So let’s talk a bit about the synthesis of Paxlovid (PF-07321332), Pfizer’s protease inhibitor drug for the coronavirus. As everyone will have seen, the US government just increased its order for this one by ten million courses of treatment, but that number is (1) not exactly huge on the absolute scale and (2) will have no immediate impact at all. That link from the New York Times says that about 350,000 courses of treatment are expected to be available to the government over the next two months, and this additional order will add about 35,000 this month and 50,000 next month. If these can reach at-risk patients they will surely help keep them out of the hospital, which is a good thing, but the need for the drug is both larger and more immediate than that.
So why don’t we have more? I did a thread on Twitter recently about this issue, but I’ll recap here. This is a fundamentally different situation than the earlier shortages of the mRNA vaccines. In the mRNA case, there were several bottlenecks, but two of the very big ones were the supply of the unusual cationic lipids used in the lipid nanoparticle (LNP) formulations, and the intrinsic difficulty of scaling that LNP formulation process up. Forming LNPs reproducibly takes some tricky flow processing through specially made chambers whose size, as far as I can tell, cannot really be increased. So you have to have a tremendous lot of them running in parallel, and you have to have the flow of ingredients into them running precisely at all times to stay within specs. It’s an odd process from several standpoints (equipment, engineering, controls) and it’s not easy to do. And that’s even if you have the lipids you need, as mentioned, which at first was a separate problem that was later resolved.
That sort of chemical supply issue, writ large and in greater variety, is what we’re looking at with Paxlovid. Small-molecule drugs are wonderful things – I’ve devoted my career to the little beasts – but they have logistic complications of their own. Let’s list some:
1. You generally need rather large quantities of the final drug. A dose of the Pfizer/BioNTech mRNA vaccine has 30 micrograms of actual vaccine in it. A few months later, you come back and get another, and then if you get a booster it’s generally the half dose of that. But Paxlovid’s course of treatment is two 150mg tablets, along with a 100mg tablet of ritonavir (more on that in a minute), twice a day for five days. That’s 1.5g of Paxlovid per patient per course of treatment, so ordering up ten million of those means that someone is going to have to make, formulate, and package 15 million grams (15,000 kilos) of the drug. We’ll concentrate here on the “make” part.
2. The synthesis of such a drug is invariably a multi-step process, and given those quantities it clearly has to be done on the true industrial scale. This is done in a chemical plant – or a bunch of chemical plants – that probably look reasonably like the ones that you might be picturing even if you don’t work in the business. Folks in hard hats, forklifts moving pallets of packaged reagents around, multifloor operations where you start on the top floor to take advantage of gravity pumping and/or filtration as you send intermediate steps from one big metal reaction chamber to another, people wearing breathing protection as they load big bags of solid reagents into hoppers, the sound of powerful pumps and stirring equipment, temperature and pressure gauges and flow meters all over, and so on. Now, one good thing about all this equipment is that a lot of it is versatile stuff. You can use those big reactors and all the associated equipment to make all kinds of stuff once you’ve worked out a good chemical process and fitted it to your facility. Indeed, while there are some chemical production sites that are built to just make Reagent X or Compound Y and do nothing but that, there are others in the fine/specialty chemical business that switch over from one sort of order to another constantly. The former generally operate on larger scale than the latter, though, as you’d imagine.
3. That multistep work means that there are a huge variety of reagents that can be involved in such syntheses, and every new drug arrives with its own list. These reagents have to come from somewhere; you’re not going to make them all yourself – that would be the equivalent of growing your own wheat to make bread. And that brings us to the supply chain, which is a phrase that everyone who so much as goes to the grocery store has become rather tired of hearing about over the last couple of years. Irritating or not, this is one of the absolutely more important parts of drug production on scale: what reagents do you need, how much, and where are you going to get them? What steps of the process are going to be done where, and by whom, and for how much money? A given drug (or its constituents) can go through two, three, half a dozen facilities in completely different countries during this work, and putting all those pieces together in an economical, reliable, safe, regulatory-compliant way is, as they say, non-trivial. What often happens is that various contract suppliers will end up being involved in making different pieces of a drug, which intermediates are then shipped somewhere else for their final organic chemistry assembly.
4. Let me spoil the suspense: if you track almost any drug’s production down to the earliest stages, you will almost invariably come to a whole bunch of offshore suppliers. These are mostly Chinese, although there are some Indian ones and other countries in the mix as well. What they tend to have in common are lower labor and capital costs and a higher tolerance for the presence of a large number of smelly, unattractive, and sometimes downright dangerous chemical production facilities. We simply do not have a lot of that capability here in the US any more; it’s been declining for decades because it’s so much cheaper to have someone else do it. These are the sous-chefs, line cooks, and kitchen help of the chemical world: they’re back there hauling sacks of potatos, washing and peeling and slicing them, while someone else then cooks them or adds them to another dish, and someone else plates it all before serving.
