Ecological Opportunity, Evolution, and the Emergence of Flea-Borne Plague

Abstract

The plague bacillus Yersinia pestis is unique among the pathogenic Enterobacteriaceae in utilizing an arthropod-borne transmission route. Transmission by fleabite is a recent evolutionary adaptation that followed the divergence of Y. pestis from the closely related food- and waterborne enteric pathogen Yersinia pseudotuberculosis. A combination of population genetics, comparative genomics, and investigations of Yersinia-flea interactions have disclosed the important steps in the evolution and emergence of Y. pestis as a flea-borne pathogen. Only a few genetic changes, representing both gene gain by lateral transfer and gene loss by loss-of-function mutation (pseudogenization), were fundamental to this process. The emergence of Y. pestis fits evolutionary theories that emphasize ecological opportunity in adaptive diversification and rapid emergence of new species.

INTRODUCTION

New infectious diseases have arisen periodically throughout history and remain an ongoing threat to human welfare. In some cases, newly emergent diseases have produced demographic and societal catastrophes with far-reaching effects on human history (1, 2). Perhaps the most famous example of this is bubonic plague, which profoundly affected Western civilization and is a subject with as much fascination for social scientists as for biological scientists. Although associated with antiquity in terms of recorded human history, plague is a recent phenomenon in evolutionary terms. Yersinia pestis, the bacterial agent of plague, is essentially an aberrant clonal derivative of Yersinia pseudotuberculosis. The two are so closely related genetically that Y. pestis is estimated to have emerged only within the last 3,000 to 6,000 years (3–6). Y. pseudotuberculosis causes relatively benign and self-limited enteric infections transmitted via the fecal-oral route and is a generalist, associated with a wide variety of animals and environmental sources. Its Y. pestis offshoot, in stark contrast, causes systemic, life-threatening disease, is primarily arthropod borne, and is a specialist that now depends on rodent-flea transmission cycles. The genetic tracks leading to these radical changes are relatively fresh and discrete and offer a compelling detective story about how new bacterial pathogens can emerge.

BACTERIA AS MODEL SYSTEMS TO STUDY ADAPTIVE EVOLUTION

Not long ago, microbiology was called the least evolution oriented of the biological sciences (7). Newly developed experimental approaches have completely changed that. The genomics revolution, in particular the availability of rapid and inexpensive whole-genome-sequencing methods, has been a major impetus, along with freely available software to deduce phylogenetic relationships and to apply complex evolutionary models to genomic data. These tools are being applied at several levels and in several different model systems. In addition, bacterial “evolution in action” experiments in which adaptive changes are tracked during long-term serial passage of an original clone are now providing empirical tests of evolutionary theory and insights into the molecular basis of adaptation (8–16).

Metagenomic analyses have disclosed phylogenetic groupings, evolutionary lineages, and geographic dispersal patterns of several pathogens over various time scales and allowed correlations between clonal genotype and host adaptation to be made (17–19). For example, the evolutionary record of Y. pestis has been documented in population genetic studies of a comprehensive sample of 148 strains collected from around the world. These studies have established the time and place of the emergence of Y. pestis in central Asia, a phylogenetic tree, and phylogeographic reconstruction of its spread throughout the world after it emerged (3–6, 20–23). An evolutionary history at this level of detail is unusual for a prokaryote but possible for Y. pestis because it is a young, genetically monomorphic clonal species with a still-extant recent common ancestor (20). The new discipline of paleogenomics has also contributed. Plague victims die as a result of severe bacteremia, and Y. pestis DNA has been PCR amplified on several occasions from highly vascularized and environmentally protected skeletal tissue samples, such as dental pulp and bone collected from ancient gravesites. Detailed genomic characterization of this ancient DNA, including whole-genome sequence reconstructions, combined with archaeological data and radiocarbon dating, have added Y. pestis strains from the 6th century Plague of Justinian and the 14th century Black Death pandemics to the phylogenetic tree (22–25). This genomic fossil record has recently been extended even further to include ancestral Y. pestis strains circulating during the Bronze Age, 4,000 to 5,000 years ago (6).

