Twenty seven hits against four tuberculosis targets

1.2 million deaths. If you did not read the title of this post carefully you may assume this statistic refers to COVID-19. In fact, it is the number of people who died of tuberculosis in 2019. Worse, drug resistant forms of Mycobacterium tuberculosis, the organism that causes TB, are spreading far faster than new treatments are being developed. Initial efforts at addressing this problem are reported (open access) in Comp. Struct. Biotech. J. by Sangeeta Tiwari (University of Texas El Paso), Vitor Mendes (University of Cambridge) and a multinational team of collaborators.

 

M. tuberculosis is capable of making all 20 amino acids. The bug can also scavenge arginine from its host, but only inefficiently: knocking out the biosynthetic pathway abolishes virulence. Thus, targeting this pathway might lead to new drugs.

 

In total eight enzymes are needed to synthesize L-arginine from L-glutamate, and the researchers targeted four of them. The proteins were screened against a library of 960 fragments (each at 5 mM) using differential scanning fluorimetry (DSF). Depending on the specific target some of the hits were validated by SPR or ligand-based NMR before being taken into crystallography, which yielded structures of all the enzymes. In total 13 fragments were found to bind to ArgB, 4 bound to ArgC, 2 bound to ArgD, and 8 bound to ArgF. All the coordinates have been deposited in the protein data bank, though they don’t seem to have been released as of June 28.

 

The paper details the binding interactions for all the hits. Most of them are quite weak, though two hits against ArgB have low micromolar dissociation constants as assessed by ITC. Tantalizingly, these inhibit the growth of M. tuberculosis, and one of them seems to be on-target (adding arginine to the media rescues the inhibition). All the ArgB fragments bind not at the active site but rather at an interface between protein subunits. Unfortunately this site is quite hydrophobic, as are the fragments, suggesting an uphill battle in optimization.

 

A good antibiotic should not hit human proteins, and neither ArgB nor ArgC have human orthologs. ArgF does, but the region where the fragments bind is quite different. ArgD, with only two crystallographically-confirmed hits and 36% identity to the human enzyme, is probably the least attractive.

 

A year before the COVID-19 Moonshot launched we highlighted the Open Source Antibiotics initiative. I don’t think that team was involved with this research, but they would seem to be a natural fit. If you have spare bandwidth and are looking to do some fragment to lead optimization, this paper provides more than two dozen starting points.

Fragments vs ASH1L: a chemical probe

It has been a while since Practical Fragments last covered epigenetic targets that bind to or remodel chromatin. The protein ASH1L (absent, small, or homeotic-like 1) is a molecular leviathan containing some 3000 amino acids and multiple domains, including three “readers” such as bromodomains that bind to acetylated lysine residues. ASH1L also contains a SET “writer” domain that transfers a methyl group to specific histone lysine residues. This domain is the subject of a new Nat. Commun. paper by Tomasz Cierpicki, Jolanta Grembecka, and a large group of collaborators at University of Michigan Ann Arbor and elsewhere.

 

Previous work implicated ASH1L in a particularly lethal type of leukemia with chromosomal translocations in a protein called MLL1. The researchers found that the ASH1L SET domain was essential for leukemogenesis, suggesting that a small molecule inhibitor could be an effective anticancer strategy.

 

A two-dimensional ¹H-¹⁵N TROSY-HSQC NMR screen of 1600 fragments in pools of 20, each at 250 µM, was conducted against the ASH1L SET domain. Compound 1 binds weakly, but chemical shift perturbations mapped the binding site near an autoinhibitory loop, suggesting the molecule could lock the protein in an inactive conformation. Fragment growing led to compound 2.

 

Further fragment growing led to AS-5, with low micromolar affinity as assessed by isothermal titration calorimetry (ITC) as well as functional activity in an enzymatic assay. At this point the researchers obtained a crystal structure, which confirmed binding of AS-5 near the autoinhibitory loop. The thioamide forms a chalcogen bond to a backbone carbonyl, a relatively uncommon interaction. Further structure-based design ultimately led to AS-99, with high nanomolar activity as assessed both biochemically and by ITC. The molecule is selective against a panel of 20 histone methyltransferases and 30 kinases.

 

Despite having relatively modest affinity, AS-99 inhibited growth in MLL leukemia cells and caused them to differentiate. Further mechanistic studies confirmed that this was due to blocking enzymatic activity of ASH1L and downregulation of target genes. Importantly, a negative control molecule in which the sulfur of the thioamide was replaced with an oxygen showed dramatically lower activity both biochemically and in cells. AS-99 has sufficiently favorable pharmacokinetic parameters for i.p. dosing, and a mouse xenograft model showed reduced tumor growth.

