Review of 2016 reviews

This year is finally coming to an end, and as we've done for the past four years, Practical Fragments will highlight some of the reviews that we didn't cover previously.

In terms of what we did cover, there were several excellent events, including the eleventh annual CHI FBDD Conference in San Diego, an inaugural meeting in Houston, and of course the first-ever major fragment event in Boston, FBLD 2016.

The twentieth anniversary of SAR by NMR was also commemorated by the eighth book devoted to FBLD, as well as a massive two volume work on lead generation. We also covered a special issue of Molecules and reviews on clinical candidates and library design.

Another review on library design was published recently in Drug Disc. Today by Ian Gilbert, Paul Wyatt, and colleagues at the University of Dundee. The researchers have built a set of 356 diverse compounds consisting of “capped” scaffolds, such that any hits could be rapidly expanded. Undergraduates did much of the actual library assembly, learning skills such as parallel chemistry and how to work with polar compounds. There is lots of nice detail in this paper, including on library storage conditions.

Targets

Practical Fragments often highlights successful fragment to lead programs, and these were the focus of a Perspective in J. Med. Chem. by Christopher Johnson (Astex) and collaborators: all 27 cases published in 2015 in which the affinity of a fragment was improved at least 100-fold to a 2 µM or better lead. Many of these were covered in Practical Fragments, including BTK, DDR1/2, ERK2, MELK, Mtb TMK, PKCθ, RET, FactorXIa, MMP-13, BCATm, PDE10A, soluble epoxide hydrolase, tankyrase, ATAD2, MCL-1, RAD51, XIAP/cIAP, and mGluR5. The paper also draws general conclusions about target types, molecular weights, cLogP values, and LE.

Targeting tuberculosis (TB) is the subject of two reviews from University of Cambridge researchers, one in Drug Discov. Today by Vitor Mendes and Tom Blundell and one in Parasitology by Anthony Coyne, Chris Abell, and colleagues. Fragment-based approaches have been more or less successful against several TB proteins, including pantothenate synthetase, CYP121, BioA, EthR, and thymidylate kinase, while other targets – such as shikimate kinase and CYP144 – have proven more difficult.

July was bromodomain month at Practical Fragments, and this target class is the subject of a review in Drug Discov. Today: Technol. by Dimitrios Spiliotopoulos and Amedeo Caflisch at the University of Zurich. The focus is on computational fragment screening methods, with examples for BRD4 and CREBBP. And while we’re on the topic of computational methods, Olgun Guvench of SilcsBio has a brief review in Drug Discov. Today on computational functional group mapping.

Rounding out target-focused reviews, Paramjit Arora and colleagues at New York University focus on protein-protein interactions (PPIs) in a Trends Pharm. Sci. paper. This covers multiple approaches to finding PPI inhibitors, including fragment-based, and also touches on hotspots and structure-based design.

Biophysics

It is impossible to imagine FBLD without biophysics, and this is the topic of an authoritative review in Nat. Rev. Drug Disc. by Jean-Paul Renaud (NovAliX), Chun-wa Chung (GlaxoSmithKline), U. Helena Danielson (Uppsala University), Ursula Egner (Bayer), Michael Hennig (leadXpro), Rod Hubbard (University of York) and Herbert Nar (Boehringer Ingelheim). In addition to covering all the major techniques, the paper does a great job of delving into some of the more obscure and emerging methods, providing an excellent discussion of the throughput and requirements for each technique as well as the kinds of information obtained. Although the review is broader than FBLD, the application of biophysical techniques to fragments is a major theme. The researchers also remind us that, “contrary to the belief that all drug discovery challenges are best solved through the introduction of new technologies, substantial advances can also be driven by innovative application.”

Individual biophysical techniques also received plenty of attention over the year, including three on NMR. The first, by Alvar Gossert and Wolfgang Jahnke (Novartis) in Prog. Nucl. Magn. Reson. Spectrosc., is a 44-page practical guide to identifying and validating protein ligands. This contains a wealth of information on most of the NMR methods you will ever likely encounter; it includes a handy chart summarizing the molecular weight and concentration limits for each technique, suggested workflows, and thorough discussions of potential pitfalls. The review may appear daunting to the novitiate – it is replete with equations and pulse sequences – but the writing is clear. In the end, much comes down to the concept of the “validation cross”, a rubric for assessing the integrity of both ligand and protein, and evaluating binding effects on both ligand and protein.

Two additional reviews, both from William Pomerantz and colleauges at the University of Minnesota, focus specifically on protein-observed ¹⁹F NMR. The first, a Perspective in J. Med. Chem., is a good general introduction. Despite being the 13^th most abundant element on our planet, only five natural products are confirmed to contain fluorine. Introducing this element into proteins – as has been done in more than 70 cases – can be a useful approach for discovering new ligands. And if you want to start doing this yourself, a paper in Nature Protocols provides practical details.