UNDERSTANDING PATHOGEN EMERGENCE VIA A SYNTHESIS OF EVOLUTIONARY BIOLOGY AND MOLECULAR PATHOGENESIS

Of course, the ultimate goal of evolutionary studies is not simply to catalog genetic changes through time in order to construct phylogenetic trees. The more challenging and elusive goals are to identify those specific genetic changes responsible for adaptive phenotypes and to determine the mechanisms by which they led to increased fitness. Inferences about these aspects of evolutionary history have for the most part been limited to associations between genetic variation and phenotype. Bioinformatics programs can identify mutations with statistical signatures of having been positively selected (26, 27), and the putative function of genes can suggest how mutations in them were adaptive. Such predictions, however plausible, remain speculative until the consequences of candidate adaptive mutations are adequately characterized.

The same molecular biology tools developed for bacterial pathogenesis studies provide the experimental means to test adaptive evolution hypotheses directly. Directed mutagenesis to eliminate gene function followed by virulence tests of the isogenic parent and mutant strains is fundamental to pathogenesis research (28). In like manner, allelic-exchange or site-specific mutagenesis can be used to compare ancestral, intermediate, and evolved alleles to test hypotheses about the evolution of virulence. Allele-specific effects on function (e.g., catalytic activity or binding affinity) can also be evaluated by biochemical and structural biology assays and related to in vivo fitness. Thus, sophisticated genomic analyses to identify candidate adaptive genetic changes and molecular pathogenesis methods to evaluate the functional significance of these changes and their effect on in vivo fitness can be used to deduce the genetic and molecular basis of adaptive evolution (29). While this can still be a tall order, the molecular biology tools and model systems are available for many bacteria.

GENE GAIN AND GENE LOSS ARE BOTH MAJOR DRIVERS OF BACTERIAL EVOLUTION

Evolution can result from the gradual accumulation of many genetic changes of small effect or by a few key genetic changes of large effect (29). Bacterial evolution is frequently characterized by the latter, often due to the very high rate of lateral gene transfer among prokaryotes. Mobile genetic elements, including plasmids, bacteriophage, and genomic islands, are constantly being exchanged among bacteria throughout the environment, in what Joshua Lederberg once referred to as the original world wide web. The impact of lateral gene transfer on bacterial evolution is exemplified by the emergence of “superbugs” that have acquired antibiotic resistance and toxin genes on plasmids or phage-related chromosomal islands (30, 31). The pathogenicity of Salmonella enterica is a consequence of many virulence genes that were introduced by lateral transfer and incorporated into chromosomal pathogenicity islands (32).

The evolutionary significance of gene gain by lateral transfer has long been recognized. In contrast, the contribution of loss-of-function mutation (pseudogenization) to adaptation has only recently been appreciated. It was initially assumed that pseudogenes would be relatively rare in bacteria because their genomes are small and gene dense. However, the recent ability to compare the sequences of orthologous genes within a clade has revealed that pseudogenes are a common feature of bacterial genomes (33). Most are due to nonsense point mutations or frame-shifting small deletions or insertions (indels) that result in truncation, to transposon insertion, or to DNA rearrangements, which in some cases can delete entire genes or operons (33, 34). Pseudogenes are particularly abundant in the genomes of recently emerged bacterial pathogens. For example, 337 pseudogenes were identified in Y. pestis strain CO92, representing about 8% of the total gene content (33).

Pseudogenes are often regarded as exemplars of neutral evolution, the passive loss of functions that are no longer needed (35). Many metabolic genes of Y. pseudotuberculosis are pseudogenes in Y. pestis, presumably reflecting the fact that Y. pestis alternates between two eukaryotic hosts and no longer needs the metabolic versatility to survive in the environment. However, if a functional gene is somehow detrimental in a new environment (a phenomenon generally referred to as antagonistic pleiotropy), its mutational loss can be positively selected. In experimental evolution studies, many mutations that eliminate or modify gene function appear to have been beneficial and fixed in the population by positive selection (8, 9, 11, 14, 16). Often these adaptive null mutations reconfigure metabolic pathways to better fit a new environment (14). Mutational alteration of central regulatory genes can have especially large fitness effects (11). Thus, a variety of simple mutations that with rare exceptions result in loss rather than gain of function can provide substantial adaptive potential (9, 14).

