5 Microbiome Myths, Busted in 2023

In 1670, Antonie van Leeuwenhoek squinted through a glass bead that formed the lens of his homemade microscope, and peered at a drop of pond water. He saw something that would change the world—something that practically defied belief: life. A whole living sea, teeming with strange organisms, so miniscule that nobody had even suspected their existence. As word of his findings percolated through society, scientists began to look closely at all sorts of mundane things, and practically everywhere they looked, the story was the same: life, uncountable in number and form.

But this cascade of new discoveries—the birth of this new field of science—was tumultuous, and often confusing. One early microscopist, a contemporary of Leeuwenhoek’s, reported observing tiny things swimming in the blood of plague patients, and ascribed the disease to their activity.¹ In reality, historians now suspect that he was actually seeing the patient’s own blood cells. 

It’s hard to blame him, though, if you put yourself in the shoes of someone living through this first bacterial revolution. Imagine being one of the first people to have to wrap your head around the fact that the gunk between your teeth is made of countless miniscule wormy things. The idea would sound like a gruesome fantasy, right up until the moment you saw the proof of it with your own eyes—and forever after, you’d question your intuition about what’s reasonable or fantastical. The discovery of bacteria was a seismic shift in our understanding of the world, and feeling the ground lurch beneath your feet is always unmooring.

With time, though, the true picture began to come into focus. Skilled craftsmen created more powerful microscopes, in the style of the ones we still use today. Biologists set to work cataloging and classifying the different types of microorganisms, identifying relationships, sketching that first dim outline of the tree of life as we now know it. Soon, others began to properly study microbes’ roles in disease, and before long, germ theory had become the dominant explanation for most illnesses—and laid the foundation of modern medicine.

We’re living through a second bacterial revolution: the dawn of the era of microbial ecology, where we’ve started to recognize that the human microbiome is an essential part of the body. It’s a lens as powerful as any microscope, and with each passing year, the research that comes from applying it seems to redefine what’s possible. Does the microbiome hold the answer to Alzheimer’s disease? Cancer? Maybe—but while 2023 brought its fair share of Earth-shaking and exciting discoveries, it was also a year of bringing things into focus, indulging skepticism, and figuring out where the record needs correcting—as in the case of the over-eager microscopist looking at a blood sample for the first time in the 1600s. Here, let’s talk about five advancements in microbiome science from 2023 that made a splash by overturning what we thought we knew. 

Science with us—nerdy reads for your inbox biome¹

Sign Up

¹By signing up, you consent to receive Seed emails.

Myth #1: We have a microbiome before we’re even born.

The discovery that a major portion of the microbiome is highly heritable was one of the foundational findings in the field—key to our understanding that these bacteria are almost as much a part of the human body as any of our own genes. Thanks to gene sequencing, we now understand that people generally possess many of the exact same strains of bacteria that their parents do—particularly their mothers. What’s more, analyzing the poops of ancient humans and neanderthals has shown that these ancestral microbiomes contained bacteria that are very similar to the symbiotic species found in the modern human gut.² But this raises questions: How do these bacteria make it from one generation’s gastrointestinal tracts to the next’s so reliably, when many of them can’t survive for long outside the human body? 

One of the ideas proposed to answer this question hinges on the idea that a child’s body might be colonized by the mother’s bacteria in utero—even before they’re born into the world. While the womb has typically been considered a sterile environment, in the past decade or so, gene sequencing studies of amniotic fluid have revealed traces of bacterial DNA, which many took as evidence for this pre-birth bacterial inoculation.

But it takes more than bacterial DNA to make a microbiome—and last year, an all-star cast of microbial ecologists, immunologists, data scientists, and other researchers came together to settle the question of the fetal microbiome, hopefully for good. Their paper, published January of this year in Nature, concludes that a healthy womb is sterile after all, and that most of the results supporting the idea of a fetal microbiome are the result of contamination during sample collection, or other methodological errors.³ It’s practically a postmortem on the theory—a sober dissection of the evidence that supported it, to see how the field can learn and avoid making similar mistakes in the future. 

Key among the team’s conclusions: The more sensitive your tools are, the more careful you have to be when you use them. The technologies of the molecular biology age make it possible to turn even a single cell’s worth of DNA into a readable signal. But the caveat is that even a single cell’s worth of contamination can make it seem like there’s something where there’s really nothing. 

