It’s a common, straightforward idea in science: When a viral, bacterial, or parasitic pathogen sneaks its way into a host and tries to infect cells, the immune system detects its presence and mounts a counterattack.

However, Harvard Medical School gastroenterologist Jonathan Kagan published a hypothesis paper in Science last week (January 26) that flips that idea on its head. In the paper, which has been downloaded over 9,000 times as of this article’s publication, Kagan asserts that the pattern recognition receptors (PRRs) that induce an immune response are not activated by the pathogens themselves. Instead, they respond to pathogen-associated molecular patterns (PAMPs), which are ligands such as short strands of genetic material or degraded cell wall proteins that are only released when the infectious agent has made an error, such as performing a low-fidelity genomic replication or dying as a result of a maladaptive or untidy mutation. By contrast, successful pathogens are able to sneak under the radar, only being targeted and destroyed when a different bacterium or virus messes up and trips the alarm.

The Scientist caught up with Kagan to talk about the origins and implications of his idea, how it’s been received despite contradicting established ideas in the immunology field, and how it could lead to newer and better immunotherapies.

The Scientist: I appreciate you taking the time to chat and let us do this article. It was a really interesting paper.

Jonathan Kagan headshot

Jonathan Kagan: Yeah, thanks. I appreciate that. In fact, since it came out last week, it’s really attracted quite a lot of attention. It’s gotten 100,000-plus views on Twitter, and I must have gotten 75 emails from random scientists around the world with comments and suggestions. And, you know, it’s funny: I usually spend most of my life writing research articles based on experiments that we do in the lab, and you’ll get your colleagues once in a while saying, ‘Hey, nice study,’ or something like that. It’s an opinion piece, but it certainly attracted a lot of attention. The goal was to synthesize a thesis that collects 1,000 little discussions that I’ve had over the years with my colleagues.

TS: I’m sure it’s nice to be able to put it to paper. This idea is pretty different from a lot of assumptions that are made about how immune cells detect pathogens. I read this as a sort of mini review, where you have several examples throughout. You said that this is something you’ve been discussing for years. But I’m curious where the idea began. What was it that clued you into the idea that this might be going on?

JK: So, in 1989, Charles Janeway illustrated a prophetic idea that there would be these receptors that our immune system uses to detect infections, and that those receptors were really operating like the windows that we have into the microbial world around us. And so, without those receptors, we wouldn’t know the microbes—the infections—that we would encounter, and we’d be in deep trouble. But with those windows, you could see the infection and fight it before it actually causes problems.

Fast forward ten years, there was an explosion of research in the late 1990s and early 2000s when these so-called windows, these pattern recognition receptors, started being identified. And these days, we know that there are a few dozen of them and their job is exactly what Janeway predicted, which was to sense molecules produced by infectious agents. Because of that, there was a very simple if-then statement that if the infectious agent is producing these molecules and causing disease, then we sense the infectious agent that’s causing disease. And it’s for this reason that we call these molecules that our innate immune system senses pathogen-associated molecular patterns.

All pathogens have these molecules. And there’s been an ongoing discussion through the years about how our immune system can recognize infections productively. Now, let’s fast forward many, many years later, and we actually have an amazing molecular insight into how these receptors recognize infectious agents, these so-called PAMPs. Most of them detect nucleic acids. And there’s no living organism that displays its nucleic acids on its surface: A bacterium hides its DNA and RNA inside of itself. Viruses do the same thing; our cells do the same thing. And so, it was kind of odd that the sensors of infection would detect molecules that were hidden from those sensors by the infectious agent itself.

I think that’s a very clever Achilles heel for our immune system to take advantage of. By waiting for a pathogen to make a mistake, which always happens during an infection, you, by definition, will be able to detect the infection itself.

