The Jeremiah Metzger Lecture: Global Warming Redux: The Disappearing Microbiota and Epidemic Obesity

INTRODUCTION

In 2004, I presented my lecture “Global Warming of the Stomach: Microecology Follows Macroecology” at the 117th American Clinical and Climatological Association meeting (1). In that lecture, I highlighted how the disappearance of Helicobacter pylori, the ancient, dominant, highly interactive bacteria that colonize the human stomach is affecting human health. Now, 7 years later, I return to that theme, and describe the relationship of our changing microbiota and obesity.

Animals have had residential microbes, probably since the days of our earliest ancestors (2). Of the residential microbial populations colonizing humans, we can consider them ancient, niche-specific, and often persisting for years or decades. Collectively, these organisms represent the human microbiome (3). Certain microbial taxa are conserved among most or all of us, whereas others may be relatively host-specific. Each particular niche in the human body has its own major bacterial populations; there is considerable intra-individual variation by site and inter-individual conservation at higher taxonomic levels (e.g., phylum) (4–6). A number of biological questions (7) are important to consider: What is the identity of these microbes? What are they doing? How is the host responding to them? What are the forces that maintain equilibrium among the populations? What are the unique characteristics of each individual?

We have developed a model of the interactions between co-evolved colonizing microbes and their host (7–11) (Figure 1). In this view, individual organisms (Figure 2A) and networks of organisms (Figure 2B) form equilibriums that may oscillate but generally remain stable for long periods within a defined range. The question we pose is what happens when these organisms are perturbed, and in the extreme, when they become extirpated (or extinct) in an individual host (Figure 2C)?

Fig. 1.

Who are we? An estimated 90% of the cells in the human body are microbial, with the human cells represented in red. An estimated 99% of the unique genes in the human body are microbial, with the human genes represented by the ant (at right). Together, the human genome and our microbiome encode a metagenome that provides metabolic functions for the host. We consider this as an integrated metabolome. Based on concepts from Blaser (2010) (7), Blaser and Kirschner (10), Plottel and Blaser (11), Blaser (2006) (18), Blaser and Falkow (19).

Fig. 2.

Schematic of interaction between co-evolved colonizing microbes and their host. In the model, the microbes and the host cross-signal and maintain an equilibrium state (A). Multiple organisms are doing this simultaneously in an integrated process; the colonization is both robust and resilient (B), but we are concerned about the effects of repeated perturbations. In C, the disappearance of a species has changed the equilibrium and the cross-talk between the microbial populations and host cells. Adapted from Plottel and Blaser (11).

H. PYLORI AS A PARADIGM

H. pylori, curved, gram negative bacteria, are ancient in humans, and colonize the gastric mucus layer (12, 13). These organisms have adapted to human biology with an intimate attachment to host cells and the injection of its own constituents (e.g., the cagA-encoded protein) into host cells, where they affect signal transduction in a variety of ways (12–15). Through interactions regulating cell cyclins, they affect cell cycle, and can alternatively signal through the SHP2/ERK and JAK/STAT pathways, which are cross-inhibitory (14, 15). In total, the interaction of H. pylori and humans is most consistent with an integrated circuit, affecting the metabolism of the bacteria and host cells and tissues (Figure 1). We can generalize from our relationship with this ancient, conserved, highly host-interactive microbe to suggest that the microbiome as a whole, in conjunction with host genes, represents a metabolome.

However, as discussed previously (1), the prevalence of H. pylori in the United States and other developed countries decreased dramatically during the 20^th century; from near ubiquity to less than 10% carriage among children in the early 21^st century (16, 17).

THE “DISAPPEARING MICROBIOTA” HYPOTHESIS

Based on our understanding of H. pylori, we have proposed the “disappearing microbiota” hypothesis (18–20) that has three tenets:

1. With changing human ecology beginning in the late 19^th century, there have been dramatic changes affecting the transmission and maintenance of the indigenous microbiota.

2. These changed circumstances have affected the composition of our indigenous microbiota.

3. The changed composition affects human physiology, and thus disease risk.

In brief, the disappearance of H. pylori fulfills each of these conditions (reviewed in 19). If it is true for H. pylori, might it not also be true for other ancient indigenous microbes (18–20). And if so, which particular consequences are of importance? We only have to consider the diseases that have increased rapidly in the past 50 years to develop a list of candidates. These include obesity, asthma, and allergic conditions, type 1 diabetes, and autism. All of these have origins in childhood, when the effects of losing our ancient constituents may have the largest effects on development. We have largely focused on one problem, the risk of childhood-origin obesity.

