Today is our last paper on high throughput screening (HTS) techniques. We’re back to discovering drugs on this one but the premise is quite different for this particular screen. Whereas other papers we’ve done so far have involved finding novel drugs for known targets or identifying drugs that produce novel behavioral phenotypes, this paper is about finding novel drugs for enzyme targets that are not fully characterized. The paper is: “Identification of selective inhibitors of uncharacterized enzymes by high-throughput screening with fluorescent activity-based probes”, Bachovchin et al., 2009, Nature Biotechnology [PMC].
The genome era has brought in a new age in terms of understanding and identifying the number of genes in mammalian and other genomes. While we have a good idea of where genes lie in genomes and what their structure looks like from all of this sequencing, we do not necessarily understand what these genes do based purely on their sequence. While we can make some good guesses (probably better than a guess) on whether such genes are enzymes, GPCRs or other types of proteins based purely on homology, we cannot necessarily understand their function within cells or whole organisms based purely on sequence data. For almost every protein, we understand how it works within complex systems because we have tools to probe its function. Sometimes these tools involve genetic manipulations or knockdown technologies but more often these tools to probe function depend on pharmacological manipulation of protein function. Therein lies the problem. If you know of an uncharacterized enzyme or receptor and have some idea of its substrates or its receptor class, but not much else, how do you screen for inhibitors or activators of that enzyme or receptor. In the case of enzymes we can generally identify at least one of their substrates based on sequence homology but that doesn’t help you to put together a functional screen in all cases. A way around that is to perform activity based protein profiling (ABPP):
A large fraction of uncharacterized mammalian proteins are enzymes. Genetic and cell biology studies have begun to link some of these enzymes to important physiological and disease processes. However, our lack of understanding of the substrates utilized by uncharacterized enzymes impedes the development of standard HTS assays for inhibitor screening. Sequence homology, on the other hand, can often assign these enzymes to specific mechanistic classes, and this knowledge has been used to develop chemical proteomic tools for their characterization. Prominent among these chemo-proteomic methods is activity-based protein profiling (ABPP).
ABPP makes use of reactive chemical probes to covalently modify the active sites of enzymes. ABPP probes typically exploit conserved catalytic and/or recognition elements in active sites to target a large number of mechanistically related enzymes. Incorporation of fluorescent and/or biotin tags into probe structures enables detection and enrichment/identification, respectively, of protein targets. ABPP has been applied to discover enzyme activities in a wide range of (patho)physiological processes, including cancer, infectious disease, and nervous system signaling. Interestingly, a large number of enzymes identified by ABPP in these studies are uncharacterized (i.e., they lack known substrates). By performing ABPP experiments in a competitive mode, where small-molecules are screened for their ability to block probe labeling of enzymes, lead inhibitors have been generated for some uncharacterized enzymes. An important feature of this approach is that the potency and selectivity of inhibitors can be concurrently optimized because compounds are profiled against a large number of mechanistically related enzymes in parallel.
But, of course, there is a problem. While these types of ABPP assays have been useful in identifying novel inhibitors for classes of enzymes, they are not well suited for HTS. This is the main idea of the paper, generating ABPP assays with HTS capabilities.
A major shortcoming of competitive ABPP studies has, however, been their limited throughput. Assays are typically readout using one-dimensional SDS-PAGE gels, which are not suitable for HTS. As a consequence, only modest-sized compound libraries (200−300 compounds) can be screened using current competitive ABPP methods. Here, we have addressed this major limitation by developing a fluorescence polarization (FluoPol) platform for competitive ABPP. We show that this platform is HTS-compatible and can be readily adapted for use with different classes of enzymes and ABPP probes. Moreover, we further report the use of FluoPol-ABPP to discover selective inhibitors for two cancer-related enzyme targets, the hydrolytic enzyme RBBP9 and the thioltransferase GSTO1.
To understand what they have done here it helps to gain some understanding of the premise behind measuring fluorescence polarization (FluoPol). With fluorescent probes we can measure two things (in general): 1) after exciting a fluorescent probe we can measure the emission spectra; this is the premise behind most fluorescent probe applications such as those you might use in an ELISA to detect the amount of protein labeled with a probe in solution, or 2) when you excite a fluorescent probe with polarized light you can measure the degree of polarization of the light emitted by the fluorescent probe. This second approach is what the authors of this paper have used. The approach is based on a simple observation: when a fluorescent molecule is in solution it will rotate or “tumble” based on the size (among other things) of the molecule. If the molecule is stationary it will emit light with the same polarization as the light used to excite it. If the molecule is tumbling, it will emit light that is “scrambled” in terms of polarization. The degree of polarization of emitted light can be measured by a machine and this can tell you about the status of the molecule in solution (e.g. is it tumbling or is it stationary) — this is explained in more detail here. In this screening technique they have taken advantage of the size principle to create what they call “fluorescent polarization activity based protein profiling (FluoPol ABPP)”. If the probe is in solution it will tumble and the degree of polarized emission will be small due to scrambling. On the other hand, if the probe is interacting with another protein (which is larger) its degree of tumbling will be reduced and it will be more stationary in solution. This will shift the degree of polarization to a more polarized emission spectra. Hence, polarized emission is indicative of interaction with a protein whereas low levels of polarization is indicative of free probe in solution. Hopefully you can see that this turns the assay into what essentially is a typical binding assay but with a different output than we typically use. This allows for competition binding assays to be performed in a HT fashion as described in the present paper.
