Redefining Bacterial Pathogenicity Using the Tools of Molecular Genetics

Among the things we have learned about bacterial pathogenicity are these fundamental characteristics:

• Pathogens are impressive cell biologists. Twenty-five years of accumulated data demonstrate that bacteria manipulate the normal functions of the host cell in ways that benefit the bacteria (Figure 3-1).
• Horizontal gene transfer via mobile genetic elements has been an extremely important force in the evolution of bacterial specialization, including that of pathogens. The genes for many specialized “bacterial” products, like toxins and adhesins, actually reside on transposons and phages.
• The inheritance of blocks of genes, called pathogenicity islands, is often the key to the expression of pathogenicity in bacteria.

FIGURE 3-1

Pathogenic bacteria interfere with or manipulate for their own benefit the normal function(s) of the host cell. SOURCE: Figure reprinted from Wilson et al. (2002) with permission from Cambridge University Press. Copyright 2002.

When we began using the tools of molecular genetics to examine virulence genes and identify their functions, we first tried to isolate particular genes for toxins and other likely virulence products. Today, we take a very different approach, which reflects our understanding of pathogen behavior. For example, unlike commensal bacteria, Salmonella breaches the host’s epithelial barrier, usually in areas of the intestinal epithelium known as Peyer’s patches, and becomes engulfed by phagocytic cells. Instead of being killed, the pathogen replicates there and is then distributed to the liver and the spleen. Eventually, in many cases, the pathogen will be shed by the host, often over long periods of time. These key events in the pathogenesis of infection, in addition to the interaction of the pathogen with the host’s innate and adaptive immune systems, are illustrated in Figure 3-2.

FIGURE 3-2

Salmonella infection. IFN-γ = gamma interferon; MLN = mesenteric lymph node. SOURCE: Monack et al. (2004b).

The difference between the pathogen Salmonella and ancestral, commensal organisms in the bowel is based on the inheritance of pathogenicity islands, which give these bacteria the ability to leave the confines of the colon for locations where other bacteria would be killed. There the evolving pathogen can act free from competition. To identify the genes that enable such incursions into the host, we use a microarray-based negative selection strategy, as shown in Figure 3-3, which allows us to screen the entire Salmonella genome for genes that are associated with different stages of infection (Chan et al., 2005).

FIGURE 3-3

Microarray-based negative selection strategy. hyb = hybridization; IP = intraperitoneal. SOURCE: Figure courtesy of Kaman Chan, Ph.D., Stanford University.

Using this strategy, one finds that while many genes are expressed within the first week of disease, there is a group of genes that are not expressed until the second week of the disease, and even others that are not expressed until the third or fourth week of disease (Figure 3-4; Lawley et al., 2006). These results indicate that particular genes are required for different stages of persistent infection within the mouse. Some of the genes are involved in the ability of Salmonella to excrete proteins that kill macrophages during initial infection, while others have evolved to allow the bacterium to replicate and persist within the vacuoles of macrophages.

FIGURE 3-4

Time-dependent selection of persistence genes. A yellow box indicates that the persistence gene is absent. A blue or black box indicates that the persistence gene is present. CFU = colony-forming units; LPS = lipopolysaccharide; PMNs = polymorpho-nuclear (more...)

Thus, it takes many genes to contribute to the ability of Salmonella to make its journey from the mouth to the Peyer’s patch in the epithelium of the small intestine, and from there into a macrophage that is spread to other sites. With the technology now available to us, we can now identify all such genes quite readily, but it will take many years before we will be able to determine their exact biological function in Salmonella-host interactions.

The Host as a Reporter of Response to Infection

The host can also tell us what happens during acute and persistent bacterial infections. When Lucinda Thompson, Denise Monack, and I examined gene expression in the peripheral blood of infected mice following Salmonella infection, we found that 344 genes were induced during acute Salmonella infection of naïve mice (Figure 3-5). These genes are associated with interferon signaling, infiltration of neutrophils, and leukocyte extravasation.

FIGURE 3-5

Response to Salmonella infection. Columns under the blue box correspond to separate naïve animals before challenge. Each column under the pink box corresponds to the gene activation in a single animal after wild-type infection (days 4–9 (more...)

We looked at specific gene pathways that were up regulated or down regulated in response to Salmonella infection, and we compared these patterns of host response with those that occurred in mice that had been previously immunized with a known Salmonella typhimurium vaccine strain, aroA. Similar genes were induced in early infection in immunized mice, as in naïve mice, and therefore constitute an “immunity signature” (Figure 3-6). Persistently-infected mice (10 months post-infection) also expressed a characteristic profile of induced genes, as compared with uninfected controls, that included genes involved in antigen presentation, defense responses, and the regulation of various cellular processes. At six months, most animals are moving toward a “normal” pre-infection gene expression profile, but some continued to express infection-associated genes.

