Prevalence of Infection-Competent Serogroup 6 Legionella pneumophila within Premise Plumbing in Southeast Michigan

The leading cause of disease due to drinking water in the United States is Legionella pneumophila (1). This bacterium naturally resides in fresh water, but the majority of Legionnaires’ disease cases originate in engineered water systems. People become infected with legionellae by inhaling contaminated aerosols generated by devices that release water vapors, such as cooling towers, hot tubs, whirlpools, decorative fountains, and showers (2). Older adults with underlying disease, immunosuppression, or a history of smoking are especially vulnerable to Legionnaires’ disease, a severe, sometimes fatal, pneumonia (3). For reasons not understood, the incidence of Legionnaires’ disease increased 3-fold from 2000 to 2011 across the United States and all age groups (1, 3, 4). Likewise, the cases reported in Europe tripled during the 1995–2014 period (5).

Clinicians typically treat community-acquired pneumonia empirically and promptly with broad-spectrum antibiotics to forgo the expense and time required for specific diagnosis (3, 6). When the infectious agent is sought, the most common diagnostic tool for Legionnaires’ disease is the urinary antigen test, a rapid low-cost assay specific for L. pneumophila of serogroup 1 (SG1) (7). Although this pathogenic type of legionellae is associated with >80% of Legionnaires’ disease cases worldwide, >60 species and 16 serogroups exist, half of which have been isolated from patients (8). Disease due to non-SG1 L. pneumophila is likely underreported, since in the United States and Europe >75% of Legionnaires’ disease cases are diagnosed by the urinary antigen test and just ~5% are confirmed by culture (8, 9). Indeed, it is estimated that only ~5% of Legionnaires’ disease cases in the United States are reported (10). Although widespread reliance on the urinary antigen test and empirical broad-spectrum antibiotic treatment are efficient and cost-effective for most patients, these clinical practices hamper not only identification of other pathogenic legionellae but also epidemiological investigations to track and eliminate the source of Legionnaires’ disease outbreaks.

Especially prone to colonization by L. pneumophila are hospitals and hotels, due to their warm temperatures, large size, and multiple partitions with irregular usage that create areas with low or no flow (11). Stagnant water favors growth of biofilms, adherent microbial communities that are difficult to eradicate (3, 12). L. pneumophila also survives and replicates within predatory free-living protozoa that graze on biofilms (11). Extermination of legionellae residing within protozoa or biofilms requires elevated doses of disinfectants (13–17). Consequently, despite remediation efforts, L. pneumophila can persist in complex engineered water systems and cause recurrent disease outbreaks for decades (18–23). Indeed, despite infection control measures, L. pneumophila was found to colonize 70% of Pittsburgh and 60% of Paris hospital water systems surveyed (24, 25).

Compared to our knowledge of the established risks of legionella colonization within large buildings, relatively little is known about exposure within single-family homes and low-rise housing units. Based on limited sample sizes, the prevalence of residences whose water is culture positive for legionellae has been reported to be 6 to 30% (26–31). L. pneumophila colonization can be suppressed by maintaining a chlorine concentration of >2 mg/liter (32, 33). However, hot water residential lines often contain little measurable levels of chlorine, due to its decay at elevated temperatures (32). Since only 4% of the Legionnaires’ disease cases reported in the United States from 2000 to 2009 were associated with outbreaks (4), sporadic or epidemiologically unrelated cases of Legionnaires’ disease due to colonization of the built environment by legionellae may be more insidious and widespread than currently documented.

In the summers of 2014 and 2015, Genesee County, MI, endured outbreaks of Legionnaires’ disease (26). Compounding the anxiety within the community, the 87 confirmed cases of disease occurred during a period of sustained damage to the municipal water system of Flint, the county’s largest city (27). When the anticorrosive agent orthophosphate was omitted from the Flint water supply, toxic lead leached from pipes and fixtures into the municipal water. The impact of the water’s altered physiochemistry on the burden, resilience, or virulence of the microbial communities in the Flint water system is an open question.

To inform the rational design of risk management strategies to keep public water supplies safe, the State of Michigan supported the Flint Area Community Health and Environment Partnership (FACHEP) research team’s analysis of legionellae within the Flint municipal system. Working closely with local health care providers, community organizations, and government agencies, we applied a multidisciplinary strategy that integrated environmental monitoring and water testing with clinical education, molecular epidemiology, and laboratory assays of L. pneumophila virulence. Here we present our laboratory analysis of the genetic diversity and virulence potential of L. pneumophila isolated from Flint premise plumbing in the fall of 2016.

