Human microbiome myths and misconceptions 人体微生物组领域相关的谜团及误解

Human microbiome research has undergone rapid growth over the past two decades and thousands of research papers on this topic are now published every year. Huge sums of money have been spent investigating the human microbiome as a cause of, or potential therapeutic solution to, a wide range of diseases, including inflammatory bowel disease and cardiometabolic conditions. Although truly exciting, the increasing focus on microbiome research has unfortunately also brought with it hype and entrenched certain misconceptions. As a result, many unsupported, or undersupported, statements have become ‘fact’ by virtue of constant repetition. Some are more widespread than others and some are relatively trivial, but, cumulatively, they highlight that misinformation is pervasive in the human microbiome literature. Given the potential importance of human microbiomes for health, it is crucial that claims are based on evidence. In this Perspective, we shine a light on persistent or emerging microbiome myths and misconceptions, outlining factual inaccuracies. We begin with relatively minor, but illustrative, points and build towards issues with greater potential impacts. We have purposefully tried to avoid unnecessary finger pointing at original sources of erroneous information, and instead hope that our insights and critical assessment are helpful to the field.

The pace of human microbiome research has greatly accelerated over the past 15 years, but the field is not in its infancy. To state so does a disservice to the excellent research that preceded the advent of high-throughput DNA-sequencing approaches. Indeed, there has been a rich history of research into human-associated microorganisms since at least the late nineteenth century. Escherichia coli was first isolated in 1885¹, bifidobacteria were described in 1899² and Metchnikoff speculated on the importance of beneficial gut microorganisms in the early 1900s³. Similarly, concepts such as the gut–brain axis have been researched for centuries⁴ and health impacts of key microbiome-associated metabolites, such as short-chain fatty acids, were first reported more than 40 years ago⁵.

Although Nobel laureate Joshua Lederberg had many notable achievements in his career, he did not invent the word ‘microbiome’. This oft-repeated claim has been thoroughly refuted elsewhere, with evidence provided that the word was used in its modern context more than a decade before Joshua Lederberg first used it in 2001⁶. Although relatively trivial, this myth serves as an example of how easy it is for falsehoods to become embedded in the human microbiome literature.

This figure is commonly mentioned in the microbiome literature, but its source has been difficult to ascertain. It may, however, have originated from dry-weight rather than wet-weight faecal cell counts. Regardless, it is incorrect. The real figure, as ascertained using various methods such as direct cell counts, fluorescence in situ hybridization, flow cytometry and quantitative polymerase chain reaction (qPCR), is typically between 10¹⁰ and 10¹¹microbial cells per wet-weight gram of faeces^7,8,9.

Although this is mentioned many times in the literature, it is often given without citation and we were unable to find an original source for this claim. Nonetheless, it is unlikely to be true in most cases. The majority of the human microbiota resides in the colon, and these microorganisms typically account for less than half of the weight of faecal solids¹⁰. The average human stool weighs less than 200 g (wet weight)¹¹, with total colonic contents ranging from 83 to 421 g in a small study of sudden-death victims¹². Therefore, aside perhaps from rare cases of severely constipated individuals with extremely compacted faecal matter in their colons, the total weight of the human microbiota is much more likely to be less than 500 g, and perhaps even considerably lower than this in some cases.

This myth is perhaps one of the most pervasive in the human microbiome literature and is one that we have also repeated uncritically in the past (we, sadly, are not immune to mythology). Excellent work¹⁰ has, however, shown that this myth seems to have originated from a ‘back of the envelope’ calculation in the 1970s. More detailed analyses indicate that the true figure, albeit still impressive, is probably closer to a ratio of 1:1. It should be noted that the ratio is likely to vary from person to person and is dependent on factors such as the host’s body size and the amount of faecal material they are carrying in their colon¹³. Current estimates are also largely based on observations from adult individuals living in urbanized high-income country settings. More comprehensive estimates will require study of individuals from lower income or rural settings, and also from across the life course.

Although variants of this statement are more often found in popular science articles than the scientific literature, it is an example of how nuance is extremely important when describing the human microbiome. Although some microorganisms are directly transferred from mother to baby during birth^14,15, proportionally few microbiota species are truly ‘heritable’ and persist through from birth to adulthood in the offspring^15,16. Indeed, most of the expansion in gut microbiota diversity occurs after birth, during the first few years of life, and increases most dramatically after weaning¹⁷ (Fig. 1). Every adult ends up with a unique microbiota configuration, even identical twins that are raised in the same household¹⁸. Therefore, although microbiota assembly is not yet fully understood, adult microbial communities seem to be predominantly shaped by prior stochastic environmental exposures, as well as multiple other factors such as diet, antibiotic therapy and host genetics, with direct ‘inheritance’ from the mother at birth playing a similarly lesser role.

Related to the previous section, this commonly used but erroneous claim stems primarily from rodent-based research, and from findings in single, or under-powered, human studies. However, as with many other studies that report links between specific microbiota profiles and disease, reproducibility is poor. Indeed, there have now been at least three meta-analyses reporting that this finding is inconsistent between human studies, and that there are, in fact, no reproducible microbial taxonomic signatures of obesity in humans^25,26,27.

