As the world knows, the mRNA vaccines for the coronavirus are proving to be very effective. That’s welcome news for the obvious pandemic reasons, but it’s also welcome validation for a technique that’s been in the works for many years now. I’ll take a moment to re-emphasize how fortunate we are that so much of this groundwork, all these false starts, re-workings, and dead ends had already been worked through to this degree before a worldwide emergency made the mRNA vaccination platform suddenly so desirable. (And I’ll also re-emphasize another issue, that we’re extremely fortunate as well that the current coronavirus has so many similarities to the 2003 SARS one, because the work that went into studying that and possible vaccines against it has been invaluable this time around).
Let’s talk more about one of the tricky aspects of getting mRNA dosing to work, because it has possible implications for where the field goes from here. As many will realize, RNA-based therapeutics have more applications than just making vaccines against infectious diseases (although that’s certainly a good one). But one of the good things about using it as a vaccine technology is that it’s “self-adjuvanted” – recall that adjuvants are substances that sets off a response of the innate immune system at the site of injection, which can potentiate that later adaptive immune response. Protein subunit vaccines (such as the Novavax candidate in the current pandemic) generally have to have an added adjuvant to get up to the needed immunogenicity. These can be as simple as aluminum salts and as complex as extracts from the rare Chilean soapbark tree.
But mRNA vaccines set off an innate immune response of their own (which is a source of the “sore arm” effect that many have noted). I should note here, by the way, that it’s tempting to decide that a strong site-of-injection reaction must lead to a strong adaptive response, but the two actually don’t seem to be related much). Whether your arm gets sore or not, though, you are getting immediate responses to an mRNA vaccine, and some of that is due to the foreign RNA payload, while some of it is also coming from the lipid nanoparticles themselves.
Balancing all this out is tricky, because while you need some of this innate response, it’s going to affect the eventual adaptive response that’s the whole point of administering the vaccine. The types of antibodies that eventually get generated and the T cell responses that are elicited can be rather different depending on how that innate response fits into the overall immune picture. What’s even more enjoyable is that animal models can only tell you so much about this: there are no animals (engineered or otherwise) that recapitulate the human immune response in all its head-banging detail. To be fair, it works both ways: humans aren’t such great models for immune responses in mice, either, although the eventual market for mouse therapeutics is considerably smaller than the human one, as these things go. But what this means is that you have to run the clinical trials in humans to see what happens.
Another quick digression: this is what gets on my nerves when I see people going on about “Ah, we had the vaccines back last March! It just took soooooo long for the big slow drug companies and the big slow FDA to get them rolled out!” Nope. What we had a year ago were vaccine candidates. Nothing is truly real in drug development until it hits human trials, because we get surprised all the time. Now, the success rate for vaccines against infectious disease is one of the best in the whole industry, true. I mean, historically, only two-thirds of them fail, which is indeed a remarkable rate of success compared to most other things we try. But still. You have to go into humans, which takes time and money (and no small amount of each).
Let’s look into some of those mRNA immune responses. The balancing act is because if you set off too strong a response, you could set off a type I interferon pathway because your immune system thinks that it’s under immediate viral attack. There is a long and complex list of “pattern recognition receptors” that are constantly on alert for signs of this, and unfortunately one of those signs is the sudden presence of odd-looking RNA. I won’t get into all the details – there are separate toll-like receptors (TLRs) that pick up on single-stranded RNAs, double-stranded RNAs, RNA species with unusual cap regions, you name it, and all of them have separate downstream signaling networks. One of the consequences is the induction (via those interferon pathways) of an enzyme called ribonuclease L. That one is a real shredding machine, ripping up every RNA molecule it can find. The idea is that it’s only unleashed when there’s a lot of viral RNA starting to pile up in an infected cell, because RNase L spares not and will tear down your own mRNA molecules almost as cheerfully as it destroys viral ones. It’s an extreme measure, and it’s pretty much the last line of defense before a the apoptotic pathways get triggered and the cell falls on its own biochemical sword in full death-before-dishonor mode.
