Legionella pneumophilaGoes Clonal—Paris and Lorraine Strain‐Specific Risk Factors

Abstract

More than 400 different Legionella pneumophila (LP) serogroup 1 strains have been described, with a minority causing most disease [1–6]. Of these, those reacting with a monoclonal antibody directed against an 8-O-acetylated lipopolysaccharide, the “Pontiac” group, cause between 70 to 95% of all culture proven cases of community-acquired Legionnaires’ disease (LD) [2, 5, 6]. Patients with nosocomial LD, or immunocompromised patients, are less often infected with the Pontiac group organisms, presumably because they can be infected with less virulent strains. Molecular typing schemes are now in common use to identify LP clones. The French National Reference Center for Legionella developed a pulse field gel electrophoresis (PFGE) method and the European Working Group on Legionella infections (EWGLI) developed a sequence based typing scheme [7, 8]. The French group discovered two novel LP serogroup 1 strains that cause a number of LD cases within France, the Paris and Lorraine strains[8–10]. The Lorraine strain, and many of the Paris strains, are in the Pontiac monoclonal antibody group [3, 9, 11]. The Lorraine strain is EWGLI sequence type 47. One study has shown that the Paris strain is EWGLI sequence type 1[11]. The Lorraine strain was described in 2008 [9]. This strain is a significant cause of LD in France, causing around 10% of cases of culture confirmed LD, most of which are sporadic community cases. Lorraine has also been shown to have caused about 25% of culture confirmed community acquired cases of LD in England and Wales from 2000–2008 [2]. Lorraine isolates have also been identified from the Netherlands and Greece [2]. The Lorraine strain may not be widespread, based on no reports of its isolation from Germany [3]. Whether Lorraine strains are found in North America is not known because the required molecular typing scheme is not widely used in this region. Despite the fact tha Lorraine strains are relatively common causes of LD, the strains are rarely isolated from the environment. Environmental isolates constituted only 0.7% and 4% of all Lorraine strains isolated in England and Wales, and France, respectively [2, 9]. Rare isolation of Pontiac group clinically dominant strains from the environment is not unusual [2, 6]. In contrast to the Lorraine strain, the Paris strain was recognized several years previously as being endemic in France, and is commonly isolated from both LD patients and the environment [8, 10]. In the ten year period between 1987 and 1997, 33% of LP clinical isolates in the Paris region were this type, as were 39% of environmental isolates [8]. The Paris strain is found widely throughout France, and has also been isolated from LD patients, or the environment, in most of Europe, North America, Australia, Japan, and Senegal [4, 10, 11]. Similar to findings in France, the Paris strain accounts for about 20% and 12%, of clinical isolates in Germany and England and Wales, respectively [2, 3]. Unlike the Lorraine strain, Paris strain isolates are not exclusively in the Pontiac monoclonal antibody group [1, 3]. Infrequent isolation of the Lorraine strain from environmental sources suggests that it is more pathogenic than those strains commonly isolated from the environment. An alternative explanation is that the Lorraine strain is common in our environment yet is not detected for methodologic reasons [12]. The apparent heterogeneity of the Paris strain in regard to monoclonal antibody group may explain why this strain is both clinically dominant yet isolated with high frequency from the environment. Non-culture based methods of detecting environmental Legionella spp. bacteria should help answer this question, as should more extensive correlative studies of PFGE, monoclonal antibody types and EWGLI sequence types of Paris strains. Ginevra and colleagues present data that may support the greater pathogenicity of the Lorraine strain [13]. Elderly women who were immunosuppressed were more likely to be infected with the Paris than sporadic strains, and were more likely to die. The Lorraine strain, on the other hand, was non-discriminatory in terms of who became infected in comparison to sporadic strains. More direct studies of pathogenicity such as animal model and cell infection studies are required to help answer the question of whether the Lorraine is indeed more pathogenic than other strains. If the Lorraine strain is indeed more pathogenic than Paris and sporadic strains, why was the fatality rate lower for infections with the Lorraine (10%) than for infections with the sporadic (26%) and Paris strains (38%)? The key is that case fatality is a product of both the organism virulence and the frailty of the host. Thus, the advanced age, immunosuppression and possibly gender of those infected with the Paris strain most likely explains this discrepancy. The elderly are more likely to die of pneumonia than are younger people, especially when immunocompromised, and for unclear reasons elderly females have a slightly higher fatality rate for pneumonia than do males [14, 15]. The statistical methods used may have also introduced some bias into the study. Population surveillance for infectious diseases can provide invaluable epidemiological evidence for tracking the clonal spread of specific pathogens and the evolution of host risk factors. However, as with all epidemiological investigations, critical design features may threaten the internal and external validity of the results. One critical feature of a surveillance network is whether the surveillance is truly population based in so far as the network is able to detect all clinically apparent cases, without systematic bias. This is likely to be the case when effective treatment requires an accurate clinical diagnosis (e.g., tuberculosis, HIV) or when testing for the pathogen is a standard component of care (e.g., blood cultures in patients hospitalized with community-acquired pneumonia to detect pneumococci). However, in the case of LD, the variability in the use of diagnostic methods and their poor performance, and the availability of widely available empiric therapy suggests that many cases go undetected. Thus, the actual set of cases that are available for analysis may be systematically biased. In the case of the paper by Ginevra and colleagues, the authors acknowledge such potential biases but do not discuss the implications for their study results [13]. For example, such sampling biases may lead to systematic over-reporting of strain specific mortality and may lead to over-representation of certain high risk populations in the data. However, such sampling bias at the population level is unlikely to vary significantly by strain type and influence the internal validity of the study unless there is substantial variability in the sensitivity of detection methods by strain type. Given the known epidemiological differences between strains discussed above, such variability in detection remains a concern. How should one interpret the results of the strain comparisons in this study? A critical point is that all of the statistical comparisons in the paper are among patients with culture proven Legionnaires’ disease. In other words, the tests of association are all conditional on disease and are best interpreted as risk factors for predicting infection with one strain or another once infection with LP is known. This is quite different than comparing strain infected patients with healthy controls in order to determine the individual risk factors for disease. Thus, while the finding that smoking is more common among patients infected with the Lorraine strain compared to sporadic strains may help us predict which strain is infecting a person with known LD, it does not mean that smokers are at increased risk of infection with Lorraine strain, compared to non-smokers. This situation is analogous to methodological concerns raised in the interpretation of epidemiological studies of drug resistant vs. drug susceptible infections and the importance of control group selection [16, 17]. While it may well be true that the factors identified in the study by Ginevra and colleagues may actually increase the risk of specific types of disease, the current study design is unable to demonstrate this point. Moreover, in the absence of a true healthy control group, we can not determine the main risk factors for infection by these strains, particularly in terms of targeting public health strategies for reducing the individual risk of disease. A different study design is required to answer that question. There appear to be no immediate practical clinical implications of different risk factors associated with specific LP strains. It may be that with more data, knowing the infecting strain could influence prognosis; however for that information to be useful strain identity will need to be known at patient presentation and not days to weeks after onset of infection. While there is good evidence that molecular methods can be used to define the infecting strain in the absence of culture data, the current technology is imperfect and is unlikely to be universally available in most laboratories in the near future [18]. As pointed out by Harrison and colleagues, knowledge of which LP strains are responsible for disease can be useful in focusing environmental control of LD [2]. Finally, knowledge of host-specific strain interactions can help dissect host-specific genetic resistance mechanisms, as well as inform studies of strain pathogenesis.