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

ABSTRACT Coinciding with major changes to its municipal water system, Flint, MI, endured Legionnaires’ disease outbreaks in 2014 and 2015. By sampling premise plumbing in Flint in the fall of 2016, we found that 12% of homes harbored legionellae, a frequency similar to that in residences in neighboring areas. To evaluate the genetic diversity of Legionella pneumophila in Southeast Michigan, we determined the sequence type (ST) and serogroup (SG) of the 18 residential isolates from Flint and Detroit, MI, and the 33 clinical isolates submitted by hospitals in three area counties in 2013 to 2016. Common to one environmental and four clinical samples were strains of L. pneumophila SG1 and ST1, the most prevalent ST worldwide. Among the Flint premise plumbing isolates, 14 of 16 strains were of ST367 and ST461, two closely related SG6 strain types isolated previously from patients and corresponding environmental samples. Each of the representative SG1 clinical strains and SG6 environmental isolates from Southeast Michigan infected and survived within macrophage cultures at least as well as a virulent laboratory strain, as judged by microscopy and by enumerating CFU. Likewise, 72 h after infection, the yield of viable-cell counts increased >100-fold for each of the representative SG1 clinical isolates, Flint premise plumbing SG6 ST367 and -461 isolates, and two Detroit residential isolates. We verified by immunostaining that SG1-specific antibody does not cross-react with the SG6 L. pneumophila environmental strains. Because the widely used urinary antigen diagnostic test does not readily detect non-SG1 L. pneumophila, Legionnaires’ disease caused by SG6 L. pneumophila is likely underreported worldwide.

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.