 

This paper is a nice example of academic lead discovery, and AS-99 looks to be a useful chemical probe. There is still much to do to optimize the molecule; the thioamide in particular will raise eyebrows among medicinal chemists. Nonetheless, as the researchers point out, AS-99 is a first-in-class molecule that should facilitate further pharmacological understanding of the role of ASH1L in leukemia.

Sociable vs unsociable fragments – and poll!

The last two posts described new fragment libraries, one of which is available for screening. But what happens once you get a hit? More than five years ago we highlighted the importance of synthetic tractability for fragments. In a recent RSC Med. Chem. paper, Jeffrey St. Denis and colleagues at Astex assess how easily (or not) researchers can work with the fragments they find.

 

The researchers coin the phrase “fragment sociability”:

 

An unsociable fragment is one that has limited (if any) synthetic methodology to enable growth vector elaboration and few commercially available close analogues. In contrast, a sociable fragment is one supported with robust synthetic methodology that enables every growth vector to be elaborated and a significant number of commercially available close analogues.

 

To illustrate, they summarize three unrelated fragment-to-lead programs from Astex and, for each case, compare an unsociable fragment with a sociable one. For the unsociable fragments, fewer than 10 analogs were commercially available, and if synthetic routes existed, they were lengthy. In contrast, the sociable fragments had from >100 to >1000 commercially available analogs and well-precedented syntheses. Three of the fragments (mexiletine, tetramisole, and efaroxan) are actually drugs targeting other proteins, but surprisingly only mexiletine is sociable. Astex advanced all three sociable fragments to nanomolar inhibitors and abandoned the unsociable fragments.

 

How can you tell if a fragment is sociable or not? The researchers use the Fragment Network (previously described here) as a computational tool to assess commercial availability of analogs. For synthetic tractability, a medicinal chemist should inspect apparently unsociable fragments, as a purely computational approach might miss “double scaffolds,” which consists of two sociable fragments linked together. While the combination may superficially appear unsociable, simple chemistry could be used to connect the two smaller fragments. Similarly, functional group transformations (such as oxidation) might make a sociable fragment appear unsociable, or vice versa.

 

The researchers analyzed the 1651 members of the Astex core screening library and found only 30 possibly unsociable members. More careful examination revealed that only 12 of these are truly unsociable, and even for these, minor changes could make them sociable.  

 

I confess that I’m both surprised and impressed at the low number of unsociable fragments, though as the researchers point out the Astex library has been continuously refined for the past couple decades. Also, if “more esoteric fragments were socialized then they would… provide an opportunity to identify novel starting points in drug discovery.”

 

I’m a huge fan of developing new fragment chemistries. But from a practical standpoint, how much of a need is there? After all, in these three fragment-to-lead examples the researchers were able to obtain potent molecules. FBLD often generates an abundance of riches: 76 and 105 crystallographic hits in two of the programs described.

 

In the spirit of inquiry, Practical Fragments launches a new poll (right side of page) asking how often you’ve encountered synthetic challenges optimizing a fragment, and whether this has impeded a fragment to lead program. Please vote and leave comments!

A minimal fragment library for maximal coverage of pharmacophore space

Last week we described a fragment library built with the aid of machine learning and designed to contain privileged fragments that should produce high hit rates. Unfortunately, only about a tenth of the library members are commercially available, so it will be some time before we know whether the design was successful. We continue the theme of fragment libraries with a just published Nat. Commun. paper by György Keserű (Hungarian Research Centre for Natural Sciences) and a large group of multinational collaborators (see also here for a nice summary by György).

 

The researchers started by analyzing more than 3300 crystal structures of protein-fragment complexes in the protein data bank. Fragments were defined as having 10-16 non-hydrogen atoms, and the computational approach FTMap was used to ensure that fragments were binding at hotspots as opposed to spurious, less ligandable sites. This exercise yielded 3584 fragments, but many of them were identical or very similar to one another. The researchers used a series of computational tools to cluster similar fragments (or pharmacophores) and choose a set that would maximize diversity. This ultimately led them to assemble a library of just 96 fragments, purchased from five vendors.

 

This SpotXplorer0 library mostly follows the rule of three, with 7 to 17 non-hydrogen atoms, MW 100-250 (or 280 for bromine-containing molecules), ≤ 3 hydrogen bond donors, ≤ 8 hydrogen bond acceptors, and ≤ 3 rotatable bonds. In addition, all members have 1-3 rings, no more than a single halogen or sulfur atom, and no PAINS. Despite the small size, this library covers most of the pharmacophores identified in the larger set, and considerably more than the F2X-Entry fragment library we highlighted last year or the top five commercial library vendors we noted here.