GENETIC AND MOLECULAR BASIS OF EVOLUTION TO A FLEA-BORNE ROUTE OF TRANSMISSION

Evolved and ancestral flea infection phenotypes.

The primary role of the flea as the vector of plague and the manner in which fleas transmit Y. pestis were described between 1898 and 1915 during the last plague pandemic (reviewed in reference 36). Unlike many arthropod-borne pathogens, which after being taken up in a blood meal disseminate to the salivary glands and are transmitted in vector saliva, Y. pestis remains confined to the flea digestive tract and is transmitted by regurgitation. This can occur the first time a flea feeds again in the week following an infectious blood meal and again after the bacteria form a biofilm in the proventriculus, a valve in the flea foregut that connects the esophagus and the midgut. As the biofilm grows and thickens, it increasingly fills the lumen of the proventriculus, which impedes and eventually completely blocks blood flow into the midgut when the infected flea feeds. Clogging up the proventricular valve greatly enhances regurgitative transmission of bacteria from the foregut into the bite site (36–39).

The flea infection phenotype of Y. pseudotuberculosis is markedly different from that of Y. pestis. Y. pseudotuberculosis causes an acute toxic response in fleas immediately after they ingest it, characterized by rapid elimination of the blood meal, prostration, and significant mortality within 24 h (40). However, many fleas that recover from the initial toxicity remain chronically infected with Y. pseudotuberculosis, with no further signs of toxicity. In these fleas, the infection is restricted primarily to the flea hindgut. The midgut is only transiently infected, the proventriculus never is, and there is no evidence of biofilm formation in the flea digestive tract (41, 42). Consistent with this picture and with earlier reports, no transmission of Y. pseudotuberculosis by infected fleas has ever been detected (42, 43). In contrast, blood meals containing Y. pestis are completely nontoxic to fleas—no initial morbidity or mortality is seen, just as in fleas that feed on sterile blood. Y. pestis never colonizes the flea hindgut but, instead, grows in the form of large multicellular biofilmlike aggregates in the midgut and proventriculus. Efficient transmission via fleabite occurs throughout the month following the infectious blood meal (42).

Acquisition of two new plasmids by Y. pestis and their effects on flea infection.

What genetic changes were behind the recently acquired ability of Y. pestis to produce a transmissible infection in the flea? To begin to answer this question, a primary strategy has been to compare the core genomes of the two species to catalog all genes either gained or lost by Y. pestis relative to the genome of Y. pseudotuberculosis. The most obvious instances of gene gain were two plasmids unique to Y. pestis that were acquired by lateral transfer from an extraneous source sometime after divergence from the Y. pseudotuberculosis progenitor. Flea infections with plasmid-cured Y. pestis strains revealed that one of these, pPCP1, which encodes a surface protease/plasminogen activator (Pla), was not required to produce a normal transmissible infection in the flea. Neither was the 70-kb virulence plasmid common to both species, which encodes a type III secretion system that is required for virulence in mammals. However, a strain lacking the other Y. pestis-specific plasmid, pMT1, was highly attenuated in its ability to colonize the flea digestive tract and produce proventricular blockage (44). This phenotype was entirely attributable to the absence of phospholipase D (PLD) activity encoded by the ymt gene of pMT1 (45). Deletion of this single gene or site-specific mutation of its PLD catalytic domain rendered Y. pestis unable to colonize the flea midgut because the bacteria rapidly lysed after ingestion. Conversely, transformation of Y. pseudotuberculosis, or even Escherichia coli, with ymt enabled the bacteria to grow to high numbers in the flea midgut and produce a chronic infection (45). The PLD activity appears to protect Gram-negative bacteria from a bacteriolytic agent that is generated in the flea midgut during digestion of the blood meal, but the biochemical details remain to be identified. Other than the gene gain represented by the two Y. pestis-specific plasmids, only eight chromosomal loci have been identified in Y. pestis that are not present in Y. pseudotuberculosis genomes, and single deletion of each of these loci had no effect on the normal flea infection phenotype (46).

Gene losses greatly enhanced biofilm formation in the flea and transmissibility.