Considering how ubiquitous bacteria are practically everywhere outside the womb, these kinds of trace contamination events are unavoidable—part of why the team also focused on the importance of interdisciplinary collaborations and critical thinking in the interpretation of results. In a key example of this, they point out that Gloeobacter, one kind of microbe purportedly found as part of the placental microbiome, is photosynthetic: like plants, it gets its energy by feeding directly on sunlight. If that doesn’t raise a red flag, consider: It’s pretty dark inside the human body. 

But that still leaves the question: How do humans get their gut bacteria?

Myth #2: Vaginal births are beneficial because of the transfer of the mother’s vaginal bacteria to the baby.

Fortunately, in March, a paper published in Cell Host & Microbe shed some much-needed clarity on the subject.⁴ 

It’s long been known that children born by C-section have less diverse microbiomes than their vaginally born peers. They are also more susceptible to certain diseases, like asthma and seasonal allergies.⁵ For years, these facts led people to assume that it’s the birthing process itself—passage through the vaginal canal and exposure to its bacteria—that provides the seeds of a healthy microbiome for the infant. But more recently, people have started to question this assumption, as evidence has emerged indicating the importance of the mother’s gut microbiome—and even new concepts like the breast milk microbiome.

To explore the relative importance of these routes of transmission, a team of researchers from institutes across Europe sequenced the genes of bacteria from mothers and their infants over the first month of life. The crux of their approach was to examine multiple body sites independently, sampling the microbiomes in stool, saliva, skin, and nose of both mother and infant—as well as the milk and vaginal microbiomes of the mother. By tracking strains that showed up in both parent and child, the researchers were able to identify which of mom’s microbiomes those strains originated in—and to examine how C-section birth altered that heritability.

The results? For one thing, a mother’s vaginal bacteria make up a relatively tiny fraction of her child’s gut microbiome. Remarkably, it’s about the same amount whether a child is born by C-section or vaginally. And while about half of each of a child’s microbiomes comes from the mother, bacteria from each of the mom’s microbiomes show up in each of the baby’s—with milk bacteria showing up practically everywhere.

One key difference between vaginally born and C-section infants revealed by this study was that vaginally delivered infants had gut bacteria that more closely resembled their mothers’ guts, while C-section babies’ gut microbiomes seemed to be more influenced by the mother’s milk microbiome. 

Their findings suggest that the ubiquity of germs in everyday life is enough to transfer important gut species from a mother to her child—and that something about C-section birth can impair this process, preventing keystone species like Collinsella and Bacteroides from properly taking root in the child’s gut. (Want to better understand how your microbiome is made? Read our deep dive.)

Myth #3: Healthy people have a “blood microbiome.”

Like we said earlier, it takes more than some bacterial DNA to make a microbiome—but what exactly does it take?

The answer to that question is at the heart of the third finding on our list, a March 2023 article in Nature Microbiology titled “No Evidence for a Common Blood Microbiome Based on a Study of 9,770 Healthy Humans.”⁶ 

Microbial ecologists draw a hard line around what counts as a microbiome, defining it in the paper as “a prevalently stable, non-pathogenic, complex microbial community that is metabolically active.” Here, much like in the “fetal microbiome” paper above, the study’s authors turn a critical eye on a spate of recent findings reporting bacteria in the blood of healthy people (which has long been thought to be sterile) and conclude that, by this definition, they don’t count as a microbiome.

It’s a complicated series of qualifiers, but words like “non-pathogenic” and “stable” are key parts of that definition. You can sometimes find bacteria in the human bloodstream, but if they’re present in any significant number, it’s usually a sign that a person is headed for the hospital. 

Many of the findings that this article refutes are based on gene sequencing, which is notoriously susceptible to contamination. Consider: In the old days, when bacteria had to be studied by growing them in culture or looking at them under a microscope, a researcher searching for bacteria in the bloodstream could avoid a false signal by steam-sterilizing their instruments, and cleaning the patient’s skin with antimicrobial chemicals like iodine before inserting a needle.

While that’s all it takes to keep live bacteria out of most samples, molecular biology tools like PCR (the method used to detect COVID in mucosal swabs) work whether a cell is alive or dead, as long as its DNA isn’t degraded beyond recognition. As a result, things which are sterile by any conventional definition can still give false signals when examined for bacteria by these methods. This isn’t a major problem if you’re looking at a sample where there are a lot of bacteria, like feces: The chemicals in DNA extraction kits often contain a few stray copies of contaminant genomes like Bradyrhizobium or Sphingomonas, but these are a drop in the bucket—negligible, compared to billions of bacterial cells in every gram of a fecal sample.

