Time for another look at the coronavirus vaccine front, since we have several recent news items. Word has come from GSK and Sanofi that they are going to collaborate on vaccine development, which brings together two of the more experienced large organizations in the field. It looks like Sanofi is bringing the spike protein and GSK is bringing the adjuvant (more on what that means below). Their press release says that they plan to go into human patients late this year and to have everything ready for regulatory filing in the second half of 2021. For its part, Pfizer has announced that they’re pushing up their schedule with BioNTech and possibly starting human trials in August, which probably puts them on a similar timeline for eventual filing.
“But that’s next year!” will be the reaction of many who are hoping for a vaccine ASAP, and I can understand why. The thing is, that would be absolutely unprecedented speed, way past the current record set by the Ebola vaccine, which took about five years. More typical development times are ten years or more. But hold that thought while you peruse another news item today from J&J. They have an even more aggressive timeline proposed for their own vaccine work: they have already announced that they have a candidate, and they say that they plan first-in-human trials in September. Data will be available from those in December, and in January 2021 they say that they will have the first batches of vaccine ready for an FDA Emergency Use Authorization. Now that is shooting for the world record on both the scientific and regulatory fronts.
So let’s talk vaccine development, because everything is going to have to work perfectly for any such timetable to be realized. Here’s a good overview of the coronavirus vaccine world in Nature Reviews Drug Discovery. The official WHO list is here, and at BioCentury they have constantly updated open-access summaries of the vaccines and other therapies that are in the clinic and the ones that are still preclinical. They have also just published this excellent overview of the vaccine issues; I recommend reading that one after you’ve picked up some background from this post.
NRDD counts 115 (!) vaccine programs, of which 37 are unconfirmed (no further information available on them) and 78 are definitely real. Of those 78, five of them are in the clinic, although that number will be climbing rapidly. You have Moderna’s mRNA1273, which as the name tells you is an mRNA candidate, and Inovio’s INO4800, which is a DNA plasmid, There are two cellular candidates from Shenzhen Geno-Immune Medical Institute: LV-SMENP-DC, a dendritic cell vaccine that’s been modified with lentivirus vectors to express viral proteins, and an artificial antigen-presenting cell (aAPC) vaccine along the same lines. And finally there’s a more traditional protein-fragment vaccine, Ad5-nCoV from CanSino.
Let’s go into what all those mean. You will note the diversity of approaches in that list, and that’s not even the whole spread. When you go back into the preclinical candidates, you have in addition “virus-like particles”, viral vectors, both replicating and non-replicating, live attenuated viruses, inactivated viruses, and more. From this you may deduce correctly that there are a lot of ways to set off the immune response. What are the differences between them?
Types of Vaccines
For starters, “Live attenuated virus” is just what it sounds like, although as always there’s room to argue about whether the word “live” should ever be used when talking about viruses at all. At any rate, this would be a real infectious virus that just doesn’t give you much of a disease but does give you immunity to the wild-type virus. The smallpox, chickenpox, rotavirus, and MMR vaccines are all of this type, and they can be very effective – in fact the most effective vaccines are mostly of this type. The protection comes on more quickly and completely, with less need for booster shots and with longer-lasting effects. The tricky part is developing one of those attenuated viruses in the range where it produces effective immunity on infection but is definitely not effective at putting people in the hospital. There is a process of getting milder with time that happens with many viruses in general as they co-exist with their hosts, and the idea here is to speed that up in the lab by passaging the virus through human cells again and again and letting it mutate. Ideally, you want a strain that has ended up with a very long path to mutating back to virulence, of course!
The next class are the inactivated virus types. In that case, even if you think virii are alive (I don’t), these are dead, having run down the curtain and joined the bleedin’ choir invisible. This was originally done by exposing pathogen preparations to high temperatures, but now is often done by through nasty denaturing disinfectants like formalin or beta-propiolactone, things that alter the proteins enough to keep the virus from working, but perhaps not so much that they don’t set off the right immune response. That’s a bit of an art form, of course, and this generally has to be tried a number of times in order to get a reproducible immune response and a reproducible way to manufacture the inactive virus. As you would imagine administering a pile of disabled protein pieces in this manner is often not as effective as the live-virus approach above, which makes the human cells crank out viral proteins on their own. You’re into big ol’ injection plus booster shot territory for the most part. The hepatitis A vaccine and the seasonal flu vaccine are of this type.
