ABOVE: © shutterstock.com, mopic

When immunologist Rachel Caspi began studying an inflammatory eye disease called autoimmune uveitis in the mid-1980s, she had to inoculate mice with a protein found in the mammalian retina to spark the disease. Injecting interphotoreceptor retinoid-binding protein (IRBP) along with an immune-stimulating compound called an adjuvant into the animals’ bloodstream prompted the mice’s own T cells to attack their eyes. That led to inflammation, tissue damage, and eventual blindness.

Caspi wanted to work with a model that better represented human uveitis, in which IRBP-specific T cells attack the retina spontaneously, without the need for any sort of exogenous immune stimulation. Ten years ago, she and her team at the National Eye Institute in Bethesda, Maryland, developed a genetic mouse model of the disease in which mice produced artificially large numbers of T cells that bound to IRBP. As in humans, these T cells traveled to...

This model allowed Caspi to explore the paradox of autoimmune uveitis: that T cells were activated by a protein that was only known to be produced by retinal cells and the pineal gland, tissues that normally have very little interaction with the immune system. The cells of the eye even typically release molecules that keep T cells out, and only T cells in an activated state can enter the immune-privileged organs. And there was no IRBP outside of the eyes and pineal gland to activate T cells in the mouse models.

Caspi reasoned that the T cells might be activated by encountering something that looked like IRBP elsewhere in the body. But she knew the protein alone wasn’t enough; T cells need a second signal to switch into attack mode. “What we think is the T cells need not only the antigenic signal, but also the immunologist’s dirty little secret—the adjuvant—the innate danger signals that are provided by bacteria,” she says.

For most people, molecular mimicry is innocuous, but in the right context it could lead to disease.

Like most immunologists, she was aware of a phenomenon in which pathogens could trigger autoimmune attacks by producing antigens that mimic those on host cells while simultaneously supplying the inflammatory signals required to activate immune cells. This process, known as molecular mimicry, had almost exclusively been studied in the context of infection, which can cause concentrations of autoantibodies, or antibodies that react to the host’s own tissues, to increase in the blood temporarily. For most people, this process is innocuous, but in the right context—for example, in someone with a genetic predisposition—it could lead to disease. Molecular mimicry is now a well-documented mechanism by which infection can lead to certain autoimmune diseases, such as rheumatic fever.

Caspi wondered if something similar could be going on in her mice. But the animals had never been exposed to infectious pathogens. “The first thing we thought about was, where do you see a lot of bacteria in a healthy individual?” she says. “And that is the gut.”

The team eliminated, as completely as possible, the intestinal microbiota in a group of the autoimmunity-prone mice by giving them cocktails of antibiotics while they were still in utero and again after birth. Sure enough, the treatment made the mice less likely to develop uveitis than their untreated counterparts, which typically develop the disease around four weeks of age. When antibiotic-treated mice did get sick, it took longer to happen and the symptoms were less severe.1 This suggested to Caspi that some microbes in the intestines were activating the IRBP-specific T cells that triggered the disease.

Links between the microbiome and autoimmunity are nothing new. Patients with various autoimmune diseases are known to carry altered communities of gut microbes. And transplanting feces from patients with multiple sclerosis (MS) into lab mice predisposed to developing autoimmune disorders can induce disease symptoms in the animals.2 Conversely, antibiotic treatments or housing in germ-free environments can treat, cure, or prevent a host of autoimmune diseases in these rodents. Some of these autoimmunity-microbiome links have been attributed to chemical signals from commensal bacteria that can push T cells into proinflammatory phenotypes. Only in recent years have researchers begun to wonder whether antigens shared by gut bacteria and humans could trigger disease.

Despite being decades old, the concept of molecular mimicry has rarely been considered in light of the growing understanding of the ecosystem of microbes living on and in the human body, says Caspi. When she and her colleagues published their results in 2015, “it was the first time we showed [that look-alike antigens] could come from our own gut flora that we cannot say goodbye to.”