FIGURE 3-6

Response to Salmonella infection in 129sv mice following immunization. The host transcriptional response in peripheral blood cells was measured in naïve mice challenged by oral infection with Salmonella typhimurium (red). The transcriptional response (more...)

Thanks to mouse models developed by Denise Monack et al. (2004a), it is now possible to look at the transmission of Salmonella from persistently-infected animals that are asymptomatic, but which excrete the bacterium for up to a year. Naïve mice placed in the same cage as persistently infected animals become infected. Some, but not all, become “supershedders,” excreting very high levels of bacteria in their stools and transmitting the infection to their uninfected cage mates (Lawley et al., 2008). Epidemiologists have identified people who are supershedders of certain infectious diseases, as well. Although they are asymptomatic, Salmonella supershedders have been found to have a pronounced inflammatory response as compared to other persistently-infected mice (which excrete relatively low levels of bacteria).

The bacteria from supershedders are no more virulent or transmissible than those from other infected mice. It was not clear what factors contributed to the supershedder property. The mice are genetically inbred so presumably an animal becoming a supershedder was governed by non-genetic factors. Mice immuno-compromised by steroid treatment were no more likely to become supershedders. However, if mice are given a small, single dose of neomycin—enough to reduce the commensal intestinal flora for less than 48 hours—and then in four days are exposed to Salmonella, almost all of them become supershedders (Lawley et al., 2008). These animals go from having low to high levels of intestinal inflammation, and at the same time, they become very efficient transmitters of the disease. One possible interpretation of this observation is that the fine balance between the commensal flora and the innate immune system is disrupted by the elimination of certain members of the normal flora by antibiotic killing and provides a window of opportunity for the establishment of larger populations of Salmonella in the bowel.

Pathogens Versus Commensals

The term pathogen is derived from a Greek phrase that means, “the birth of pain.” A successful pathogen must do the following:

• Acquire virulence genes
• Sense the environment
• Switch virulence genes on and off
• Move to the site of infection
• Become established
• Acquire nutrients
• Survive stress
• Avoid the host’s immune system
• Subvert host cytoskeleton and signaling pathways
• Disseminate to other sites and/or hosts (in some cases)
• Be transmitted to a new susceptible host

Successful commensals also do many of these things, but unlike a pathogen, a commensal does not have an inherent ability to cross anatomic barriers or breach host defenses that limit the survival or replication of other microbes. As previously noted, pathogens (but not commensals) choose to live in dangerous places to avoid competition and get nutrients. These invasive properties are essential to the pathogen’s survival in nature.

Nevertheless, several members of the human bacterial flora that usually live uneventfully in the human nasopharynx—including Streptococcus pneumoniae, Neisseria meningitidis, Haemophilus influenzae type b, and Streptococcus pyogenes—sometimes cause disease. These microbes, which I call “commensal pathogens,” have virulence determinants that suggest that they regularly come into intimate contact with elements of the innate and adaptive immune system (Falkow, 2006). Recent history shows that immunization against these pathogens not only prevents human disease but also eliminates the microbes’ ability to colonize the human host efficiently. Such commensal pathogens persist in a significant proportion of the human population, the vast majority of whom are asymptomatic carriers.

This raises several questions: Are such microbes pathogens, or are they commensals? Are “virulence” factors often actually a subset of adaptive factors that allow microbes to exploit a particular niche, but not necessarily designed to cause disease? Are they virulence factors, or would a better term be colonization factors? I submit that medicine’s focus on disease may sometimes distract us from understanding the biology of pathogenicity. Of course, it is important that there be a medical definition of a pathogen that is based on disease. On the other hand, I would argue that disease does not encompass the biological aspects of pathogenicity and the evolution of the host-parasite relationship. Many pathogens cause persistent infections in humans (e.g., Mycobacterium tuberculosis, Treponema pallidum [the cause of syphilis], Chlamydia, S. typhi, and Helicobacter pylori), and most such infections are asymptomatic. Might these microbes be considered commensal pathogens?

Moreover, as Figure 3-7 suggests, the decline of many infectious diseases in industrialized countries has been accompanied by an increase in immune disorders, such as multiple sclerosis, Crohn’s disease, and asthma. The eradication of Helicobacter has been associated with an increase in esophageal cancer and, more recently, asthma (Blaser et al., 2008). Perhaps one third or more of the world’s population is asymptomatically colonized with Mycobacterium tuberculosis. Over the course of the last century, as Western society has eliminated this pathogen through basic public health measures, it seems that we may actually have eliminated a microbe with which our species had a dialog, a microbe that “talked” to the human immune system and kept it primed for defensive action. Perhaps, as Pogo cartoonist Walt Kelly has suggested, “we have met the enemy and he is us.”

FIGURE 3-7

The decline of many infectious diseases in industrialized countries. (A) has been accompanied by an increase in immune disorders (B), United States, 1950 to 2000. SOURCE: Bach (2002).

