The ABCs of SFBs
As a fairly recent arrival in the field of gut microbiology, I am still constantly surprised by discoveries that, although not new to the field, are new to me. Among these are the important role played by the segmentous filamentous bacteria in controlling gut immunology. One of the things I like about this group is the very descriptive nature of their name – there certainly won’t be any surprises when one sees a picture of them (below). Although unsurprising in their appearance, they are revealing a whole host of novel functions and in the process serving as a reminder of how little we know when it comes to the interactions between mammalian hosts like us and our unseen microbial partners.
Discovery of segmented filamentous bacteria
The presence of these morphologically unique components of the gut microbiota have been noted for over 100 years but very little was known about what, if any, effect they played in the actual biology of the host. Like many things in biology those days, the ability to answer these kinds of questions were limited by the technology of the time and progress had to wait until our current age of next-generation sequencing was developed.
SFB attached to the ileum of a mouse (copyright Elsevier Inc)
Role in the innate immune system
Although unculturable, several groups established protocols 20 years ago in which germ-free animals were stably monoassociated with SFBs. The work of these early groups demonstrated that SFBs affect the innate immune response and development of germinal centers in Peyer’s patches as well as the production of IgA in the intestinal mucosa. In early 2008, Ivaylo Ivanov in Dan Littman’s group published a paper that clearly showed a difference in TH17 responses and resistance to bacterial gut pathogens between genetically identical mice from different suppliers. This difference was due to the presence of SFBs in the gut microbiota of C57BL6 mice from Taconic labs that were absent in mice from Jackson Labs. This difference in susceptibility to pathogens in genetically identical mice sourced from different suppliers had been noted by other groups and was linked to the microbiota, but the reason for the difference wasn’t appreciated until Ivanov and coworkers published this manuscript.
The ability to link the innate immune response to particular, specific components of the microbiota required novel methods that simply weren’t available earlier and the ability to conduct these experiments required significant amounts of both technical know-how and sheer luck to uncover. I won’t go into the difference between skill and luck here, but perhaps another day I’ll muse a little on the relative importance of each of these to the results. The TH17 and pathogen resistance phenotypes were transferrable between suppliers by providing the Taconic microflora to the Jackson lab mice and the authors could demonstrate clearly that the presence of SFBs in the small intestine correlated very well with the production of a host signaling compound called serum amyloid A. Providing this molecule in tissue culture could restore some of the functions seen in vivo in response to the SFBs themselves, suggesting that the ability of the SFB to induce this molecule was, at least in part, responsible for this phenotype.
Genome sequencing
Although it was clear that these bacteria had a profound effect on the host immune response, the bacteria themselves were (and still are) unculturable. The development of next-generation sequencing provided a means to get at genomic data without actually having to establish in vitro culture systems. In a recent issue of Cell Host and Microbe, two papers describe the genome sequences of the mouse and rat SFB species. The paper from the Ivanov group describes sequencing 1.57 megabases (Mb) of the murine SFB in five contigs using Roche 454 sequencing. Using a combination of 454 pyrosequencing combined with Sanger sequencing for gap closure, the Hattori group was able to completely close the genomes of the rat (1.52 Mb) and mouse (1.59 Mb) SFB strains. These analyses have uncovered a great deal about these organisms, but in many ways, they also illustrate just how little we know about the biology of these host-adapted endosymbionts.
As predicted based on 16S ribosomal sequencing, SFB are most closely related to Clostridial species (which may be familiar as the causative agents of diptheria, tetanus, and C. difficile-associated gastroenteritis). Although most closely related to these organisms, pathway analysis suggests that SFBs form a phylogenetically and functionally distinct clade. However, the genomes of both strains showed evidence for reductive evolution in which large number of genes associated with metabolic plasticity have been lost. In fact, analysis showed that the SFB are functionally (note the distinction from phylogenetically) related to several endosymbiotic genera with minimized genomes, including Mycoplasma, Borrelia, and Ureaplasma. SFB appear to be auxotrophic for most amino acids (although pathways for amino acid interconversion are maintained) and genes required for nucleotide and cofactor biosynthesis pathways are also underrepresented in the genome. Overrepresented pathways included membrane transporter proteins and regulatory proteins suggesting that the bacterium obtains most of these metabolic components from the nutritionally rich ileal lumenal enviroment.