Unlike the modest number of new genes acquired by Y. pestis since its divergence, there are hundreds of examples of gene loss from mutation and deletion (33). Evolutionary theory predicts that many of these mutations would not have been adaptive but selectively neutral and present in the population by chance because of random genetic drift (12). Genetic drift effects are particularly strong following a large reduction in effective population size associated with the transition from a free-living to a host-dependent lifestyle (47–49). This increases the difficulty of determining which of the hundreds of gene loss events in Y. pestis might have been adaptive. Most of the genetic variation among Y. pestis strains consists of different pseudogene profiles, typical of a clonal species in which little recombination occurs among strains (50, 51). Therefore, as the genome sequences of more and more strains of Y. pseudotuberculosis and Y. pestis become available, it should be possible to winnow down the list of candidate gene losses that were positively selected by concentrating on orthologs pseudogenized or absent in all Y. pestis isolates but still functional in all Y. pseudotuberculosis isolates.

Because of the central importance of the bacterial biofilm phenotype to the flea-borne transmission mechanism, genetic pathways related to outer surface characteristics and biofilm development were of obvious interest. Although Y. pseudotuberculosis does not form a biofilm in the flea, it does so under other environmental conditions (41), and the operon responsible for the synthesis and export of the extracellular polysaccharide matrix required for biofilm formation is identical to that of Y. pestis. The regulatory networks leading to production of the extracellular matrix and biofilm development are complex and differ among bacteria, but genomic comparisons of the Yersinia species identified two regulatory pathways of particular interest. One was the Rcs signal transduction system of the Enterobacteriaceae, which controls capsular polysaccharide synthesis and regulates genes with roles in biofilm synthesis (52). Screening transposon mutants for aberrant in vitro production of the extracellular polysaccharide matrix indicated that the Yersinia Rcs system acts to repress biofilm formation and that one of the five rcs genes (rcsA) is a pseudogene in Y. pestis but not in Y. pseudotuberculosis (53).

A second highly conserved system that regulates biofilm development in Gram-negative bacteria involves signaling via the bacterial second messenger cyclic di-GMP (c-di-GMP) (54, 55). A complex and diverse variety of pathways that cause a shift from planktonic to biofilm growth are induced when the intracellular concentration of c-di-GMP increases. Cyclic-di-GMP is itself under complex regulatory control. Most Gram-negative bacteria encode several GGDEF-domain diguanylate cyclase (DGC) enzymes that synthesize it and EAL- or HD-GYP-domain phosphodiesterase (PDE) enzymes that degrade it. Differential levels of production of these two enzyme families in response to environmental cues control the c-di-GMP concentration and, in turn, biofilm development (54, 55). Comprehensive bioinformatics and biochemical analyses determined that Y. pseudotuberculosis contains three functional PDE genes but that two of these are pseudogenes in Y. pestis (56, 57).

The intact Rcs system of Y. pseudotuberculosis represses the transcription of hmsT and hmsD, the two genes that encode c-di-GMP-synthesizing DGCs, and of hmsHFRS, the biofilm extracellular matrix operon (58–60). Thus, the original function of all three pseudogenes (rcsA and the two PDE genes that function to degrade c-di-GMP) involve pathways predicted to suppress biofilm development, suggestive that their loss would have been beneficial to the biofilm-dependent mechanism by which fleas transmit Y. pestis. Support for this hypothesis came from experiments showing that allelic exchange replacement of each of the three pseudogenes with the functional ortholog from Y. pseudotuberculosis resulted in attenuation of the Y. pestis in vitro biofilm phenotype (42). However, biofilm formation is highly dependent on environmental conditions, and biofilm regulation is different in the flea digestive tract than in vitro (57). To avoid the risk of “adaptionist storytelling” (61), therefore, it was essential to evaluate the fitness effects of these genetic changes in the flea. To do this, fleas were infected with standard doses of wild-type or defined-mutant strains of Y. pestis and Y. pseudotuberculosis and the infection phenotype was monitored. Individually restoring functionality to each of the three pseudogenes did not affect the ability of Y. pestis to colonize the flea digestive tract but did significantly reduce the rate of proventricular blockage (42, 53). Even more striking was the fact that just four discrete genetic changes to the Y. pseudotuberculosis strain IP32953—a loss-of-function mutation of rcsA and the two PDE genes plus the addition of the ymt gene, to match the Y. pestis genotype—were sufficient to confer full flea-borne transmissibility in fleas that survived the initial acute toxicity (Fig. 1) (42, 53). In essence, the four changes allowed Y. pseudotuberculosis to move up the digestive tract from the hindgut to the midgut (gain of ymt) and then to stably colonize and occlude the proventricular valve in the foregut (loss of rcsA and two PDE genes), the staging area for regurgitative transmission.

