You don’t want a PAMP-er to put a damper on hopes for an mRNA vaccine – enter scientist Katalin Karikó to the scene! 

Synthetic mRNA technology is awesome – you can put messenger RNA (mRNA) copies of a protein recipe into cells and get those cells to make the protein. This can either replace a missing or defective protein, or it can get cells to make something “foreign” that the immune system will learn to recognize and make antibodies, etc. against. The latter is the basis of Pfizer/BioNTech’s & Moderna’s coronavirus vaccines, which use mRNA to deliver the instructions for making the coronavirus Spike protein (the one that juts out from the membrane and allows the virus to dock to our cells). Those vaccines have proven wildly successful, beyond many people’s expectations. But the technology behind them was almost abandoned because early attempts to use mRNA therapeutically ran into the problem that the mRNA was acting as a Pathogen Associated Molecular Pattern (PAMP) – the mRNA was being seen as foreign and leading to a nonspecific immune response that destroyed the mRNA before it could be useful and caused damaging inflammation. Scientists weren’t quite sure why, and most people wanted to abandon the work, but Karikó believed in it and was willing to take a demotion in order to pursue it. Today she’s a senior VP of BioNTech and her findings are saving countless lives around the world. 

originally posted March 19, 2021. Video added October 9, 2021

Karikó’s key finding was that, if you made some modifications to the mRNA to make it more “human-like,” cells would use it without setting off those generic immune system alarm bells.  RNA only has 4 genetically-encoded letters (nucleotides) – A, C, U, & G – but those letters can get modified, such as with the addition of methyl (-CH₃) groups, or the rearrangement of some of the atoms (isomerization). Bacteria don’t modify their RNA letters very much, but our cells do. Therefore, if our immune system sees un-modified RNA, it might get suspicious that bacteria’s around. But how, at the biochemical level, does an immune system “get suspicious”? And how does it relay that message? One way is through proteins that act as receptors for Pathogen Associated Molecular Patterns (PAMPs). PAMPs can be anything from bacterial toxins to double-stranded RNA (characteristic of some viruses) to, as is relevant here, unmodified RNA. Different Pattern Recognition Receptors (PRRs) bind to different types of PAMPs and trigger signaling cascades, the release of chemical messengers called cytokines, etc. to set the immune system in motion.

You probably hear a lot about adaptive immunity. This side of the immune system involves antibodies, some T-cell responses, and other ways the body learns to recognize specific features of a specific invader. For example, an adaptive immune response to infection with, or vaccination against, SARS-CoV-2, the coronavirus that causes the disease COVID-19, involves the production of little proteins called antibodies that can bind to the Spike protein and prevent it from docking onto our cells. Finding well-matching antibodies involves a lengthy trial and error process by the immune system, so it takes a while. But your body doesn’t just wait around and let invaders build up before then. Instead, there’s another aspect of the immune system called the innate immune system, which recognizes more generic features of pathogens (disease-causing microbes like viruses, bacteria, and fungi). 

One type of PAMP-recognizer is a family of proteins called Toll-Like Receptors (TLRs), some of which specialize in binding to “suspicious” RNA. 

Scientists had known for a while that double-stranded RNA (dsRNA) could trigger an innate immune response. And this made sense because dsRNA is a characteristic feature of some viruses. 

Quick molecular biology background note so no-one gets lost: DNA & RNA are both a type of molecule called nucleic acids, and they’re really similar, but they have slightly different backbone sugars (ribose in RNA vs deoxyribose in DNA). DNA & RNA each have 4 letters – A, C, G, & either T (in DNA) or U (in RNA). Each letter specifically pairs with one other letter (C to G and A to T or U) and therefore, if you have the right reaction-helpers (enzymes), you can use a strand of DNA to make a complementary strand of RNA (a process called transcription). This is super useful to our cells because we store our permanent genetic blueprint (genome), including the genes containing recipes for making proteins, in the form of DNA, which is more stable than RNA. We can keep that DNA locked up safe in a membrane-bound compartment of cells called the nucleus and then, when we want to make a protein, our cells can transcribe messenger RNA (mRNA) copies of the corresponding gene and ship them out of the nucleus so that protein-making complexes called ribosomes, located in the general cellular interior (cytoplasm) can use them to make the corresponding protein in a process called translation. 