If the nature of microbial pathogenicity is schizophrenic—characterized by inconsistent or contradictory elements—then it is important to study every aspect of its biology, and not be distracted by its role in causing disease. Instead, we must follow the instruction of Thomas Huxley, as expressed in a quotation from Huxley’s letter to Charles Kingsley on September 23, 1860, that both Lederberg and I have treasured: “Sit down before fact as a little child, be prepared to give up every preconceived notion, follow humbly wherever and to whatever abysses nature leads, or you shall learn nothing” (Huxley, 1901).

Acknowledgments

Josh Lederberg was a remarkable man with a remarkable intellect. His legacy is yet to be fully appreciated. He influenced me from the outset of my career, and although I told him so in person on several occasions, I think it appropriate to express my gratitude to him here as well.

My remarks at this symposium were summarized by Eileen Choffnes and her staff and formed the basis for this paper. I am grateful to Lucinda Thompson and Denise Monack for permitting me to present here some of their previously unpublished results on the host response in naïve and immunized mice to Salmonella infection. I also thank Denise Monack and Trevor Lawley for discussing with me their recent exciting findings on Salmonella transmission.

Bacterial Virulence as an Immune Response

We define virulence as the magnitude of the morbidity and the increase in the likelihood of mortality resulting from the colonization and proliferation of bacteria in or on a host. To facilitate our consideration of this virulence and its evolution we use the gross simplification, a cartoon, of the bacterium-host interaction presented in Figure 3-8. Bacteria enter a site, the blue box, where they replicate and establish a population and colonize the host, but in which they do not generate perceptible symptoms. Virulence requires their passage into a second site, the red box, where the presence of bacteria (or their products) can, but need not, cause symptoms, e.g., for a Streptococcus pneumoniae bacteremia the blue site is the nasopharynx and the red is the bloodstream. In this model the red site needn’t be a different physical location. It could be a different state of the bacteria in the site of their colonization, e.g., for a Staphylococcus aureus skin infection, the blue site would be the skin and the red a boil.

FIGURE 3-8

The artist’s conception of the infection process and the host’s immune response and overresponse. Blue—site where the presence of bacteria does not result in symptoms—asymptomatic. Red—site/state where the presence (more...)

In Figure 3-8 as well as in mammals, virulence occurs in two ways, both of which require the bacteria to enter the red, potentially symptomatic site or state: (i) direct damage to the host tissue is caused by the replication of the bacteria and/or the production of specific products (toxins), or (ii) indirect damage to the host occurs through an inappropriate or overresponse of the immune system. Both types of damage are represented by the “!” within a triangle. In this scheme the immune defenses can prevent virulence in one or more of seven related ways:

1. Limiting the entry of bacteria into the asymptomatic site
2. Limiting the proliferation of the bacteria in the asymptomatic site
3. Increasing the rate of clearance of bacteria and their products from the asymptomatic site
4. Preventing entry of bacteria or their products into the potentially symptomatic site or state
5. Reducing the rate of proliferation of the bacteria within the potentially symptomatic site or state
6. Increasing the rate of clearance of bacteria and their products from the potentially symptomatic site or state
7. Preventing an immune overresponse to the bacteria or their products in the potentially symptomatic site or state

The first three of these immune responses maintain the density of bacteria and concentrations of their products in the asymptomatic site at levels where they are unlikely to spill over or otherwise enter the site or state where they can generate symptoms. Whether they do generate symptoms and the magnitude of those symptoms given passage into the red site or state also depends on how well the immune system limits their densities and the concentrations of their products. In Figure 3-8, the number 8 is the infectious transmission of bacteria promoted by the generation of symptoms, and the number 9 is the transmission of bacteria from the asymptomatic site. All of these enumerated steps in which the immune system limits the virulence of the bacteria can be classified as appropriate responses. However, inappropriate responses, for which we use the term overresponse when they lead to host damage, may occur due to defects in one of these steps or as a secondary consequence of mounting an immune response.

In the following we focus primarily on the virulence resulting from the over-response of the immune system. There are, however, examples of virulence that can be attributed to the direct damage of host tissue by the replication of bacteria or the secretion of their products. Included among these are (i) dental caries, resulting from the acid produced by metabolizing Lactobacillius acidophilus or Streptococcus mutans (Gibbons, 1964), (ii) paralysis due to the neurotoxins secreted by Clostridium botulinum or Clostridium tetani acting on the nerve and motor endplates (Schiavo et al., 1992), and (iii) diarrhea resulting from enterotoxins that inhibit resabsorption of sodium chloride or promote its secretion. Examples of virulence being a direct product of the interaction between bacteria and host cells appear to be rare relative to those in which the morbidity and mortality can be attributed to the indirect damage due to an immune overresponse.