FIG 1.

How to become an arthropod-borne pathogen in a few easy steps: reconstruction of the evolutionary path leading to the emergence of flea-borne plague. The sequence of important genetic changes as they appear in the phylogenetic tree and their incremental effect on transmission and disease are depicted. Black arrows indicate transmission of bacteria between mammal, flea, and the external environment, and their weights indicate ecological importance; dotted arrows indicate occasional transmission; green arrows indicate gene gains; and red arrows indicate gene losses. The R₀ value indicates the probability that a flea feeding on a bacteremic host will subsequently transmit Y. pestis to at least one naive host. Gain of ymt and loss of ureD, rcsA, and the two PDE genes are sufficient to enable Y. pseudotuberculosis IP32953 to produce a transmissible infection in fleas (42, 65). In addition to gain of pla, other genetic changes were required for increased invasiveness and virulence in the mammal. Extensive gene loss (pseudogene accumulation) during the emergence of Y. pestis has also led to reduced fitness in the external environment. (Adapted from reference 42 with permission of the publisher.)

Further genetic changes appear to have reinforced the proventriculus-blocking phenotype (42).The extracellular matrix of Yersinia biofilms is composed of poly-β-1,6-linked N-acetylglucosamine (62, 63). Y. pseudotuberculosis encodes a β-N-acetyl glucosaminidase (NghA), but the corresponding gene is nonfunctional in Y. pestis due to a small frameshifting insertion. NghA inhibits Yersinia biofilm formation in vitro, and transformation of Y. pestis with nghA inhibits proventricular biofilm formation in fleas, although this effect requires nghA in high copy numbers (63). Loss of nghA could also have been selectively favored because it resulted in a structurally more stable and coherent biofilm in the flea.

Elimination of flea toxicity by a single gene loss.

The inherent oral toxicity of Y. pseudotuberculosis to fleas represented a separate impediment in the path to the new transmission route, because it would eliminate a significant percentage of potential vectors before they ever had a chance to transmit. More than 20 putative insecticidal toxin genes are annotated in the Yersinia genomes, and suspicion initially fell on those that differed between the two species. For example, a pathogenicity island-encoded operon containing orthologs of the Tc family of potent insect toxins showed evidence of gene loss, pseudogenization, and disruption by transposon insertion in Y. pestis but not in Y. pseudotuberculosis (64). However, deletion of Tc genes or other putative insecticidal toxin genes did not affect the toxicity of Y. pseudotuberculosis to fleas (40). A different investigational strategy revealed that the toxic factor was present only in the membrane fraction of Y. pseudotuberculosis cell lysates and highly enriched in a soluble subfraction of membrane-associated proteins (65). Comparative proteomic analyses revealed that several components of the urease enzyme were present in this subfraction in Y. pseudotuberculosis but not in Y. pestis. Unlike Y. pseudotuberculosis, Y. pestis is urease negative due to a frameshift nonsense mutation in ureD, one of the seven genes required to produce a functional urease enzyme (66). Surprisingly, deletion of the Y. pseudotuberculosis urease operon, or just ureD, resulted in complete loss of toxicity to fleas. Conversely, restoration of ureD was sufficient to make Y. pestis toxic (65). The urease-dependent toxicity is apparently due to activation of the enzyme in the low-pH environment of the flea gut and the release of cytotoxic levels of ammonia from degradation of the urea present in the blood meal. Loss of urease activity was likely subject to positive selection because increasing the pool of potential vectors helped to maintain stable transmission cycles (65).

