Permeability of phosphine oxides

Following the discovery of brigatinib (JMedChem2016), phosphine oxides have been showing up in more and more clinical candidates over the past 10 years. As a substituent, dimethyl phosphine oxide certainly stands out in a Craig plot as one of the most electron withdrawing and among the less lipophilic functional group (link to downloadable version of this plot).

Luckily for the medhcem community, Boehringer Ingelheim scientists decided to take a thorough look at phosphorus-based groups in a medchem context (and share their results in JMedChem2020). I cannot stress enough how useful their review is, especially as they synthesised and fully profiled a lot of analogs of drug-like structures. Below are a set of analogs of imatinib measured in logD and Caco-2.

What is pretty striking is that dimethyl phosphine oxide seems to "drop-off" the correlation. In other terms, given its lipophilicity, we would expect it to be more permeable. One likely reason for this is that phosphine oxides are extremely good hydrogen bond acceptors. This has been shown to negatively impact permeability (BioorgMedChemLett2012). It can be measured and is called pKBHX (or pKβ) (for more information/data on this, check out an original compilation of measured pKBHX here:JMedChem2009; Peter Kenny has an updated (and downloadable) compilation here). Phosphine oxides are 10 times better hydrogen bond acceptors than amides, 100 times than sulfones. Definitely something to keep in mind when adding them to your molecules.

We then have to wonder how big a problem this is for the permeability of phosphine oxide containing drugs. It is interesting to see that most of the most drugs/clinical candidates bearing a phosphine oxide had the phosphine oxide tied up in a Intra Molecular Hydrogen Bond (Ariad's ALK inhibitor brigatinib JMedChem2016; Syros' CDK7 inhibitor JMedChem2022; AZ's EGFR inhibitor JMedChem2021; the GLP-1R agonist aleniglipron). An IMHB would definitely shield the phosphine oxide and improve permeability.

This begs the question, is an IMHB an absolute prerequisite? To try to answer this, I turned to the literature. As a preamble, I should mention that although there is a good amount of reports of phosphine oxides being used in medchem, I couldn't find any measured permeability data (or logD for that matter). So in order to try to answer the question, I looked at enzyme/cell drop-off where I could and compared to other functional groups in the articles - there are obvious limitations to this approach (there are multiple other possible explanations for cell-drop off) but it is still a reasonable place to start.

As a high-level summary, the good news is that "free" phosphine oxide bearing compounds -as in not tied up in an IMHB- can have perfectly adequate permeability...but you still have to be careful. As highlighted by Mike Waring, 15 years ago (!), in his seminal paper on permeability ( BioorgMedChemLett2009), if you want to maintain good permeability, there is a limit to how much hydrophilicity you can add to molecules while increasing MW.

The first example comes from the Shangai Institute of Materia Medica and the study of PARP inhibitors (JMedChem2023). In this case, it is clear that the drop off of the phosphine oxide derivative is much greater than other neutral analogs like the oxetane and alcohol ones, and more in line with the one seen for the charged azetidine analog. As mentioned above, it is probably not too surprising that strongly hydrophilic substituent will be particularly not tolerated from a permeability point of view on such big molecules (MW 546).

The next example comes from Jacobio Pharmaceuticals and their SHP2 inhibitor program (JMedChem2024); thanks Praful Chovatia for highlighting this one). The drop-off is again significantly higher for the phosphine oxide derivative, and probably logical given the size (MW 530) exacerbated, in this case, by the number of HBD.

Moving to smaller and/or more lipophilic molecules, we have Cephalon's JAK 2 inhibitors (JMedChem2012). Although a bit higher than for the sulfone derivative, the drop-off is probably more manageable in this case.

As early adopters of phosphine oxides, Ariad scientists have successfully used them in multiple projects, including the following BCR-Abl example (JMedChem2009). In this case, there is no difference in drop-off vs the pyridine analog. Although pretty big ( MW 590) the compounds are so lipophilic (XlogP of the phosphine oxide > 5) that they can accommodate the phosphine oxide.

The last 2 examples come Bayer scientists and their ATR inhibitor program (MedChem2020) and University of Michigan scientists (again!) and their embryonic ectoderm development inhibitor program (JMedChem2020). For these 2, I was able to plot all the compounds from the publications. I think this shows pretty convincingly that there is no difference in behaviour of the phosphine oxide derivatives for these programs as far as cell drop-off is concerned (ie the green star is not an outlier). The molecules are smaller (MW<450) with only 1 HBD which probably puts them in an ideal permeability space.

I found the last 2 examples quite reassuring. Yes, phosphine oxides are polar and very good HBA but you can still get good permeability when keeping your properties in check. However, adding them to big, not too lipophilic molecules will lead to predictable loss of permeability. The silver lining is that you can probably mitigate that with a carefully crafted IMHB....

But wait there's more...

Phosphine oxides (and other phosphorus groups) can really stimulate medchemists' creativity. Boehringer Ingelheim scientists used the high polarity/low permeability properties to their advantage in the discovery of BI 1265162, an inhaled ENaC inhibitor requiring low permeability (EuropJMedChem2024; thanks Pierre Sierocki for highlighting this one). Roche scientist have used phosphine oxides in a brilliant way on their TLR7 inhibitors. For instance, a 2-sulfonyl pyrimidine analog would not be chemically stable but a phosphorus one is.... certainly worth remembering (WO2024/013205; highlighted in ACSMedChemLett2024). Finally, although not by design, the strong HBA capacity of the phosphine oxide is essential to create a 9-membered IMHB in the bioactive conformation of AZ's BFl-1 inhibitor (ACSMedChemLett2024).

This would not be a proper linkedIn post without a little CCDC torsion plot... Just thought I would end this by sharing a quick torsion plot of phosphine oxides found in the small molecule X-ray world. Contrary to amides, which tend to lose co-planarity with aryl groups when substituted (tertiary amides are not never coplanar), phosphine oxides seem to favor coplanarity...although my guess is that this is more for steric than electronic reasons.

#medchem #drugdiscovery #CCDC #pKBHX #phosphorus #measureyourlogD