Somewhat surprisingly (at least to me) is the high number of regulatory proteins encoded in the genome. In most endosymbiotic species, genome reduction (and presumably, niche stability) serves to reduce the requirement for the bacterium to respond to changing environmental pressures. For example, in Mycoplasma, very few regulatory genes appear to be encoded and this type of metabolic reduction has also been observed in insect endosymbionts like Buchnera. In contrast to these examples, the ~4% of the genome devoted to regulatory proteins seems to match quite well with patterns observed correlating genome size with regulatory complexity observed previously. If this correlation between endosymbiosis and loss of regulatory control is not maintained in SFBs it suggests that the bacteria may respond to some component of the host intestinal environment, perhaps related to the diet of the host or to the host immunological state. This further suggests that the ability to modulate the activity of SFB using diet or small-molecule approaches may be a fruitful avenue for further research.
Metabolically, the bacteria lack a TCA cycle and appear to derive most of their energy by glycolysis and pentose phosphate utilization. Of particular interest given the tight association of these bacteria to the ileal epithelium (with its associated mucosal layer) is the presence of large numbers of genes encoding proteins for the degradation, importation and utilization of host-derived carbohydrates. The bacteria possess a complete sialic acid utilization genes, as well as those for the utilization of cellobiose, mannose and N-acetylglucosamine – also components of host mucin. They also have transporters and degradatative ability for other simple carbohydrates like ribose, fructose and ascorbate – compounds that are readily available in the intestinal milieu. Several adhesin-like proteins were found in the genome, although the contribution of these ORFs to the tight epithelial association will require the development of genetically pliable systems.
What stimulates TH17 responses?
The most important phenotype linked to the presence or absence of SFB is the production of TH17-associated innate immune responses. However, very little is known about which gene products may be responsible for this activity. The genome encodes several proteins that appear to be cell-surface associated, although the means of surface attachment appears to be unique. Interestingly, since electron microscopic observation has never demonstrated the presence of flagella in ileal-associated SFB, the genomes contain multiple flagellin gene clusters. These flagellin genes appear to contain motifs that are responsible for TLR5 interactions – motifs that are missing from the flagellin protein of other commensal clostridal flagella. As TLR5 responses have been implicated in the control of IL-17 and IL-22, this is a very intriguing observation that will require further investigation.
What does this mean for human health?
Despite the long history of SFB observation and the robust data supporting a role for SFB in host immunocompetance, the manuscripts reached different conclusions regarding a role for SFB in human health. In the Ivanov manuscript, the authors conclude that SFB are not present in humans. Using what appears to be (at least to this non-bioinformaticist) a robust analysis, the authors failed to find any 16S rRNA sequences in the MetaHIT human intestinal metagenomic database nor did they find any other region of the murine SFB genome in this dataset. Absence of evidence is not evidence of absence however, although the authors speculate that other human-adapted commensal species may fill the immuno-ecological niche that is occupied by SFBs in rodents.
The Hattori group used a seemingly simpler analysis to search for homologous sequences to the 16S sequence of mouse and rat SFB and found several, one from gorilla feces, one from monkeys, one from a human skin sample, and one from dog intestine. These sequences and the rat/mouse-associated genes clustered separately from the commensal Clostridial species, suggesting that they can be found in other non-rodent mammalian species, although their role in human biology remains an open question. The reasons for the discrepancy is unclear, but since genetically identical C57BL6 mice from Taconic and Jackson have different SFB status, it seems to demonstrate that in some instances at least, these components of the microbiota can be gained or lost. The relative frequency of these in other mouse strains or in humans is an open question, and one that may have a profound influence on the study of gut pathogens in specific models of pathogenesis.