So, we have: gene (in double-stranded DNA in the nucleus) gets transcribed and exported -> mRNA (single-stranded RNA in the cytoplasm) gets translated -> protein

Nowhere in that healthy-cell process should there be double-stranded RNA anywhere, or DNA outside of the nucleus. But if a virus infects a cell, there could be, because viruses have their own genetic material. Some viruses have DNA genomes (single- or double-stranded) and others have RNA genomes (single-or double-stranded). Even the single-stranded RNA genomes end up producing dsRNA because, once inside the cell, they have to make copies of their complementary strand, in order to make copies of the original strand and/or use the complementary strand if the virus is a so-called negative-stranded virus. This is getting too complicated to get into here, sorry, but the key point to take away is that viruses often cause there to be dsRNA in cells. Therefore, cells have evolved proteins, including TLRs, that can bind to dsRNA and set off immune system alarm bells. 

So, scientists knew that, if they wanted to try to use RNA to deliver protein recipes, they should probably avoid dsRNA. They used a process called in-vitro transcription to make copies of RNA from DNA “in a test tube,” but when they introduced that RNA into cells, the alarm bells were still going off. Some of that alarm-raising, they would discover was due to dsRNA impurities produced as a byproduct of the transcription process, but even if they purified out that dsRNA, they still saw the cytokine release and other signs characteristic of an innate immune response. So, what was going on?

Karikó wanted to find out, but most people (including her boss as I’ll get into later) wanted to move on. 

Why was she so persistent? From early on she saw the huge potential for mRNAs, with her interests primarily in protein replacement. Say, for example, someone has a disease that causes them to not make functional protein X. Scientists could try to make protein X recombinantly (express it in cells like insect or mammalian cells), purify it, and deliver it to patients. Some drugs work this way, but it’s an expensive, lengthy, and resource-intensive process. Furthermore, the proteins might not be fully “normal” if they’re expressed outside the person. Lots of proteins get modified post-translationally, through glycosylation (addition of sugar chains), phosphorylation (addition of negatively-charged phosphate groups), etc. and these modifications might not get added if expressed in different cells. The proteins might not even fold properly without the aid of chaperone proteins that often help out in human cells. 

If you give patients these abnormal proteins they might not work properly and they might even cause damaging immune system responses. But, if you deliver the mRNA, the patients’ own cells will make the protein, so the protein will be made “on-site” just like the protein is in healthy cells. Plus, the process of making mRNA is much cheaper and more easily adaptable for different proteins since you mostly just need to change the genetic instructions rather than have to figure out a whole new purification scheme. 

So all that sounds great, but it was all just pie-in-the-sky theoretical unless scientists could figure out how to get the mRNA into cells without the body attacking it (and attacking the body’s own tissues in the process). 

In an aha-moment paper, Karikó outlines how she, in collaboration with a UPenn immunologist named Drew Weissman who I’ll tell you more about later, solved the mystery of the RNA’s immunogenicity. They took a type of immune cell called an MDDC (monocyte-derived dendritic cell, if you really want to know what that stands for) and they added various types of RNA, including RNAs from bacteria and mammalian cells. When they added total bacterial RNA (i.e. a combination of all the types of RNA in bacteria), the cells set off immune system alarm bells (instead of ringing loudly, they secreted high levels of immune system signaling molecules called cytokines). But if they added total mammalian RNA, this response was much lower. In jargon, we can say that the bacterial RNA was much more immunogenic. 

Now they needed to figure out what was different about the two (bacterial vs mammalian) and they found a key clue when they divided those two groups of “total RNA” into their different subtypes and tested those subtypes individually. For example, although the total RNA fraction from bacteria was highly immunogenic, bacterial tRNA was much less so (tRNA stands for transfer RNA and it’s a form of RNA that brings amino acids to ribosomes to be added to a growing protein chain during translation). And, although total RNA from mammalian cells wasn’t very immunogenic, mitochondrial RNA was. 

