Microbial genes outnumber human genes by 100 to 1 in the intestinal microbiome, leading some to propose that it is a “microbial organ” that performs important functions for the host, such as nutrient harvesting and immune development 11 . However, as with any complex and intimate interaction, there is a mixture of common and divergent interests with opportunities for mutual benefit 11 and manipulation 12 . Fitness interests of gut microbes are also often not aligned, because members of the microbiota compete with one another over habitat and nutrients. This means that highly diverse populations of gut microbes may be more likely to expend energy and resources in competition, compared to a less diverse microbial population. A less diverse microbial population is likely to have species within it that have large population sizes and more resources available for host manipulation. Moreover, the larger a particular microbial population is, the more power it would have to manipulate the host through higher levels of factor production or other strategies (see below) and large scale coordination of these activities (e.g., through quorum sensing). Therefore, we hypothesize that lower diversity in gut microbiome should be associated with more unhealthy eating behavior and greater obesity (i.e., decreased host fitness).

Conflict over resource acquisition and resource allocation can occur as a result of conflict between different genetic interests within an organism. For example, genetic conflict between maternal and paternal genes is hypothesized to play a role in the unusual eating behavior that characterizes the childhood genetic diseases Beckwith–Wiedemann syndrome and Prader–Willi syndrome. These syndromes are characterized by altered appetite and differences in infant suckling that can result from overexpression of genes of paternal or maternal origin, respectively 9 , 10 . In parent‐of‐origin genetic conflict, paternally imprinted genes are thought to drive increased demands for extracting resources from the mother, and maternally imprinted genes tend to resist these effects. Metagenomic conflict between host and microbiome can be considered an extension of this genetic conflict framework, but one that includes other genomes (i.e., microbes in the gut) with genes that affect the physiology and behavior of a host organism, potentially altering host eating behavior in ways that benefit microbe fitness.

The struggle to resist cravings for foods that are high in sugar and fat is part of daily life for many people. Unhealthy eating is a major contributor to health problems including obesity 1 as well as sleep apnea, diabetes, heart disease, and cancer 2 - 4 . Despite negative effects on health and survival, unhealthy eating patterns are often difficult to change. The resistance to change is frequently framed as a matter of “self‐control,” and it has been suggested that multiple “selves” or cognitive modules exist 5 each vying for control over our eating behavior. Here, we suggest another possibility: that evolutionary conflict between host and microbes in the gut leads microbes to divergent interests over host eating behavior. Gut microbes may manipulate host eating behavior in ways that promote their fitness at the expense of host fitness. Others have hypothesized that microbes may be affecting our eating behavior 6 - 8 , though not in the context of competing fitness interests and evolutionary conflict.

Evidence indicates many potential mechanisms of manipulation

There is a selective influence of diet on microbiota Individual members of the microbiota, and consortia of those microbes, have been shown to be highly dependent on the nutrient composition of the diet. Prevotella grows best on carbohydrates; dietary fiber provides a competitive advantage to Bifidobacteria 13, and Bacteroidetes has a substrate preference for certain fats 14. Some specialist microbes, e.g. mucin degrading bacteria such as Akkermansia mucinophila, thrive on secreted carbohydrates provided by host cells. Other butyrate producing microbes, e.g. Roseburia spp., fare better when they are delivered polysaccharide growth substrates in the diet. Specialist microbes that digest seaweed have been isolated from humans in Japan 15. African children raised on sorghum have unique microbes that digest cellulose 16. Many other examples exist 17. Even microbes with a generalist strategy tend to do better on some combinations of nutrients than others, and competition will determine which microbes survive 18, 19.

