A cell-active fragment targets the microRNA-372 hairpin precursor

Fragment-based lead discovery against RNA has been a theme on Practical Fragments for well over a decade. Unfortunately, most of the resulting hits are either weak or non-druglike. A paper published late last year in J. Am Chem. Soc. by Matthew Disney and collaborators at the Scripps Research Institute in Florida provides a nice counter example.

 

The researchers started by building a new fragment library based on known RNA-binding molecules – a venerable approach we first described back in 2009. The most common scaffolds differ somewhat from the most common scaffolds found in drugs (see here), with pyrimidines, triazoles, furans, and benzimidazoles over-represented. A set of 2500 fragments mostly conforming to the rule of three was purchased from ChemDiv, and the structures of all of them are helpfully provided in the supporting information.

 

Most fragment screening efforts start with a target of interest, but here the researchers chose a “library-versus-library” format, in which they screened all 2500 fragments against 5120 different RNAs. Each RNA molecule consisted of an identical 40-nucleotide hairpin containing randomized 3 x 2 or 3 x 3 internal nucleotide loops. Fragments were immobilized onto an agarose-coated microarray and radiolabeled RNA was added. Interestingly, only 19 fragments were found to bind the RNA. Competition experiments with other RNA sequences and DNA reduced the number of specific binders down to three. RNA sequencing experiments revealed that two of these fragments were fairly promiscuous, each binding over 100 different RNAs, but compound 3 bound just 28 RNAs with the 3 x 2 internal loop and none with the 3 x 3 internal loop.

 

Having identified specific RNA sequences bound by compound 3, the researchers searched human microRNAs (miRNAs) and found that the pre-miR-372 contains a bulge predicted to bind. When this RNA is processed it produces miR-372, which represses translation of the tumor suppressor LATS2, thereby increasing cellular proliferation. Perhaps the binding of compound 3 to the pre-miRNA would impede its processing.

 

The affinity of compound 3 for pre-miR-372 was measured to be 300 nM, giving it an impressively high ligand efficiency. In vitro experiments showed that the compound blocked processing by the enzyme Dicer. The researchers conducted a series of cellular experiments showing that compound 3 decreased levels of miR-372 and increased pre-miR-372m while having no effect on 379 other miRNAs. Encouragingly, compound 3 also increased levels of LATS2 mRNA and protein and decreased cell proliferation. Additional experiments with siRNA, mutated versions of pre-miR-372, and cell lines with lower levels of miR-372 all support the on-target mechanism.

 

This is a lovely paper, and compound 3 appears to be a good starting point for further optimization. However, the work also suggests why finding lead-like RNA binders may be so difficult. In the library-vs-library approach described, the researchers studied 12,800,000 different molecular interactions and came up with just three (somewhat) specific binders. The ligandability of RNA – or at least the small internal loops studied here – appears to be low. To prospectively find hits against specific RNAs may require much larger fragment libraries than are typically used. Perhaps this could be an application for DNA-encoded fragment libraries, which we wrote about here.

FragLites and PepLites meet bromodomains

The last two Practical Fragments posts focused on bromodomains, epigenetic readers that recognize acetylated lysine residues. Today’s post could thus be considered part of a trilogy, though the focus is less on bromodomains themselves than a specific type of fragment library.

 

In 2019 we highlighted FragLites, small fragments containing pairs of hydrogen bond acceptors and/or donors along with a bromine or iodine atom. FragLites were designed to assess ligandability as well as identify what types of interactions would be favorable at various sites. The original test protein was the kinase CDK2. In an open-access paper published late last year in J. Med. Chem. by Martin Noble, Michael Waring, and colleagues at Newcastle University, FragLites are screened against two members of the bromodomain family.

 

The first bromodomain (BD1) of BRD4 is considered highly ligandable, with multiple inhibitors disclosed (see for example here). In contrast, ATAD2, a bromodomain in another subfamily, is more challenging, in part because it lacks a hydrophobic region useful for increasing affinity for small molecules. Thirty-three FragLites were individually soaked at 50 mM into crystals of either bromodomain. The halogen atom on each FragLite facilitates analysis by anomalous dispersion, allowing more sensitive detection of low-occupancy binders. This, along with Pan-Dataset Density Analysis (PanDDA), was used to identify specific protein-ligand “binding events.”