 

The researchers then screened this library against eight targets. Three GPCRs (the serotonin receptors 5-HT_1A, 5-HT₆, and 5-HT₇) were assessed in a cell-based radioligand displacement assay with fragments at just 10 µM. Despite the low concentration, 4-11 hits were found. Biochemical screens conducted at 800 µM against the proteases thrombin and Factor Xa yielded 7 and 8 hits respectively. Further analysis revealed that the SpotXplorer0 ligands sampled a majority of the pharmacophores found in published fragment hits against theses five targets.

 

Next the researchers screened their library against the histone methyltransferase SETD2, an oncology target with few known attractive ligands. An enzymatic assay yielded two hits, with IC₅₀ values between 300 and 500 µM.

 

Finally, the SpotXplorer0 library was part of the XChem crystallographic screens against the SARS-CoV-2 main protease (M^pro) and Nsp3 macrodomain, which we discussed here and here. For M^pro, just a single hit was found. This is only half the overall hit rate for noncovalent fragments in the crystallographic screen against this target, but the hit is functionally active and has a high ligand efficiency.

 

The screen against NSP3 yielded five hits binding at two different sites, for a hit rate of 5.2%. The overall hit rate against this target was 8%, but that encompasses screens against two crystal forms of the protein. The crystal form used for SpotXplorer0 had a hit rate of 21%.

 

In summary, SpotXplorer0 is new fragment library that gives high coverage of experimental pharmacophore space. Laudably, structures of all 96 fragments are provided in the Supplementary Information. But the jury remains out on how hit-rich the library will be. Interestingly, the F2X-Entry library we highlighted last year gave considerably higher hit rates of 21% and 30%, albeit against two different targets. SpotXplorer0 is being screened crystallographically against multiple targets at XChem, and it will be interesting to see how it performs in the long run.

New fragments suggested by machine learning

Machine learning has become a hot new thang in drug discovery, attracting massive attention and investment. While easy to parody, artificial intelligence techniques are becoming increasingly powerful. A new paper in J. Chem Inf. Mod. by Angelo Pugliese and colleagues at the Beatson Institute applies the methodology to generate a new fragment library.

 

Machine learning entails collecting large amounts of data, passing that through various neural networks, and obtaining recommendations. In this case, the researchers wanted to generate “privileged fragments” that would hit in multiple assays. (Of course, the idea would be to make genuinely privileged fragments, such as 4-azaindole, rather than PAINS.) The researchers used a training set of 66 fragments that hit in at least three of 25 screens done at the Beatson, for which the average hit rate was 2.18%.

 

First though, the researchers needed to teach their model how to generate chemically valid fragments in the first place (for example, fewer than 5 bonds to carbon). To do this they used both SMILES (simplified molecular-input line-entry system) and chemical fingerprints from a set of 486,565 commercially available fragments. They then combined this model with the privileged fragments. Extensive details are provided; as they go well beyond my expertise I won’t even attempt to summarize them. (For example, “the classifier for the smi2smi model comprised sequential 64-unit and 32-unit dense ReLU layers followed by a single sigmoid output neuron.”) At the end of the exercise, and after triaging by medicinal chemists, the researchers came up with a set of 741 fragments.

 

What are their overall properties? For one thing, generated fragments tend to be more planar (as assessed by PBF) and have lower Fsp³ values than the nearly half-million fragments used for training. The researchers acknowledge that this could reflect the historical composition of the Beatson fragment library, although as we argued here it could also be true that flatter fragments just give higher hit rates.

 

Molecular complexity is a fundamental but poorly defined aspect of fragment-based lead discovery, and the researchers have come up with their own metric, called feature complexity (FeCo), which incorporates information on rotatable bonds, numbers of halogens, hydrogen bond donors and acceptors, charged groups, aromatic rings, and hydrophobic elements, all normalized by the number of heavy atoms. Hopefully this will be explored more fully in a dedicated publication.

 

What do the individual fragments actually look like? Five examples are shown in the paper, and nearly 200 more are provided in the supporting information. Below are seven chosen arbitrarily from that list (sampling every 30 structures).

 

Of course, the question remains as to whether these fragments will truly turn out to be privileged. As might be expected given the vastness of chemical space, only 78 of the 741 are commercially available. The researchers note that they are acquiring some of these, and it will be interesting to see how they perform in the screens to come.