The stuff nightmares are made of…
When the story first broke back in 2009 that a researcher at the University of Chicago had died due to a laboratory-acquired plague infection, my supervisor at the time was visibly shaken. The entire crux of our ability to work on the pathogen rested on the notion that a well-defined mutation rendered it avirulent. The fully-virulent plague pathogen (the thing most people think of and the species responsible for wiping out a third of Europe in the 1300s) is simply too dangerous to work with on an open lab bench and although people do great work on the virulent strain, the expense and difficulty of carrying it out means that fewer researchers could or would undertake it. In addition, the Centers for Disease Control (or Health Canada here at home), strictly regulate research on non-attenuated level 3 pathogens. That a healthy adult could both contract and ultimately succumb to the very bacterium he was researching called into question the nature of the regulatory process as well as the presumptions many people (myself included) were working under. Recently however, the cause of the researcher’s death was reported in the New England Journal of Medicine.
Like virtually all living organisms, the plague bacterium (Yersinia pestis, for those of you keeping track of such things) strictly requires iron in order to live. Inside our bodies, most of the iron we need is actively bound up to specific proteins in our blood that keep it in a form that isn’t readily usable by most bacteria. This does a reasonable job of keeping most bacteria at bay. Pathogens though, have evolved countermeasures to stay a step ahead of us – they secrete molecules called siderophores that bind to iron very tightly and then specifically bring that iron back to the bacteria. In the attenuated strains of Y. pestis, the loss of virulence is because one of these iron-transport molecules is missing from the genome. This means that the bacteria can’t acquire enough iron inside a human to cause disease.
Y. pestis (green) infecting immune cells (Grabenstein et al, Infection and Immunity, 2006 Jul;74(7):3727-41)
At least, that’s what we assumed – since the CDC (and anyone with a conscience) frowns heavily on injecting plague into humans to test this, the data from animal models of disease served as a proxy for the exemption to category 3 classification. The fact that a researcher had acquired this infection in the lab while using a supposedly attenuated strain led to a lot of questioning about whether our presumptions were correct and whether they should be reconsidered.
What followed would make a great episode of House (although I suspect that no doctors had to break into the researcher’s lab/apartment to steal samples). An obvious place to look was at the bacterium itself – did the researcher manage to get the wrong strain somehow – thereby actually becoming infected with wild-type plague? Genome analysis showed that this wasn’t the case and he was, in fact, working with the well-known attenuated D27 strain. Samples collected at autopsy showed that the researcher had very high levels of iron deposited in his liver. This is a feature of a disorder known as hereditary hemochromatosis – which is essentially an iron overload disorder in which the body can’t process its iron effectively, resulting in higher levels of the metal in the body. Thus – the attenuated strains’ Achilles Heel was undone by a genetic Trojan Horse. Genetic analysis showed that he possessed a known mutation in a gene associated with the disease – a most unusual confluence of events, with a tragic outcome.
It isn’t entirely clear how the researcher was exposed to the disease, but it serves to reinforce what we’re told during orientation and annual refresher courses on biological safety. Gloves on, lab coat on, frequent hand washing, all the usual suspects because as with many things – sometimes what we don’t know really can hurt us.
The difficulty of defining disease
After popping into the ER for yet another bout of incredibly debilitating (but apparently, benign) vertigo, I got to thinking about how it is that the medical community diagnoses disease. It’s something that seems to be a complicated interplay of symptomology, statistics and more than a little intuition. Really, when you get right down to it – there are many paths that can usually lead to a given current state. Imagine driving up on a single-vehicle car accident – clearly, something went terribly wrong, but what? Was the driver drunk? Did the brakes fail? Did he fall asleep? Did a deer jump in front of him? We have ways to answer this question, necessarily conducted by a bevy of CSI-type hotties using high-tech techniques that absolutely no municipal labs could ever need/afford (I digress here). When it comes to infectious disease, we have Koch’s postulates.