 

In total, 26 binding events at five sites were identified for BRD4; four ligands bound at more than one site. Of these, 17 FragLites bound at the orthosteric site of BRD4 (which recognizes N-acetyl lysine). In contrast, ATAD2 displayed 16 binding events total over seven sites; only three bound at the orthosteric site, consistent with its lower ligandability. ATAD2 had previously been screened crystallographically against the 776-membered DSI-poised fragment library, and this effort also identified seven ligand-binding sites, six of which were common to those discovered here, suggesting that the small FragLite set is able to identify most pockets.

 

As far as specific types of interactions, the average FragLite made 1.1 hydrogen bond, suggesting that the second donor or acceptor is often not engaged. In contrast, the bromine or iodine atom makes protein contacts in 33 of 42 binding events. In half a dozen cases no hydrogen bond to the protein was observed, with the primary interaction being a halogen bond.

 

The FragLites are small, relatively “flat” aromatic molecules, but of course most proteins interact with other proteins. To try to explore such interactions, the researchers developed a library of “PepLites:” N-terminally acetylated amino acid residues with a C-terminal bromopropargyl group. These were also screened crystallographically against the two bromodomains and produced considerably lower hit rates, with six bound to BRD4 (all at the orthosteric site) and nine bound to ATAD2 (of which five bound to the orthosteric site). Reassuringly, the N-acetylated lysine PepLite bound to both proteins in a similar manner as seen in larger peptides.

 

The researchers conclude that FragLites and PepLites “represent highly valuable components of a larger crystallographic screen, and we anticipate that this is where they will fit into most drug discovery programs.” Indeed, this is already happening; last year we wrote about how FragLites were screened against the bromodomain PHIP2 as part of a larger screen, and I was surprised this paper was not mentioned here. Laudably, all the atomic coordinates have been deposited in the Protein Data Bank, so folks are able to do their own analyses.

 

As FragLites and PepLites are screened against ever more targets, it will be fun to see what they can teach us about intermolecular interactions and starting points for new leads.

Fragments vs PBRM1 bromodomains revisited, more selectively

Last week we highlighted the discovery of a selective inhibitor for the BET family of bromodomains. The 61 human bromodomains fall into eight subfamilies, of which the BET family has probably been most heavily studied. In contrast, family VIII has received less attention, in part due to the lack of selective inhibitors. This deficit is beginning to be addressed by Shifali Shishodia, Brian Smith, and collaborators at Medical College of Wisconsin and Purdue University in J. Med. Chem.

 

The researchers were particularly interested in the aptly-named protein Polybromo-1 (PBRM1), which contains six of the 10 family VIII bromodomains. The protein has normally been considered a tumor suppressor, but it has also been implicated as a tumor promoter in prostate cancer. Chemical probes would be very useful to unravel the complicated biology. A few pan-inhibitors of family VIII have been developed, one of which we wrote about back in 2016, but none of these are selective for the PBRM1 protein.

 

The researchers started with an NMR screen of the second bromodomain of PBRM1, BD2, the structure of which had previously been solved by NMR. A ¹H-¹⁵N SOFAST-HMQC screen of 1968 fragments (all rule of three compliant, from Maybridge and Zenobia) in pools of 12 ultimately yielded a dozen hits, all of which are shown in the paper. Of these, compound 5 was the most potent, with dissociation constants of 45 µM by NMR titration and 18 µM by isothermal titration calorimetry (ITC). 

 

 

One of the previous pan-family VIII inhibitors described in the literature was structurally similar to compound 5, and borrowing a chlorine atom from this led to compound 11, with improved affinity. Further exploration around both phenyl rings ultimately led to compound 16, which displayed low micromolar affinity by ITC and high nanomolar activity in an inhibition assay.

