MOLECULAR MECHANISMS CONTRIBUTING TO FUZZY EPIDEMICS CAUSED BY GROUP A STREPTOCOCCUS, A FLESH-EATING HUMAN BACTERIAL PATHOGEN

Abstract

Epidemics caused by microbial pathogens are inherently interesting because they can kill large numbers of our brethren, cause social upheaval, and alter history. Microbial epidemics will likely continue to occur at unpredictable times and result in poorly predictable consequences. Over a 30-year period, we have used the human bacterial pathogen group A streptococcus (also known as Streptococcus pyogenes) as a model organism to gain understanding of the molecular mechanisms contributing to epidemics caused by this pathogen and attendant virulence mechanisms. These epidemics have affected tens of millions of individuals worldwide and were largely unrecognized until revealed by full-genome sequence data from many thousands of isolates from intercontinental sources. Molecular genetic strategies, coupled with extensive use of relevant animal infection models, have delineated precise evolutionary genetic changes that contribute to pathogen clone emergence and successful dissemination among humans. Here, we summarize a few key findings from these studies.

INTRODUCTION

For millennia, humans have fretted about microbial epidemics because of the death and mayhem they cause. It is well known that epidemics have affected the course of history by decimating human and domesticated animal populations and detrimentally altering crop yields. Consequently, there is considerable interest in understanding the molecular events that contribute to pathogen emergence and the ability to cause geographically widespread epidemic or pandemic disease. Due to space constraints, we will focus on bacterial pathogens and not discuss viral or eukaryotic pathogens.

The great majority of pathogenic bacterial species are characterized by very abundant diversity in gene content and allelic variation (1-4). The genetic diversity found in bacterial pathogens is far greater than that found in higher eukaryotic species such as Homo sapiens. This substantial genetic diversity commonly results in differences in molecular interactions between pathogen and host, which contributes to changes in disease frequency and character, including severity. Thus, understanding the contribution an infecting bacterial pathogen makes to a pathogen-host interaction outcome requires detailed knowledge of gene content, allelic variation, and differences in gene transcriptome profile.

Historically, many techniques were used to detect diversity among bacterial isolates for studying the relationship between strain genotype and disease phenotype (1,2). Commonly used techniques included bacteriophage typing, antigen typing, resistotyping, and metabolic profiling, but these phenotypic techniques were limited because many had restricted resolving power and the phenotype being assessed generally could not be equated with precise genetic differences between strains (1,2). Subsequently, a few population geneticists began to apply the techniques and theory of contemporary population genetics to assess the degree of variation present in natural populations of bacteria (1,2,5,6). Early analyses focused on Escherichia coli and employed techniques such as multilocus enzyme electrophoresis, pulsed-field gel electrophoresis, and multilocus sequence typing (5-8). Although these techniques represented significant advances at the time, they were all limited because they assessed variation at only a relatively small number of chromosomal loci and, thus, substantially underestimated the magnitude of genetic diversity among isolates.

In approximately the last 15 years, the confluence of high-throughput, low-cost genome sequencing has revolutionized the study of bacterial population genetics (now commonly called population genomics) and permitted very large samples of strains to be interrogated at the full-genome level, including at the individual nucleotide (9-17). This has opened the door to several types of analyses, including the molecular events contributing to epidemics caused by many different bacterial pathogens, including Streptococcus pyogenes [group A streptococcus (GAS), see below]. These are termed “fuzzy epidemics” because they were not recognized and shown to be caused by recently evolved clonal progeny, arising from a common ancestor until full-genome sequence data were available, to unambiguously demonstrate shared recent common ancestry.