Robert Koch formulated a series of rules to determine whether a particular organism causes a specific disease
Robert Koch was a German physician and microbe-hunter working in the second half of the 19th century. He is one of the founders of bacteriology (along with Louis Pasteur) and his studies on the transmission of anthrax and tuberculosis are what gave us the set of rules that we typically use to demonstrate that a particular organism causes a particular disease. These are summarized as follows:
1. The disease causing microbe is only found in diseased subjects and is not found in healthy ones
2. The disease causing microbe must be grown in pure culture
3. The cultured microbe must cause disease when introduced to a healthy subject
4. The infectious agent should be reisolated from the infected and diseased subject and shown to be identical to the original causative agent.
This has laid the framework for much of the last hundred years of infectious disease research, which has been devoted to the paradigm of ”one bacterial species = one disease”. This model of the world has served us well, leading to the development of most of our current understanding of bacterial disease but it is an imperfect model because it does not accurately reflect the fact that host-derived factors are also important for the determination of disease status.
We have had a strong notion for many years that a healthy gut requires a diverse set of commensal gut microbes and that in some individuals, the lack of this normal diversity is associated with disease states, but this has never been terribly well-described, owing both to the complexity of the communities that we are trying to manipulate as well as the difficulty of developing good animal models of gut disease. Indeed, a large chunk of the probiotics market is devoted to restoring this healthy balance through the application of live bacterial cultures to the diseased gut. A recent publication, however, has done a remarkable job of dealing with both of these issues, and as such has successfully applied Koch’s postulates to polymicrobial inflammatory bowel disease.
The review
The paper from the lab of Thaddeus Stappenbeck at the Washington University School of Medicine with Seth M. Bloom as the lead author, utilized a genetically modified mouse strain that lacks the receptor for the anti-inflammatory cytokine IL-10 as well as encoding a dominant-negative transgene for another anti-inflammatory molecule, TGF-beta. These mice, called dnKO mice, are especially susceptible to inflammatory colitis but importantly, treatment of the mice with an antibiotic cocktail can cure the mice of this disease. When the antibiotic cured mice are infected with bacteria from the colon of sick mice, those mice go on to develop disease, demonstrating that there may be a specific infectious agent in the mice with disease.
The authors then went on a bit of a sleuthing expedition, using selective growth media to isolate different types of the gut microbiota from the diseased animals. Utilizing media to select for both Gram-positive or Gram-negative anaerobic bacteria (the type that typically colonize the large intestine) the authors use these selectively enriched mixtures to infect non-diseased animals and show that the Gram-negative enriched cultures contain the infectious agent. Using further molecular techniques to identify six individual species of Bacteroidetes (an abundant group of anaerobic Gram-negative bacteria) any one of which could be the infectious agent. Using antibiotic cured dnKO mice, the authors were able to show that two of these species, Bacteroides vulgatus and B. thetoiotaomicron, were capable of inducing the symptoms of gut disease in these mice – therefore fulfilling all four of Koch’s postulates in the context of this model.
Bacteroides can cause gut inflammation in susceptible hosts
As with all well-conducted research, the study raises as many new questions as it answers. It has been well-known for quite some time that when the gut is inflamed, there is a huge increase in the number of Enterobacteraciae (think E. coli and related species) in the gut. This relationship has been used to infer a causative relationship between the Enterobacteraciae themselves and the inflammation – a reasonable notion, but not necessarily the correct one. In the Bloom paper, the authors also observe this inflammation-associated increase in Enterobacteraciae, but they use their model to test whether or not the Enterobacteraciae are capable of inducing disease and they found that they were not – suggesting that the high levels of Enterobacteraciae found in IBD patients may be a consequence of the inflammatory process rather than a cause of it.
Why should I care about any of this?