 

Differential scanning fluorimetry (DSF) is commonly used to measure binding of small molecules to bromodomains, and the researchers tested some of their best compounds in a panel of bromodomains that included 9 of the 10 family VIII members. Encouragingly, compound 16 only showed a strong thermal shift (ΔT_m = 5.4 °C) to PBRM1-BD2 and moderate shifts (ΔT_m = 1.8 °C) to PBRM1-BD3 and PBRM1-BD5. No significant stabilization of the 18 other bromodomains was observed.

 

A series of shRNA experiments by the researchers revealed that the prostrate cancer cell line LNCaP was dependent on PBRM1, and compound 16 was active against these cells, albeit weakly (EC₅₀ ~ 9 µM). In contrast, the compound did not show activity against two other cancer cell lines that do not seem to be dependent on PBRM1.

 

This work is a nice example of academic fragment-based lead discovery. Although the cell activity of compound 16 is probably insufficient for a serviceable chemical probe, it does show that selectivity is possible. Hopefully these researchers, or others, will continue improving it.

Efficiency metrics in action for a bromodomain inhibitor

The metrics ligand efficiency (LE) and lipophilic ligand efficiency (LLE or LipE) are frequently used during fragment-to-lead optimization. A recent paper in J. Med. Chem. by Philip Humphreys and colleagues at GSK describes how they were useful in developing an “oral candidate quality” inhibitor of BET-family bromodomains.

 

Practical Fragments has written frequently about bromodomains, which bind to acetyl-lysine residues in histones to epigenetically modulate gene transcription. Some 17 bromodomain inhibitors have entered the clinic, of which at least three (pelabresib, PLX51107, and ABBV-744) came from fragments. GSK was an early pioneer in the field, and researchers there were interested in using fragments to develop a differentiated class of molecules that would inhibit all four members (BRD2, BRD3, BRD4, and BRDT) of the BET family, each of which contains two separate bromodomains designated as either BD1 or BD2.

 

GSK already had BRD4 BD1 binding data for 50,000 compounds, and these were analyzed to find molecules with LE>0.3 kcal/mol per atom that were structurally differentiated from known bromodomain binders. Compound 9 was quite potent and had high LE as well as respectable LipE. (As the researchers note, LE is “the more relevant metric” for fragments, with LipE becoming increasingly important during later optimization.) A crystal structure of this molecule superposed onto another bromodomain inhibitor suggested that adding a methyl group to fill a small pocket could boost affinity, and this was confirmed by compound (R)-10. This molecule showed cell activity and good permeability, although hepatocyte stability was poor, likely due to the two methoxy groups. Removing these led to compound 12, the most ligand-efficient compound that had been seen. (All values in the figure below are for binding to BRD4 BD1.)

 

Compound 12 mimics the N-acetyl lysine residue of the natural ligand, and previous research had revealed two additional regions of the bromodomain that could be targeted for enhanced affinity, the so-called “WPF shelf” and the “ZA channel.” Structure-based design was used to independently explore both areas, leading to compounds such as 24 and 31. In addition to assessing LE and LipE, the researchers paid close attention to other factors such as permeability. Virtually combining the best moieties that bind at the WPF shelf and ZA channel led to 770 potential molecules to make, which were winnowed down to just 40 on the basis of predicted lipophilicity (specifically chromLogD_pH7.4), molecular weight, and TPSA. The best of these were more extensively profiled, including in pharmacokinetic studies. I-BET432 emerged as the winner.

 

I-BET432 binds tightly to both bromodomains of the four BET family proteins and is at least 80-fold selective against two dozen other bromodomains. It shows excellent oral bioavailability in rats and dogs, does not inhibit hERG, is not mutagenic in an Ames test, and does not inhibit CYP3A4. The molecule is also clean in a panel of four dozen off-target proteins. Human oral dose predictions come in at 5-18 mg per day. A crystal structure of the molecule bound to BRD2 BD2 showed the expected binding pose, and that the two alcohol substituents may be forming an intra-molecular hydrogen bond, which could explain the high permeability.

 

This is a nice case study in metric-driven optimization. As the researchers note, I-BET432 “has the highest LipE (6.2) and LE (0.43) of the candidate quality GSK pan-BET inhibitors disclosed to date.” Although the molecule does not seem to have gone forward into development, the story is nonetheless worth reading to see how metrics can yield quality molecules.