Caspi is still looking for definitive proof that commensal mimicry is causing autoimmune uveitis in her mice. Her team is trying to find out which bacterial antigens are training IRBP-reactive T cells and how they are programming the cells to attack the eye. Since she began this project, however, the idea has been buttressed by the work of several other groups. In the last four years, researchers have tied commensal microbes to autoimmune diseases including MS, lupus, glaucoma, type 1 diabetes, and rheumatoid arthritis, showing that bacterial antigens can activate human immune cells to attack their own body’s tissues.

Recent research has demonstrated that, like pathogens, commensal microbes cause the immune system to “cross-react” to host tissues, says Martin Kriegel, an immunologist at Yale University. The challenge now, he says, is to figure out under what circumstances, molecular mimicry leads to autoimmunity.

Commensal Mimicry in Autoimmune Disease

Recent work indicates that antigens originating from the microbiome may look, from the perspective of immune cells, like proteins found in the human body, and may therefore trigger an autoimmune response. In several studies, researchers have found that the B and T cells that attack the body bind human proteins as well as mimics of those proteins made by commensal bacteria.


Commensal bacteria from the mouth, skin, and gut produce an ortholog of the human protein Ro60. Some of these bacteria and bacterial Ro60 activated T cells from the blood of a patient with lupus. Ro60-specific antibodies from lupus patients also bound to bacterial Ro60, suggesting commensals could have a hand in activating antibody-producing B cells involved in the autoimmune disease.


T cells isolated from the cerebrospinal fluid of several multiple sclerosis patients reacted to both the human and the bacterial versions of the protein guanosine diphosphate (GDP)-L-fucose synthase. Researchers think it’s possible that gut bacteria that make the protein activate T cells that go on to attack the central nervous system.


Several commensal gut bacterial proteins share sequence homology with two proteins isolated from rheumatoid arthritis patients’ blood and joint fluid. Antibodies and T cells from patients, but not healthy controls, reacted to both the human and bacterial peptides.

See full infographic: WEB | PDF
© mica duran

Bait and switch

The earliest demonstration that similarities between molecules produced by human cells and those made by bacteria—pathogenic or commensal—could cause autoimmune disease involved a condition called rheumatic fever, with symptoms that include arthritis, heart failure, and a neurological condition that leads to uncontrolled twitching and trouble with mood and balance. As far back as the 1950s, researchers suspected that infection with group A Streptococcus bacteria could somehow trigger the body to attack heart tissue. The link remained tenuous until 1962, when two Harvard researchers, Melvin Kaplan and Mary Meyeserian, studied the heart of an 11-year-old boy who had died of rheumatic fever following a Streptococcus infection. They found immune cells and deposits of antibodies embedded in his heart tissue.3 They also immunized rabbits with streptococcal cell walls and found that something in the rabbits’ blood—later discovered to be antibodies—reacted with human heart tissue.

Subsequent research in mice and humans revealed that during an infection with group A Streptococcus, B cells make antibodies that bind to a sugar group on the surface of the bacteria, tagging the microbes for destruction. This same sugar group is also found on proteins in the heart, skin, and brain, meaning that activated immune cells will attack these tissues as well. Meanwhile, T cells carrying receptors that recognize the sugar group multiply and, along with the antibodies, induce inflammation that can damage the heart and incite neurological symptoms. The immune system’s response to the infection becomes a destructive attack on self.

This type of mimicry has since been recognized as a possible trigger of several autoimmune diseases. For example, similarities between cell membrane components of nerve cells and of the bacterium Campylobacter jejuni is a well-documented cause of the neurological autoimmune disease Guillain-Barré syndrome. Another established example is the connection between MS and Epstein-Barr virus. The same T cells that attack myelin in MS patients recognize Epstein-Barr viral antigens, and carrying the virus is considered a risk factor for the disease.

One study reported that humans share more than 99 percent of our peptides shorter than eight amino acids long with at least one bacterial species.

