Medicinal chemistry of Boron-Nitrogen heterocycles

There is no more debate on the druglikeness of boronic acids, benzoxaboroles/benzoboroxoles or closely related B-O containing compounds. Pfizer clearly put this question to rest by acquiring Anacor for over $5Billion a few years back.

As reversible covalent binders, benzoboroxoles are now part of many serine protease programs, have shown up in a dozen clinical candidates and are the linchpin of some screening decks.

I was intrigued by a couple of articles highlighting the lesser known B-N heterocycles as possible rings of interest for MedChem. Before going any further, I should mention that most of what I will discuss below has been reviewed recently as part of a broader Boron review (BioorgMedChem2022) and covered in, yet another, essential Nicholas Meanwell review (written with Murugaiah?Subbaiah; JMedChem2021).

Of all the different B-N heterocycles synthesised to date, the main one to have been studied in a medicinal chemistry context is the 1,2 azaborine.

A few things stand out from the get-go. First, the polar B-N bond induces a dipole moment in this unit, which is predicted by DFT calculations to be 2.2 Debye (similar to pyridine, check out the dipole wheel if you are interested!).

Second, as it takes part in the aromatic character of the ring, the boron atom will not show the same acidity/reactivity as in boronic acids/benzoboroxoles and will not act as a reversible covalent inhibitor.

Finally, the ring can display a N-H bond. Schoichet and Liu published an exhaustive study (including the co-crystal structure on the right) to determine whether it could act as a Hydrogen Bond Donor with a protein... and the short answer is yes. This is quite interesting as, apart from pyridones, there are no other six membered heteroaromatic rings containing a HBD. Given the prevalence of pyridones in MedChem, this could hint to a bright future for these rings.

Both academic and industrial groups have published studies on these moieties, making an effort to gather more than "just" primary pharmacology. Out of these, 2 really stand out.

Janssen scientists looked at benzofused version of the 1,2 azaborine (benzazaborinine) and their potential use as naphthyl bioisostere (JMedChem2015) applied to the β-blocker propranolol.

Both isomers are tolerated from a primary pharmacology point of view but to the credit of the authors, they extensively profiled the compounds in a battery of assays (and shared the results with the rest of the community!). We can see that the added dipole does not make the benzazaborinine analogs significantly less lipophilic (even when taking into account the additional methyl group). It is very gratifying to see that there is no detrimental impact on microsomal metabolism, permeability and P-gp recognition (even with the added HBD). Protein binding and brain binding are similar (roughly tracking with logD) and most importantly, the in vivo behaviour is quite similar with good cover and brain penetration (the full profile is available in the article). Of note, there is a chemical instability of the compounds in acidic media which could make oral development a bit trickier - but far from being a show stopper. Gratifyingly, the authors also profiled the compounds in Ames (compounds were negative) and in cytotoxicity assay (a slight signal was detected).

The other study comes from the Shin-Yuan Liu's lab (ChemMedChem2017) in collaboration with Novartis. The authors, looked at 3 different matched pairs to probe the benefits of replacing a phenyl unit by a 1,2 azaborine. They applied this to a D3 receptor antagonist, a PPARγ antagonist and a CDK2 inhibitor. The latter was profiled the most and is depicted below.

In this case, the azaborine compound is less lipophilic (of note, this is not the case of the other compounds in the publication, suggesting more subtle effects depending of the B-phenyl substituent). As a consequence, solubility is improved. Rat microsomal Clint remains high despite the lowered logD and, encouragingly, in vivo profile is quite similar to the phenyl analog. This suggests once again that azaborines do not display a striking PK liability. Finally, the azaborine compound slightly increases potency, which the authors rationalise by a putative additional H-bond to the protein (docking study).

There are a few other noteworthy studies, one on PDE10 inhibitors and one on Retinoid agonists.

The synthesis of other B-N heteroaromatic rings has been described (BioorgMedChem2022) and there seems to be a massive interest for these rings in the electroluminescence field.

The search for an isoxazole bioisostere (OrgBiomolChem2018) is a pretty stark reminder that chemical stability can be quite limiting for these rings. To the authors credit, they generated small molecule X-ray of both molecules showcasing a nice isosterism of the 2. However, fast and irreversible hydrolysis in wet DMSO precludes any future MedChem application...

All this being said, it is striking that 8 years after the initial disclosure of these rings and their properties, almost no patents have come out containing these structures (exceptions being WO 2021/011873 and WO 2019/068910). This could be due to intrinsic synthetic complexity (some approaches still require a glovebox) or that maybe they do not deliver on the early promises... literature alerts are set up ...I'll keep you posted if something shows up!

Regarding the synthetic accessibility, this is still an active field of research in academia (JAmChemSoc2018). I have to admit that when I see the kind of structures made by the Molander lab (JOrgChem2017), it certainly makes me want to add them to a HTS deck...