16 October 2017

Docking for finding and optimizing fragments

Docking can sometimes seem like the Rodney Dangerfield of FBDD: it don’t get no respect. In last year’s poll of fragment finding methods, computational approaches ranked in seventh place. This partly reflects the largely biophysical origins of FBDD, but it is also true that ranking low affinity fragments is inherently challenging. Still, the continuing rise in computational power means that methods are rapidly improving. A recent paper in J. Med. Chem. by Jens Carlsson and collaborators at Uppsala University, the Karolinska Institute, and Stockholm University illustrates just how far they can take you.

The researchers were interested in the enzyme MTH1, whose role in DNA repair makes it a potential anti-cancer target. The crystal structure of the protein bound to an inhibitor had previously been reported, and this was used for a virtual screen (using DOCK3.6) of 300,000 commercially available molecules, all with < 15 non-hydrogen atoms, from the ZINC database.

Finding fragments is one thing, but one really wants slightly larger, more potent compounds to begin lead optimization. Thus, the top 5000 fragments were analyzed to look for analogs with up to 6 additional non-hydrogen atoms among the 4.4 million commercial possibilities. This led to 118,421 compounds, each of which was then virtually screened against MTH1. Of the initial 5000 fragments, the top 1000 that had at least 5 analogs with (predicted) higher affinity were manually inspected. Of 22 fragments purchased and tested in an enzymatic assay, 12 showed some activity, with the 5 most active showing IC50 values between 5.6 and 79 µM and good ligand efficiencies.

Since each of these fragments had commercially available larger analogs, the researchers purchased several to see if these did indeed have better affinities. Impressively, this turned out to be the case: both compounds 1a and 4a bound more than two orders of magnitude more tightly than their fragments. Interestingly, while the researchers were unable to obtain crystal structures of fragments 1 and 4 bound to MTH1, they were able to obtain crystal structures of 1a and a close analog of 4a, and these bound as predicted.


Of course, not everything worked: in the case of one fragment, among 19 commercial analogs purchased, the best was only 7-fold better. The crystal structure of this initial fragment bound to MTH1 was eventually solved, revealing that it bound in a different manner than predicted, thus explaining the modest results. In another case the most interesting commercial analogs turned out not to be available after all, but during the course of the study a different research group published a low nanomolar inhibitor with the same scaffold.

One notable aspect of this work is going from fragments to more potent leads without using experimentally determined structural information, something the majority of respondents in our poll earlier this year said they would not attempt. Although such advancement is not unprecedented, published examples are still rare.

In some ways this work is similar to the Fragment Network approach we highlighted last month, the key difference being that while Fragment Network was focused on looking for other fragments, this is focused on finding larger molecules. But how general is it? The researchers found that, while there are a median of just 3 commercial analogs in which a fragment is an exact substructure of a larger molecule, this increases to 700 when the criterion is relaxed to similarity (for example compound 1 and 1a). These numbers undoubtedly become even more favorable for organizations with large internal screening decks.

Eight years ago I ended a post about another successful computational screen with the statement that “the computational tools are ready, as long as they are applied to appropriate systems.” This new paper demonstrates that the tools have continued to improve. I expect we will see computational fragment finding and optimization methods move increasingly to the fore.

09 October 2017

Fragments vs ketohexokinase (KHK) deliver a chemical probe

Anyone who has paid any attention to health news will be aware of concerns over high fructose corn syrup. Just a few years ago the stuff was ubiquitous. Today, due to consumer backlash, it is less common, though still widely used as a cheap sweetener in foods and beverages.

In humans fructose metabolism, unlike glucose metabolism, is not regulated by feedback inhibition, so the sugar is metabolized preferentially. Overconsumption of fructose has been correlated with all sorts of metabolic disorders, from insulin resistance to obesity. But even if you avoid consuming any fructose, your body can still convert glucose into fructose.

The rate-determining step in fructose metabolism is the enzyme ketohexokinase (KHK). Mice lacking this enzyme are healthy and resistant to metabolic diseases. Could a pill do the same thing? Although previous KHK inhibitors have been reported – one starting from fragments discussed here – these did not seem suitable for in vivo studies, not least because they are considerably less potent on rat KHK than human KHK. In a recent J. Med. Chem. paper, Kim Huard and her colleagues at Pfizer describe a chemical probe for KHK.