Using a multidisciplinary approach coordinated with local partner health care providers, community organizations, and government agencies, the Flint Area Community Health and Environment Partnership (FACHEP) research team investigated the presence of L. pneumophila in Flint’s premise plumbing 1 year after Legionnaires’ disease outbreaks in two consecutive summers that coincided with the Flint water crisis. To do so, we integrated environmental monitoring and water testing with clinical education, assessment of social needs and resilience, population-based and molecular epidemiological surveillance, and laboratory assessments of virulence. In the molecular and phenotypic analysis of L. pneumophila strains isolated in Southeast Michigan reported here, we found little genetic overlap between the environmental and the clinical strains. All 33 clinical isolates are of SG1, consistent with the widespread use, both in Southeast Michigan and globally, of diagnostic tests that are specific to serogroup 1 L. pneumophila. In contrast, only two of the 18 premise plumbing isolates are SG1 strains. Instead, 89% of the 18 household strains are SG6 L. pneumophila of ST367 and ST416, two closely related STs. The virulence of the SG6 strains isolated from premise plumbing in Flint and Detroit resembles that of the SG1 clinical strains, as judged by their capacity to infect and replicate in cultured macrophages, a hallmark of L. pneumophila pathogenesis. As expected, SG6 L. pneumophila ST367 and -461 strains did not bind an SG1-specific antibody (Fig. 3). Since the widely applied diagnostic urinary antigen tests also rely on serotype-specific antibodies, it is likely that SG6 strains are not readily detected (42).

Across the globe, the most widespread L. pneumophila strains are SG1 ST1 strains. In China, SG1 ST1 strains represented 49% of the 164 L. pneumophila isolates collected from natural and engineered water supplies from 2005 to 2012 (43). Likewise, a U.S. survey analyzing 571 sporadic SG1 clinical isolates and 149 environmental isolates unrelated to disease cases that were collected from 1982 to 2012 found that ST1 L. pneumophila isolates were not only of the most prevalent serogroup (SG1) in both potable and nonpotable water (49% of all samples) but also the most frequent cause of sporadic Legionnaires’ disease (25% of cases) (18). Of the 4,785 clinical strains submitted from Europe to the EWGLI database, 10% are of ST1 (44). In France from 2008 to 2012, SG1 ST1 strains represented 21% of environmental isolates and 9.1% of clinical cultures yet accounted for 31% of nosocomial cases of Legionnaires’ disease (45). In a 2000–2008 surveillance study in England and Wales, L. pneumophila was cultured from 3.2% of samples collected from premise plumbing (46). Compared to 167 isolates from unrelated clinical cases, there was very little genetic overlap, as judged by ST (P < 0.0001). The only exception was SG1 ST1 strains, which comprised 20% of the premise plumbing isolates and 4.8% of the patient isolates (46). Likewise, in Southeast Michigan, SG1 ST1 was the only strain type isolated from both premise plumbing (1 of 18 samples; 5.6%) and patients (4 of 33 patients; 12%) (Fig. 2). Analysis of the whole-genome sequences of 229 L. pneumophila ST1 isolates from community- and hospital-acquired cases of Legionnaires’ disease and from water systems of the corresponding hospitals determined that ascribing an epidemiological link between ST1 strains requires a combination of molecular genetic and clinical epidemiological data, because levels of genetic diversity and stability differ across locations, and L. pneumophila spreads internationally (47). The molecular mechanisms and epidemiological processes that contribute to SG1 ST1 L. pneumophila prevalence worldwide remain to be understood.

Among the 33 SG1 clinical isolates collected in Southeast Michigan from 2013 to 2016, the majority belonged to one of four clonal complexes (Fig. 2). None of these clinical isolates were of ST35, ST36, or ST37, three strain types that caused a total of 10 outbreaks and 64 sporadic disease cases in the United States during 1982 to 2012 (18). In Southeast Michigan from 2013 to 2016, the largest clonal complex of clinical isolates included strains of ST222 and ST213, which are identical at six of the seven typing loci. In the United States from 1982 to 2012, L. pneumophila ST222 strains were among those most frequently associated with outbreaks, and strains of ST222 and ST213 were often associated with sporadic cases of Legionnaires’ disease (18). The ST222 genetic lineage has been prevalent in the northeastern regions of the United States and Canada but rare elsewhere (18, 48). After emerging in Ontario in 1999, over the next 8 years, ST222 L. pneumophila strains comprised 11 to 15% of clinical isolates in this Canadian province, including those responsible for the large 2005 outbreak at a long-term-care facility (48). The 2012 Calgary, Canada, outbreak of eight Legionnaires’ disease cases is an interesting exception, as it occurred during winter months with temperatures between 7°C and −26°C (49). Although environmental surveillance did not identify the reservoir, epidemiological and genome sequencing data predicted a likely point source: water contaminated with L. pneumophila of ST222 sprayed to extinguish fires or to minimize dust at construction sites in this city in western Canada (49).