Mitochondria are often described as the “powerhouses of the cell” – they’re membrane bound rooms inside of cells where the cellular energy currency of ATP gets made. Evolutionarily, they came from an early cell swallowing a bacterium, adopting it as their own, and adapting it to form as an intracellular powerhouse. I’m not just telling you this because it’s cool (although it certainly is!). Instead, I’m telling you this evolutionary background because one of the consequences of it is that mitochondria have their own genomes, and thus make their own RNAs and, since they originally come from bacteria, these RNA’s are more “bacteria-like,” and therefore it made sense that they could trigger an immune response. (Note: since these RNAs are usually confined to the mitochondria, they don’t trigger a response unless the mitochondria get damaged or something and the RNA gets out into the cell and, in that case, your cells would want to know!)

But, what does it mean at the biochemical level for RNA to be more “bacteria-like?” A key thing to know about bacterial RNA is that, apart from tRNAs, it contains very few post-transcriptional modifications (methylations, etc.). This is in stark contrast with mammalian RNAs which, apart from mitochondrial RNAs, are extensively modified post-transcriptionally. 

Lack of modification could explain why most bacterial RNAs, as well as mitochondrial RNAs, were immunogenic. And presence of modification could explain why most mammalian RNAs, as well as bacterial tRNAs, were much less immunogenic. It was a beautiful, sensible, hypothesis, but a hypothesis is just an educated guess. Could they prove it?

To do this, Karikó would need to show that RNA modification could prevent RNA from triggering immune system activation. Instead of isolating natural RNAs from cells like she’d been doing, she’d now turn to in-vitro transcribed RNAs (RNAs made in a test tube using an RNA polymerase to make RNA copies from a DNA template) which would allow her to add specific modified RNA bases. Normally, these modifications get added post-transcriptionally, but they can be added during transcription in vitro if you throw them into the transcription mixture. 

But RNAs contain hundreds of different types of mutations and she didn’t know which ones might matter. She wouldn’t be able to test them all, but she could test out a bunch, with help from a hunch… She’d come across a study showing that un-modified uridine (U) could set off particular immune receptors, so she included several different forms of modified U’s, and she found that many of the modifications decreased the immunogenicity (as evidenced by lower cytokine levels and other activation markers in the cells they stuck the RNA in). They tested various U modifications and found that one of the “best” was a natural U modification called “pseudouridine” in which the ring of the base is “rotated” and attached differently to the sugar backbone (best explained visually).

Further cementing the relationship between U modification and immune response, she did a series of experiments in which she set up in-vitro transcription reactions with a mixture of normal and modified U’s. By increasing the fraction of modified U’s compared to normal U’s in the reaction mix, she was able to increase the amount of modification in the transcribed RNA. She made a series of RNAs with the same sequence but different fractions modified and showed that, the more modification there was, the lower the immunogenicity.

But what was going on mechanistically? How was the cell distinguishing between the different RNAs? A paper had come out showing that certain RNAs bound to a Toll-Like Receptor (TLR) called Toll3. So, they thought that maybe TLR3, and/or one of the other TLRs might be involved. To test this, she did a series of experiments in cells that don’t normally express (make) TLRs. These cells didn’t show a strong response to normal or modified RNAs. But that didn’t prove that TLRs were involved because there are other proteins not made in those cells. To show that TLRs are involved, she’d need to add the relevant TLR(s) and see a response. 

She could get the cells to make one TLR at a time and see if cells would react to normal and modified RNAs. If there was a response to one of the normal RNAs (and not the modified), it would indicate that the TLR they added was at least partially responsible. 

Technical note: all cells have the genes for making TLRs, because all cells have a full genome, but in the genome, genes are preceded by sequences called promoters which help the cells know what to make when. The promoters in front of TLRs lead them to be “turned off” in these cells, but she could add extra copies of the genes “exogenously” (from the outside) in pieces of DNA called plasmids, with promoters in front of them that are “turned on” so the TLRs would get made

She added the genes for TLR3, TLR7, and TLR8 one at a time, and measured the release of a cytokine called IL-8. She found that TLR3, 7 & 8 where all capable of sensing the RNAs but that TLR7 & 8 in particular where very sensitive to modified U’s, showing a markedly smaller response to modified RNAs.