Problems arise when the material you’re looking at is largely sterile, and those contaminant cells in the DNA extraction kit are the only bacterial genes present. The problem is so common that researchers have recently coined a term to describe the phenomenon: the “Kit-ome.” Although these false signals are easy enough to distinguish from the real thing (most of them aren’t native to environments anything like the human body), many researchers seem unaware of the problem, and don’t think to filter them out before reporting their results.

That’s not to say that contamination was the only source of bacteria in people’s bloodstreams discovered in this paper, though. While more than 80% of healthy subjects showed no trace of microbial life, the remainder had a small smattering of bacteria commonly found in the gut, mouth, or other body sites. However, one key characteristic of a microbiome is consistency across people—something noticeably lacking from these blood samples. There’s remarkable diversity in the human microbiome, but there are still core groups like Bacteroides and Bifidobacterium that show up at some level in the guts of nearly every healthy person. The same is true on the skin, where Staphylococcus and Cutibacterium abound. In the blood, though, there are no such reliable residents—a nail in the coffin for the idea of the blood microbiome.

Myth #4: Nearly all cancers have a microbiome signature.

It’s a known fact that certain microbes play a role in some types of cancer. Over 90% of cervical cancers, for instance, are caused by the human papillomavirus (or HPV)⁷. Helicobacter pylori, a bacterium known for causing gastric ulcers, is responsible for most cases of stomach cancer: The inflammation and tissue damage that it causes increase the risk of DNA mutations.⁸

So it was a big deal when, in 2020, the Knight Lab at UCSD reported distinct bacterial signatures in 32 different kinds of cancer, including leukemia, breast cancer, and melanoma.⁹  Their analysis pulled gene-sequence data from over 18,000 samples of tumor tissue and patients’ blood, logged in a public dataset known as The Cancer Genome Atlas. After first filtering out all the human DNA, they turned a machine-learning algorithm loose on the remaining data, with the goal of discovering rules that could help differentiate one type of cancer from another—or even tell healthy samples apart from cancerous ones—based purely on the types and numbers of microbes present in a sample.

The result was a smashing success—or so it seemed. Their program could differentiate tumors from healthy samples in 15 of the 33 types of cancer. When comparing one type of cancer to another, it could reliably tell which kind it was looking at in 32 of the 33 types. The implications were enormous. Most obvious among them: a distinct microbial signature in diseases like breast cancer would enable earlier and more reliable diagnosis, saving lives. Even beyond that, it raised fundamental questions about these cancers: Could certain viruses or bacteria somehow be responsible for melanoma? For leukemia? This one extraordinary paper seemed poised to revolutionize how we think about a whole suite of deadly diseases. 

But extraordinary claims, as the saying goes, require extraordinary evidence. This year, a team of researchers led by Steven Salzberg, Ph.D., at Johns Hopkins University subjected the Knight Lab’s 2020 study to the kind of close scrutiny it deserved; and in the process, they unearthed major errors in the earlier study’s methodology that invalidate its findings. Their analysis, published in October of this year, doesn’t pull any punches: “The microbiome-based classifiers for identifying cancer presented in the study,” they state flatly, “are entirely wrong.¹⁰” 

Here, as in the “blood microbiome” work discussed above, the authors were motivated to do some deeper digging when a common-sense check of the headline results left them scratching their heads. Hepandensovirus was purportedly the strongest identifying feature in adrenal cancer—a surprise, considering it’s a virus that typically only infects shrimp and other crustaceans. In kidney cancer, the machine learning algorithm rated the abundance of the bacterium Thiorhodospira as being highly important—despite the fact that it’s native to environments a hundred times more alkaline than any part of the human body.

When Salzberg and his colleagues took a look “under the hood” of the computer-generated classifiers, they discovered two major issues—either of which might have been enough to render the study’s results useless. The first problem was in the Knight study’s process for deciding whether a given “read” (a short sequenced strand of DNA or RNA) had come from a human or a microbe, and—if the latter—which microbe it had come from. 

Although the original study’s authors filtered their data by throwing out anything that appeared to match the human genome, that matching process isn’t perfect (it can’t be, because every human’s genome is slightly different). As a result, millions of human-genetic “reads” were left in the data from each sample when they were compared against a database of microbial genomes, to discern which microbes they had come from. This might not have been a problem in itself, but the database that the 2020 study’s authors used was enormous and messy: Many of the microbial genomes in it contained a few erroneous gene sequences from those microbes’ human hosts.

The upshot? Millions of misclassified reads meant that, in some cases, the original study’s estimates of microbial abundance were 67,000 times higher than they should have been.