Update: the relative efficacy of all these types are broad categories that can overlap – see for example the polio vaccines, where the best inactivated ones can offer more immunity in some ways than the attenuated ones. But the latter can be dosed orally, which is a big help in a planetwide vaccination campaign, which tells you that there are several factors to consider in a real-world deployment. . .
Yet another common sort of vaccine uses just a particular protein, protein fragment or subunit piece of a pathogen. (For some bacterial diseases, you can also try to raise antibodies to some protein toxin that the bacteria produce, rather than to the bacteria themselves). The key is to pick one that provokes a strong immune response, and since there are a lot of possibilities, working through them can be a process all its own. The good part is that you can then produce the protein recombinantly and in quantity, once you’ve narrowed down. There are other possibilities, of course – this could be a glycoprotein, or even just a piece of polysaccharide from an organism’s outer coating, since those can be quite distinctive. The tricky part here is getting enough response – the immune system can be very sensitive to pathogen attack, but these pathogen pieces can be less effective in triggering antibody production, and generally need adjuvants to work well (see below!) Vaccines of this class include ones for shingles, hepatitis B, HPV, meningococcus, and more.
Update: there’s a sort of intermediate step between protein subunits and inactivated or attenuated viruses, the “virus-like particle” (VLP) vaccines. That’s an assembly of several recombinant protein units into something that has the broad size, shape and organization of the real virus, but is not infectious. You can mix and alter the subunits however you like to get the best effect, with the complicating factor that there are a lot of potential choices. The HPV cervical-cancer vaccine (Gardasil) is a well-known example, and there’s a hepatitis B vaccine in this class as well.
A more recent approach is a DNA vaccine. This uses a circular DNA plasmid, coding for some antigen protein, which has been engineered with strong promoter signals and stop signals at both ends of the sequence. The plan is that this will be taken up by cells, where the DNA may well then be transcribed into RNA and that then translated into protein, which sets off the immune response. A nice feature, as with the attenuated-virus technique, is that you’re taking advantage of all the cellular machinery to make your antigen proteins for you, so they come out folded correctly and with the necessary post-translational modifications already done for you. There is no human vaccine yet that uses any DNA technique, although there is a Zika DNA vaccine for horses. Some candidates have been tried, but haven’t elicited enough of a response. Another tricky part is stability of the DNA plasmid, both on storage and on injection, but these problems have had a lot of money poured into them from the gene therapy end, and the situation has improved over the years. Overall, though, I would say that a DNA vaccine for SARS-CoV2 would be a real come-from-behind story.
Similary, the mRNA vaccine idea has had a great deal of work put into it in recent years. That’s conceptually similar to the DNA vaccine idea, only you’re jumping in at the messenger RNA stage. I wrote a bit about it in the CureVac post – basically, the immunogenicity was noticed as an unexpected side effect in experiments giving mRNA to animals, and people have gradually taken it from there. As with the DNA vaccines, you can actually get two kinds of immune response – the innate immune system can recognize foreign nucleic acid sequences floating around as a sign of infection, and the adaptive immune system can generate antibodies to the resulting proteins. One of the challenges has been getting a bit less of the innate response and a bit more of the adaptive one (which is what counts for the long-term immunity that you want from a vaccine). The mention the other day of younger recovering Covid-19 patient who don’t seem to have developed antibodies is an example of that very problem: a really robust innate response could clear the virus in an infected person, but leave them without much long-term immunity.
mRNA has some potential advantages over DNA, and (perhaps) over all the virus and protein techniques laid out above. It’s pretty much the most stripped-down vector that you can imagine, so you don’t run into so much immune-response-to-the-vector trouble, which can be a problem on repeat dosing with other vaccine technologies, and it can’t possibly be inserted into the genome. A big problem over the years has been getting the mRNA species to last long enough on dosing, to be taken up into the cells efficiently, and to be well translated into protein once that happened. The first link in the preceding paragraph has a great deal of information on this, with links to yet more reviews, and I won’t even try to summarize it all. But there have been extensive modifications made to the RNA sequences themselves and to the formulations that they’re dosed in (a lot of this by pretty brutal trial-and-error work), and the technique might be ready for prime time. We don’t quite know that yet, though. The DNA vaccines have been around longer and (as mentioned) haven’t produced a human therapy yet. Are the mRNA ones better, or is it that we just don’t know about the disappointments to come? We’re going to find out more quickly than we had planned.