This, and later work, showed that modifying RNAs could prevent that general immunogenicity. But would it affect translation (protein-being made from them)? You might think that modifying the RNAs would hinder translation, but remember that normally mRNA is extensively modified. And, in fact, they found that modified mRNA got translated even better than non-modified RNA. Even in actual mice, not just cells in a dish. 

In a 2010 paper, Karikó showed one reason why this was the case. Unmodified synthetic mRNA, in addition to binding TLRs, could bind to another immune system watchdog called PKR (protein kinase RNA-activate). As the name suggests, PKR is a kinase (a phosphate-adder) and when it binds to RNA that it likes, it autophosphorylates (phosphorylates itself) which activates it and then it phosphorylates, among other things, one of the proteins needed to initiate translation, eukaryotic translation initiation factor 2α (eIF2α). That phosphorylation shuts down translation, so the proteins don’t get made. Adding modifications, Karikó showed, prevented the synthetic mRNA from binding to, and activating, PKR, allowing translation to proceed normally and lots of protein to get made. 

There still remained key hurdles to taking this synthetic mRNA technology out of the lab and into the clinic. Some of these hurdles were technical, such as figuring out how to deliver the synthetic RNAs into live humans. But the main hurdle Karikó faced was a lack of support…

To explain, I now I want to tell you more about Karikó (and I’ll list some sources for you to check out at the end to learn more)

Katalin Karikó was born January 17, 1955 and grew up in Hungary, in a town called Kisújszállás. She went to the University of Szeged, where she heard about possible therapeutic uses of mRNA and got intrigued. But she found there were limited opportunities to study the topic in Hungary (where people seemed more excited about DNA therapies). Therefore, she moved to the US in 1985 to work at Philadelphia’s Temple University. She took with her her husband; a daughter, Susan, who was 2 at the time and would later go on to become an Olympic gold medalist in rowing; and a teddy bear stuffed with the money they got from selling their car on the black market (there were restrictions at the time on Hungarians bringing things to the US). 

At Temple, instead of mRNA, at this point she was working with dsRNA. And, somewhat ironically, she was working on capitalizing on the features of dsRNA that make it bad for introducing genes. One of her projects was trying to get dsRNA to amp up the immune system in AIDS patients so that HIV might get caught in the crosshairs. 

In 1990, she moved to the University of Pennsylvania, accepting a tenure-track position. She really wanted to establish mRNA as a therapeutic tool, and one of her projects was trying to figure out a way to introduce mRNA to transplanted blood vessels to get those vessels to make proteins that would help them survive and thrive. In another project she was trying to deliver mRNAs to nerve cells. 

Other groups’ findings that you could inject mRNA into mice’s muscles and they’d make the corresponding proteins (JA Wolff et al., Science, 1990) gave her hope that it could work in humans. And she even had similar successes of her own, delivering mRNA for complex proteins to mice and having them make functional proteins. But the human work wasn’t working because of that immune system stuff. 

Karikó believed the challenges could be overcome, but her bosses at UPenn didn’t, at least not without substantial investment they (and countless grant providers she applied to) didn’t feel was worth it. In 1995 Penn gave her a choice – switch research projects and keep her tenure-track position, or keep working on the mRNA but accept a demotion. In what was a horrific week for her (she had just been diagnosed with cancer and her husband was locked out of the country for 6 months on a visa issue) she took the demotion and kept working tirelessly on the project, on a small salary, limited budget, and without much help.

Without much help that is, until 1997, when, at the copier, she met a newly-hired immunologist named Drew Weissman. After she told him about what she was trying to do, he was hooked on the idea too and willing to help. Importantly, he was well-respected and able to get the funding they’d need. But, as an immunologist, he was more interested in mRNA vaccine technology. So that’s what they’d focus on, initially working on an HIV vaccine.

