absent were some of the fragments known to
have affinity for the S1 pocket, such as several guanidines. This is not
surprising: for a molecule to be processed as a substrate it needs to be able
to fit in the S1 pocket as well as to position the cleavable bond near the
catalytic machinery – subtle changes in geometry will prevent processing. This
got the researchers thinking about alternative ways to use their library.
The approach they came up with, “modified
substrate activity screening” (MSAS), starts by first looking for inhibitors rather than substrates, since
a poor substrate can behave as an inhibitor. The idea is to incubate library
members with the enzyme and a single potent substrate. This allowed the
researchers to reduce the enzyme concentration by 10-fold and run a much
shorter assay (10 minutes). It also reduces the risk that contaminating enzymes
will be responsible for activity, though of course inhibition assays are
susceptible to all sorts of other artifacts.
When the researchers applied MSAS to uPA, not
only did they rediscover the 11 hits they had identified as substrates previously,
they also identified 17 additional molecules, including all the
guanidine-containing fragments.
The researchers propose a flowchart for
MSAS in which compounds are first screened for inhibition. These hits are then
followed up using SAS to determine whether some of these inhibitors are substrates
too. Any substrates thus identified can be readily transformed into inhibitors
by adding an appropriate warhead. Inhibitors identified in the first step
that aren’t substrates can also be useful to provide structure-activity relationships and
new fragments to take forward.
Of course, one could argue that if you are
doing inhibition assays, there is no point in going to all the trouble of
making custom libraries for MSAS. That said, if you’ve already got the substrate-based
libraries, doing an initial inhibition screen is probably a good idea.
I'm biased because I worked on the phosphatase/SAS methodology in Jon Ellman's lab and my laboratory recently reported SAS methodology for kinases (Angewandte 2014), but:
ReplyDeleteThe idea that you need to use a lot of enzyme, but that low levels of a contaminating enzyme pose a problem is more than a bit contradictory. If you are looking for substrates for any enzyme, you are going to want to use pure enzyme. But SAS can clearly use low enzyme concentrations, as found with the low amount of contaminating protease giving "false" posititives.
In my experience with SAS, the same concentrations used to screening for enzyme inhibitors are sufficient to screen substrate libraries (for example: we use 30 nM kinase with both SAS and traditional inhibitor screening methods). If you cannot find substrates with a reasonable amount of enzyme, then you need to improve your substrate library.
Screening the substrate libraries as inhibitors should always be done to avoid false negatives that can occur SAS. In fact, we did this with both phosphatase and kinase libraries.
We tried this with a iso-peptidase and ended up with an expensive library with no hits.In the end it turned out the enzyme had an odd biding mechanism where it first bound close to the N-terminal of the ~10 kDa artificial substrate we had been using initially, before the catalytic pocket opened allowing entry of substrate. Our fragment based substrates stood no chance of making this interaction and thus we only got a few false positives.
ReplyDeleteJust goes to show that if the enzymology and crystallography are done well before you start your proposed screen you can avoid slip ups like the above.
To date, not too many teams have provided accounts of their experience with SAS in the literature and there are I think still a number of open questions. Please note that as the authors of the publication mentioned in the post, we do support the rationales behind the original SAS methodology, and share the opinion that screening a SAS-library is a good approach to obtain “hit” fragments and inhibitors of enzymatic targets. The main points we try to put forward in the publication are related to designing a screening experiment that renders a maximal amount of information with a minimal amount of resources. This experiment is what we provisionally call MSAS. Our manuscript covers only one target (urokinase), but we have meanwhile significantly expanded our SAS-library and garnered additional experience with two other (cysteine) proteases. In all of these cases, our findings have been fundamentally identical: when screening only for substrates, you overlook a significant number of valuable fragments.
ReplyDeleteOur latest experiments with the cysteine proteases also did not change our view regarding the issue brought up by Matt Soellner. We do need rather large amounts of enzyme to generate a reliable fluorescence signal during a substrate screen, even for the “best” substrates in the library. Examples from the publication include the alkylguanidines that were inspired by the side chain of arginine: one could reasonably argue that urokinase has evolved to specifically accomodate this functionality and to efficiently process substrates containing it. You would therefore expect the corresponding SAS-members to be good substrates. This was not the case. Evidently, the proteases we selected might have lower intrinsic catalytic turnover efficiencies than the targets that Matt Soellner worked on. Soellner’s targets might also be less dependent on processing assistance by other structural features present in the target’s natural substrates, but not in the fragments. Once more, it would be good to learn about other researchers’ experiences.
Finally, we are still a bit puzzled by the inhibition experiments mentioned by Soellner. It seems that these were carried out on a subset of the substrates/library (?) to assess the possibility of false negatives, and not on the full library for the sake of identifying inhibitors. All in all, the publications do not go in much depth on these experiments and provide little experimental details for the experiments (substrate/enzyme) concentrations used. It is therefore not evident to compare whether Soellner and our results are unisonous for these experiments.