The other day I had a look at the process used to make the mRNA vaccines, so I thought it would be a good idea to do the same for the adenovirus vector ones, such as J&J, Oxford/AstraZeneca, CanSino, Gamaleya et al. It’s a different system, with its own advantages and disadvantages, and that’s the broad story of scale-up manufacturing all the way: tradeoffs at every turn. It’s always tough to break out of the constraints of the Engineer’s Triangle: “Fast, Cheap, Good: Pick Any Two”. That is, if it’s good and cheap, it’s unlikely to be fast, and if it’s fast and good it’s unlikely to be cheap. And of course, if it’s fast and cheap, it’s unlikely to be any good!
Intro: Adenoviruses and Infection
Adenoviruses are extremely common double-stranded DNA-containing pathogens, and it seems like a sure bet that every single person reading that has been infected with several of them over the years. They tend to cause mild respiratory symptoms and sometimes show up as ear infections or conjunctivitis. There are no antiviral drugs that target them, and no adenovirus-targeting vaccines that are available to the general public (although there have been anti-adenovirus vaccination programs in the military for a couple of subtypes). Subtypes, indeed: there are at least 88 of them that infect humans (more since that paper was published!), divided into several related groups. Among the more common ones is adenovirus 5 (Ad5), and in some regions of the world you find 80 to 90% of the population is already seropositive to it (the US figures seem to be close to 40%).
The adenoviruses have long been used as tools in molecular biology, since they have plenty of room to carry a modified DNA payload, show no tendencies to integrate DNA into the host genomes, and can infect both dividing and nondividing cells. But as those figures just cited show, a downside of using them as human therapies is that many people may well already have antibodies to your viral vector tool right at the start, which will surely knock down its effectiveness. For this reason, there has been a long-running search for rare and unusual Ad forms to use as platforms, which explains why you see J&J and Gamaleya using Ad26, Oxford/AZ using a virus from chimpanzees that’s not in the human population, ReiThera using a gorilla adenovirus, etc. And it’s also why people wonder about CanSino’s efficacy in general, since they stayed with Ad5.
No matter what variety you use, though, you also have to wonder about what happens if you administer a second shot of the same vaccine: how much makes it through? The current wave of trials, I have to say, is going to provide more real-world data to answer this question than we could ever have imagined having so quickly. That’s closely related to an even larger question: once you’ve had a vaccine (or even a gene therapy?) with a particular adenovirus vector, what happens if you want to get another vaccine for a different disease that uses the same vector? Are you going to cross off large numbers of people from being dosed in Ad5 space, Ad26 space, and so on? Another unanswered question, as of yet.
No matter what the adenovirus types, the end result of such a vaccination is rather similar to what happens with an mRNA vaccine like Moderna or Pfizer/BioNTech. The adenovirus goes in and does its normal infection route; all that machinery is intact.. But in this case, the DNA payload that’s delivered into your cells is not a big set of instructions for making more adenoviruses, it’s a much shorter sequence that codes for the coronavirus spike protein instead. So the modified DNA gets transcribed to messenger RNA in your cells (and that’s the exact step that the mRNA vaccines jump in at if you take them), and this mRNA is taken up by ribosomes and translated into the Spike protein itself. And production of that foreign protein sets off your immune system, which to be sure has already been ringing alarm bells because it is sensing that a foreign virus is attacking. That’s why you don’t need an adjuvant added to either the viral vector vaccines or the mRNA ones; they set off the various Intruding Virus Detected machinery very well on their own, whereas just injecting you with the Spike protein itself skips past some of those warning systems. Those are generally set up around sensing foreign DNA and RNA, and jumping past them with an injected protein leads to a less vigorous response (thus the need for an adjuvant in the mix).
Making an Adenovirus Vector
So let’s move on to how you make such vaccines. You’ll need to produce a large mass of infectious viral particles, each with the modified stripped-down DNA that you want to target to a patient’s cells. You especially want to take out the part of the adenovirus genome called E1; removing that makes it impossible for the virus to replicate. If you need more room for your own payload, you can delete the E3 region as well. This stuff has all been explored some years back; there are number of regions in the viral genome that have been shown to be suitable for splicing in your own sequences.