LESSONS FROM AGRICULTURE

Antibiotics came into widespread use after the large-scale industrial development of penicillin in the 1940s. Within several years, tetracyclines, macrolides, chloramphenicol, and streptomycin, among other antibiotics, came into medical practice. Their use was miraculous in many cases with cures of acute, formerly fatal, bacterial infections. Since that time, there have been multiple uses and indications for antibiotics, including many newly discovered classes.

Beginning in the late 1940s, farmers made a striking observation: adding antibiotics in low doses to the food of farm animals promoted their growth (21, 22)! Such use of sub-therapeutic antibiotic treatments (which we call STAT) became a widespread practice because it was a very profitable form of growth promotion (23, 24). Although its mechanism of action was not well-studied, several observations became clear:

1. The use of the agents promotes absolute growth and also improved feed efficiency — the ability to convert food calories into body mass.

2. The effects are seen in many different animal species, ranging from chickens to swine and cattle. This is a wide swath of vertebrate evolution.

3. Many different types of antibiotics, varying in their chemical class, mode of action, and spectrum of activity, are effective.

4. Antibacterials (including chemical agents [e.g., the ionophore, monensin]) are effective, but not antifungals or antivirals.

5. The younger in life that the agents are started, the more profound the effects on both growth and feed efficiency.

The utility and profitability of this approach was so great, that an estimated 70% of all antibiotics produced in the United States were used for growth promotion of farm animals (24)! Because of issues involving the spread of antibiotic resistance in bacteria from farm animals to humans, such use by food producers has come under sustained attack (25). Antibiotic use for growth promotion was banned in much of Europe (26), but continues in the United States, although there are increasing limits (27). This is a separate and important issue, but perhaps the most interesting question is: Why does it work? And a second, immensely important question is: If exposure of young animals to antibiotics is changing their metabolism and development, what are the effects of all the antibiotics that we are purposively giving our children?

THE MODERN PRACTICE OF PEDIATRICS

Although we are not giving STAT to our children, they are highly exposed to antibiotics. Because of their utility treating bacterial causes of otitis media, pharyngitis, and bronchitis, up to 80% of young children with such conditions leave the doctor's office with a prescription for antibiotics (28). We now know that this represents enormous over-usage because most acute infections are of viral, not bacterial, etiology, and there has been substantial progress in recent years to curtail such usage (29), but the cumulative exposure is great. Based on surveys in the United States and in Europe, it may be estimated that the average child receives 10 to 20 courses of antibiotics before the age of 18 years; much of it during the first5 years of life, when their development is most rapid (30, 31).

What are the effects of such usage? I contend that we simply do not know. When antibiotics were first introduced, their effects were often so dramatic that they were considered miraculous. This still is often the case! Aside from the very rare side effects of allergies, such as the idiosyncratic reactions to chloramphenicol, no serious long-term effects were observed. Short-term side effects are common; including skin rashes, thrush, and gastrointestinal upset, but when the offending drug is stopped, the problem usually clears up shortly. The widespread notion among physicians and the lay public that there were no long-term effects of antibiotics came from such observations. The idea that antibiotic use could permanently change the composition of the microbiota was simply not in the conventional wisdom. It was not even discussed.

Yet, it is far from impossible. We know that the microbiota consists of hundreds, if not thousands of species in each host, and that within these, there is enormous strain variation (6, 32, 33), even early in life (34). We know that antibiotic use selects for resistance, and we now know that such resistant organisms can persist for years in the absence of any further antibiotic exposure (35, 36). We have long known that antibiotic use changes the composition of the microbiota (37), and now with more recent tools, we can observe large shifts in microbial composition after use (38–40). Although the concept that antibiotics could cause oscillations in microbial abundances was widespread among microbiologists and physicians, the notion that use may lead to extinctions was not in the conventional wisdom.

I have proposed that early life use of antibiotics in human children leads to changes in microbial composition, including extinctions, at a critical time in development — a time when immune and metabolic capabilities are being established. I have hypothesized that, as a result of such changes, early life development has been altered, affecting adult height, weight, and susceptibility to allergic disorders, including asthma, allergies, and atopy (18–20, 41–44).

ANIMAL MODELS OF EARLY LIFE EXPOSURES TO ANTIBIOTICS

These are provocative, but fundamentally important questions about the future of human health. Much work must be done to examine their veracity. First, epidemiologic studies are needed to address whether markers of microbial disappearances or of antibiotic use per se can be linked to disease occurrence. Such studies are underway, and there already are some linkages between antibiotic use and the development of obesity (45). A growing body of evidence is linking the loss of H. pylori to the development of asthma and related disorders in both epidemiologic studies (reviewed in 46), and in animal models involving immune cell function (47–49).

