Here’s a ring system that you’ve never used before – the cyclopropyl system in purple at the end of the row in the diagram at right. It’s described in this paper from GSK-Stevenage as a new morpholine isostere. A 4-morpholino-pyrimidine hinge binder core is preferred in many PI3K and PIKK inhibitors, but the team was looking for a replacement to try to get better metabolic stability. As they describe it, brainstorming sessions led to someone drawing that ring system up on the board, so they took a closer look.
That’s a DFT energy plot calculated as you rotate the bond between the two rings, and you’ll see that the morpholine itself is strongly predicted to be coplanar to the pyrimidine (energy minima at 0 and 180). The tetrahydropyran (in orange) isn’t nearly so picky, apparently, and those analogs lose affinity because they don’t like being in the conformation that gives them a crucial hydrogen bond to the ring oxygen. Meanwhile, the dihydropyran (blue) has a profile a lot more like the morpholine, and indeed there are analogs known for these kinase inhibitors where it is a perfectly good substitute – until you start looking beyond the binding assay, that is. It’s not a particularly stable group, metabolically, so it’s not the answer to the question, either. You can put a methyl on the THP ring to kick it back into a better conformational space, but you’ve now introduced a chiral center and (as it turns out) have also not solved the metabolic stability problem.
As shown, the bicyclo ring compound is predicted to be coplanar in the DFT calculations, but interestingly, a force-field minimization (MMFF94X) predicted strongly that it would be (uselessly) 90 degrees out of plane (just like the plain tetrahydropyran in those calculations comes out, too). Sometimes there’s no alternative to actually running the experiment, you know, so the group took the dihydropyran-pinacolboronate intermediate that they had on hand and cyclopropanated it using chloroiodomethane and diethylzinc. I mean, you never want to miss a chance to use diethylzince, am I right? Especially if you’re into sudden flames, although they did use just the 1M solution in hexane, which is (fortunately) a far less fraught experience than the neat liquid. It took some hammering (multiple additions of both reagents over an extended period), but they pulled about a 40% yield out eventually and used the corresponding trifluoroborate in the subsequent coupling reaction.
Well, the DFT is right in this case and the molecular-mechanics calculations are wrong, which is the sort of thing that’s worth remembering. The bicyclo analog is indeed a perfectly good replacement for the morpholine and the 3-methyl THP, and what’s more, it shows greatly improved metabolic stability and clearance. An X-ray crystal structure confirmed the coplanar geometry. So congratulations to whoever it was on the list of authors who put that one up on the whiteboard. The sharp-eyed will have noticed, though, that this compound (like the methyl-THP) is now chiral, and the GSK team admits that so far they haven’t been able to come up with a chiral synthesis of the intermediate (everything’s been brute-forced by chiral chromatography at the end). That’s of course doable while you’re optimizing a compound series, but no one wants to go that way on scale – for one thing, you’re throwing away half your time, money, and effort, and for another, scaling up those chiral separations adds a whole new set of problems. So that’s still an unsolved problem. (Update: for more on this issue, see Henry Rzepa here.)
And while we’re talking small saturated heterocycles, this paper is worth a look, too. It’s an extension of some work that I blogged about last year, about how adding a bridging group in rings like morpholine, piperazine, or piperidine doesn’t always have the effects that you’d predict. And now comes work that the small spiro replacements for those have their oddities, too. For one thing, their measured logD values show that they’re almost all significantly more polar, even though you’ve added another carbon to form the bicyclo system. That seems to be down to increased basicity, and I’m not sure that everyone appreciates either of those effects. And something that I’m sure that people haven’t realized is the later claim in the paper that the diaza [3.3] systems isn’t a good isostere for piperazine in the first place. You have that basicity change, but also the distance between the two nitrogens is increased, and they’re now forced ninety degrees around from each other compared to the starting heterocycle. As mentioned in the discussion of the previous paper above, this is surely going to mess up any well-formed hydrogen bond that the original compound had.
So these compounds may have their own virtues – that polarity and that geometry might be just what your binding site has been pining for – but you should make them for chemical diversity, not because you’re trying to mimic something else. . .