Given the increasing number of links between infectious pathogens and autoimmune disease in their hosts, the conceptual leap to consider commensals as a source of mimicry is natural, Caspi says. And molecular similarities between proteins made by microbes and those made by mammals are incredibly common. One study reported that humans share more than 99 percent of our peptides shorter than eight amino acids long with at least one bacterial species.4

In 2016, Yale School of Medicine clinician and immunologist Li Wen and colleagues found that mice genetically predisposed to develop type 1 diabetes display major alterations in their gut microbiome composition. The research team compared the sequence of a pancreatic protein called islet-specific glucose-6-phosphatase catalytic subunit–related protein (IGRP), which is known to activate pancreas-attacking T cells in mice and humans with type 1 diabetes, to a database of bacterial proteins and identified three commensal bacterial mimics. When the researchers injected T cells stimulated with the IGRP mimics into disease-prone mice, levels of glucose in the animals’ urine began to increase, a sign that they were developing diabetes sooner than they normally would. Feeding the mice one of the bacterial species that makes an IGRP mimic similarly resulted in earlier diabetes onset. The study was the first to demonstrate that a commensal mimicking a mammalian peptide could directly activate the T cells responsible for causing type 1 diabetes.  

The list of diseases that might have a link to commensal mimicry in humans is quickly growing. In March of last year, Kriegel’s team published evidence that commensal mimicry may have a role in lupus, a disease in which the immune system attacks a variety of organs, including the skin and kidneys. Knowing that antibodies against a human protein called Ro60 are a risk factor for lupus that can be present in patients’ blood years before their diagnoses, the researchers looked for commensal bacteria–produced proteins that had similar structures. As they expected, the team found protein mimics of human Ro60 that were made by bacteria in the gut, mouth, and skin microbiomes of healthy individuals and lupus patients. Only lupus patients, however, carry antibodies and T cells specific to Ro60. The researchers isolated Ro60-specific T cells from patients and found that the cells bound and reacted to Ro60 mimics made by the gut commensal Bacteroides thetaiotaomicron6 and the oral and skin commensal Propionibacterium propionicum.  Ro60-specific antibodies also bound both the human protein and bacterial mimics. Kriegel says this at least shows it’s possible that commensals could activate Ro60-specific B and T cells in the body.

In unpublished work, Kriegel’s group has also observed that T and B cells specific for an antigen involved in the autoimmune disease antiphospholipid syndrome cross-react with a bacterial antigen found in the gut. And in studies of rheumatoid arthritis, a team of researchers at Boston University and Harvard University have reported overlap in the sequences of peptides from several different commensal bacterial species and two antigens found in fluid from patients’ inflamed joints. In lab experiments, the bacterial peptides activated T cells from about half of patients, but not from healthy individuals.7

Balancing Act

Even if commensal bacteria do trigger immune attacks on our own bodies, researchers agree that eliminating the microbiome is not the answer to treating autoimmune disease. Resident bacteria are essential for developing an immune system, and in some cases are responsible for keeping potentially damaging immune responses at bay. “We are much worse without our commensals than with them,” says Rachel Caspi, an immunologist at the National Eye Institute in Bethesda, Maryland.

One recent study even found that mimicry can play a role in commensals’ immunosuppressive activities. An antigen made by several species of the gut microbe genus Bacteroides looks like the pancreatic protein IGRP and activates a subset of cytotoxic IGRP-specific T cells. Those T cells travel to the intestines and protect mice from colitis by killing other immune cells known to cause gut inflammation (Cell, 171:655–67, 2017).

Still, removing certain microbes could be a strategy to consider for patients with a high risk of developing autoimmunity or in the earliest stages of disease, says Yale University immunologist Martin Kriegel. “It’s theoretically feasible to select those patients, identify the cross-reactive triggers, and remove the inciting agent,” he says. “The hard part is how do you remove them?” In the case of lupus, he says, it might be possible to use a topical antibiotic to remove the source of Ro60 mimics in the skin microbiota.

Caspi is somewhat skeptical of targeting commensals to treat a disease that’s already underway. “A lot of these things are inductive events,” she says. “They trigger something. That doesn’t mean that by eliminating the inductive event you can stop it.” Instead of removing bacteria from the body, Caspi suggests that adding them in the form of probiotics to, for example, manipulate abundances of gut bacteria that display antigen mimics. “By using the appropriate probiotics, we still might be able to affect the progression of disease,” she says.