To better understand the role of antibiotics in childhood-onset obesity, animal models are needed. We have begun to develop two models. The first is called the STAT model, which mimics the use of antibiotics on the farm. In that model, mice are exposed to STAT doses in their drinking water beginning early in life. In an alternative version of the STAT model, exposure begins during the late days of pregnancy, so that the microbiota transferred inter-generationally (vertically) after birth already has been altered, and the antibiotics that the pups receive in their mother's milk and afterwards continue their selective actions.

The second model is called PAT (pulsed antibiotic treatment), and is intended to mimic the exposures that human children receive early in life. In that model, the antibiotic doses are calculated to achieve systemic levels comparable to those that children receive when they are receiving treatment for their ear or upper respiratory infections. The mice are exposed to 3- to 5-day pulses followed by no antibiotics. In both models, we focus on beta-lactam antibiotics, the type most commonly used in childhood, and we also study macrolides, which are widely used. The work is in progress, but preliminary data indicates that the antibiotic exposures are changing microbiota compositions at critical times in mouse development, and that there are metabolic, hormonal, immunologic, and morphometric changes in the mice. The work is preliminary, but we are providing experimental support for the notion that early life antibiotic usage is changing development in their recipients.

CONCLUSIONS

Continuing with the theme begun in my ACCA presentation in 2004, I am concerned about the consequences for human health because of the antibiotic-induced changes in our ancient, conserved, interactive microbiome and its inherent metagenome. It is possible that these effects are cumulative across generations, reflecting the microbiota that girls pass on to their children when they become mothers (50, 51). In any event, we must study these issues in greater detail. Our preliminary studies have been launched, and with favorable winds, we soon should be able to confirm or refute the hypotheses raised.

DISCUSSION

Nathan, Boston: When a hematologist rises to ask questions about metabolism and microbiota, my advice to the audience is run, do not walk, to the exits. This fascinated me, Marty, and here's my question. I absolutely see the relationship between the antibiotic use and growth, and that there is a clear shift in the organisms, but the question is, why do they get fat and do they get fat because they are not utilizing fat. In other words, is the metabolic wheel changing so they utilize less fat and therefore gain fat and the caloric expenditure is actually the same or it's just that fat utilization is down and protein utilization and glucose utilization is up. So, I'd like to ask about caloric expenditure in these mice. What really happens to them? My second question is: antibiotics could be playing a role on the cells themselves quite independent of their effects on the organism, so, can you switch the organism load by loading with organisms and changing how these mice behave without the antibiotics? Forgive a hematologist for a stupid question.

Blaser, New York: Excellent questions and in time-honored tradition, I will answer the second question first. We are in the middle of conducting experiments, in which we are transferring microbiome from control and STAT mice to germ-free mice, and I can tell you that the preliminary results are promising. So, that's kind of a Koch's postulates for transfer. We presume it's the microbiome because going back to the old literature on the farm, this is happening with so many different classes of antibiotics. We don't think it's any kind of side effect or specificity. Now, David's second question is what are the mechanisms and particularly about caloric balance. As a part of our ongoing work, we are putting mice in metabolic cages so that we can answer some of the questions that you suggest. However, we thought that we were going to get growth phenotypes with mice because farmers get it with all these other vertebrates, so, it's almost a no-brainer. It took a little while to work it out, probably more than it should have, but giving antibiotics is changing early mouse development and we are working to understand whether it is metabolic, whether it is immunologic, whether it is behavioral, et cetera, and we have a number of clues that we are pursuing.

Glass, Bethesda: That was lovely. In the 1970s, a group of infectious disease doctors looked at that animal data that you presented at the beginning and suggested for malnutrition in developing countries, children should be put on low-dose antibiotics as a trial and I don't think the trial was ever done so that would be an interesting follow-up. My real question is whether you played with antibiotics in another context. Part of the obesity epidemic has been linked to sugary foods, McDonald's or whatever, and I wondered if in your models you've tried to feed these mice Coca-Cola or Big Macs?