As illustrated in Table 3-1, the morbidity and mortality of bacterial infections can be attributed to the host’s immune system operating in one of three inappropriate ways: (i) being more vigorous than needed, (ii) being incorrect for that pathogen, or (iii) responding to the wrong signals. The best-investigated example of the immune system responding too vigorously is bacterial sepsis, where the entry of cytokines and bacteria into the bloodstream brings about widespread blood vessel injury and multiple organ failure (impaired pulmonary, hepatic, or renal function). Here the response to the bacteria is at one level appropriate, as the cytokines released play an important part in attracting neutrophils (immune cells that phagocytose bacteria) to the local infection site, but is also excessive (Kurahashi et al., 1999). The distinction between an inappropriate and appropriate immune response can be seen in the spectrum of illness associated with Mycobacterium leprae (Modlin, 2002; Sieling et al., 1999). Hosts that respond to infection predominantly with antibodies and very few CD4 T cells have infectious sites with large macrophages that contain numerous mycobacteria. These macrophages are responsible for the multiple skin lesions and nodules seen in lepromatous leprosy, while a host with T helper 1-type response (high interferon-γ production and low interleukin-4 [IL-4]) has numerous well-formed granulomas with very few mycobacteria that form minor skin lesions. Superantigens provide an example of the immune system responding to an incorrect signal. Superantigens are bacterial products that stimulate a large number of T cells (1–40% of T cells will react) by binding to major histocompatibility complex class II molecules and T cell receptors (beta chain) independently of their specificity for antigens (Rott and Fleischer, 1994). S. aureus, Streptococcus pyogenes, Mycoplasma arthritidis, and Yersinia pseudotuberculosis are among the bacteria that produce superantigens. In the case of toxic shock syndrome, the superantigens produced by S. aureus induce the indiscriminate and overwhelming activation of T cells leading to the production of cytokines that mediate shock and tissue injury. In all three of these cases the morbidity and mortality of the host can be attributed to an apparently misguided response of the immune system, which we refer to as an overresponse.

TABLE 3-1

Some Examples of Virulence Resulting from an Immune Overresponse.

The Evolution of Bacterial Virulence as an Immune Response

How does the observation that much of the morbidity and mortality can be attributed to a host overresponse to bacteria help in understanding the evolution of virulence? In a perspective written a decade ago, one of us (Levin, 1996) listed four hypotheses that account for the evolution of virulence: (i) the conventional wisdom, (ii) epidemiological selection, (iii) coincidental evolution, and (iv) short-sighted, within-host evolution. Since that time, although there have been a number of theoretical, experimental, and speculative articles on the evolution of virulence (for a small and admittedly biased sample see Andre and Godelle, 2006; Bonhoeffer and Nowak, 1994; Brown et al., 2006; Bull, 1994; Ebert and Bull, 2003; Ebert and Herre, 1996; Frank, 1996; Grech et al., 2006; Lipsitch et al., 1995, 1996; Lipsitch and Moxon, 1997; Mackinnon and Read, 2004; Regoes et al., 2000), we do not know of studies that have rejected any of these hypotheses and only one adding what may be a new hypothesis: quasispecies evolution (Pfeiffer and Kirkegaard, 2005). This fifth hypothesis may only apply to viruses with high mutation rates and arguably could be subsumed under the broader rubric of within-host evolution. In this section we consider how the observation that morbidity and mortality of bacterial infections can be attributed to the hosts’ immune overresponse fits each of these hypotheses for the evolution of the virulence of bacteria.

The Conventional Wisdom

This phrase, which the late John Kenneth Galbraith coined to describe ideas and explanations that are widely accepted as true by the public, was applied by Bob May and Roy Anderson (May and Anderson, 1983b) to describe the then-prevailing view of the evolution of the virulence. According to that wisdom, virulence is an artifact of the relative novelty of parasite’s association with its host. As the relationship between the parasite and host matures, natural selection in either the parasite or host population or both will lead to the extinction of one or the other species or the evolution of symbiosis or mutualism.

While the original theory behind this hypothesis for the evolution of virulence of infections amounts to little more than the adage “don’t bite the hand that feeds you,” the evidence in support of it was and remains compelling. Many of the bacteria responsible for morbidity and mortality of humans were acquired from other species in the not-so-distant past (after the advent of agriculture), and some are continuously acquired in this way. Included among these zoonotic (and protozoonotic) infections are plague, tuberculosis, Legionnaires’ disease, botulism, anthrax, brucellosis, tularemia, Rocky Mountain spotted fever, cholera, and other diarrheal diseases. The bacteria responsible for some of these infections, such as Mycobacterium tuberculosis, are transmitted between humans and can be maintained without the animal source. Others such as Legionella pneumophila are not. Also consistent with the conventional wisdom is the correlated observations that only a very small minority of the vast numbers of species of bacteria that colonize mammals cause disease.