This misconception also reflects an unhelpful tendency to examine sequence-based microbiota profiles at very broad taxonomic levels, such as phyla. Although this is appealing from a data simplification point of view, it fails to incorporate the huge and inherent variability within individual phyla. To draw a crude analogy, humans, birds, fish, reptiles and even sea squirts are all members of the phylum Chordata, yet clearly have very different physiologies, lifestyles and impacts on their environments.

Moreover, this claim was also based on relative-abundance-based DNA sequence data. Compositional data are still useful, and can correlate well with absolute quantifications obtained with techniques such as qPCR^28,29. However, some studies have suggested that relative-abundance-based correlations can lose significance when absolute microbial abundances are also factored into analyses⁹. Moving forwards, increased incorporation of absolute quantification data may help to make conclusions based on compositional analyses more robust.

This claim derives from studies showing that, whereas the taxonomic composition of human metagenomes can vary hugely, functional gene prediction profiles remain remarkably consistent. We contend that this is at least partly artefactual, as these functional comparisons are typically carried out after discarding the large proportion of metagenomic data that does not map to reference databases³⁰. Much of what does map to those databases is likely to be derived from common housekeeping and/or well-characterized genes, which are found across many different bacteria and are also relatively well represented in reference databases. These comparative analyses therefore fail to accurately capture specialist, or less well-characterized, functions. As such, the truth is more nuanced. Although there are important functionalities that are conserved across many different human microbiota species, such as short-chain fatty acid production²⁹, there are many key functions that are only carried out by a relatively small number of microbiota species. Examples include oxalate³¹ and resistant-starch³² degradation. In the absence of key species, functionalities such as these may not necessarily be fully replaced by other members of the microbiota.

Although sequence-based methods have been transformative for microbiome research, they are not perfect. Biases can be introduced at every step of sequence-based studies, from sample collection and storage, through laboratory-based steps such as DNA extraction, to choice of bioinformatic pipelines and reference databases used to analyse the data³³. Comparisons of sequence-based versus culture-based studies of the microbiota have shown that sequence-based approaches completely failed to detect some species that were only recovered using traditional culturing methods³⁴. Modern sequence-based approaches are hugely powerful but, like all techniques, they are not unbiased. For optimal interpretation of results, it is important to be aware of the inherent limitations of any given method.

This opinion is prevalent in the microbiome field and is sensibly grounded in a desire to make it easier and more robust to compare results from different studies. However, as outlined above, there are no methodologies that are perfect, and all are biased in some way. If everyone in the world is using the same method, then everyone is equally blind to the limitations of that particular approach. There is also the problem of deciding which protocol everyone should use. For example, comparisons of results from the Human Microbiome Project with the MetaHIT project showed stark differences in microbiome profiles and indicated that the Human Microbiome Project protocol was less effective at extracting DNA from eukaryotes and Gram-positive bacterial lineages³⁵. The truth is that the ‘best’ method fundamentally depends on the underlying structure of the microbial community in a given sample and this can vary hugely between individuals and between body sites. For these reasons we argue, as others have, that optimization and verification of sequence-based results with additional approaches are preferable to asking everyone to adopt the same method³⁶. An additional advantage of multi-faceted studies using different methods and research platforms is that they can enable more mechanistic understandings of associations between microorganisms and host phenotypes^37,38,39. Increased transparency when reporting methodology choices would be helpful for comparing results from different studies. The recently published STORMS (Strengthening the Organization and Reporting of Microbiome Studies) guidelines⁴⁰, for example, could greatly aid this process if adopted widely.

The adoption of high-throughput sequence-based technologies has also been mirrored by claims from some quarters that these methods must be used because most human-associated microorganisms cannot be cultivated in the laboratory. In fact, a reasonably large proportion of the bacterial and archaeal component of our microbiota has already been cultured⁴¹ (viruses and fungi remain under-represented), with pioneering work from as early as the 1970s establishing the cultivability of a broad diversity of species from the human gut microbiota⁴². Many more species continue to be cultured as laboratory-based efforts have increased throughput^43,44. This implies that current gaps in culture collections are at least in part attributable to a lack of previous effort rather than an inherent ‘unculturability’. Although cultivation is undeniably labour intensive, has its own biases and often requires expensive specialist equipment and media, there are clear advantages to having microorganisms in culture. These include enabling mechanistic experiments, verifying genomic predictions, and developing them as novel therapeutics⁴⁵. Given the importance of continued cultivation-based work for the progression of microbiome research, it is gratifying that this myth has become less prevalent in recent years following the publication of the aforementioned high-impact studies that demonstrate it to be false. However, it serves as an excellent example of how previously widely accepted dogma is sometimes simply not true. There are important lessons for many other myths and misconceptions that have yet to become as widely rejected.

The microbiome field is broad, and there are many other controversial topics that might also have been included here. However, knowledge is still evolving on many of these; consequently, we have largely focused on concepts where we believe there is a strong evidence base for rejecting myths and misconceptions. Although some of the points above may seem trivial, we argue that the accuracy of details such as these matters. If we are consistently repeating falsehoods about minor details, can our accuracy be relied upon when covering more important matters? We hope that, by illustrating just a few examples of microbiome myths and misconceptions, we can draw increased attention to the potential problems of over-simplification and insufficient critical assessment in the microbiome literature.

Given the many potential health impacts, the huge amount of funding and the keen public interest in microbiomes, rejection of unfounded assertions is crucial if we wish to avoid expending finite resources researching unproductive avenues and undermining public confidence in our conclusions.