Turning to other biophysical techniques, surface plasmon resonance (SPR) continues to be very popular, and is reviewed by Alain Chavanieu and Partine Pugnière in Expert Opin. Drug Discov. The paper provides a good general overview on using SPR for FBLD, covering the theory, history, various screening strategies, comparison to other methods, recent applications to a variety of different targets, and a suggested workflow.

Calorimetry is less commonly used for fragment screening, even though it can provide thermodynamic data. Michael Recht and collaborators at the Palo Alto Research Center and Zenobia discuss both enthalpy arrays as well as more conventional isothermal titration calorimetry (ITC) in a Methods Enzymol. chapter.

Chemistry

But while biophysics is important, FBLD would be nowhere without chemistry. In MedChemComm, Stefan Kathman and Alexander Statsyuk (then Northwestern, now University of Houston) review one chemical approach, covalent tethering. This touches on the original reversible (thermodynamically-controlled) disulfide tethering approach developed back at Sunesis but is primarily focused on irreversible (kinetically-controlled) methods. The paper does an excellent job summarizing challenges, potential pitfalls, design rules, and recent successes. As of early this year the Statsyuk lab had sent their 100-member covalent fragment library to nine different research groups, three of which had already identified hits. The review ends with some provocative questions, and it will be fun for practitioners to answer them as covalent approaches garner increasing attention.

Another chemical technique we’ve touched on is substrate activity screening (SAS), and this is reviewed in ChemMedChem by Pieter Van der Veken and collaborators at the University of Antwerp. All published examples are summarized, including the modified approach developed by the Van der Veken lab; some unpublished data are also discussed. The paper also includes a good general section on the subtleties and complexities of transforming substrates into inhibitors.

Finally, if all this is a bit too much, a good general review on FBLD was published in Pharmacol. Ther. by Martin Scanlon and colleagues at Monash University. This concise but thorough paper covers theory, history, library design, hit finding and characterization, and select clinical success stories. The longest section is devoted to chemical strategies for elaborating fragments, and includes some of the less commonly used methods such as target-guided synthesis, Tethering, and off-rate screening.

And that’s it for this year. Thanks for reading, and especially for commenting. Take care, do important work, and may 2017 be better than we can reasonably hope.

Fragments vs Lp-PLA2: A new hope

A few months ago we highlighted work out of Astex and GlaxoSmithKline describing the fragment-based discovery of inhibitors of lipoprotein-associated phospholipase A2 (Lp-PLA₂), an inflammatory disease target. Although low nanomolar compounds were identified, they had high clearance in rats. In a new J. Med. Chem. paper the team – led by Alison Woolford of Astex and Vipul Patel of GlaxoSmithKline – describes a completely new series of molecules with better pharmacokinetic properties.

Recall that the researchers had previously solved the crystal structures of 50 fragments bound to the “canyon-like” active groove of Lp-PLA_2. Hydantoin 3 was one of these, and although it had no detectable activity in a biochemical assay, it did make contacts with residues in the catalytic site of the enzyme. A virtual screen of 16,000 related compounds identified 33 potential hits, and crystallographic and biochemical screening of these led to compound 5, with low micromolar activity.

The researchers were able to trim back the cyclohexyl group and remove one of the carbonyls with only a modest loss in affinity. They could also take advantage of extensive structural information from other fragment hits. For example, adding a nitrile from another fragment produced compound 13, with improved affinity.

Next, the researchers turned to the left side of the molecule, adding substituents to make a stacking interaction with a tryptophan residue in the protein – an interaction seen previously with a uracil fragment. Simple aromatic rings worked, but aliphatic heterocycles such as amines and sulfones were even better, with compound (S)-23 being among the best.

Although compound (S)-23 has only high-nanomolar potency in a biochemical assay, it is equipotent with darapladib in a whole plasma assay – despite the fact that darapladib is a picomolar inhibitor in the biochemical assay. The researchers attribute this difference to the fact that darapladib, which reached phase 3 trials, is a poster child for molecular obesity, while (S)-23 comes in with a svelte molecular weight below 400, very low plasma protein binding, and a solubility of at least 3.5 mM. The molecule is also permeable, does not inhibit CYP450s, is selective against the closely related PLA2-VIIB, and has low clearance in dogs. The clearance is higher in rats, but a closely related compound is better and also has high oral bioavailability.

This paper provides another example of finding a fragment with no detectable activity and advancing it to an attractive series. It illustrates the power of crystallography to reveal useful fragments as well as the importance of crystallography during lead optimization. Darapladib failed in two massive phase 3 clinical trials for cardiovascular disease, which probably poisoned GlaxoSmithKline's appetite for Lp-PLA₂. Still, if future biological discoveries suggest new indications for this target, molecules from this series may provide a path back into the clinic.

Fragments vs COMT revisited

Catechol O-methyltransferase (COMT) metabolizes neurotransmitters such as dopamine and is a validated target for Parkinson’s disease. In theory other diseases could be treated with COMT inhibitors too, but most of these contain – like dopamine itself – catechol moieties, and thus have lousy pharmacokinetics and poor brain penetration. Catechols in general are best avoided, and in a recent paper in J. Med. Chem. María Sarmiento and colleagues at Roche have found an alternate scaffold.