GAS IS A NUMERICALLY IMPORTANT HUMAN PATHOGEN

GAS is a gram-positive, human-specific pathogen that is responsible for greater than 700 million infections annually (18,19). This organism causes unusually diverse infections including pharyngitis and/or tonsillitis, skin infections such as impetigo and erysipelas, acute rheumatic fever (ARF), scarlet fever, poststreptococcal glomerulonephritis, sepsis, puerperal sepsis (commonly referred to as childbed fever), toxic shock syndrome, meningitis, and necrotizing fasciitis (18-20). Despite a century of study, for reasons that remain obscure, poststreptococcal glomerulonephritis, ARF, and subsequent rheumatic heart disease (RHD) occur in a subset of individuals as post-infection sequelae. Although antibiotic treatment has largely eliminated ARF and RHD in the United States and Europe, these diseases remain a very substantial public health problem globally, especially in certain areas of Africa and Oceania (18,19). Unfortunately, there is no licensed vaccine against GAS infection, although substantial effort has been directed toward this goal in the last 40 years and some important progress has been made (21,22). The direct and indirect costs of human GAS infections are estimated to be billions of dollars annually.

Historically, GAS has been subtyped based on serologic variation in the aminoterminal part of a bacterial cell surface-anchored protein known as M protein, a highly polymorphic protein that is a major virulence factor since it is anti-phagocytic (20,23). M protein serotyping has been largely replaced by sequencing of the hypervariable part of the emm gene encoding M protein. More than 240 emm types are now recognized, and this substantial allelic variation has proven useful for characterizing strains in epidemiologic studies. Importantly, no single emm type is solely responsible for any one type of GAS infection; however, strains expressing particular M serotypes have been repeatedly and nonrandomly associated with specific diseases over many decades of study (20,24). For example, emm1 and emm3 strains are leading causes of severe invasive infections in many parts of the United States and other Western countries, emm18 strains have been repeatedly associated with rheumatic fever outbreaks in the United States, and emm28 strains have been nonrandomly associated with cases of puerperal sepsis and neonatal infections (24-33). As a consequence of these epidemiologic associations, the concept arose that there are distinct molecular underpinnings to the nonrandom associations of emm type and disease phenotype. This thinking has resulted in substantial genome sequencing and other analyses strategies designed to identify molecular underpinnings responsible for the disease associations, including epidemics. Some important findings related to several distinct GAS emm types will be discussed below.

RESULTS AND DISCUSSION

GAS as a Useful Model Pathogen for Studying Epidemics

GAS is a useful model pathogen for studying human epidemic disease at the full-genome level for several reasons. First, the ability of GAS to cause human epidemics has been appreciated for more than a century. Second, the organism has a relatively small genome size, with the genome of most strains in the range of only 1.8 to 1.9 Mb (9,12-16,32). This is approximately half the size of many other common bacterial pathogens such as Staphylococcus aureus and E. coli, only about one-fifth the size of many fungal pathogens, and much smaller (one-tenth the size) than many other eukaryotic pathogens such as Plasmodium species. Third, GAS severe invasive infections are reportable in many countries, and large culture collections spanning decades exist, most notably in the Nordic countries. Some of these culture collections have linked patient metadata available, which greatly assists and enriches analysis of strain genotype-patient phenotype relationships. Fourth, the organism can be readily manipulated genetically, which means isogenic mutant strains can be precisely constructed and used to test focused mechanistic hypotheses emerging from population genomic studies. Similarly, relevant animal models of many types of GAS infections exist, and these can be exploited to test hypotheses bearing on the molecular genetic basis of many infection types, including pharyngitis, necrotizing myositis, female genital tract interactions, and others (12-14,22,35-37). Finally, as noted previously, GAS is a human-specific pathogen, which means molecular events linked to clone emergence and epidemics occur only in the human host (i.e., not in a nonhuman host or elsewhere in the environment).

Serotype M1 Epidemic Disease

Type emm1 GAS is a common cause of pharyngitis and severe invasive infections in the United States, Canada, western European countries, and elsewhere. In the late 1980s and early 1990s, type emm1 GAS was reported to be responsible for increased disease frequency and severity in many of these countries (34,38-40). The infection resurgence captured public attention in part because of the death of the Muppet creator, Jim Henson, due to severe emm1 disease. Severe GAS infections also occurred in well-known public figures in other countries such as Canada (41). Some early data suggested that the increase in disease frequency and severity was due to emergence of a new emm1 clone (34,38-40,42).