Another key to understanding the present paper is knowledge of how they chose their activity-based probes. They look at two main enzymes in the paper: 1) an enzyme called RBBP9 and 2) an enzyme called GSTO1. Neither of these enzymes are fully characterized in terms of in vivo substrates (much more is known about GSTO1 than RBBP9 and they spend most of their time on RBBP9) or discovery of selective inhibitors. However, for both enzymes, at least one interacting compound is known for the generation of a fluorescent probe. In the case of RBBP9, structural genomic and proteomic studies have indicated that this protein is a member of the serine hydrolase superfamily and this led to the discovery that the enzyme interacts with fluorophosphonate (FP) in solution. Hence, the authors conjugated FP to rhodamine (a common fluorescent molecule) and used this for the basis of their HTS. A key factor here is that FP interacts with most serine hydrolases so their screen is by no means specific but they have ways to get around this to look at specificity, as we will discuss shortly.
Okay, onto the screen. With recombinant RBBP9 in hand, and after determining equilibrium conditions for FP-rhodamine interactions with RBBP9 in solution (as measured by FluoPol), the authors used FP-rhodamine and RBBP9 to screen 18,974 compounds for RBBP9/FP-rhodamine disruption of FluoPol (a competitive binding assay) in 384 well plates. They found that 35 compounds from the screen disrupted FP-rhodamine/RBBP9 interactions suggesting that these compounds might be inhibitors of RBBP9. They then used the more traditional “in gel” ABPP assay to verify these hits and found that 20 of them held up under these conditions with IC50s in an acceptable range (~20uM or less). From this they conclude that they have discovered 20 potential inhibitors of RBBP9. All of these experiments were done with recombinant RBBP9 in solution so these experiments tell us little about specificity or whether these inhibitors are capable of interacting with RBBP9 in a native environment.
This problem was addressed by either mixing RBBP9 with mouse brain proteins or looking at RBBP9/compound interactions in proteins from Cos-7 cells expressing RBBP9 through transfection. Looking at Fig 4 we can get a better idea of how this works.
In A and B the control lanes are DMSO and MAFP is used as a general interacting compound with other serine hydrolases but not RBBP9. Of the 5 compounds tested (1-5) in this assay most of them interact with other proteins in the proteomes (you can see this as lighter bands at MWs compared to DMSO) in addition to RBBP9 but one of them appears to be specific for RBBP9 interactions. This compound (1) is emetine, a natural product that is known to have effects on protein synthesis (which may or may not be related to its interactions with RBBP9 but this is unlikely based on some SAR done in the paper). They go onto show that emetine is a competitive inhibitor of RBBP9 and that it is fairly potent (IC50 ~ 5uM). Hence, this HTS assay for RBBP9 inhibitors yields a potent and specific inhibitor of RBBP9 that might be useful in gaining a better understanding of the function of this previously uncharacterized enzyme, that is associated with a number of disease states, in cells.
They go on to do more or less the same thing with GSTO1 and, in this case, they discover another compound which is selective and even more potent (IC50 ~ 400nM). Hence, they conclude that FluoPol-ABPP HTS can be used as a general method to discover novel inhibitors of uncharacterized enzymes:
Complete genome sequences have revealed that eukaryotic and prokaryotic organisms universally possess a huge number of uncharacterized proteins6. Even for proteins that may be considered ‘annotated’, we have yet, in most instances, to achieve a complete understanding of their biochemical, cellular, and physiological functions. A central component of efforts to annotate the proteome is the development of selective pharmacological probes to perturb the function of individual proteins in native biological settings. HTS has assumed a prominent role in small-molecule probe development in both academia and industry, as exemplified by the National Institutes of Health Molecular Libraries Screening initiative. The success of such endeavors hinges on the advancement of high-quality screens, which is particularly challenging for proteins of poorly characterized biochemical function. Here, we have introduced FluoPol-ABPP as a general solution to this problem for a potentially wide range of enzymes.
Bachovchin, D., Brown, S., Rosen, H., & Cravatt, B. (2009). Identification of selective inhibitors of uncharacterized enzymes by high-throughput screening with fluorescent activity-based probes Nature biotechnology, 27 (4), 387-394 DOI: 10.1038/nbt.1531