Mireia Sospedra Ramos, an immunologist at the University of Zurich, says there is already some evidence that altering the gut microbial assemblage could benefit multiple sclerosis patients, pointing to studies that have shown that taking probiotics can reduce markers of inflammation and some symptoms of the disease. She adds that the same commensals may not benefit all patients, however, because people are likely to carry immune cells that react to different antigens.

The solution is unlikely to be as simple as eradicating or altering a single species of bacteria, says Caspi. “I don’t think that we can draw a straight line to treatment from [our findings on molecular mimicry]. We may need to call upon other methods of restoration of immune balance rather than trying to eliminate what caused it in the first place.”

There is some evidence that commensal antigen mimics could provide the stimuli that immune cells need to get access to immune-privileged sites such as the eyes and the central nervous system (CNS). Recently, University of Zurich immunologist Mireia Sospedra Ramos isolated T cells from an MS patient’s cerebral spinal fluid and brain lesions that react to an enzyme called guanosine diphosphate (GDP)–L-fucose synthase. This enzyme is involved in producing cell-surface proteins that mediate cell-cell interactions throughout the body, particularly in the brain and gut. GDP–L-fucose synthase, it turned out, is also made by several species of commensal bacteria.8

Sospedra Ramos and her colleagues found that about 40 percent of MS patients in their study had T cells in the cerebral spinal fluid that react to human GDP–L-fucose synthase. The researchers found that T cells from four out of the seven patients they tested reacted strongly with a version of the enzyme made by two bacterial species that are overabundant in the guts of MS patients.5 She says it’s possible that activation by bacterial GDP–L-fucose synthase in the gut could allow T cells to pass through the cellular defenses surrounding and lining the blood vessels of the CNS called the blood-brain barrier.

“We know that in order to cross the blood-brain barrier, these cells should be activated from the outside of the brain,” she says. “We think maybe T cells in the gut could recognize this protein in the bacteria and later, for reasons that we still don’t understand, can migrate into the brain, cross the blood-brain barrier, and in the brain, recognize the human enzyme. This is the hypothesis.”

Although the idea that commensal bacteria enable chronic autoimmunity remains speculative, experiments like these convince Kriegel that the link between commensal mimicry and autoimmune disease is real. “Overall, I think there’s no question that cross-reactivity occurs and that it can contribute to disease,” he says. “It’s just very hard to prove definitively.”


Several species of commensal bacteria make proteins that share peptide sequences or structures with human proteins that are bound by autoimmune T and B cells. In lab experiments, these bacterial mimics can activate disease-causing immune cells collected from people with autoimmune diseases and induce or speed up disease development in mice. This makes researchers think that antigen mimics made by gut bacteria can somehow activate immune cells that then travel to and attack target tissues, causing autoimmune disease.

Commensal bacteriaHuman antigen being mimickedDisease link

Propionibacterium propionicumRo60Lupus
Bacteroides thetaiotaomicronRo60Lupus
Roseburia intestinalisβ2-glycoprotein IAntiphospholipid syndrome
Prevotella spp. and Parabacteroides spp.N-acetylglucosamine-6-sulfataseRheumatoid arthritis
Prevotella spp. and Butyricimona ssp.Filamin ARheumatoid arthritis

Akkermansia spp. and
GDP-L-fucose synthaseMultiple sclerosis
Bacteroides thetaiotaomicronRo60Lupus
Unknown (likely many)interphotoreceptor retinoid-binding proteinAutoimmune uveitis
Leptotrichia goodfellowiiislet-specific glucose-6-phosphatase catalytic subunit–related proteinType 1 diabetes
Unknown (likely many)Heat shock proteinsGlaucoma

Second signals

Colonizing germ-free mice with bacteria that produce self protein–mimicking antigens has certainly hinted at a causal relationship by increasing the animals’ risk of autoimmunity, but typically the presence of the bacteria or their antigens alone is not sufficient to cause disease. Just as Caspi had to use her “dirty little secret” to initiate uveitis in her mouse models of the 1980s, Kriegel’s team had to treat mice with an adjuvant, in addition to feeding the animals a bacterium that produced Ro60-mimicking proteins, to spur kidney inflammation—a core symptom of lupus.3

I don’t think that we can draw a straight line to treatment from our findings on molecular mimicry.