In an effort to gain new understanding of the molecular events contributing to the rapid and recent upsurge in emm1 infections, we assembled a large international collection of 3,615 emm1 strains and sequenced the genomes of all of them (12). These strains were assembled from comprehensive, population-based collections from decades of cases. At the time it was conducted, this study represented the largest bacterial genome study of this type. Genome sequence analysis permitted us to unambiguously demonstrate that this epidemic was a consequence of three distinct genetic events that caused the emergence of a new, hypervirulent clone of emm1 GAS. Moreover, we were able to precisely define the order of these three genetic events, which involved acquisition of a bacteriophage (bacterial virus) that encodes scarlet fever toxin variant 1, also known as streptococcal pyrogenic exotoxin A1 (SpeA1), a superantigen. Subsequently, SpeA2 evolved from SpeA1 by a single nucleotide change that resulted in one amino acid replacement in a functionally important region of the toxin. The third “big bang” genetic event was acquisition by horizontal gene transfer of a large chromosomal region containing the nga-ifs-slo three-gene operon that encodes two secreted toxins, NAD⁺-glycohydrolase and streptolysin O, a membrane lytic toxin. Thus, we demonstrated that the molecular evolutionary events occurring in clonal progeny of just one bacterial cell ultimately produced tens of millions of human infections globally (12). Our analysis also resolved a decades-old controversy about the sequence of genomic alterations producing the explosive global epidemic (34,38-40,43-45). Importantly, using nonhuman primate models of necrotizing myositis and pharyngitis, we showed that the contemporary clone is more virulent than a pre-epidemic reference strain (12,46,47). Quite unexpectedly, we discovered that the contemporary clone rose to global prominence in only a three-year period (1987–1989), essentially displacing antecedent emm1 strains over very broad geographic regions (12).

Although the population genomic study described above was of critical importance in precisely defining the genetic steps involved in evolution to a more virulent state, it did not address all questions. For example, it left open the exact molecular pathogenesis processes underpinning how the final genetic event (horizontal acquisition of the nga-ifs-slo gene region encoding the NAD⁺-glycohydrolase and streptolysin O toxins) triggered the global spread of progeny of the contemporary clone and increased infection frequency and severity. These matters were addressed in two subsequent studies (46,47). Using molecular genetic tools and animal infection models, we used isogenic mutant strains to discover that only three polymorphisms occurring in this toxin gene region increased toxin production, increased resistance to killing by human polymorphonuclear leukocytes, increased bacterial proliferation in vivo, and increased virulence in animal models of pharyngitis and necrotizing fasciitis (46,47). We also showed that production of both the NAD⁺-glycohydrolase and streptolysin O toxins was required for full virulence (47). That is, genetic inactivation of either the nga or slo gene decreased virulence to an approximately similar degree, and genetic inactivation of both genes greatly decreased virulence in relevant animal infection models. We did not expect to discover that an analogous recombinational replacement event also contributed to an intercontinental epidemic of type emm89 GAS infections and to the emergence and spread of progeny of a type emm28 GAS clone (13,14; see below). Thus, in the aggregate, our studies delineated the molecular changes in GAS that enhance upper respiratory tract fitness, increased resistance to innate immune defenses, and increased tissue destruction. A recent study (15) reported that recombinational replacement involving the nga-ifs-slo region is more widespread than emm1, emm28, and emm89 GAS. However, the role in virulence of other emm types was not addressed by construction and analysis of isogenic mutant strains in animal infection models of virulence.

Serotype M3 Epidemics

Type emm3 GAS also is a common cause of severe invasive infections in many geographic regions (25-27,30-33). Several epidemiologic studies have shown that type emm3 strains cause a disproportionate number of severe invasive infections such as bacteremia, streptococcal toxic shock syndrome, and necrotizing fasciitis (25-27). Importantly, large population-based studies conducted in Canada and the United States have reported that emm3 strains cause a higher rate of lethal infections than strains of other emm types (25-27). Like type emm1 GAS, emm3 strains can also undergo very rapid shifts in disease frequency. That is, they experience epidemic behavior, but the molecular causes are not known.