This will be considered a seminal paper moving forward both because it provides a framework for uncovering novel IBD-associated infectious agents but also because it describes a very robust model of disease that does not depend on raising germ-free or specific pathogen-free mice to study IBD (another useful, but expensive, model). I suspect that we will see a number of papers using similar approaches moving forward in different animal models of IBD and that this will greatly increase our understanding of the processes affecting this disorder.
Like anything in science (or in life for that matter), definitions are important. We usually have a pretty clear and mutual understanding of what a particular word or phrase means. Sometimes though, misunderstandings arise. In the case of IBD, it’s pretty clear what the phrase means; it’s right there in the name – inflammatory bowel disease – an inflammation of the bowel. However, that doesn’t tell us anything about what causes that disease. In this paper, the authors take a large and important step forward in defining an actual cause for it in a specific type of model – it’s future applicability to human disease (we’re not trying to help mice here, after all) will likewise be very important.
Bacteriology 2.0
Microbiology, like many scientific disciplines, is something that is not always well understood by the public. This is not a large problem when things are chugging along nicely and we’re all happily disease-free. The problem, however, lies where that ignorance extends to the corridors of power and to those who are meant to represent our best interests. The first few days of the recent E. coli outbreak in Europe is an excellent case in point. In early May, surveillance programs throughout the EU started showing a higher rate of atypical EHEC infections, particularly in the area surrounding Hamburg, Germany – atypical both for the disease itself and for the types of people it was making sick. In contrast to previous outbreaks in Walkerton, Ontario or Sakai, Japan, where mostly children were sickened, this outbreak seems to strike down healthy adults, particularly women. With several hundred cases by late May, the finger-pointing began. Initial media reports of epidemiological screening homed in on one or all of spanish cucumbers, lettuce or tomatoes as a likely cause of the outbreak which was quickly followed by a German group reporting that E. coli had been found on cucumbers.
Spanish cucumber were mistakenly blamed for the European outbreak of O104:H4 E. coli (photo credit: Kittikun Atsawintarangkul)
The response of the market was swift and merciless; economic damage to Spain’s vegetable industry has not yet been fully calculated, but is estimated at greater than 200 million euros per week. The EC has requested compensation of 210 million euros and this number will almost certainly rise as the full scale of the damage comes to light. That people would shun products that are known to be unsafe is hardly surprising – but there’s the rub. The science certainly wasn’t anywhere near where it needed to be in order to make such irresponsible statements. The German pronouncements of the source of the seemed to be based more on political expediency than it was on actual science. Considering the geographic clustering of the nationalities of the victims, a local source in the Hamburg region was more probable than a foreign source of infection.
Bean sprouts from a German farm were positively identified as a source of the novel O104:H4 outbreak (photo credit: hinnamsaisuy)
Time and science have borne this out with the identification today of bean sprouts from a German farm as the source of the O104:H4 pathogen. Fortunately, we’re getting close to a point where science will actually be able to function as quickly as politics can. Scientists in Shenzhen, China in collaboration with German scientists sequenced the genome of the O104:H4 outbreak strain and released their data to scientists around the world. Annotation began immediately and the result was somewhat surprising. Rather than a typical EHEC strain as might have been suspected, the culprit turned out to be an enteroaggregative E. coli (EAEC) strain that had picked up Shiga toxin somewhere in it’s evolutionary history. This information was then used to develop a high specificity test for the strain in question that will be used going forward to positively identify suspected new cases of disease. With over 3000 confirmed cases of disease and more than 31 deaths, this information will prove critical moving forward. The identification of the strain, the generation of the genome sequence and the ability to formulate hypotheses about its exceptional virulence within one month of the beginning of the outbreak marks the first time the crowd-sourced approach has been brought to bear on bacterial infectious diseases.
The cooperation demonstrated by the scientific community in this case is certainly not new, but it should stand as a model for how our politicians could behave. In an outbreak situation health trumps economics and the public will always scramble for answers, but it is surely better to be cautious and correct than rapid and wrong. In this case, as in so many others, the wisdom of the crowd is more than just an idiom.