Microbes can manipulate host behavior There is circumstantial evidence for a connection between cravings and the composition of gut microbiota. Individuals who are “chocolate desiring” have different microbial metabolites in their urine than “chocolate indifferent” individuals, despite eating identical diets 20. There is also evidence for effects of microbes on mood. A double‐blind, randomized, placebo controlled trial found that mood was significantly improved by drinking probiotic Lactobacillus casei in participants whose mood was initially in the lowest tertile 21. There are many other examples of microbes affecting their hosts' mood and behavior, mostly from animal studies (Fig. 1). Butyrate, a short chain fatty acid largely produced by the microbiota, has been shown to have profound central nervous system effects on mood and behavior in mice 22. Microbiota transfer to germ free mice leads to timid behavior when fed feces from mice with anxiety‐like behavior. When germ‐free mice from an anxious strain were fed with a fecal pellet from a control mouse, the inoculated mice exhibited behavior that was more exploratory, and more like their fecal donors 23. In addition, a probiotic formulation with Lactobacillus helveticus R0052 and Bifidobacterium longum R0175 alleviated psychological distress 24. This effect can be altered by diet and inflammation 25. If one feeds Lactobacillus rhamnosus (JB‐1) to mice, not only does it reduce their stress‐induced corticosterone hormone levels, but it also makes them more dogged: L. rhamnosus (JB‐1) fed mice swim longer than the control fed mice when put in a glass cylinder filled with 15 cm of water and no means of escape 26. This effect disappeared when the experimenters severed the vagus nerve, suggesting a role for the vagus nerve in microbial manipulation of host behavior. In contrast, severing the vagus nerve had no effect on swimming behavior of control mice that were not fed L. rhamnosus (JB‐1) 26. In a widely cited example of microbes affecting behavior, Toxoplasma gondii suppresses rats' normal fear of cat smells, often to the detriment of the rats, but to the benefit of the microbes that are ingested into their new feline host. T. gondii infected rats are reported to become sexually aroused by cat urine 27, a propensity that promotes transmission of T. gondii at the expense of the fitness of the rat. Figure 1 Open in figure viewerPowerPoint Like microscopic puppetmasters, microbes may control the eating behavior of hosts through a number of potential mechanisms including microbial manipulation of reward pathways, production of toxins that alter mood (shown in pink, diffusing from a microbe), changes to receptors including taste receptors, and hijacking of neurotransmission via the vagus nerve (gray), which is the main neural axis between the gut and the brain.

Microbes can induce dysphoria that changes feeding behavior Although certain Lactobacillus appear to reduce anxiety, colonization of the gut with the pathogen Campylobacter jejuni increased anxiety‐like behavior in mice 28, raising the possibility that microbe‐induced dysphoria might also affect human behavior. Recent studies have linked the inconsolable crying of infant colic with changes in gut microbiota including reduced overall diversity, increased density of Proteobacteria and decreased numbers of Bacteroidetes compared to controls 29. Colic has been reported to result in increased energy delivery to infants, sometimes resulting in accelerated weight gain 30. If infant crying has a signaling function that increases parental attention and feeding 31, 32, colic may increase the resource delivery to the gut and hence microbial access to nutrients. One potential mechanism by which dysphoria can influence eating involves bacterial virulence gene expression and host pain perception. This mode of manipulation is plausible because production of virulence toxins often is triggered by a low concentration of growth‐limiting nutrients. Detection of simple sugars and other nutrients regulates virulence and growth for a variety of human‐associated microbes 33-37. These commensals directly injure the intestinal epithelium when certain nutrients are absent, raising the possibility that microbes manipulate behaviors through pain signaling. In accord with this hypothesis, bacterial virulence proteins have been shown to activate pain receptors 38. Moreover, pain perception (nociception) requires the presence of an intestinal microbiota in mice 39 and fasting has been shown to increase nociception in rodents by a vagal nerve mechanism 40.

Microbes modulate host receptor expression One route to manipulation of host eating behavior is to alter the preferences of hosts through changing receptor expression. One study found that germ‐free mice had altered taste receptors for fat on their tongues and in their intestines compared to mice with a normal microbiome 41. In another experiment, germ free mice preferred more sweets and had greater numbers of sweet taste receptors in the gastrointestinal tract compared to normal mice 42. In addition, L. acidophilus NCFM, administered orally as a probiotic, increased intestinal expression of cannabinoid and opioid receptors in mouse and rat intestines, and had similar effects in human epithelial cell culture 43. These results suggest that microbes could influence food preferences by altering receptor expression or transduction. Changes in taste receptor expression and activity have been reported after gastric bypass surgery, a procedure that also changes gut microbiota and alters satiety and food preferences (reviewed in 44).

Microbes can influence hosts through neural mechanisms Gut microbes may manipulate eating behavior by hijacking their host's nervous system. Evidence shows that microbes can have dramatic effects on behavior through the microbiome‐gut‐brain axis 6, 45, 46. The vagus nerve is a central actor in this communication axis, connecting the 100 million neurons of the enteric nervous system in the gut 47 to the base of the brain at the medulla. Enteric nerves have receptors that react to the presence of particular bacteria 48 and to bacterial metabolites such as short‐chain fatty acids. Evidence suggests that the vagus nerve regulates eating behavior and body weight. For example, blockade or transection of the vagus nerve has been reported to cause drastic weight loss 49, 50. On the other hand, vagus nerve activity appears to drive excessive eating behavior in satiated rats when they are stimulated by norepinephrine 51. These results suggest that gut microbes that produce adrenergic neurochemicals (discussed below) may contribute to overeating via mechanisms involving vagal nerve activity. Together these results suggest that microbes have opportunities to manipulate vagus nerve traffic in order to control host eating. Intriguingly, many practices that are known to enhance parasympathetic outflow from the vagus nerve, e.g. exercise, yoga, and meditation, are also thought to strengthen willpower 52 and improve accuracy of food intake relative to energy expenditure 53. However, increased vagus activity is not always associated with health. One study linked parasympathetic vagus activity with weight loss in patients with anorexia nervosa 54, suggesting that vagus nerve signaling is important in regulating body weight, and sometimes can lead to pathological anorexia.