RETRACING THE EVOLUTIONARY HISTORY OF Y. PESTIS

Piecing together clues from paleogenomics, population genetics, phylogenetics, and molecular pathogenesis studies, it is possible to deduce the major steps along the evolutionary path from Y. pseudotuberculosis to Y. pestis and at what points they occurred (Table 1; Fig. 1). The earliest Y. pestis strains for which genome sequences are available are from 4,000- to 5,000-year-old Bronze Age human skeletal material from Eurasia (6). The six most ancient of these strains contained sequences of the two plasmids acquired by Y. pestis after its divergence. One (pPCP1) was essentially identical to that of modern Y. pestis, but the pMT1 plasmid lacked a 19-kb segment containing ymt (6). The missing 19-kb segment also contained the plasmid partition locus parABS (67), suggesting that pMT1 was not replicating autonomously in these ancient strains but had integrated into the chromosome. Conceivably, part of pMT1 was originally acquired in the form of a chromosomal pathogenicity island that later integrated with an incoming small ymt-encoding plasmid. Chromosomal integration of pMT1, or portions of it, occurs at a high frequency even in modern Y. pestis strains (68, 69). Thus, another possibility is that the Bronze Age strains originally harbored a complete pMT1 but the segment containing ymt and the par locus was deleted during or after recombination into the chromosome and not maintained.

TABLE 1.

Adaptive gene changes that enabled transmission of Y. pestis by fleas

Flea-related phenotype of:		Genetic change	Effecta
Y. pseudotuberculosis (ancestral)	Y. pestis (adapted)
Acute oral toxicity	Nontoxic	Loss of ureD	Eliminated urease activity responsible for ammonia cytotoxicity to the flea digestive tract
Hindgut-only infection; no transmission	Midgut and foregut infection	(i) Gain of ymt	Ymt phospholipase D activity allowed survival in the flea midgut, enabling foregut infection and rudimentary transmissibility by flea bite
		(ii) Unknown	Loss of hindgut-specific adhesin?
No biofilm development in the flea	Biofilm development in foregut	(i) Loss of rcsA	Derepression of two DGC genes and the hmsHFRS biofilm matrix operon; increased c-di-GMP, enhanced biofilm formation and transmissibility
		(ii) Loss of PDE2 gene	Eliminated PDE activity; increased c-di-GMP, enhanced biofilm formation and transmissibility
		(iii) Loss of PDE3 geneb	Eliminated PDE activity; increased c-di-GMP, enhanced biofilm formation and transmissibility
		(iv) Loss of nghA	Eliminated glycosyl hydrolase activity that degrades the extracellular matrix of the biofilm
Dissemination from intradermal inoculation site halted in the draining lymph node	Systemic spread from intradermal inoculation or flea bite site	Gain of pla	Enabled bacterial replication and survival in the draining lymph node required for systemic infection

^aPDE, phosphodiesterase enzyme; DGC, diguanylate cyclase enzyme.

^bLoss of PDE3 function occurred in two discrete steps: a promoter mutation (present in some Y. pseudotuberculosis and all Y. pestis strains) that reduced transcription and a point mutation that introduced a premature stop codon (present in all Y. pestis strains except for some basal Pestoides group strains) (42).

Y. pseudotuberculosis infects the intestinal tract of many different animals following ingestion of contaminated food and water. The infection is usually relatively benign, spreading no farther than the mesenteric lymph nodes before it is resolved. The type of disease associated with the Bronze Age Y. pestis is uncertain, but the terminal stage must have been sepsis, given the facile detection from vascularized peripheral tissue. Since these strains had acquired the virulence factor Pla, required for bacterial survival and extensive multiplication in and subsequent dissemination from lymph nodes (70), the disease may have been a severe, disseminated form of food- and waterborne yersiniosis. Alternatively, the acquisition of pla would also have enabled pneumonic plague and direct human-to-human transmission via the aerosol route (71, 72). It is unlikely that these strains were maintained primarily by flea-borne transmission cycles. Besides lacking ymt, these strains contained functional, Y. pseudotuberculosis-type alleles of ureD, rcsA, and the two PDE genes that are pseudogenes in modern Y. pestis. However, some transmission by fleabite might have occurred, because the early transmission potential that fleas have immediately after an infectious blood meal does not require ymt (73) or biofilm production (74). Modification of Pla function by an I259T substitution in modern pandemic Y. pestis strains has been hypothesized to have been important or even required for bubonic plague pathogenesis (6, 72, 75). However, some present-day Y. pestis strains still have the ancestral version of Pla and cause bubonic plague from a subcutaneous injection route, at least in rodents.