But getting this done and into a system that will make a pile of new virus requires some genomic dance steps. The most common way to do this, in broad outlines, is to make a bunch of (linear) adenovirus DNA, with your own modifications, and then get that into a big reactor full of human cells. And (this is key) these human cells have already been engineered to make the proteins that the viral E1 region makes, the ones the virus needs to replicate. This complementation trick allows the modified adenovirus to replicate away in the human cells and give you a much-increased yield of new infectious virus particles, but ensures that these viruses themselves are still unable to replicate. The E1 proteins they’d need are not coded for in their own genomes (you took that part out), but were just present for them inside the human cells. And when injected into a patient, they will most definitely not be encountering any other human cells that are cranking out viral E1 proteins for them.
So the first step in this process is to engineer the viral DNA that you need and make a lot of it. This step has often been done in bacteria because bacterial DNA is relatively straightforward to handle and to get replicated. There are actually commercial systems you can buy to do this on a laboratory scale – that is, you can purchase plasmids (the circular DNA molecules used by bacteria) that already have the A5 adenovirus genome in them, with the E1 and E3 regions already removed, and with the sequences set up for easy insertion of whatever DNA you want. Another way to do that is with a variety of plasmid called a bacterial artificial chromosome (BAC), and you can buy those with the features you need for modification. But you’ll recall that J&J (and Gamaleya) are both using Ad26, while Oxford/AZ are using a chimpanzee adenovirus, so the commercial Ad5 reagents won’t be of any use – the teams involved have been working up their own tools for the job. Earlier on, CanSino reported using the commercial AdMax system from Microbix (Toronto) for their plasmid work. In any case, though, you’re forcing the bacteria (often good old E. coli) to make copies of these plasmids, as many as it can stand. You then lyse (break open) the bacteria, isolate your DNA, and then break open the circular plasmids to get a linear DNA molecule. Some of the BACs can be engineered so that they do that to themselves, saving you a step.
You need the linear DNA because it turns out that in that form It can directly infect human cells (with the help of some additives in the cell culture to get it through the cell membranes more efficiently). Update: you need the linear DNA form for adenovirus replication/packaging once it’s in the cell. The earlier Oxford papers reference this book chapter for their methods. As for J&J, this patent would appear to have some of the details of their system. Updated: you can get adenovirus DNA in as the circular plasmid form, too
From this paper and this one, it appears that the Oxford/AZ team is using a BAC, engineered from combination of two different plasmids through “recombineering” to bring in their sequence for the Spike protein into what used to be the E1 region of the adenovirus sequence. (I’m skipping the details of that process to save time, space, and patience). Meanwhile, you can read about what appears to be the J&J plasmid system here and here (that last one detailing another adenovirus subtype, but apparently with similar techniques used for Ad26).
Now it’s time to get those linear DNA molecules into human cells. Here we get into some controversy, depending on your beliefs. It looks like Oxford/AZ is using a complementation-engineered version of HEK293 cells for this process, as are Gamaleya and CanSino, while J&J is using a line called PER.C6. These two have both been around for a while. The HEK initials stand for “human embryonic kidney”, and it was indeed first isolated from aborted fetal tissue in the early 1970s at Leiden University. PER.C6 as a complementation strain goes back to 1998, but the origin of the cell line is back in 1985 in Leiden as well, also from aborted tissue, with the “ER” part standing for “embryonic retinoblasts”. As you can well imagine, people with strong anti-abortion beliefs are not enthusiastic about taking vaccines that touch on this area in any way for their production, while other with different beliefs are not bothered at all. No matter what, though, it seems crucial for the linear DNA to be transfected into some sort of human cell complementation line; that’s the only way you’re going to get amplified yields of the final viral particles used in the vaccine.
As you read about vector vaccine production, you’ll sometimes see the phrases “virus seed stock” and “host cell bank”. You’ll see below that there are manufacturing sites all over the world for these vaccines, and the last thing you need is for everyone to be out there going it alone. Batches of the plasmids, the linear DNA, the complentary cells, and the final adenovirus are all going to be stored for future reference and/or distribution, and exhaustively characterized. You definitely want to keep a close eye on the batches of these things to make sure that you’re dealing with the same stuff at all the production sites.
Human cell culture – any cell culture – is simultaneously a scientific process and an art form. Ask anyone, literally anyone who’s done it, and if you can find someone who’s worked on it at an industrial scale, they’ll confirm that all the more vigorously. This is (or can be) the weak point of the entire viral-vector production process. When everything is working, this method for infecting living cells and turning them into virus factories is hard to beat. But it doesn’t aways work the way it’s supposed to. It appears that AstraZeneca has been having problems because one of their largest production facilities has been experiencing problems with low yields of virus, even though everything should be the same (same viral DNA, same cell line, etc.)