Let’s look at the Paxlovid potatoes, then. I obviously don’t know the details of what’s actually being done on production scale, but what we do have is the route (at right) used on the 150-gram scale that the team published last year (see the SI file). If you’re a synthetic organic chemist, you’ll break things down pretty quickly. The final molecule is stitched together by two key amide bonds, and those steps in the first row make the right-hand side of the final molecule, while the steps starting in the second row show you how the middle and left-hand sides are put together. In the bottom row, that step where T13 (from the top row) is a reagent is the final connection, and the last step forms the key nitrile warhead that inactivates the viral protease. A lot of the other steps in the scheme are the sort of protecting-group functional manipulation that most organic syntheses have in them: hydrolyzing esters to expose reactive carboxylic acids (T14 to T15), removing Boc protection from a nitrogen atom (T15 to T16), putting a trifluoroacetyl on another nitrogen (T16 to T17). It’s a bit unusual to have that group in your final molecule, but as the paper illustrates, there used to be a sulfonamide on that nitrogen and it was inferior in every way (potency, pharmacokinetics) to the trifluoroacetyl, and you can bet that they tried quite a few possibilities before settling on that one.
But this route, as straightforward as it is in the scheme and as well-described as it is in the supplementary information file, sets off a whole list of questions on the productions scale. Remember, every single thing that’s over one of those reaction arrows is going to have to be sourced somewhere. Now, a lot of those things on the scheme are available in huge amounts, things like diisopropylethylamine (DIEA), methanol (MeOH), methyl t-butyl ether (MTBE), lithium hydroxide (LiOH), methyl ethyl ketone (MEK) and so on. But others will take some attention. For example, that Boc-protected t-butylglycine that’s used to make T14; that’s not coming down out of the sky or being fished out of the lake. Someone’s going to have to supply that at scale, in acceptable purity (chemical and enantiomeric). It’s the sort of thing you can order from a number of suppliers on a gram scale, but ordering up tons, well, that’s another matter. In fact, when you start looking at the (few) people who can deliver those quantities of a given intermediate or reagent, you often find out that all those other small-scale suppliers and buying the material from just one or two folks in the background and reselling it under their own label.
Even what look like common reagents can be trouble. There are a lot of Boc groups in that scheme, for example, and they’re shown just appearing in the starting materials. But Boc groups have to be put on, generally with “Boc anhydride“, aka DIBOC. A longtime friend in the business tells me that under current conditions, that’s not as simple as it looks. DIBOC is pretty much a commodity, but when you look closely at the supply chain, you find that there are not as many original suppliers of it as you might have thought. And there are, in fact, some supply problems on scale right now because making DIBOC needs (among other things) another reagent called sodium t-butoxide. Well, that’s another commodity – I’ve used that stuff every so often since the 1980s and never gave a moment’s thought to where it comes from. But to make that, you need t-butanol and sodium metal, and it turns out that there is, of all things, a bottleneck for sodium t-butoxide because there’s not quite enough sodium to go around. Sodium metal is produced in a brute-force, energy-intensive electrolysis process that goes back to 1924, and there have been electricity supply problems that have interfered with the plants making it.
So there’s a shortage of the stuff, that’s used to make the stuff, that’s used to make the stuff, that’s used to make two of the starting materials for Paxlovid. And that’s just one of the reagents. Not being a chemical manufacturing supply-chain guy myself, I don’t know what happens when you look under the hood on some of those other steps. I never would have thought that Boc groups would be something to worry about, so that just tells you to take nothing for granted. For example, that first compound in the upper left of the scheme. Let’s ignore the Boc group on it and ask where the rest of that compound (an unnatural amino acid derivative) comes from. I have no idea. Similarly, that fused dimethylcyclopropylpiperidine compound that starts off the second row is another one that is surely experiencing a sudden upsurge in demand. At least in that case we have some precedent – that one was part of the Schering-Plough and then Merck Hep C protease inhibitor boceprevir (no longer even on the market after having been blown out by Gilead’s drug combos, just like its onetime Vertex competitor telaprevir) so it’s had attention paid to its scaleup some years ago, work that is surely now coming in handy.
That’s how it goes in the fine chemical business – there’s a compound that no one really cares much about, until they do and they care hugely, and then suddenly no one cares about it again, until some other bizarre reason emerges to put it back into demand. Multiply that by the thousands of things that are or have been commercially available chemicals at one time or another. And there are other issues that I’m sure the Pfizer folks, who are quite capable, have been resolving. You’ll note the presence of Burgess reagent in the scheme, and while that’s a fine way to convert a primary amide to a nitrile, it is probably not such a fine way to do it on a multiton scale. There are an awful lot of dehydrative methods to run that transformation, though, and I wouldn’t be at all surprised if a decent alternative is running right now (and I would be surprised if it’s really Burgess reagent on scale; I wouldn’t think that there’s enough of it in the world).
OK, there’s a fairly superficial look at what’s going on, and I hope it gives people a view of what this sort of production involves. Remember, as complicated as this may seem, there’s a lot more to setting up a production supply chain that this! What it all means is that when someone says “Oh, we can just make Paxlovid in plants all over the world”, they have left out the rest of the sentence, which is “. . .if we can get the starting materials”. And for now, supply of those starting materials is going to be tight.