To give you a very simple example of something that’s contemporary, SARS-CoV-2 has its RNA, which our immune system detects. But it’s, of course, inside the virus, not on the outside of the virus, and Salmonella, tuberculosis, E. coli, the bacterium that causes plague, all of these bacteria have DNA and RNA inside of them. Yet, at the same time, our immune system detects them. And so, that was kind of intellectual disconnect number one.

Intellectual disconnect number two came from the discovery, over many labs over the years, that many of these DNA and RNA sensory proteins can be found in lysosomes. And so, if they’re inside of lysosomes, that means then that not only are these molecules sensing nucleic acids, but they’re sensing nucleic acids in an organelle that’s designed to destroy the bacteria and the viruses that we see. Successful pathogens would have avoided being killed, and [therefore] avoided being sensed. And so, you have this very odd disconnect number two: You have sensors of infection that are placed in regions of the cell that are only inhabited by infectious agents that made a mistake; infectious agents that tried to infect their cells but got themselves killed.

TS: This is kind of a semantic question for me. But when you were explaining the if-then statement, that all pathogens have these PAMPs, and therefore they’re all detected. But even in your paper, you give quite a few varied examples, such as bacterial flagellin proteins, viral RNA, and so on. Is that just a semantic category that these are the pathogenic molecules that are detected, or are we talking about an actual unique set of molecules here?

JK: The examples that I gave are all examples of molecules that are largely—but I’ll say for conversational purposes—only produced by microbes; we don’t make them. Flagellin proteins, cell wall components of bacteria; we don’t have a cell wall, so we don’t make those molecules. With that in mind, it also was emerging around the same time that our innate immune system does not detect the intact proteins from the cell wall. What they often detect are pieces of these proteins that would only be revealed if they fell apart. And so, you have another example: The successful pathogen actually has a very healthy and robust cell wall, but if your cell wall was falling apart, then our immune system can detect it. So, when would your cell wall start falling apart? Probably when you’re being degraded. And, by definition, that means that this an unsuccessful pathogen.

TS: You talk about successful versus unsuccessful infections and about errors, with these PAMPs coming from pathogens that make mistakes. And in context, that would be a poorly executed molecular process that gives off these fragments or from pathogens that die off from other reasons and leave behind parts that can be degraded?

JK: I would say all of the above. There are years of research showing that not all infectious agents are capable of actually causing infection. So, for example, if you take 100 bacteria, or 100 viruses, and you add them to cells, some of them—I’ll make up the numbers, let’s say 80 percent of them—will be able to successfully infect a cell, but two out of ten will actually fail in their infectious attempts and be killed by the host. Sometimes the host wins, and sometimes the pathogen wins. And once the pathogen makes a mistake, that is when, I would argue, your immune system is able to dispense the entire infection.

TS: One point that you made in your paper that I wanted to ask about is how, as you mentioned, this idea essentially means that the immune system is defending against a pathogen’s ability to rapidly evolve. I assume that’s connected to how evolution is not a tidy process—maybe more errors happen as it adapts to a pressure. So I’m curious what that could mean for long-term immune system interactions, and the idea of the immune system and pathogens being locked in an arms race.

JK: Yeah, for me, that is a very interesting concept to consider. I mean, the truth is that we know, biologically, that pathogens don’t have to make the mistakes that some make. I’ll give you one simple example: Viruses have polymerases that replicate their DNA or RNA, just like we do. Our polymerases work with almost perfect—nothing is absolutely perfect—but almost perfect fidelity, which means that every time that your or my cells multiply, we don’t make mistakes. But viruses have extremely low-fidelity polymerases, meaning that they could be more efficient at multiplying their genomes, and they [aren’t]. The question has always been ‘Why?’.

Every field has a tipping point where enough people and enough data emerge, where you can completely shift the way you think.