It may seem that the proposition that the virulence of bacterial infections can be attributed to host immune overresponse fits quite well with this conventional wisdom. To wit, the immune system has not yet had the time to evolve to moderate the response to these novel bacteria and their products and/or these bacteria have not yet evolved into being nice. Eventually, or as it was once referred to, on “equilibrium day,” (Levin et al., 2000), mutualism will prevail and the immune overresponse will be tempered.

Epidemiological Selection

The conventional wisdom is an observation rather than a mechanism, an observation that focuses on the interactions between bacteria and the individual hosts they colonize. To fully understand the evolution of commensal and pathogenic bacteria, however, it is necessary to consider their lifestyle outside the host and, in particular, their transmission between hosts. One approach to this more comprehensive picture of the evolution of parasitic microbes has been to draw inferences about the nature and direction of selection from epidemiological models (Levin et al., 1982; Levin and Pimentel, 1981; May and Anderson, 1983a). In accord with this perspective, the fitness of a particular strain of bacteria is given by its basic reproductive number, R₀, the number of secondary infections caused by a single infected individual in a wholly susceptible population of hosts; the higher the value of R₀, the greater the fitness of the bacteria. In these traditional epidemiological models, virulence is only expressed as mortality. Morbidity and other more subtle effects of infections are not directly considered in these epidemiological models.

As long as the transfer to new hosts requires viable hosts, selection will favor bacteria that are not only infectiously transmitted at ever-higher rates but also persist longer in colonized hosts (i.e., are less likely to kill the host). In other words, selection will favor ever-more-benign, symbiotic, or better yet, mutualistic bacteria. Evolution in the host population will also be for reduced virulence; hosts that are less subject to infection-associated morbidity and mortality will be favored. As long as transmission occurs from the blue asyptomatic site (8 in Figure 3-8) rather than the red site (9 in Figure 3-8), these epidemiological models can be seen as the theoretical basis of the conventional wisdom (also see Lenski and May, 1994). If, however, transmission and the morbidity and mortality of the host are coupled so the more virulent bacteria are transmitted at higher rates than the more benign, there is a trade-off between the loss of the host and gain to the bacteria; virulence would be favored in the bacterial population (Ebert and Bull, 2003).

On first consideration it may seem that this transmission and virulence tradeoff is inconsistent with the proposition that the morbidity and mortality of the infection is a product of the host’s immune overresponse. We suggest this is not necessarily the case. The host overresponse could be a by-product of selection operating on individual bacteria to promote their transmission. While we don’t know of overwhelming, quantitative, empirical evidence of this being the case for any pathogenic bacteria (viruses are another matter; see Fenner and Ratcliffe, 1965), this interpretation is supported by reasonable plausibility arguments. Here we consider two of the more compelling of these examples of pathogenic bacteria of humans.

The first is the diarrheal diseases in which humans play a significant role in the transmission process. Because of the massive output of bacteria, diarrhea is likely to increase the density of bacteria in water and food products and thereby the transmission rate of these bacteria. Thus, as long as transmission is promoted by diarrhea, selection in the bacterial population will favor mechanisms that cause diarrhea. In some cases the induction of diarrhea is attributed to what can be seen as immune overresponse. The dysentery bacteria Shigella flexneri induces the release of the cytokine interleukin 1 (IL-1) in infected macrophages, which leads to extensive injury of the colon mucosa, which in turn results in fluid and protein loss into the intestinal lumen and the ensuing diarrhea (Hilbi et al., 1997). This hypothesis for the evolution of diarrhea to increase transmission requires that the transmission advantage more than makes up for the loss in transmission due to host mortality. To our knowledge, there are no quantitative empirical studies demonstrating that this trade-off obtains for any diarrheal disease.

The second example is plague. Albeit not yet as well documented as the oft-told mother of all trade-off stories, myxoma and the Australian rabbits (Fenner, 1965), the emerging tale of the evolution of the virulence of the plague bacillus, Yersinia pestis, has parallels to that story. There is compelling evidence that this flea-transmitted pathogen evolved from a not very virulent enteric, oral-fecal transmitted Yersinia relatively recently by the acquisition of a couple of plasmids and a few chromosomal genes (Achtman et al., 1999; Carniel, 2003). Since fleas acquire these bacteria from the blood of rodents, the density of bacteria in circulating blood would be directly associated with the likelihood of their transmission to other rodents (or humans). Also directly associated with this density of bacteria in the blood is sepsis, the virulent manifestation of Y. pestis infections. Elisabeth Carniel (personal communication) has suggested that the capacity to generate lethal sepsis is not just a by-product of the proliferation of bacteria in the blood, but may be selected for in the bacterial population. Although the cost-benefit calculation has not been made, it may be that the rate of transmission of the bacteria is augmented by their killing infected rodents, as fleas move to new hosts when their original host dies. For both diarrheal diseases and plague, the virulence resulting from the host overresponse is associated with transmission. Clearly more empirical work would be necessary to confirm the existence of a trade-off between bacterial transmission and an immune overresponse and the postulated exploitation of this overresponse for the epidemiological advantage of the parasite.