You clearly don’t want to trigger this process if you’re trying to administer an mRNA, because you will have sent most all of your attempted therapy molecules straight into a cellular buzzsaw instead. To pile on, that interferon response also sets off hundreds of genes that (among other things) damp down translation of mRNA into protein and the rate of protein synthesis in general. Which is more of what you don’t want.
That’s why there are so many modifications made to the RNA molecules in any attempted therapy. Bert Hubert takes you through these for the Pfizer/BioNTech sequence here, and it’s a great look at the subject. Modifications get made all up and down the sequence – the cap region, the 5′ untranslated region that comes next, the codons that make up the sequence for the antigen protein itself (the open reading frame) and the nucleosides themselves, the 3′ untranslated region after that, and the “poly-A” tail at that end of the whole thing. This is one of the things I’m talking about when I say I’m glad that we had a chance to explore this stuff over the last ten or twenty years before we needed it for the pandemic: there are a lot of changes that you can imagine making to all these regions, and there’s been a lot of trial-and-error to see which ones are beneficial.
Another thing you have to watch out for is the production of some double-stranded RNA while you’re making your mRNA payload. You will of course have optimized your RNA production platform away from this sort of thing, but the problem is that even very small amounts of dsRNA can set off the toll-like receptor systems and undo everything after your dosing. dsRNA is an unusual chemical species found in very low levels in human cells under normal conditions, and its increased presence is generally interpreted as “funky virus detected”. A further purification step may well be needed, even with an optimized process. Here’s a paper from Moderna talking about just that problem, as well as the use of pseudouridine species to replace native uridine whenever possible. The immune response you get in human patients is very sensitive to both of those factors.
But we haven’t even gotten to the effects of the formulation yet, and here’s where things might get complicated. (I realize that this phrase might cause some people to bury their heads in their hands or go off and clean their basements, but, well, immunology). As is now well-known, the current mRNA vaccines are formulated as lipid nanoparticles, and the lipids themselves include at least one cationic species and at least one that has a polyethylene glycol chain on it. Jonas Neubert lists these for the Pfizer/BioNTech and Moderna vaccines in this great post. Now, there’s a potential problem, in that your adaptive immune system can raise a response to the ingredients of these formulations themselves. That leads to the phenomenon of “accelerated blood clearance” (ABC) after repeated dosing, and this is described in detail in this 2019 Moderna paper, which is open-access.
From that work, it looks like what happens is that many people have antibodies (IgM ones) that already recognize some phosphocholine motifs. That sends some of this antibody-bound form to the spleen, where it’s torn up and the PEG-lipids get presented to B cells. Some of those then get activated to this as an antigen, and that primes the immune system to really rip into another dose with this same formulation (see their Figure 5). Anti-PEG antibodies are indeed a thing (and in fact, there are people who have them even without ever having been exposed to an mRNA formulation).
Here’s yet another one of those balancing acts, though: for a vaccine, some of the adjuvant effect of the mRNA vaccines may well be through the lipids in the formulation. In fact, lipid nanoparticles themselves are being investigated as adjuvants for non-mRNA vaccines. What happens, then, if you’ve been dosed with an LNP-formulated vaccine and then (perhaps a few years later) need to take another LNP-formulated therapy? Does its efficacy decrease because of your prior immune response? On the other side of the question, would an mRNA vaccine whose lipid formulation was somehow completely immunologically silent even work as well? The situation is reminiscent of the potential problems with (for example) adenovirus vaccine vectors. You have populations with pre-existing antibodies to some of those, of course, but after you’ve vaccinated a large population you’ve just created a bunch more of them. There are reports of RNA therapies in animal models that do not elicit immune responses on repeated dosing, but the ABC literature shows that it can be a real problem. And you can also see that people are working on the problem by coming up with formulations featuring “PEG-shedding” to ameliorate the immune response.
I’m not sure yet how this all works out. I’m not sure if anyone is sure. With some of these new therapies and new vehicles, we are leaving immunological footprints, and we’re going to have to keep that in mind. Some of them might end up being like footprints on the beach, and wash away after a while – but others are going to be like the footprints we’ve left on the Moon. . .