As it became clear that the mRNA itself was immunogenic (regardless of what protein recipe it held), Karikó became determined to find out why. Some immunogenicity would be okay for a vaccine (but not so much that it causes damaging inflammation and/or stops the mRNA from being translated), but for a therapy like protein replacement, any immune system activation would be a deal-breaker. 

The two paired up and did their amazing work.  They started their own company called RNARx, but that didn’t last long, and UPenn licensed their technology to, among other places, a fledgling German biotech company called BioNtech.

Despite her in-retrospect ground-breaking discoveries, UPenn told her she wasn’t faculty material and refused to promote her back to a tenure track position in 2013. So she left, taking a position as senior vice president of BioNTech. There, in addition to helping lead the development of the widely successful Pfizer/BioNtech coronavirus vaccine, she helps direct a wide-spanning range of RNA therapeutics. 

Note: UPenn still of course, touts her accomplishments as a “University of Pennsylvania mRNA Biology Pioneer”… 

Karikó is really excited about the non-vaccine applications of mRNA technology as well – as she describes in a great talk I’ll link to, which I encourage you to watch, her main interest from the beginning has been more therapeutic uses – vaccines were more fundable and meshed with Weissman’s passions. She’s obviously still super stoked about the vaccine, but she is also excited about the vaccine as a sort of “proof of concept” of the feasibility and vast potential utility of mRNA technology – we’re talking protein replacement for diseases, immunotherapies, etc. So stay tuned! 

One of the biggest hurdles for the technology is getting it to the desired cells. Only cells that take in the mRNA will make the protein and, because the mRNA doesn’t integrate into our genetic material, it doesn’t get passed on to future cells. If you just want something that acts locally, it’s not a big issue

For proteins that get secreted into the bloodstream, it’s not as big an issue because, although only some cells take up the mRNA and make the protein, they can provide that protein to the rest of the body. But, for non-secreted proteins, the mRNA is only going to help the cells it gets into. To help the mRNA get into more cells, scientists are experimenting with different delivery mechanisms, often various formulations of lipids (fat/oils) that coat the mRNA in a sort of shell called a lipid nanoparticle (LNP). And to help the mRNA get into *specific* cells, scientists are modifying those LNPs to contain cell-type specific molecules on their surface – these molecules bind to receptors on the targeted cell type and get taken in.  

Another area of optimization is the cap and the tail of the mRNA – mRNAs usually get a backwards-facing modified G cap at their 5’ end (the starting end) and a long string of A’s called a poly-A tail at the 3’ end (the ending end). Scientists working with mRNA therapeutics often do a lot of pre-clinical optimization to find the ideal tail length, etc. They can also fiddle around with what codons to use – each protein letter (amino acid) is spelled by one or more 3-letter mRNA words (codons). Since some amino acids are spelled by multiple letters, you can change the codons without changing the protein and this can help the protein be made better. There are some other things you can change too, like what “splice variant” to use, etc. but that’s outside the scope of the post. I just didn’t want to give you the impression that it’s as simple as “you only need to know the sequence” to get a functional product and you can “just swap out the sequence” – in theory, yes, but in practice you’re going to need to put in a bit of work specific to each protein you’re trying to make (though not as much as you’d have to if you were trying to express it in vitro and purify it!).

Many people were shocked by mRNA vaccines’ successes. Katalin Karikó was not. She’s believed in them from the start. And, without her unwillingness to give up on mRNA, we might be in an even worse situation today!

more on mRNA vaccines: 

more on in-vitro transcription: 

More on Karikó:

Aaron Brown, Pioneers in Science: Katalin Karikó, Advanced Science News, 01/25/21, 

David Cox, How mRNA went from a scientific backwater to a pandemic crusher, Wired, 12/02/20 

Aria Bendix, BioNTech scientist Katalin Karikó risked her career to develop mRNA vaccines. Americans will start getting her coronavirus shot on Monday, Business Insider, 12/12/20 

That video I was telling you about, where Karikó and Weissman give talks at a virtual “mRNA Day 2020” event: 

Key papers:

This is one of the papers she credits with giving her the most hope to keep going – she delivers synthetic mRNA for a complex protein, urokinase receptor, into mammalian cells. And those cells make lots of functional urokinase receptor proteins. This gave her a lot of hope because this was a tricky protein to make because it’s post-translationally modified, membrane-incorporated, etc. But the cells did it. Successfully! 