Coincidental Evolution

In accord with this hypothesis there is no advantage to the bacteria to make the host sick and certainly no advantage for the host to be ill; virulence is a consequence of the bacteria being in the wrong host or in a wrong site in the right host (Levin and Svanborg Eden, 1990) (the arrow above 7 in Figure 3-8). The bacterial products responsible for the morbidity and/or mortality of the host, virulence determinants, evolved in response to selection for some function other than virulence.

Reasonable candidates for coincidental virulence due to an immune over-response are diseases associated with Helicobacter pylori. These bacteria colonize and maintain populations in the stomachs of the majority of humans for most of their lives without generating symptoms and appear to have done so since prehistoric times (Falush et al., 2003). However, it wasn’t until Marshall and Warren (1984) presented evidence that a curved bacteria we now know as H. pylori was an etiologic agent for gastric and peptic ulcers that this seemingly commensal bacteria was elevated to the status of pathogen. This distinction was further enhanced by evidence that H. pylori was also associated with gastric cancers (Moss and Blaser, 2005; Tatematsu et al., 2005). H. pylori colonization can result in a chronic inflammatory state that is generated when the host responses (such as the release of IL-8 and other chemokines, the attraction of neutrophils, macrophages, and the local stimulation of T cells) fails to clear the bacteria and lymphoid aggregates form in the lamina propria of the stomach and duodenum. This continued stimulation of the immune and inflammatory cells (termed chronic atrophic gastritis) results in the destruction of the gastric epithelium, formation of peptic ulcers, and increased risk for gastric cancers. Presumably, but not yet formally demonstrated, the induction of the inflammatory response and the subsequent diseases provides no advantage to H. pylori in a colonized host or its transmission to new hosts. In this sense, the virulence of H. pylori in colonized humans is coincidental.

While they are commonly described as pathogens, especially in grant proposals and by people suffering from the symptoms they can generate, a number of bacteria responsible for morbidity and mortality in humans also have good credentials as commensals. Like H. pylori they are carried asymptomatically by many and cause disease in few. Included among the more prominent of these commensal pathogens for humans are S. aureus, Haemophilus influenzae, S. pneumoniae, and Neisseria meningitidis. From an evolutionary perspective, invasive disease seems to be the wrong thing for these bacteria to do—dead ends. The sites of their virulence, blood and meninges, are certainly not good for their transmission to new hosts by their normal route, through respiratory droplets. The rare virulence of these commensal bacteria can be accounted for by an immune overresponse in these sites (Bergeron et al., 1998; Braun et al., 1999). The occasional movement of bacteria into a site where they can cause disease (the red in Figure 3-8) may be due to chance or coincidental evolution or as we argue below may be a consequence of within-host evolution of the bacterial population.

Within-Host Evolution

In accord with this hypothesis, the virulence of bacteria is the product of selection favoring more pathogenic members of a population colonizing an individual host (Levin and Bull, 1994). The advantage gained by the bacteria by generating symptoms in a colonized host is restricted to that host and may be to its disadvantage in its transmission to a new host; this evolution is short-sighted. A mutant commensal bacterium with the capacity to establish and maintain populations in normally sterile sites, cells, or tissues could be favored within a colonized host because in those sites there is less competition for nutrients and/or those mutant bacteria are somewhat protected from the host immune defenses.

Although we can make a good case and even cite evidence for the virulence of some viruses, such as poliovirus and Coxsackievirus, being the product of within-host evolution (Gay et al., 2006;Levin and Bull, 1994), for bacteria the best we can do at this stage is present arguments founded on plausibility and consistency with observations (see, for example, Meyers et al., 2003). Central to these arguments are the results of studies with mice and rats demonstrating that the bacteria responsible for invasiveness (blood infection) are commonly derived from one of very few cells (Meynell, 1957; Moxon and Murphy, 1978; Pluschke et al., 1983; Rubin, 1987). One possible explanation for these observations is that the bacteria responsible for the blood infections are the products of single, mutant cells with an enhanced capacity to invade and proliferate in blood.

While supporting the within-host evolution hypothesis for virulence, these observations are also consistent with the coincidental evolution hypothesis: that, by chance alone, only one or a few cells establish blood infections can be attributed to very small holes in the host’s defenses through which only one or very few bacteria traverse the arrow above 7 in Figure 3-8. Although the coincidental and within-host hypotheses could be distinguished by demonstrating that the bacteria establishing a blood infection have an inherited propensity for the invasion of blood, to our knowledge there are no published studies that have done this test. However, whether the invasiveness of the blood or other normally sterile sites is coincidental or due to within-host evolution, the virulence of bacteria in these sites can be attributed to a host’s immune overresponse.