ECOLOGICAL CONTEXT OF THE EMERGENCE OF FLEA-BORNE PLAGUE

Genetic changes supply the raw material for selection, but the emergence and establishment of any new species or infectious agent always depends on a particular set of ecological conditions. Similar to the invasion of an alien plant or animal species into a new geographic area, two ecological issues are involved: how the invader got there, and how it was then able to successfully establish itself (76). With respect to the insect host, the first question is easy for the Y. pestis ancestor. Y. pseudotuberculosis usually does not disseminate beyond the mesenteric lymph nodes but will spread into the bloodstream opportunistically. When this occurs in rodent hosts, which typically support a permanent ectoparasitic community of fleas, the bacteria will be ingested in flea blood meals. This would have occurred with regularity after the dissemination factor Pla was acquired, as evidenced by the septicemia associated with the Bronze Age yersiniosis. At these first stages, infection of the flea would have been confined to the hindgut, which Y. pseudotuberculosis can stably colonize, but from which transmission can occur only to the environment via flea feces (42).

The emergence of plague as a vector-borne disease from this primordial Yersinia-flea interface can be interpreted retrospectively in the light of general and epidemiologic evolutionary theory. The process of adaptive radiation, the rapid emergence of new species, is widely believed to be initiated by ecological opportunity. Such opportunities often arise following the evolution of a “key innovation” that enables the colonization of a previously unoccupied environmental niche. The selective pressures acting on the population in the new environment are different, leading to adaption in order to exploit the new resources (77, 78). In the history of Y. pestis, the acquisition of ymt could be considered a key innovation. The transformation of Y. pseudotuberculosis with this single gene has a profound effect, enabling not only high-density colonization of the midgut and shedding in flea feces but also some localization to the foregut and modest transmissibility by flea bite (42). A slightly different viewpoint is presented by the “ecological fitting” theory (79). Rather than a de novo key innovation, this theory proposes that some preexisting genotype in the population may serendipitously already be fit and able to survive in a newly encountered environment. This is closely related to the exaptation concept of Gould and Vrba, defined as an existing trait that is co-opted for a new and different function (80). The emphasis here is that Mother Nature did not invent Ymt so that a Gram-negative enteric bacterium could survive in the midgut of a flea. Ymt is a member of a large family of PLD enzymes found in all kingdoms of life that presumably serve a variety of utilitarian functions (81). Interestingly, Y. pestis Ymt is most similar to PLDs of Photorhabdus and Arsenophonus, two other members of the Enterobacteriaceae family that infect insects and soil invertebrates. Y. pseudotuberculosis is able to infect the same nematode hosts as Photorhabdus (82), and horizontal transfer of ymt may originally have provided some benefit to Y. pseudotuberculosis in encounters with soil invertebrates before later proving useful in adapting to the flea vector.

The gain of ymt and pla by horizontal transfer was essential for transmission from the flea and subsequent dissemination from the flea bite site to the bloodstream, respectively. Thus, even at this stage, flea-rodent-flea transmission was possible but, evidently, not sufficiently efficient to permit persistent, ecologically stable transmission cycles. The key concept relating to the persistence and spread of a pathogen is its basic reproductive number, R₀, the average number of secondary infections that stem from a single infected individual (76, 83). If R₀ is <1, transmission chains stutter and die out. This is the stage exemplified by certain avian influenza viruses that occasionally infect humans but whose subsequent human-to-human spread is very inefficient. Epidemiologic theory warns, however, that following an initial mutational innovation that potentiates some level of transmissibility (R₀ is <1 but >0), the probability of further adaptive mutations that increase R₀ is high (84, 85). For Y. pestis, this occurred via the sequential loss of genes that function to repress biofilm formation in the flea, culminating in greatly increased transmissibility and permanent flea-mammal cycles (42).