We conducted several studies (9,48-50) designed to delineate molecular genetic processes contributing to emm3 clone emergence and disease epidemics. Analysis of 255 emm3 strains cultured from patients in Ontario, Canada, during two peaks of severe infections occurring over 11 years found that distinct emm3 genotypes experienced rapid population expansion and caused infections that differed significantly in character, including severity. A four-amino-acid duplication in the extreme aminoterminus of M protein was proven to be a contributor to one of the epidemic waves. This is the region of M protein against which the host generates opsonizing antibodies. Importantly, this four-amino-acid change caused a difference in serologic reactivity and resulted in a significant difference in the ability of human polymorphonuclear leukocytes to phagocytosize and kill GAS expressing these variants. Given that the aminoterminal region of M protein is a well-known vaccine candidate, it is reasonable to think that this finding has implications for vaccine development and deployment, including the potential for rapid selection of immune escape variants.

Serotype M28 Epidemics

Type emm28 GAS strains are of importance to human infection biology for several reasons. First, emm28 strains are among the top five emm types that cause invasive GAS infections in the United States and many European countries (25-27,30-33). In some countries, emm28 strains are the most common cause of invasive infections and pharyngitis (30). Second, as noted above, type emm28 strains are repeatedly significantly overrepresented among cases of puerperal sepsis and neonatal infections (24,28). Third, the great majority of emm28 strains are a chimera in that they contain a 37-kb segment of DNA [termed region of difference 2 (RD2)] that was likely acquired from group B streptococcus, a very common cause of neonatal infections (51,52). This segment of DNA encodes at least seven secreted extracellular proteins, some of which are made during the course of human infections, as shown by serologic studies (53). A recent study reported that deletion of this entire region altered pathogen-host interaction in vitro (54).

Knowing that important new information about epidemic behavior could be acquired by molecular population genomic analysis coupled with bacterial pathogenesis techniques, we tested the hypothesis that new understanding of emm28 infection biology would be obtained by integrated analysis of the population genomics, transcriptomics, and virulence (13). We assembled an international collection of 2,101 emm28 strains recovered in comprehensive population-based surveys of invasive infections. The genomes of all strains were sequenced. We conducted RNA-seq transcriptome analysis on 492 phylogenetically diverse strains, and 50 genetically representative strains were tested for virulence in a mouse model of necrotizing myositis (13). We made new discoveries about the molecular pathogenesis of type emm28 GAS. For example, we found a striking magnitude of transcriptome variation among a relatively closely related genetic clade of organisms. In addition, application of statistical methods and machine learning led to the discovery of a new molecular genetic process that underpins enhanced virulence in some GAS strains (13). We determined that the RD2 region noted above has a homopolymeric tract of T residues that is highly polymorphic in number, resulting in variation in expression of a transcript that encodes a regulatory protein that controls expression of one of the secreted extracellular proteins. Finally, we found that one of the genetic subclades has increased in frequency rapidly in the United States (i.e., has experienced a recent epidemic wave) (13). Strains of this epidemic subclade were associated with puerperal sepsis, neonatal infections, and female genital tract infections and were also significantly more virulent in the mouse model of necrotizing myositis (13). Just like the type emm1 epidemic clone, the emm28 epidemic subclade has experienced a reciprocal recombination event involving the nga-ifs-slo region. The acquired nga-ifs-slo region is essentially identical to the analogous three-gene operon present in Streptococcus dysgalactiae subspecies equisimilis. Moreover, the acquired variant region produces significantly higher transcript levels of nga-ifs-slo, resulting in increased production of the two secreted toxins, NAD⁺-glycohydrolase and streptolysin O (13). Molecular genetic studies using isogenic mutant strains have confirmed that the upregulation of this three-gene operon is responsible for enhanced mouse and nonhuman primate virulence (13).