Many evolutionary biologists have suggested that the reason why is the same reason why people get a liberal arts education as opposed to going to a specialized school. If you go to a specialized school for college, you know when you’re 18 years old that you’re going to be, for example, an engineer. You’re going to take engineering classes, and that’s the only thing you’re going to do. As long as society still needs engineers, you will be successful in it, and hopefully, you’ll enjoy your life. In a liberal arts education, on the other hand, you are trying many different things in college, and eventually you find something that works and maybe you’ll change your mind later, because you took a few classes on a different topic. That same logic applies to why viruses have low-fidelity RNA and DNA polymerases—because they’re effectively allowing themselves to make mistakes, because one of those mistakes may actually result in new opportunity. They may create a mutation by accident that allows them to now infect a different host. And so, if you are highly specialized, where you never make a mistake, that means that you are absolutely dependent on the habitat—that you are locked in for life. There’s no chance of evolvability and if habitat goes away, so do you. But if you are liberal in your, let’s say, exploration of your genome by having mistake-prone virulent strategies, once in a while, one of those mistakes may allow you to jump into a different species as a host and allow you to continuously adapt to the world around you.

So what that means, then, is that pathogens are always allowing themselves to make mistakes for the long-term survivability of the species. But because it’s a mistake that’s being made, it’s very difficult for the pathogen to control it. I think that’s a very clever Achilles heel for our immune system to take advantage of. By waiting for a pathogen to make a mistake, which always happens during an infection, you, by definition, will be able to detect the infection itself. A successful pathogen, which may have infected the cell right next to you, is still going to be eliminated because all of our immune responses and the immune responses of plants are systemic. A chain is only as strong as its weakest link; if you have a chain of 100 pathogens, and one out of that makes a mistake, that one is the one that’s detected by the immune system. And now you have inflammation and defense against the entire population of infectious agents.

TS: It’s easy to ascribe agency to these things: The pathogen is trying to change and the immune system is trying to detect it. But assuming this hypothesis is correct, and this is happening constantly and randomly, it seems like an interesting challenge to adapt successfully while avoiding the immune system.

JK: Yeah, I would agree with that. I mean, what’s interesting is despite the fact there’s so much literature to support this contention that what we’re really detecting are PAMPs that could never be revealed by a successful pathogen, the general belief has always been that we detect pathogens. So, if you want to say that it is true that pathogens exist and they cause infection, and it’s also true that the PAMPs that we know of are the real ones, there must be a way to accommodate both of those statements. And the way to accommodate both those things is to propose that there are indeed mistakes made during an infection. And we call these mistakes, for scientific reasons, infectious infidelities, which means that if it was a higher-fidelity infection, all the pathogens would be able to succeed. If it’s a low-fidelity infection, sometimes they’re going to make a mistake.

TS: You were talking earlier about the reception to this hypothesis paper. It doesn’t have any brand new experiments associated with it, though you certainly cited a bunch and you talk about how there’s been plenty of publications sort of supporting this idea. But there’s this common perception that the immune system targets the pathogen itself. This is a pretty bold assertion—it’s a significantly different idea. How was it been putting that out there into the literature?

JK: The response has been phenomenal. Many folks have emailed me over the last week, giving me other examples of what they considered to be pathogen infidelities that drive immunity. And you start realizing that even with what I consider to be comprehensive documentation of infection infidelity examples I provided in this essay, they were more out there.

See “Prokaryotes Are Capable of Learning to Recognize Phages

So far, the community has really said, “This makes a lot of sense. We’ve always kind of discussed it in this way, but never formally documented the thesis.” Every field has a tipping point where enough people and enough data emerge, where you can completely shift the way you think. Certain ideas, over the years, are not effectively ready for wholesale acceptance by the community. And that’s one of the reasons why people often say ideas were ahead of their time. But in this instance, I think that this idea was proper, it was ripe to this time, because of how many examples that are in the literature, which really counter the idea that the pathogen is the entity that pattern recognition receptors sense. Now, I should say that in Science, they don’t give you that many words to write. But there is a sister of the pattern recognition receptors, and these proteins are called Guards [or GarDs]. And guard proteins are almost certainly capable of directly sensing the pathogenic entity. The challenge is that that work largely originated in plants, and only in recent years have examples of guards emerged in mammals; humans and mice, for example. But my prediction would be that over the years, we’re going to learn that your immune system effectively uses two sisters to protect themselves. The common one, the most ancient one, the one that’s found in every single multicellular organism that’s ever been sequenced, are the pattern recognition receptors, and those rely on infection infidelities to detect the PAMPs. But the sisters of those are these guard proteins that seem to have the ability to detect pathogens very, very specifically.