The Evolution of Virulence Determinants

Not all bacteria or even all members of the same species of bacteria capable of colonizing mammals are responsible for disease. One explanation for why some bacteria cause disease and others do not is what have become known as virulence factors or virulence determinants, the expression of which are, by definition, essential for that bacteria to cause disease in (or on) colonized hosts (Finlay and Falkow, 1989). Included among these are characters that facilitate adhesion to host cells, evade the host constitutive and inducible immune defenses, and produce toxins. Appropriately, much of contemporary bacteriology is devoted to understanding the molecular biology, genetics, evolutionary origin, and mode of action of virulence determinants as a way to understand bacterial diseases and ideally prevent or treat them. While virulence determinants (factors) are almost certainly the products of adaptive evolution in bacterial populations, not so clear are the selection pressures responsible for their evolution and maintenance. Are they favored because of virulence, i.e., the morbidity and mortality of the host promotes the colonization, persistence, and infectious transmission of bacteria that express these determinants? Are virulence factors by-products of selection for other functions, e.g., their expression provides protection against grazing protozoa (Wildschutte et al., 2004) and/or facilitates competition with other microbes? Or is the virulence attributed to these factors an inadvertent by-product of their normal function in a host, a primitive character that will be lost on or before equilibrium day. While these hypotheses may be mutually exclusive for any specific bacterium-host and virulence factor, they are clearly not so collectively. Whether they evolve in response to selection for virulence or not, some of these virulence factors are responsible for triggering the immune overresponse.

Why Does the Immune System Overrespond?

In the preceding, we have portrayed the host immune system as misguided, overresponding in ways that cause rather than prevent the morbidity and mortality of a bacterial infection. From the perspective of evolutionary biology, however, “misguided” is hardly an explanation. Colonization by bacteria is not a rare event but rather something mammals confront all the time, and overresponding in a way that results in their morbidity and mortality would almost certainly be selected against. In their review of “immunopathogy,” Graham and colleagues postulated a number of reasons for this transgression of the immune response (Graham et al., 2005). Here we offer our perspective on this issue.

As we see it, there are two general classes of explanations for the maintenance of an overresponse of the immune system. (i) While infectious disease may be a major source of morbidity and mortality (Haldane, 1949), disease-mediated selection can be relatively weak, and extensive amounts of time would be required to evolve mechanisms to modulate the immune response to specific bacterial infections. (ii) Functional constraints on the immune system limit the ability of natural selection to totally prevent and maybe even partially mitigate an immune overresponse to bacterial infections.

(a) Even if selection universally favors tempering the immune overresponse to infections, and the favored genotypes could be generated (which we question below [b]), the time required for temperance to evolve could be considerable, especially if the overresponse is specific for particular bacteria and/or their products. This is due to two factors. (a) At its maximum the intensity of selection for modulating the immune overresponse to an infection would equal the fraction of the population with that infection. It would be substantially lower if the symptoms of the infection were not expressed in all colonized hosts, were rarely lethal or sterilizing, or were primarily manifest after reproductive years or if the magnitude of the reduction of the overresponse of the favored genotype was less than absolute. For most of the diseases listed in Table 3-1 virulence is a rare occurrence in colonized hosts (less than 1%), and therefore the intensity of selection against an immune overresponse would be relatively weak. (b) It can take a considerable amount of time for a rare beneficial mutant to ascend to substantial frequencies. For example, if the selection for a reduced overresponse is operating on genotypes at a single locus (the best case), the initial frequency of a favored allele is 10⁻³, the favored genotype has a 1% selective advantage, and there is no dominance, it would take 1,381 generations (more than 20,000 years for humans) for that gene to reach a gene frequency of 50%. If the favored genotype is recessive, the corresponding number of generations would be 100,491 (Crow and Kimura, 1971).

What about the role of the bacteria in the evolution of a more temperate immune system? As a consequence of their vastly shorter generation times, haploid genomes, and propensity to receive genes and pathogenicity islands by horizontal transfer, it seems reasonable to assume that bacteria would have an edge in an evolutionary arms race with their mammal hosts. We suggest, however, that this edge contributes little if anything to the slowing pace at which mammalian evolution could modulate the immune overresponse. Although there maybe situations where virulence is positively correlated with the infectious transmission of bacteria, in most of these cases the morbidity and mortality associated with their transmission is not to the bacteria’s advantage and may be to their disadvantage. Even greater transmission of these bacteria would be possible if the hosts were not debilitated or killed as a result of diarrhea or if the bacteremias required for vector-borne transmission did not result in sepsis. In this interpretation evolution in the bacteria population would not oppose the evolution of a more temperate host immune system. Of all the examples considered in this chapter, the only one in which evolution in the bacterial population might favor an immune over-response is Carniel’s suggestion that by killing their host, Y. pestis acquires a transmission advantage.