To make matters worse, the process they used to “normalize” the data—making it so that different sample types and different studies could be compared to one another—introduced tiny, distinct irregularities to each sample’s data. The machine-learning algorithm picked up on these, and discovered that it could use them to tell which dataset a given sample had come from. In short, the program figured out how to cheat at the task of classifying cancers.

Machine learning is impressive in its ability to find patterns in huge datasets. It can extract information that ordinary analysis would miss entirely, via a process that’s so complicated it can often be hard to understand. That complexity is the source of its power, but also the greatest pitfall in this kind of work. The authors and their peer-reviewers should have caught these errors before the original paper was published, but it’s all too easy to think, “Who cares how? It gets the right answer every time!”

In some ways, it’s troubling that it took so long for someone to turn a critical eye to these findings. The Salzberg article points out that, since the erroneous 2020 study was published, more than a dozen papers have come out which build on its findings—and all of them are likely invalid as a result. 

Even so, the recent paper is a heartening—and even entertaining—example of science working as it should. One of the computational tools used in the 2020 study was originally developed in Salzberg’s lab, and you can almost sense the personal indignation in the patient, methodical evisceration of the study that so misused his work—it’s rare for a scientific article to communicate so much human drama in its subtext.

Myth #5: Alzheimer’s begins in the brain.

Alzheimer’s is a devastating disease—one that, despite the best efforts of modern medicine, we still have very little insight on. Patients lose their memory and their faculties as tangled clumps (or “plaques”) of a protein called beta-amyloid accumulate in the brain. So far, much of the effort to treat Alzherimer’s has focused on developing drugs that prevent or remove those plaques. 

Some of these drugs work—but only in the sense that they stop the plaques, while doing little or nothing to impact the course of the disease.

For a long time, there have been hints that, although this disease manifests in the brain, it may begin elsewhere. In 2010, researchers reported that amyloid beta—the normal function of which is still somewhat unclear—has antimicrobial properties, and may be part of the body’s natural defense system against pathogens.¹¹ A number of studies found that diet has an impact on a person’s risk of developing the disease, and in animal models of the disease, the microbiome plays a significant role.^12,13

Animal models aren’t perfect, though. Much like the amyloid hypothesis, they involve a lot of assumptions about how the disease works. But over the past few years, the human evidence has started to roll in. Metagenomics studies looking at the gut microbiome in people with dementia have found significant differences between the microbiomes of Alzheimer’s patients and healthy control subjects—things like the suppression of key butyrate-producing species. This is consistent with what we understand about butyrate’s role in our biology: It helps the body control epigenetic programming by acting as a histone deacetylase inhibitor, or HDAC.¹⁴ Synthetic drugs which act via this same mechanism have been shown to enhance synaptic plasticity, and are currently considered leading candidates to treat Alzheimer’s.¹⁵

But correlations and plausible mechanisms can only take you so far in science—the strongest evidence comes from interventional data. In 2020, a scrap of this emerged when a physician published a case report: A patient who had Alzheimer’s became ill with Clostridioides difficile colitis—the notoriously lethal “C. diff”—and was treated with a fecal transplant from his wife. The C. diff was cured…and then the patient’s memory started to improve. By four months after the transplant, the report states, “The patient now remembered his daughter’s birthday, which he had not been able to recall previously, and was able to correct the physician’s recollections of his symptoms.”¹⁶

That brings us to October of this year, when scientists investigating the gut-brain axis published an article describing the mirror-image of that case report’s findings: It’s possible to cause an Alzheimer’s-like syndrome in lab rats simply by replacing their microbiome with that of an Alzheimer’s patient.¹⁷ Using samples from multiple different patients, they found that the worse a person’s disease was, the worse the symptoms it induced in the rodent—while gut bacteria from cognitively healthy people had no such effect, suggesting that there truly is something unique about the microbiome in Alzheimer’s that drives the disease’s pathology.

It’s not yet possible to point to any one kind of bacterium as being responsible for the effect, although the paper notes statistically significant increases in Desulfovibrio and decreases in Coprococcus in patients—and in the rodents post-transplant. In some ways, these results make it look like a cure has never been closer, but if we’ve learned anything from the history of Alzheimer’s research—or indeed the other stories in this year’s roundup—it’s that looks can be deceiving.

In a lot of ways, the past two decades have been like those early years after Leeuwenhoek’s first glimpse of the microbial world.  But now, as 2023 winds down, some of the year’s most prominent findings seem to suggest that the field is finding its footing, sorting out the signal from the noise, and getting a grip on the true scope of this new scientific frontier.