It is possible that SG6 strains of L. pneumophila are endemic in the Great Lakes region. Of the 18 environmental isolates obtained in the fall of 2016 from Flint and Detroit premise plumbing, 89% were SG6 L. pneumophila strains. Among these, strains of the same ST, ST367, were isolated from 11 premise plumbing sites in Flint and one residence in Detroit, which did not receive treated Flint River water in 2014 to 2015. A 1981 survey of 52 sites in the Great Lakes, which border Michigan, yielded only two L. pneumophila strains, but both were of SG6 (50). In a 1984 analysis of water from 31 buildings in Ontario, a Canadian province adjacent to Southeast Michigan, SG6 strains were isolated slightly more frequently (12 of 19 buildings) than SG1 L. pneumophila strains (10 of 19 buildings). Elsewhere, SG6 strains are also prevalent. In the water distribution systems of nine hospitals in Poland, SG6 strains comprised 33% of the 11 environmental isolates, whereas SG1 strains accounted for 24% (31). In the Attica region of Greece, SG6 strains represented 62% of 37 L. pneumophila environmental isolates (51). In contrast, SG6 strains comprised only 15% of the L. pneumophila strains collected in England and Wales (46) and 11% of those isolated in France (52). More extensive surveillance studies can determine whether the predominance of SG6 L. pneumophila strains observed in Flint in the fall of 2016 is typical of the Great Lakes region or instead reflects a microbial response to the altered physiochemistry of the Flint municipal water system during 2014 and 2015. Another factor that might contribute to the predominance of SG6 strains in surveillance studies is L. pneumophila’s capacity to differentiate into a viable-but-nonculturable state. By this model, if, compared to other legionellae, SG6 strains are more readily cultured from water samples, their apparent prevalence would be elevated. Whether exposure to corrosive water characteristic of the 2014–2015 Flint water crisis alters the culturability, resilience, or virulence of L. pneumophila is one focus of ongoing analysis.

The genetic homogeneity of the 16 SG6 L. pneumophila strains cultured from Southeast Michigan premise plumbing is also striking: 12 strains are of ST367 and 4 others are of ST461, which share six of seven typing loci (Fig. 2). In contrast, of the 410 SG6 L. pneumophila strains submitted to the EWGLI database, only 5 are of ST367 and 9 are of ST461, and they originated from a variety of countries and continents. Whether the climate or water chemistry in Southeast Michigan contributed to the prevalence of this lineage during the fall of 2016 requires more detailed studies.

Although very few studies have reported sequence type analysis of SG6 L. pneumophila strains, the stability of particular genetic lineages has been suggested. In one hospital in Sweden, SG6 L. pneumophila isolates from three different patients and seven water samples were each of ST1392, as were eight of nine control clinical isolates from unrelated locations (53). In the Attica region of Greece, among 25 SG6 isolates, 28% were of ST68 and 20% were of ST461 (51), a type prevalent in Flint premise plumbing (Fig. 2) and also present in hospital water systems in Poland (31). In contrast, a 2-year surveillance study of a larger territory in Greece identified 10 different STs among 28 SG6 L. pneumophila strains isolated (51). More detailed genetic and phenotypic analyses of SG6 ST367 and ST461 strains from Flint, Detroit, Ontario, and Spain (33, 35) can evaluate whether this L. pneumophila lineage is either endemic to Southeast Michigan, more readily recovered from water samples, or more fit in the physiochemical conditions peculiar to the 2014–2015 Flint municipal water system.