Efficient transmission from a flea to a mammal represents only half of the Y. pestis life cycle. It is also necessary to produce a transmissible infection in the mammal. The ability to produce a high-density bacteremia, invariably leading to septic shock and high mortality, is required for this, because the 50% infective dose of Y. pestis for fleas is high and fleas take small blood meals (86). Thus, evolving a hypervirulent phenotype in the mammal was also fundamental to the transition to the flea-borne transmission route. Here also, Y. pseudotuberculosis did not have very far to go. Y. pestis inherited most of its essential virulence factors, most notably the virulence plasmid, from Y. pseudotuberculosis. By intravenous injection routes, Y. pseudotuberculosis is as virulent as Y. pestis (87, 88). The major new requirement for bubonic plague pathogenesis was the ability to overcome containment and elimination in the draining lymph nodes following fleabite (intradermal) transmission. The acquisition of pla was essential for this. Pla is not required for flea-borne transmission per se, but subsequent disease progression to the bacteremic stage required for mammal-to-flea transmission depends on it. Pla is necessary but not sufficient for systemic dissemination—gene losses, such as epistatic mutations of O-antigen genes and other genetic changes yet to be identified, likely contributed (70, 89). Negative regulators of pathogen hypervirulence (antivirulence genes) whose loss contributes to pathogen emergence are being increasingly identified (90, 91). For example, the E. coli lysine decarboxylase gene cadA has been mutated or deleted from all descendant Shigella and enteroinvasive E. coli strains, and the loss of this gene, in addition to the acquisition of a virulence plasmid, was essential for increased invasiveness and virulence in humans (90). The genotype of the Bronze Age Y. pestis intermediate strains and the pathology associated with them suggest that the evolution of hypervirulence predated the evolution of the flea-borne transmission route (Fig. 1). If so, adopting an arthropod-borne transmission route presumably increased ecological stability, because the Bronze Age genotype is now apparently extinct.

EVOLUTIONARY THEMES ILLUSTRATED BY THE EMERGENCE OF FLEA-BORNE PLAGUE

The transition of Y. pestis to an arthropod-borne transmission route provides a real-life illustration of several principles of bacterial adaptive evolution and the emergence of novel pathogens. (i) Gene loss is an important engine of evolutionary change. Null mutations from simple insertions or deletions occur frequently over a large genetic space and can be explored rapidly by an evolving population (9, 14, 16, 33). (ii) Loss of genes can be positively selected because their absence is beneficial, not neutral (8, 9, 11, 14, 16). (iii) Gene gain by genetic exchange among unrelated bacteria can represent key innovations that open up new ecological niches and instigate adaptive diversification. (iv) Genetic changes that alter gene regulation or that reconfigure metabolic pathways can have large adaptive effects (11, 13, 14, 92). For the evolution of flea-borne transmission, the primary target was c-di-GMP metabolism to upregulate the preexisting biofilm-forming ability of Y. pseudotuberculosis—gene loss led to gain of a phenotype. This is reminiscent of a seminal laboratory evolution experiment in which a divergent biofilm-forming Pseudomonas niche specialist emerged via loss of genes that modulate c-di-GMP (93, 94). (v) Adaptation can proceed by a few mutations of large effect. (vi) Disease emergence also may require only a few discrete genetic changes that can accrue in a gradual, additive fashion in which sequential changes confer incremental increases in R₀ (84, 85). (vii) Adaptive diversification typically results in an ecological change from a generalist to a specialist species (13, 77, 92). Once a stable flea-mammal life cycle was established, mutations in genes with functions that are beneficial for survival in the external environment but detrimental or superfluous in the insect or mammal could accumulate and eventually lead to host dependency.

The story of plague is a sobering example of how, under the right circumstances, just a few simple genetic changes separate a relatively benign pathogen from one responsible for the worst pandemics in human history. The periodic emergence of new infectious diseases reminds us that evolution is continuous and its outcomes often unpredictable.

Funding Statement

This work was funded by NIH, NIAID (1 ZIA AI000796-19; B.J.H.), and by the National Basic Research Program of China (973 no. 2015CB554200; Y.-C.S.).