There’s another related technique that has been used successfully in humans, though. If you want to really stack the deck for protein production, you can take a known virus (which doesn’t have to be related to the pathogen you’re vaccinating against) and re-engineer its nucleic acid payload (DNA or RNA) to deliver just the piece you want. In that case, you’re back into the “live attenuated virus” technique, but by sort of cobbling one together from different parts. This may sound pretty similar to gene therapy, which also generally uses viral vectors, and if so your intuition is right on target – the two fields have had a lot to teach each other. The Ebola vaccine uses this method, with a livestock virus as the vector, and there are many other virus types under investigation for this sort of delivery. Update: broke this into a separate category for clarity.
There’s another key vaccination technique that I haven’t mentioned, and it applies to all of the techniques above: adjuvants. Obviously, the big thing you want from a vaccination is a robust, long-lasting immune response, and it turns out that various additives can provoke just that. These are all about that balance between the innate and adaptive immune response mentioned above; the idea is to get the best carryover from the immediate innate mechanisms to drive the antibody-centric adaptive ones. See this post for a quick immune-system primer, and there are of course many other places to learn about this – the key here is the handoff to the antigen-presenting cells and the helper T cells.
The adjuvant field started out, frankly, as about the closest thing to voodoo that you’ll find in infectious disease treatment. Antibodies were generated by injecting horses and extracting their plasma, and a veterinarian (Gaston Ramon) noticed in the 1920s that the yields were higher from animals that had developed a strong reaction at the original injection site. He started experimenting with additives to induce such reactions, including things like tapioca starch. In the same era, Alexander Glenny was formulating various diphtheria vaccines and noticed that the ones that included aluminum salts were much more effective. No one really knew the details of how these things did what they did, but aluminum salts are still very common in vaccines nearly a century later. We’ve learned more about what’s going on – in the 1990s, the first new adjuvants in decades began to show up, and more have been added. For example, the GSK shingles vaccine (Shingrix) has lipoproteins from Salmonella bacteria added to it along with terpene glycosides from the Chilean soap-bark tree, which seems to be an especially powerful combination. I can tell you that the reaction at the site of injection for that one is very impressive, especially on the second shot! GSK’s expertise in this field is in fact what they’re bringing to the collaboration with Sanofi mentioned in the first paragraph, and they’re collaborated with many others as well.
Developing a Covid-19 Vaccine: Efficacy
OK, back to the broad picture of developing a coronavirus vaccine: the question is, which of all these possible techniques is the most effective and safe? That we are only going to find out, in the end, by dosing people. Lots of people. With therapies targeting the immune system, there is in the end no other way to know, because of the complexities of the human immune response and its wide variation in the human population. Rushing the process is going to take a vast amount of effort, and some of the steps are going to have to be done on a scale never before attempted. There’s another point that can’t be ignored, either: if we want this done as quickly as we would like, there are going to have to be some shortcuts.
To that point, one reason that the Moderna vaccine got off the mark so quickly is that the mRNA route can be intrinsically faster, but a bigger reason is the step of seeing how well it works in animals was entirely skipped, a very unusual step indeed. That’s partly because it’s still unclear which animal model will be the most informative. We have a bit of a head start thanks to the work that’s been done on the earlier human coronavirus pathogens for SARS and MERS, but you may recall Monday’s post talking about how SARS and the nCoV-19 virus do show real differences in various tests (there are many lines of evidence for that). We can expect those differences to carry over to the animal models as well. One approach that I know that people are taking is to breed animals that have been engineered with the human form of the ACE2 protein which seems crucial for viral entry – one way or another, we should be able to find a small animal (mouse, hamster, etc.) that can be useful, but will it be found in time to actually be useful? My guess is that several other clinical vaccine candidates will end up going the same route as Moderna’s, and skip past animal efficacy entirely. Believe me, that’s a shortcut, and there will be others.
Fortunately, testing for vaccine efficacy can be (fairly) straightforward, and it involves many of the same issues that are being frantically beaten on for antibody testing: does a vaccinated patient develop antibodies? How many? Are they the right kinds to neutralize the virus? And how long do they last? Those first three are the subject of a huge amount of work right now, and although it’s nerve-wracking at the moment I have no doubt that these are questions that can be and will be resolved. We’re going to have a lot to think about with what endpoints we’ll be measuring for efficacy, to be sure – surrogate ones will be faster, but will regulatory agencies want to see more patient-focused clinical endpoints as well?