TS: How much room is there for shades of gray in this idea? It sounds like you’ve certainly had a great reception for this among people who are saying, “Yes, we see these examples.” Has there been pushback? Are there counterexamples that people have been pointing out too?

JK: There haven’t really been counterexamples. I would argue that the closest are the ones that I referenced at the end of the essay: There are some pattern recognition receptors, and particularly those that bind to [lipopolysaccharides] (LPS), that probably detect all bacteria, regardless of whether they are a pathogen or not. So, they would seemingly violate this thesis. But I also highlighted that 90 percent of the multicellular life on our planet does not have LPS receptors.

TS: One of the things that you mentioned within your essay is the idea that this could drive experimentation into new categories of immunotherapies, or antibiotics that target cell walls to degrade them and induce an immune response. Are there obvious or feasible next steps towards testing that?

JK: There are! Yeah, that idea, I would argue, is quite ripe for testing by the biotech and pharmaceutical communities, and even the academics who are medically inclined. Because what we typically do when we screen for anti-infectants is screen for them in a very minimalistic model. So for example, antibiotics that target bacteria: We grow these bacteria on a petri dish, and we see if an antibiotic will stop them from growing. But I would predict, based on this idea, that some antibiotics target the cell wall and cause cell wall components to be released, effectively creating a scenario where you effectively get a two-for-one deal. Your antibiotic will also will not only kill the bacteria, but also induce innate immunity to the killed bacteria because, in that instance, the killed bacteria is the equivalent of a failed infection.

You can also think about this from the perspective of the numerous viral entry inhibitors that are out there on the market today, and particularly the ones for SARS-CoV-2. The strategy used to screen for those is again in a very minimalistic system, not an immune cell. The prediction would be that if you blocked viral entry, you will of course prevent viruses from multiplying. But if you did that screen for drug development in an immune cell, the prediction would be that if you blocked viral entry, that bug will get stuck in a lysosome and killed, and you would then activate your pattern recognition receptors.

So one very simple twist would be to use the exact same chemical screening pipelines for antibacterials and antivirals, but instead of using minimalistic systems, using immune cells as the responder to the dying pathogen.

TS: And at that point, you could just check whether these mechanisms would be activated, or activated by a very specific stimulus?

JK: You can actually use the cancer chemotherapy field as a frame of reference. Many years ago, many chemotherapies were identified because they kill cancer cells, but we now know that one of the ways in which chemotherapies actually work is by not only killing cancer cells but doing so in a way that results in the release of DNA, so your immune system thinks there’s a virus around. So, now we know that your innate immune system needs to respond to the dying cancer cells in order to protect you from the cancer itself. The same exact logic applies here. Perhaps the most effective antibiotics already on the market today are so effective not because only they kill the bug, but because they kill the bug in a way that releases a PAMP.

TS: That sort of preclinical test would also test this overall hypothesis, right?

JK: You could almost work backwards, which is what we just discussed, and go right into clinical development. And say that if it is true that most effective anti-infectives on the market are so effective because they not only kill microbes but also stimulate pattern recognition receptors, well that’s a win in terms of validating this idea.

But from an academic perspective you could also be a bit more precise, because one of the predictions here is that if you did an infection of human cells in a lab, and you actually monitored at the single cell level which cells get infected and which cells win the infection fight, and then determine which of those cells are the ones that get their pattern recognition receptors stimulated, you could leverage the next generation of single-cell analyses that are so dominating in the academic community to pressure-test these ideas.

Editor’s note: This interview has been edited for brevity.