 —Rachel Caspi, National Eye Institute

This is likely because the immune system has many measures in place to prevent autoimmunity. One major immune control is T cells’ requirement for stimulation through at least two different receptors before they can mount an attack. “Normally, in order to activate naive T cells, you need two signals: the antigen and costimulation,” says MIT immunologist Jianzhu Chen whose team recently uncovered a role for commensal mimicry in a mouse model of glaucoma.9 Most often, T cells receive costimulation from immune cells called antigen-presenting cells (APCs) that act as first responders to injury or infection. These cells consume and process things such as damaged cell components or bacterial antigens, sense the environment, and grant T cells permission to attack if needed.

In the gut, APCs usually do the opposite, supplying signals that tell T cells to tolerate commensal microbes. Something in this process must go haywire to allow commensal mimics to drive autoimmunity. “You can imagine how quickly this could go wrong if someone has the wrong genetic wiring so that not a regulatory response but a pathogenic response occurs,” Kriegel says.

The difference between people who experience autoimmunity and those who don’t may be tied to the presence or abundance of certain microbes, but it also lies in how people’s immune systems handle the cross-reactive antigens those microbes carry. “Any disease is multifactorial,” Kriegel says. “You have the wrong genes plus the wrong bugs together, then you are prone to develop disease.”

Caspi is trying to suss out the components of that recipe in her mouse model of uveitis. In addition to identifying the bacterial mimic of IRBP, her team is looking at metabolites and other molecules made by commensals that may have a hand in directing IRBP-reactive T cells toward a proinflammatory phenotype. “It’s very complex environment,” she says. “There are opposing influences that, when integrated, will ultimately result in disease or no disease.”

Amanda B. Keener is a freelance science journalist living in Littleton, Colorado.


  1. R. Horai et al., “Microbiota-dependent activation of an autoreactive T Cell receptor provokes autoimmunity in an immunologically privileged site,” Immunity, 43:343–53, 2015.
  2. K. Berer et al., “Gut microbiota from multiple sclerosis patients enables spontaneous autoimmune encephalomyelitis in mice,” PNAS, 114:10719–24, 2017.
  3. M.H. Kaplan, M. Meyeserian, “An immunological cross-reaction between group-A Streptococcal cells and human heart tissue,” Lancet, 279:706–10, 1962.
  4. B. Trost et al., “No human protein is exempt from bacterial motifs, not even one,” Self Nonself, 1:328–34, 2010.
  5. N. Tai et al., “Microbial antigen mimics activate diabetogenic CD8 T cells in NOD mice,” JEM, 213:2129, 2016.
  6. T.M. Greiling et al., “Commensal orthologs of the human autoantigen Ro60 as triggers of autoimmunity in lupus,” Sci Transl Med, 10:eaan2306, 2018.
  7. A. Pianta et al., “Two rheumatoid arthritis–specific autoantigens correlate microbial immunity with autoimmune responses in joints,” J Clin Invest, 127:2946–56, 2017.
  8. R. Planas et al., “GDP-l-fucose synthase is a CD4+ T cell–specific autoantigen in DRB3*02:02 patients with multiple sclerosis,” Sci Transl Med, 10:eeat4301, 2018.
  9. H. Chen et al., “Commensal microflora–induced T cell responses mediate progressive neurodegeneration in glaucoma,” Nat Commun, 9:3209, 2018.

Interested in reading more?

living with bacteria 2019 the scientist june issue

Become a Member of

Receive full access to digital editions of The Scientist, as well as TS Digest, feature stories, more than 35 years of archives, and much more!
Already a member?