Microbes can influence hosts through hormones Microbes produce a variety of neurochemicals that are exact analogs of mammalian hormones involved in mood and behavior 8, 55-57. More than 50% of the dopamine and the vast majority of the body's serotonin have an intestinal source 58, 59. Many transient and persistent inhabitants of the gut, including Escherichia coli, 8, 55, 56 Bacillus cereus, B. mycoides, B. subtilis, Proteus vulgaris, Serratia marcescens, and Staphylococcus aureus 60 have been shown to manufacture dopamine. Concentrations of dopamine in culture of these bacteria were reported to be 10–100 times higher than the typical concentration in human blood 60. B. subtilis appears to secrete both dopamine and norepinephrine into their environment, where it interacts with mammalian cells. Transplant of the microbiome from a male to an immature female mouse significantly and stably increases testosterone levels in the recipient 61. In turn, host enzymes are known to degrade neurotransmitters of bacterial origin. For instance, mammals use monoamine oxidase to silence exogenous signaling molecules, among other functions 62, 63. This may be evidence for selection on hosts to counteract microbial interference with host signaling. Certain probiotic strains alter the plasma levels of other neurochemicals. B. infantis 35624 raises tryptophan levels in plasma, a precursor to serotonin 64. The lactic acid producing bacteria found in breast milk and yogurt also produce the neurochemicals histamine 65 and GABA 66. GABA activates the same neuroreceptors that are targeted by anti‐anxiety drugs such as valium and other benzodiazepines. Appetite‐regulating hormones are another potential avenue for manipulation of mammalian eating behavior. In mice, treatment with VSL#3, a dietary supplement consisting of a mixture of Lactobacillus strains, reduced hunger‐inducing hormones AgRP (agouti related protein) and neuropeptide Y in the hypothalamus 67. Germ‐free mice were also shown to have lower levels of leptin, cholecystokinin, and other satiety peptides 41, hormones that control hunger and food intake partly by affecting vagus nerve signaling. Numerous commensal and pathogenic bacteria manufacture peptides that are strikingly similar to leptin, ghrelin, peptide YY, neuropeptide Y, mammalian hormones that regulate satiety and hunger 68. Moreover, humans and other mammals produce antibodies directed against these microbial peptides, a phenomenon that could have evolved as a mammalian counter‐adaptation to microbial manipulation. Anti‐hormone antibody production may be important in maintaining the fidelity of host signaling systems. However, these antibodies also act as auto‐antibodies against mammalian hormones 68. This autoimmune response implies that microbes have the capacity to manipulate human eating behavior (i) directly with peptide mimics of satiety regulating hormones, or (ii) indirectly by stimulating production of auto‐antibodies that interfere with appetite regulation. The antibody response to microbial analogs of human hormones supports the hypothesis that conflict between host and microbiota influences the regulation of eating behavior.

Mucin foraging bacteria control their nutrient supply Several commensal bacteria are known to induce their hosts to provide their preferred nutrients through direct manipulation of intestinal cells. For example, Bacteroides thetaiotaomicron is found on host mucus, where it scavenges N‐glycated oligosaccharides secreted by goblet cells in the gut. B. thetaiotaomicron induces its mammalian host to increase goblet cell secretion of glycated carbohydrates 69, 70. Investigators have shown that another mucin‐feeding species, A. muciniphila, also increases the number of mucus producing goblet cells when inoculated in to mice 71. On the other hand Faecalibacterium prausnitzii, a non‐mucus‐degrading bacterium that is co‐associated with B. thetaiotaomicron, inhibits mucus production by goblet cells 70. These species provide a proof of principle that gut bacteria can control their nutrient delivery, involving a mechanism that is energetically costly for the host 72.

Intestinal microbiota can affect obesity Evolutionary conflict between the gut microbiome and host may be an important contributor to the epidemic of obesity. In a landmark paper, Backhed and colleagues showed that mice genetically predisposed to obesity remained lean when they were raised without microbiota 73. These germfree mice were transformed into obese mice when fed a fecal pellet from a conventionally raised obese mouse 74. Inoculation of germ‐free mice with microbiota from an obese human produced similar results 75. Mice lacking the toll‐like receptor TLR5 became obese and developed altered gut microbiota, hyperphagia, insulin resistance, and pro‐inflammatory gene expression 76. Fecal pellets from these TLR5 knockout mice, when fed to wild type mice, induced the same phenotype. The gut microbes of obese humans are less diverse than the microbiota of their lean twins 77, consistent with the hypothesis that lower diversity may affect eating behavior and satiety.