COMT is a magnesium-dependent enzyme, and the catechol binds to the magnesium ion. The researchers decided to target the pocket that binds the cofactor, S-adenosyl-L-methionine (SAM). They screened 6000 rule-of-three compliant fragments at 200 µM using surface plasmon resonance (SPR). First they examined wild-type enzyme in the absence of magnesium and SAM, and they also counter-screened against six variants containing mutations in the SAM binding site to exclude fragments that bound elsewhere. Even after this specificity profiling 600 hits remained. Dose-response curves whittled the number down to 200, all of which were examined using ligand-detected (CPMG) NMR. Hits from CPMG NMR were further studied using protein-detected (¹H/¹⁵N HSQC) NMR. Finally, all 600 of the hits from the initial SPR screen/counter-screen assays were tested in an enzymatic assay. Only four fragments made it through all of these filters, three of which were pyrazoles such as compound 1.

Two years ago Teddy highlighted a paper from Takeda also focused on COMT, and there too pyrazoles predominated – an observation that didn’t escape the Roche researchers. In fact, compound 1 in the current paper is almost identical to compound 5 in the Takeda paper. Substructure searching and screening led to compound 4, which is identical to compound 7 in the Takeda paper. Whether COMT is really this choosy when it comes to fragment hits, or whether this reflects similarities in fragment libraries remains an open question.

But happily there’s more. The researchers used an iterative structure-guided fragment-growing approach to improve affinity. This ultimately resulted in compound 24, which is competitive with SAM and has mid-nanomolar activity. The solubility could be improved, and no other biological data are presented, but at least this paper demonstrates that it is possible to find potent inhibitors of COMT that are not phenols or catechols. 

Molecules special issue: Developments in Fragment-Based Lead Discovery

Last December the first-ever Pacifichem symposium on FBLD was held in Honolulu. Two of the organizers, Martin Scanlon and Ray Norton, invited participants to submit manuscripts to a special issue of Molecules, which has now published.

The collection starts with a very brief Foreword by me describing the Symposium itself. The first actual paper, from Qingwen Zhang and collaborators at the Shanghai Institute of Pharmaceutical Industry, WuXi AppTec, and China Pharmaceutical University, focuses on kinase inhibitors. The researchers examine fragment-sized substructures of 15 approved drugs that inhibit kinases and use these to design a high-nanomolar inhibitor of the V600E mutant form of BRAF, which modeling suggests should bind to the protein in the “DFG-out” conformation.

Next comes a fragment-finding paper from Thomas Leeper and collaborators at the University of Akron and the University of North Carolina, Chapel Hill. The researchers were interested in finding inhibitors of the glutaredoxin protein (GRX) from the pathogen Brucella melitensis, which causes brucellosis. An STD NMR screen of 463 fragments (each at 0.5 mM in pools of 5-7) resulted in 84 hits, though 75 also hit human GRX. Subsequent experiments including chemical shift perturbation and modeling identified a mM binder with modest selectivity over the human enzyme. Next, the researchers introduced several covalent warheads (including a rather exotic ruthenium analog), one of which led to improved affinity, though the stoichiometry was not determined.

The remaining papers are all reviews, starting with one on native mass spectrometry (MS) by Liliana Pedro and Ronald Quinn at Griffith University. This provides a good historical, theoretical, and practical overview of the technique generally, as well as various applications for fragment-screening. It also covers most of the published examples and discusses both the strengths (such as speed and low protein consumption) as well as the weaknesses (false positives and false negatives) of native MS.

NMR is up next, with a paper by Pacifichem organizer Ke Ruan and colleagues at the University of Science and Technology of China, Hefei. This provides a concise but detailed description of library design, ligand- and protein-detected fragment screening, structural model generation, and hit to lead optimization.

Protein-directed dynamic combinatorial chemistry (DCC) is tackled by Renjie Huang and Ivanhoe Leung, both at the University of Auckland. In addition to summarizing the theory and various literature examples, the authors do an excellent job covering the pros and cons of different types of chemistries and analytical techniques.

Next comes a review by Begoña Heras and collaborators at La Trobe University and Monash University on the subject of bacterial Dsb proteins, which are essential for disulfide bond formation in virulence factors. The review covers the biology as well as several approaches to finding inhibitors, some of which we’ve previously covered (here and here). There is much more to do: as the researchers conclude, “the development of Dsb inhibitors is still in its infancy.”

Finally, Ray Norton and colleagues at Monash University discuss applications of ¹⁹F NMR for fragment-based lead discovery. In addition to covering fluorine-containing fragments, the researchers also discuss using fluorine-containing probe molecules and – even more unusual – fluorine-labeled proteins, in this case using 5-fluorotryptophan. The paper includes previously unpublished results on how these latter two approaches can be used to understand protein-ligand interactions.

One nice feature of this journal is that it is open-access, so if you are lucky enough to be back in Hawaii this December you can pull up the papers on your smartphone while lying on the beach.