To give you an idea, HEK293 cells themselves come in varieties that grow on the surfaces of a culture vessel (adherent, HEK293A) or grow floating around in suspension (HKE293S). You may well want the latter for serious scaleup (not least because you’re growing in three dimensions instead of two), but it can be done either way. Adherent cells grow until they touch and form a confluent layer on their surface, and some lines are OK when that happens and some aren’t (or gradually become less happy about it). Suspension cell lines divide and make a thicker, more concentrated suspension, and all of them react somewhat differently to that process, too. You have to think about what media all these things are growing in and what nutrients to provide (and in what concentration), the buildup of waste products (and debris from dead cells), the washing of adherent lines with fresh media and the stirring rates and techniques for suspension ones. . .oh, it’s glorious.
For example, when using engineered cells to make modified human proteins (an extremely common task in both academic and industrial molecular biology), I have been on a project where the yield of protein changed dramatically using the same damn cells grown in cylindrical “roller bottles” which were stirred (as the name implies) by slow rotation (rather like a convenience store hot dog machine), versus being grown in “shaker bags”, a more free-form affair that was sloshed around slowly by rotary oscillation. Why did the cells care? You tell me – but under one set of conditions they made a lot more protein than the other. Why is one of AZ’s plants making less virus than it should? Who knows?
Purification and Packaging
Isolation of the viral particles is likely pretty similar for all of these vaccines. I’ve been unable (no great surprise) to find detailed production information for any of the current vaccines, but this was likely one of the less stressful parts of the process optimization, given all the work that has already been put into adenoviruses over the years. You’ll lyse the cells in the cultures and start with some rough filtration to pass the viral particles and retain the cellular debris. From this AstraZeneca page, it looks like they’re using a series of filtration steps, followed by membrane chromatography (likely some sort of ion-exchange technique, in this case, based on the charged residues of the viral surface proteins), followed by an ultrafiltration step. You can bet that the organizations involved already had a pretty clear idea of what steps they’d be taking, although all of this stuff needs some tweaking for optimization and also validation at every step. The regulatory agencies involved will have seen these details, but I don’t think we’re going to.
And then you have to formulate the viral particles, which is a much less fraught process than it is with the mRNA vaccines. The other ingredients for the vaccine itself are going to be pretty innocuous stuff, no weird lipids as needed for the lipid nanoparticles. Here’s the list for the Gamaleya vaccine (see the first page of text); there’s nothing on it that looks to be any sort of supply problem. Now it’s time for fill-and-finish, which has been a common step for everyone, rounding up enough production-line capacity for filling and capping sterile vials.
I see that the earliest batches of the Oxford/AZ vaccine were produced at Oxford itself, and later on were manufactured and packaged by a company called Advent (in Pomezia, Italy) and by COBRA Biologics (in Keele, UK) with vial-filling by Symbiosis (in Sterling, UK). They’re working with the large contract firm Catalent in both the US (Harmans, MD) for production and Europe (Anagni, Italy) for fill-and-finish. There is production in the Netherlands (Halix) and Belgium (Novasep, in Seneffe). The last one is apparently the site with the yield problems. It’s also being packaged in Dessau, Germany by IDT Biologika. Russian manufacturer R-Pharm has a plant in Germany that’s in production for export back into the CIS countries (they’re also producing the Gamaleya vaccine). Insud in Spain is involved as well, as is a new plant of theirs in Argentina. AZ also has a big production deal with India’s Serum Institute, and WuXi is involved in China and at a plant in Wuppertal, Germany. And I’m sure I’ve missed some deals.
J&J, for their part, has a lot of capacity in the Netherlands (such as in Leiden), and they have signed deals with Emergent to produce the vaccine in Baltimore (who are also working with AstraZeneca, and indeed with Novavax, producing their protein vaccine at a separate Maryland plant). They’re also working with Catalent (at their Bloomington, Indiana plant and also at the Anagni site in Italy), Reig Jofre in Barcelona, Aspen Pharmacare (in Port Elizabeth, South Africa), Biological E in India (who just bought another facility in Himachal Pradesh), and with PCI Pharma for cold storage and shipping. No doubt there are more deals out there, too.
So there you have it, in outline form anyway. Any one of these steps can be zoomed in on to reveal a forest of further details, but that should give you an idea of what’s happening (and in many cases may provide even more than you ever wanted to know!) As you can see, it’s a fundamentally different process than the mRNA vaccines, with its own features (good and bad). That may well become important if we have to retool the existing vaccine candidates for new variants, but that’s a post for another day!