A critical question is whether the SG6 L. pneumophila strains present in premise plumbing in Southeast Michigan put residents at risk. The virulence potential of environmental SG6 L. pneumophila ST367 and ST461 strains was comparable to that of SG1 clinical strains, as judged by quantifying entry, survival, evasion of digestion, and replication within cultured mouse macrophages (Fig. 5 and 6). In Israel, Spain, and Greece, SG6 strains of ST367 and ST461 have been isolated from patients as well as from corresponding environmental surveillance efforts (32, 33, 51). Moreover, in a phylogenetic analysis of 73 clinical non-SG1 L. pneumophila isolates in Ontario, which borders Southeast Michigan, SG6 was most prevalent (47%) (35). In a 1990 study in Sweden, SG6 was also the most common non-SG1 L. pneumophila serogroup isolated from Legionnaires’ disease patients (54). In the United States, a 1980–1989 survey of 327 Legionnaires’ disease patients attributed 6.4% of disease cases to SG6 L. pneumophila, second only to SG1 (62%). However, the 64% mortality rate of patients with SG6 infection exceeded that for patients infected with SG1 L. pneumophila (37%), even after controlling for specific underlying conditions, age, nosocomial acquisition, and other risk factors (39). On the other hand, a 2013 laboratory analysis discovered that the SG6 L. pneumophila Thunder Bay strain replicates more efficiently in Acanthamoeba castellanii and human monocyte-like cells than SG1 strains yet is more sensitive to killing by the innate immune system’s alternative complement pathway (35). In a mouse model of infection, these SG1 and SG6 strains replicate to similar numbers in the lung, but the SG6 Thunder Bay strain was less proficient at disseminating into the blood and other tissues (35). Accordingly, infection and replication in cultured mouse macrophages may not accurately predict virulence in human lung.

Other key unanswered questions concern L. pneumophila fitness in water and aerosols, the major route of human infection. It is not known whether survival in water aerosols by SG6 L. pneumophila is comparable to that of SG1 strains, whose unusual LPS includes a lipid A moiety that is especially hydrophobic (55). Since the LPS of SG1 L. pneumophila also contributes to lysosome evasion (36), SG1 and -6 strains may differ in their capacities to survive within phagocytic protozoa, their natural predators (56).

Because the widely applied urinary antigen test is specific for SG1 L. pneumophila (42), Legionnaires’ disease due to SG6 and other non-SG1 strains is more difficult to diagnose. Compared to the 60 to 80% sensitivity of the urinary antigen test, the sensitivity of respiratory specimen culture methods ranges from 20 to 80% (57). Furthermore, obtaining sputum specimens from pneumonia patients is challenging, and diagnosis by culture takes several days, requires specialized media and trained personnel, and is hindered by prior empirical treatment with antibiotics (7, 57). Consequently, <5% to 12% of Legionnaires’ disease cases in Europe and the United States are confirmed by culture (4, 5, 9). Instead, urinary antigen testing has been enlisted to confirm diagnosis of >90% of Legionnaires’ disease cases in both the United States and Europe (7, 9). Indeed, although pneumonia due to non-SG1 L. pneumophila likely accounts for >10% of total Legionnaires’ disease incidence (51), it is generally understood that non-SG1 cases are underdiagnosed and underreported (3, 7, 9, 10). In Flint’s Genesee County in 2014 to 2015, the average number of deaths due to pneumonia increased 43% above the average of the previous 5 years; in contrast, pneumonia deaths across Michigan increased only 6% above the state’s 5-year average (58). During this period in Southeast Michigan, it was extremely rare that a Legionnaires’ disease case that was negative by the urinary antigen test was also analyzed by either culture or direct fluorescence antibody testing, and none were analyzed by serology or PCR. Accordingly, during the 2014–2015 Genesee County outbreak, there is no clinical microbiology laboratory evidence of Legionnaires’ disease due to non-serogroup 1 L. pneumophila. Whether SG6 L. pneumophila strains present in Flint premise plumbing contributed to the 2014–2015 increase in pneumonia in Genesee County could be investigated by an epidemiological study that incorporates serological testing for SG6-specific antibodies in patients with community-acquired pneumonia of unknown etiology. Until rapid, sensitive tests are developed to identify and type non-SG1 strains of L. pneumophila, clinicians and public health professionals must rely on culture-based methods to identify cases or sources of Legionnaires’ disease, a critical step in protecting the public from this illness, the most common associated with drinking water in the United States.

We are grateful for the contributions of Joanne Smith-Darden, Mark Zervos, Lara Treemore-Spears, Mary Beth Perri, and other members of the Flint Area Community Health and Environment Partnership, whose surveillance efforts and laboratory assistance were a foundation for this study.

Our work was supported by the Michigan Department of Health & Human Services (MDHHS) under contract number 20163753-00.

As contractually mandated, the manuscript was submitted to the MDHHS for review more than 30 days in advance of being submitted for publication. The content is solely the responsibility of the authors and does not necessarily represent the official views of the MDHHS.

This article is a direct contribution from a Fellow of the American Academy of Microbiology.