The researchers used STD NMR to screen their 2592-fragment library in pools of 4 or 10 compounds, with each at 240 µM. This resulted in a formidable 451 hits, of which 448 were screened in full dose response curves using SPR. Of these, 179 confirmed, and 114 had affinities better than 100 µM. All of the SPR-validated hits were tested in an enzymatic assay, leading to 23 fragments with IC50 values from 46 to 439 µM. All 23 of these were soaked into crystals of KHK, and all of them yielded structures showing them bound in the ATP-binding pocket. (Incidentally, this is a lovely example of a successful screening cascade using multiple orthogonal methods, though it would be interesting to know what the outcome would have been had the researchers jumped directly to the X-ray screen.)

But what do you do with 23 fragment hits, all with decent ligand efficiencies and experimentally determined binding modes? Rather than focusing on a single fragment, the researchers noticed that many shared common features, for instance a central heterocycle surrounded by various lipophilic substituents, as in compounds 4 and 5. Many, such as compound 4, also contained a nitrile that made a hydrogen bond to a conserved water molecule.


Next, the researchers combed the full Pfizer screening library for compounds that merged common elements of the fragment hits. This led to more potent inhibitors, such as compound 9 (which was present in the library as a racemate – make sure to vote in the poll on the right!). Parallel chemistry around analogs of this and another hit led to compound 12. In contrast to previously reported molecules, this compound is equipotent on rat and human KHK. It also has decent pharmacokinetics, is orally bioavailable, and is quite selective against a broad panel of off-targets. Rat experiments revealed that the compound inhibits fructose metabolism in vivo.

This story is a nice illustration of how lots of different crystal structures can enable fragment merging. There is still some way to go – the potency in particular could be improved. Also, there are actually two human isoforms of KHK, and compound 12 hits both equally – which may or may not be desirable. Nonetheless, this chemical probe should help further elucidate KHK biology, and help to address whether the enzyme is druggable, or merely ligandable.

02 October 2017

Dynamic combinatorial chemistry revisited: why it’s so difficult

Last year we discussed the application of dynamic combinatorial chemistry (DCC) to fragment linking. The idea is that a protein will shift the equilibrium of a reversible reaction, selecting the tightest binder. Over the past twenty years practitioners of DCC have generated plenty of papers, some quite nice, but I do not recall seeing examples of the technique generating novel and attractive chemical leads. A new paper in Chem. Eur. J. by Beat Ernst and colleagues at the University of Basel explains why it is so difficult.

The researchers were interested in the bacterial protein FimH, which helps microbes colonize the urinary tract by adhering to human proteins that are decorated with mannose. The chemistry the researchers decided to explore for DCC was the reaction of aldehydes with hydrazides to form acylhydrazones. This reaction is slowly reversible at pH 7, allowing exchange between library members to occur, but it can be essentially frozen by raising the pH.

To try to understand every aspect of their system, the researchers focused on a tiny library. Two aldehydes were chosen, one based on mannose, the other based on glucose. Four (quite similar) commercially available hydrazides were purchased.

The researchers made and tested the affinity of each of the eight possible library members using surface plasmon resonance (SPR). The four acylhydrazones based on mannose had dissociation constants (KD) ranging from 0.33 to 0.76 µM, while the mannose aldehyde came in at 3.2 µM. In contrast, the four acylhydrazones based on glucose had KD values between 152 and 735 µM, comparable to the glucose aldehyde itself (194 µM). Since mannose is the natural ligand for FimH while glucose is not, this was expected.

One challenge of DCC is separating library members from the protein for analysis; releasing bound ligands can be particularly challenging if they bind tightly to the protein. A variety of methods were tested, including microfiltration, but this gave “massive alterations in composition.” Various attempts at protein denaturation and precipitation using organic solvents or heat also failed. The fact that this step was so difficult, even for closely related ligands (the difference between mannose and glucose is the stereochemistry around a single hydroxyl group) underscores the challenge of analyzing DCC mixtures.

The problem was finally solved by using a biotinylated version of FimH which could be captured using commercial streptavidin agarose beads.

The most general approach works as follows.

1. Incubate 100 µM FimH protein with library (with each aldehyde and hydrazide at 50-200 µM) at pH 7 for 3 days in the presence of 10 mM aniline, which catalyzes the acylhydrazone exchange.

2. Raise the pH to 8.5 to stop the reaction, add streptavidin agarose, centrifuge, and discard the supernatant containing the unbound molecules.

3. Resuspend the agarose beads containing the protein, add a competitor to release bound ligands, increase the pH to 12 to ensure release, and analyze the product ratios using HPLC.

Although cumbersome, this protocol does work: mannose-derived compounds were enriched relative to glucose-derived compounds, as expected due to their higher affinities, and the most potent compound was enriched over the less potent ones. That said, the robustness of the results were dependent on the ratios of library components.

So will DCC ever be practical? I’m not so sure. But, as the researchers end hopefully but not hypefully, their work “is a contribution to this challenge.”