(ii) While the above realities of the ecology and genetics of natural selection may be part of the answer to the question of why evolution has not eliminated the immune system’s overresponse to bacterial (and other) infections, we suggest it is not the most important reason. We conjecture that the primary reason mammalian evolution has not tempered and perhaps cannot temper the immune overresponse to bacterial and other infections is functional constraints that limit the extent to which the immune system can be modified. The immune system has roles other than clearing bacterial infections. It has been postulated that these other roles dominated the evolution of the mammalian immune system (Burnet, 1970). These different roles as well as the extraordinary diversity of organisms colonizing mammals, bacteria, viruses, fungi, and worms of various ilks and the variety of sites of colonization impose different and potentially conflicting demands on the immune defenses, phenomena referred to as antagonistic pleiotropy. An appealing hypothesis for the immunopathogy known as allergies is an overresponse of those elements of the immune system that in less-pristine times would otherwise be occupied with the control of helminth infections (Wilson and Maizels, 2004).

There is a fine line between responding (1–6 in Figure 3-8) and overresponding (7 in that figure), which may be difficult for the systems regulating the immune response to perceive, much less avoid. As suggested by Frank (Andre et al., 2004), the intensity of an immune response may be determined by a tradeoff between increasing the strength and rapidity of an immune defense and the virulence from an immune system overresponse.

Is there evidence in support of these two hypotheses for why evolution has not eliminated the virulence resulting from the immune overresponse? Not much—at least not yet. We suggest, however, that some of the considerable amount of inherited variability in the susceptibility to infectious disease in human populations (Bellamy and Hill, 1998; Bellamy et al., 2000; Segal and Hill, 2003; Sorensen et al., 1988) can be interpreted as support for these hypotheses. To be sure, there is good and even overwhelming evidence that some of this variation is maintained by disease-mediated balancing or frequency-dependent selection, but this is not the case for all or even the majority of it. We suggest that much of the standing genetic variation in disease susceptibility in human populations is a reflection of the myopia and limitations of natural selection: (i) the relative weakness of selection for modulating the immune overresponse and (ii) even more, the impotency of natural selection due to the constraints on the immune system—antagonistic pleiotropy. Genetic variation that is not or is poorly perceived by natural selection will build up and persist (Crow and Kimura, 1971).

Implications

While the morbidity and mortality of most bacterial infections can be attributed to an immune overresponse, virtually all of our efforts to treat these infections are directed at controlling the proliferation and clearing the bacteria, primarily with antibiotics. This approach has been and continues to be effective, but not completely so. Antibiotic treatment commonly fails, and patients die or remain ill for extended periods. Resistance of the pathogen to the antibiotics employed for treatment is only one of the reasons for this failure and for some infections is not the major one, at least not yet (Levin and Rozen, 2006; Yu et al., 2003).

The obvious alternative approach to treating infections is to reduce the morbidity and prevent the mortality by modulating the immune system’s over-response. There have been attempts to do just that for the treatment of bacteria-mediated sepsis. Clinical trials have evaluated the use of glucocorticoids (Bone et al., 1987), drugs designed to neutralize endotoxins (Ziegler et al., 1991), tumor necrosis factor α (Fisher et al., 1996), and IL-1β (Fisher et al., 1994), but none of these treatments was effective. The most successful trials in humans to date have been with a component of the natural anticoagulant system, activated protein C, which has substantial anti-inflammatory properties along with being a potent anticoagulant (reduces the formation of clots that are responsible for organ failure in late stages of sepsis) (Fourrier, 2004). In addition, new agents redirect the immune response and hold promise as effective future therapies for sepsis, such as IL-12 (O’Suilleabhain et al., 1996) and antibodies against complement (C5a) (Czermak et al., 1999). However, understanding the specifics of the immune overreaction and the intricacies of the feedback mechanisms that control an immune response is necessary for therapies to be directed at enhancing or inhibiting the patient’s immune response.

At this time, taken at large, the success of these immune modulating methods in preventing the morbidity and mortality of bacterial infections can at the very best be described as modest. However, in maintaining the speculative nature of this rant, and desiring an optimistic conclusion, we suggest that as we learn more about the regulation of the immune response and develop procedures to monitor as well as administer regulatory immune molecules in real time, these methods will become increasingly effective for the treatment of bacterial infection.

Acknowledgments

We thank Elisabeth Carniel for sharing her ideas about the evolution of the virulence of Y. pestis. We are grateful to Jim Bull and Harris Fienberg for insightful comments and suggestions. B. R. L. acknowledges his continuous gratitude to Fernando Baquero, for inspiration, ideas, never-ending whimsy, support, and friendship. This endeavor was supported by a grant from the NIH, AI40662 (B. R. L.), and an NIH Training Grant (E. M.).