Blaser, New York: So, that was in one of the slides I took out due to limitations of time. If we have a chance to put my slides back on, toward the end, we can get back to that point. What we've done is an experiment in which after the STAT experiment, mice are on long-term antibiotics and then at a certain point, half get a high fat diet and half remain on a normal diet, and if I can show you the slide, the results are pretty dramatic. There is synergy: so antibiotics have an effect, high fat has an effect, you put them together, and it's greater. [Hold on. Let me see if I can fast forward to that particular slide. There it is.] So, here is where we put males on high fat and here is the total mass in the mouse; here is high fat along with antibiotics control (normal) diet and antibiotics. We can see that this is the differential effect of the antibiotics on fat levels. Here it is in females. It's really quite dramatic.

Palmer, New York: This is a fascinating subject. It has lots of ramifications. I've been particularly interested in the work that's been emerging that I'm sure you're familiar with about the ability of changes of the microbial flora to influence behavior, particularly anxiety and transplanting through fecal inoculation, the anxiety factors. I am wondering if it would make sense then to think that if antibiotics have this kind of effect on the bacterial flora that influence obesity, is it also possible to imagine that antibacterial use may effect psychological traits, such as anxiety?

Blaser, New York: As Dr. Oates pointed out, part of my training was as an epidemiologist, and I am very interested in the diseases that seem to be rising rapidly that have early life origins and this “disappearing microbiota” hypothesis is not specific to obesity. It's about all disease consequences as a result of changing our ancestral microbiota that we've had for hundreds of thousands if not millions of years, over a very short period of time. We are interested in all of those rapidly changing diseases, like obesity, childhood onset asthma, type 1 diabetes, inflammatory bowel disease, celiac disease, and autism. All of this remains to be tested.

Martin, New Orleans: So, that was a fabulous talk, Marty, and it's my misfortune to have to follow you up but you've made my task a little bit easier by explaining some of the more complex issues and technology. We in infectious diseases have been concerned about the issue of antibiotics in cattle feed and animal feed for a long time, and it would sound like, from your data, that the primary effect is to accumulate fat in the animals, and so I wonder if this might be a weapon that we could include in our arsenal to point out, to make it clear that the nutritional value of weight gain in these animals is really not worthwhile. In that context, do you know of anyone who actually is doing microbiome or microbiota work in cattle to look at the parallels between what's happening there and what you're showing in your mouse model?

Blaser, New York: Thank you, David. What Dr. Martin points out is the fact that antibiotic use in livestock is so widespread and may have a number of pernicious effects. In Europe, in fact, its use has been banned for growth promotion for almost 40 years now, and meat is a little more expensive there, but it is antibiotic-free. I want to emphasize that our work is quite preliminary. We have preliminary data in a number of different directions, and I would not say that the phenotypes that we are observing are fixed because under slightly different experimental conditions, we are seeing variation. However, we are consistently seeing differences in bone or lean or fat, or combinations of those, so we have consistent evidence that we are changing early life development. What are the consequences of eating antibiotics in foods of animal origin? I am not certain. The FDA requires that there is a washout period so the antibiotics don't get into human food. Whether or not that actually occurs, I am not certain, but we are interested in farm use of antibiotics as an analogy, as a metaphor. If feeding antibiotics early in life changes development of farm animals, what are we doing to our children?

Abboud, Iowa City: This is a follow-up on David Nathan's question. So, the concept that obesity is a result of an imbalance between energy intake and energy metabolism or utilization is being challenged by what you are presenting to us. You suggested that there is lipogenesis or a turnover of cells or a proliferation of cells. Is there any evidence that the microbiome alters the rate of cell generation, proliferation, and apoptosis, particular in their size, no matter where they are?

Blaser, New York: Yes. It's a great question. We don't know the answer to that yet. Our goal is to establish the model and put it out there so that people who have skillsets with regard to these questions can explore that. We don't know yet, but it's a wonderful question.

Thorner, Charlottesville: That was a beautiful presentation, very provocative. Looking at this particular slide that you've got up here, it strikes me that if on a high fat diet, then the change in the microbiome has a fatty modest effect. Would you like to comment on that, and is really the diet as important as the change in the microbiome?

Blaser, New York: What we find is that the effects are not completely consistent between males and females and from experiment to experiment, although they all are pointing in the same direction. In this case, and I point out that this is total mass in grams, we can see, that there actually is an effect here in the males although we are not seeing much of an effect in the females. Here we are looking at body fat percentage, and in this case, the antibiotic didn't have an effect in the absence of the high fat but it did magnify the effect of the high fat. Here, we are seeing it again in the females. So, there are a lot of variables in our models, including the timing of intervention and exactly how we are doing it. We have definitely not worked out all of the bugs, but we have enough evidence that we can see the broad pattern.