Basic Science and Public Health Utility of the Availability of Large and Well-Characterized Strain Samples with High-Quality Genome Sequences

One of the initial goals of our study of the molecular genetic underpinnings of GAS epidemics was to generate sufficiently robust datasets from large strain samples that could be used to analyze matters of basic science and public health utility, as need arises. The datasets arising from our study of emm1, emm28, emm89, and other emm types GAS strains have indeed proved to be useful in an analysis of recently identified nonsynonymous (amino acid-changing) mutations in the pbp2x gene associated with decreased susceptibility to beta-lactam antibiotics (55,56). Many generations of physicians, clinical microbiologists, and others have been taught that GAS strains are universally highly susceptible to beta-lactam antibiotics. However, a recent report (55) describing two GAS strains with decreased susceptibility to several beta-lactam antibiotics was very concerning and compelled us to rapidly bioinformatically interrogate our >7,000 genomes of type emm1, emm28, and emm89 strains for nonsynonymous mutations in pbp2x (56). We found that approximately 2% of organisms had amino acid-conferring mutations in pbp2x, usually one amino acid replacement per strain. The strains with pbp2x mutations were grown and examined by standard clinical laboratory methods for decreased susceptibility to benzylpenicillin, ampicillin, and several other beta-lactams. Because the strains had been archived over several decades and obtained from patients in multiple countries, we now know (unfortunately) that GAS strains with decreased susceptibility to beta-lactam antibiotics are neither exceedingly rare nor geographically restricted (56). These findings illustrate the importance of ongoing routine monitoring of beta-lactam susceptibility phenotypes in GAS strains.

In an analogous use of genomic and phenotypic data generated from banked strains, Lynskey et al. (57) discovered that emm1 organisms associated with a recent upsurge of scarlet fever activity in the United Kingdom represented a new emm1 clone that is characterized by increased production of scarlet fever toxin (SpeA).

CONCLUSIONS

Studies addressing the molecular genetic processes contributing to GAS strain emergence and epidemics have been a fruitful area of investigation over approximately 30 years. Some of these distinct processes include seemingly relatively minor genetic changes such as nucleotide mutations that upregulate toxin gene expression and en bloc horizontal gene transfer involving acquisition of one or more toxin genes that influence pathogen-host interactions. Delineation of the molecular contributors to bacterial pathogen strain emergence and epidemic spread is of fundamental biomedical importance and has substantial relevance to public health and downstream clinical and translational research, including vaccinology. The strategies used to investigate GAS are generally applicable to any pathogen for which there are available appropriate strain collections, genetic manipulation techniques, and animal infection models that mimic human disease.

DISCUSSION

Hook, Birmingham: Thank you for your presentation and for your work. Something you said challenged a misperception I may have and that has to do with super spreaders. You are saying that super spreading is a function of the virulence and pathogenicity of the organism. I always thought that super spreaders were, for lack of a better term, special people or maybe it's a combination. I would love to hear your thoughts.

Musser, Houston: We believe that super spreaders occur in certain hosts who have very high CFUs in the upper respiratory tract, caused in part by a lack of sufficient immunity, coupled with genetic attributes of the infecting GAS strain. Many of you in the audience—especially the young folks who know the literature from the 1950s—know about the studies done by a husband/wife team that really showed that once you get above about 10⁷ CFUs in the upper respiratory tract then you can effectively spread. Those of course were done in the military populations. So we really think of it as certain individuals who have the ability—for reasons that we don't understand but I have some ideas about—to develop very high levels of CFUs and then spread to others quite effectively. The ability to upregulate the nga and slo toxins rapidly, as I showed and we have published on, we think basically trumps any preexisting host immunity in the upper respiratory tract. We actually have shown that as well in the nonhuman primates. But it is clearly an intricate sort of tango between the host and the pathogen. I don't know if that answers your question, but I am happy to talk more about it.

NOTE: This paper was presented in 2018 at the ACCA meeting in Sarasota, Florida, but the manuscript was not submitted for editorial review until the Fall of 2019, so it is being included in this volume of our Transactions.