Here is a review from the dear, long-gone days of 2016 of the standard development process for a new preventative vaccine. Take a look at the lengthy, detailed, overlapping, interlocking system of trials that such vaccines have undergone in the past, and reflect that we’re not going to be able to do all of that if we want a vaccine on the timelines stated at the beginning of this post. Ideally, you want to study these efficacy questions in Phase II trials in different populations (age, gender, pre-existing health conditions and range of medications being taken), all with different dosing schedules, and carefully tune things up for bigger Phase III runs. We’ll be able to deal with some of that by running a lot of simultaneous trials instead of doing things more sequentially, but that’s not going to cover every issue. Not by a long shot. Remember, there are at least 78 of these things under development right now – there will be fierce attrition, and only a few (low single digits) will make it deep into the process, but it’s still a fearsome process to get all this organized.
And some things cannot be accelerated by any means known to humanity. The last point above, how long immunity lasts, is a big question for both people naturally infected by SARS-Cov2 and for those given a vaccine, and unfortunately there is no way to answer that one other than time, which is in short supply these days. The field provides many examples of vaccines whose protection has not held up as well as expected as the years went on. My guess is that we may end up with a first-round vaccine that doesn’t last as long as it might, but will provide enough immunity to do the job and provide cover for us to collect more data on an optimized candidate.
Developing a Covid-19 Vaccine: Safety
But that takes us to the second question for any new therapy: safety, and its balance with efficacy. This is an especially fraught question with any therapy that’s targeting the immune response, because the downsides are gigantic: a runaway immune reaction can disable someone for life or even kill them within minutes where they stand. Guillain-Barré syndrome is an example: your body reacts to an antigen (a viral infection or a vaccination) by deciding that the myelin sheaths around your nerves are also the enemy, and starts destroying them. Very bad news, and although most people recover, a few die. Roughly estimated, even a seasonal flu vaccine might kill about one out of every ten million recipients though such a reaction – we give it to everyone possible, though, because far more people will die if we don’t. The 1976 swine flu debacle shows what can happen, both in perception and in reality, when you get this balance wrong. But you can’t avoid the problem: the huge person-to-person variation in everyone’s immune system means that these severe events can never be ruled out at some low level if you’re dosing enough people.
Now you see the exact bind that vaccine development has always been in, because the whole point is to treat millions, even billions of people who are not currently sick, to protect them against disease while not doing more harm along the way by setting off the body’s fiercest and most alarming biological responses. I have no doubt that the companies and regulatory agencies involved will be doing everything they can to address safety issues, but if you’re looking at a vaccine getting an EUA early next year, well. . .
Developing a Covid-19 Vaccine: Logistics
Another big problem is going to be manufacturing and distribution. Many readers will have heard about the difficulties that sometimes occur during the flu-vaccine production process, leading to shortages. Depending on what vaccine technology comes out on top, manufacturing enough doses in a reproducible fashion could be quite challenging – space and finger fatigue don’t permit going into all the details, but they are many and complex. Keep in mind as well that many vaccines need “cold chain” distribution and storage, which is always a layer of complexity. What if an eventual vaccine needs more than one round of administration, as many of the adjuvant-formulated ones do? Keeping track of that and following up on it is yet another issue.
My guess is that scale-up and manufacturing could well be the biggest chance for the timelines mentioned earlier to blow up, so there is going to be a massive effort to front-load the work on these problems – this is why, for example, Bill Gates has already indicated willingness to fund factories for up to seven vaccines up front. The live-attenuated virus, inactivated virus, recombinant protein, and nucleic acid vaccines will all involve completely different production methods and formulations, and since we don’t know which way we’ll be going, this would seem the only way to address the issue. Pfizer and others have already said that they’re going to be working on production even before the efficacy data come in, which needless to say is not the usual business practice. I think we’ll get vaccine efficacy, one way or another, although it sure won’t be characterized as thoroughly as it normally would. And I think we’re already agreeing to cut corners on safety, whether anyone says so in as many words or not. But producing the vaccine on scale could be a bigger issue yet, and as the process goes on, that’s where I would keep an eye out for trouble.
It is a tightrope, folks, and we’re going to be trying to run across it. Watch closely; with any luck we will never see anything quite like this again.