Karikó, K., Kuo, A. & Barnathan, E. Overexpression of urokinase receptor in mammalian cells following administration of the in vitro transcribed encoding mRNA. Gene Ther 6, 1092–1100 (1999).

Here they show that unmodified mRNAs set off TLR3:

Katalin Karikó, Houping Ni, John Capodici, Marc Lamphier, Drew Weissman, mRNA Is an Endogenous Ligand for Toll-like Receptor 3, Journal of Biological Chemistry, Volume 279, Issue 13, 2004, Pages 12542-12550, ISSN 0021-9258, 

Here they show that modified mRNAs don’t trigger TLRs – this is the paper I talked most about:

Katalin Karikó, Michael Buckstein, Houping Ni, Drew Weissman, Suppression of RNA Recognition by Toll-like Receptors: The Impact of Nucleoside Modification and the Evolutionary Origin of RNA, Immunity, Volume 23, Issue 2, 2005, Pages 165-175, ISSN 1074-7613, 

Here they show that pseudouridine-modified mRNA is less immunogenic and better-translated than non-modified mRNA in cells and mice: 

Katalin Karikó, Hiromi Muramatsu, Frank A Welsh, János Ludwig, Hiroki Kato, Shizuo Akira, Drew Weissman, Incorporation of Pseudouridine Into mRNA Yields Superior Nonimmunogenic Vector With Increased Translational Capacity and Biological Stability, Molecular Therapy,

Volume 16, Issue 11, 2008, Pages 1833-1840, ISSN 1525-0016,

Here they show that modifications increase translation (protein-making) by preventing activation of PKR (which, when activated, phosphorylates eIF2α to shut down translation):

Bart R. Anderson, Hiromi Muramatsu, Subba R. Nallagatla, Philip C. Bevilacqua, Lauren H. Sansing, Drew Weissman, Katalin Karikó, Incorporation of pseudouridine into mRNA enhances translation by diminishing PKR activation, Nucleic Acids Research, Volume 38, Issue 17, 1 September 2010, Pages 5884–5892,

Here they show that super-purifying in-vitro transcribed RNA to remove contaminants like dsRNA eliminates its immunogenicity:

Katalin Karikó, Hiromi Muramatsu, János Ludwig, Drew Weissman, Generating the optimal mRNA for therapy: HPLC purification eliminates immune activation and improves translation of nucleoside-modified, protein-encoding mRNA, Nucleic Acids Research, Volume 39, Issue 21, 1 November 2011, Page e142, 

Here’s the paper on Pfizer/BioNTech’s coronavirus vaccine using the technology she developed: 

Sahin, U., Muik, A., Derhovanessian, E. et al. COVID-19 vaccine BNT162b1 elicits human antibody and TH1 T cell responses. Nature 586, 594–599 (2020). 

Here’s a good review article on mRNA-based therapeutics: 

Sahin, U., Karikó, K. & Türeci, Ö. mRNA-based therapeutics — developing a new class of drugs. Nat Rev Drug Discov 13, 759–780 (2014).

And one on mRNA vaccines:

Giulietta Maruggi, Cuiling Zhang, Junwei Li, Jeffrey B. Ulmer, Dong Yu, mRNA as a Transformative Technology for Vaccine Development to Control Infectious Diseases, Molecular Therapy, Volume 27, Issue 4, 2019, Pages 757-772, ISSN 1525-0016, 

For the super-nerds, here’s a really interesting talk by a biotech company called TriLink about how synthetic mRNA is made at scale and different choices people can make about caps, tails, and other nitty gritty details! 

photo: Jessica Kourkounis for the Boston Globe

Leave a Reply

Your email address will not be published.