RESULTS
Legionella screening in Flint, MI, and neighboring census tracts. To determine the prevalence of legionellae in Flint premise plumbing, a random sample of Flint's households was surveyed from 6 September to 29 October 2016, 1 year after the 2015 summer outbreak. In Southeast Michigan, this period corresponds to the end of the typical Legionnaires' disease season, which peaks annually during warm-weather months. From 130 residences, water samples were collected from the kitchen faucet, shower, and hot water heater and then cultured to isolate viable legionellae. Overall, L. pneumophila was cultured from 13 Flint and 2 Detroit residences (data not shown). Water samples from two other Flint residences contained non-pneumophila legionellae; these strains were not analyzed further. In one home, L. pneumophila was isolated from the hot water heater, shower, and kitchen faucet; in another residence, a positive culture was obtained from the shower and another from the kitchen faucet (Table 1). In total, we obtained a single L. pneumophila isolate from each of 16 positive premise plumbing samples from 13 Flint residences and a single L. pneumophila isolate from each of 2 premise plumbing sites in two Detroit residences ( Fig. 1; Table 1). A similar incidence of culture-positive premise plumbing was obtained in Flint and non-Flint residences; however, the small sample size of non-Flint residences precludes statistical tests of significance. Of the hot water heaters that were sampled and noted, 125 were gas and only 1 was electric; therefore, our residential survey cannot address whether hot water heater design contributed to the risk of premise plumbing colonization by L. pneumophila, as reported previously (28).
Phylogenetic analysis of Southeast Michigan clinical and environmental L. pneumophila isolates. To gain insight into the genetic diversity of the L. pneumophila strains obtained from premise plumbing in Flint in the fall of 2016, we determined the phylogenetic relationships among the 18 environmental isolates. In parallel, we also analyzed 33 clinical isolates that hospitals in three Southeast Michigan counties (Genesee, Wayne, and Oakland) submitted to the Michigan Department of Health and Human Services Bureau of Laboratories from 2013 to 2016 (Table 1). We first applied a DNA sequence-based typing (SBT) scheme that is standard practice in the international Legionella community (29). Using the DNA sequences of seven chromosomal alleles, SBT can discriminate among more than 2,000 different sequence types (STs) of L. pneumophila strains present worldwide, increasing the power of outbreak investigations (18). To identify related genotypes within the typed environmental and clinical L. pneumophila isolates from Southeast Michigan, we applied the eBURST algorithm (30). We defined a clonal complex as those strains that shared at least four of seven typing alleles with at least one other member of the group.
SG1 and SG6 L. pneumophila isolates verified by immunofluorescence microscopy. ST367 and -461 strains have been isolated previously from both patients and environmental surveillance operations during outbreaks in Poland, Israel, and Spain (31)(32)(33). Notably, these ST367 and -461 L. pneumophila strains were demonstrated to be of SG6. To determine whether the 16 Southeast Michigan premise plumbing L. pneumophila ST367 and -461 isolates were also of SG6, serotype-specific antibodies were used to perform immunocytochemistry (Pathogen Control Associates, Peachtree Cor- ners, GA [data not shown]) and immunofluorescence microscopy. As expected, each of the ST367 and -461 L. pneumophila isolates from Flint and Detroit residences reacted with an SG6-specific antibody, whereas none cross-reacted with an antibody specific to SG1 antigens ( Fig. 3 and data not shown). Some SG6 L. pneumophila strains are sensitive to human serum. L. pneumophila serotyping is based on the structure of its lipopolysaccharide (LPS) O antigen, a feature of the outer membrane that also affects complement-mediated killing. Serogroup 1 L. pneumophila strains are resistant to killing by complement in human serum (34), whereas the serogroup 6 L. pneumophila Thunder Bay strain is sensitive, as judged by direct binding of the complement components C 3 and C 9 , as well as decreased bacterial  Table 1. The scale bar indicates the distance separating isolates that differ at one of the seven typing alleles. viability (35). To assess whether serogroup or strain origin affected the serum sensitivity of our Southeast Michigan strains, we incubated representative strains in nonimmune human serum or heat-inactivated human serum for 30 min at 37°C and then quantified viability by enumerating CFU. Consistently with the growth phase regulation of its LPS (36), the capacity of L. pneumophila to resist complement-mediated killing was growth phase dependent; serum killed Ͼ50% of exponential-phase cells of the serogroup 1 laboratory strains Lp01 and Lp02 but had little effect on the viability of cells cultured to the postexponential (PE) phase (Fig. 4). Complement-mediated killing accounted for most of the toxicity, since heat-treated serum had little effect on the viability of PE-phase bacteria. Although variability between strains and between experiments was observed, a few consistent patterns were noteworthy. Survival of~90% or more was observed for PE-phase cells of six of the seven clinical serogroup 1 isolates (C53, C32, C35, C46, C43, and C56) and the serogroup 1 premise plumbing isolate from Detroit (ED21). In contrast, three of the five ST367 serogroup 6 strains were serum sensitive: two of these isolates were from Flint premise plumbing (EF2, EF23), and the third (ME4774) was isolated from a water facility during the 2001 outbreak in Murcia, Spain (37,38). On the other hand, serum resistance was observed for two other ST367 strains (EF6, ED20) and three of the closely related ST461 serogroup 6 strains (EF11, EF14, EF15). Whether genetic differences within the genes encoding the LPS O antigen account for the degrees of serum sensitivity observed among and between the serogroup 1 and serogroup 6 strains analyzed here can be evaluated by genome sequencing.
Both SG1 and SG6 L. pneumophila isolates infect macrophages efficiently. According to an analysis of 3254 Legionnaires' disease cases from 1980 to 1989 (39), mortality after infection with SG6 L. pneumophila exceeds that caused by SG1 strains (63% versus 37%, respectively). To cause disease, L. pneumophila survives and replicates within macrophages (40). Therefore, as a first step to gauge the virulence of the SG6 L. pneumophila strains isolated from Southeast Michigan premise plumbing, their capacity to survive ingestion by macrophages was quantified.
Macrophages derived from mouse bone marrow were infected for 1 h with representative clinical (C) and environmental (EF, from Flint; ED, from Detroit) isolates, fixed, , clinical strains ("C" followed by a number), and environmental strains from Flint ("EF" followed by a number), Detroit ("ED" followed by a number), and Murcia, Spain (ME4774), to survive treatment with 90% human serum without (black circles) and with (gray circles) heat inactivation of complement were quantified by enumerating CFU before and after the 90-to 120-min incubation at 37°C. Shown are the means (solid or dotted horizontal lines) calculated in 2 to 4 independent experiments Ϯ SD. Brackets indicate STs within clonal complex A or E (subscripts). For all strains, PE-phase cells were analyzed. Where indicated, exponential-phase cells (E) were also studied. Serogroup 6 L. pneumophila in Premise Plumbing ® and labeled with serotype-specific anti-L. pneumophila antibodies, and their DNA was stained. The infected macrophages were then examined by fluorescence microscopy. As expected, 95% of cell-associated PE-phase cells of L. pneumophila laboratory strain Lp02 remained as intact rods, whereas nearly 60% of the exponential-phase bacteria were degraded by macrophages ( Fig. 5A) (41). As with the virulent PE-phase Lp02 control bacteria, for each of the clinical SG1 L. pneumophila strains, whether of ST1, ST213, or ST222, Ͼ85% of the PE-phase bacteria associated with macrophages remained rod shaped after a 1-h incubation. Likewise, Ͼ90% of the inocula of either SG1 or SG6 environmental strains of L. pneumophila evaded digestion by macrophages. The capacities to avoid degradation were similar for environmental strains of ST1, ST367, and ST461, as well as for isolates obtained from Flint and Detroit premise plumbing. The capacity of representative clinical isolates ("C" followed by a number, white bars) and environmental isolates (light-gray bars) from Flint ("EF" followed by a number) and Detroit ("ED" followed by a number) to avoid degradation within mouse macrophages 1 h after infection with PE cells was determined by fluorescence microscopy using serotype-specific antibodies and PE-and exponential-phase cells (E) of strain Lp02 as positive and negative controls (dark-gray bars), respectively. Shown are means Ϯ SD calculated from triplicate cell preparations from which at least 100 bacteria were scored as either rod or non-rod shaped. The data are representative of at least two independent experiments. (B) The capacity of each isolate indicated to bind to and survive on or within macrophages was determined for PE-phase cells by enumerating cell-associated CFU 1 h after inoculation of macrophage cultures. PE-phase wild-type Lp02 and dotA mutant cells served as positive and negative controls, respectively. Shown are means Ϯ standard errors (SE) calculated from triplicate cell cultures. The origin, sequence type, and serogroup of each strain are indicated. One-way analysis of variance was calculated by comparing each strain to the WT laboratory strain: *, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.001; ****, P Ͻ 0.0001. The means for the two ST461 and four ST367 strains were compared using Welch's t test (**, P Ͻ 0.01).
As a more rigorous test of infectivity, the capacity of representative clinical and environmental isolates to survive on or within macrophages was determined by enumerating cell-associated CFU 1 h after inoculation of the cultures. Four of the five clinical SG1 L. pneumophila strains survived their initial interaction with macrophages as well as or better than did the laboratory wild-type (WT) control strain (Fig. 5B). For example, 24% of the inoculum of the virulent PE-phase WT laboratory strain was cell associated and viable 1 h after addition to cultured macrophages, compared to 33 to 51% of the SG1 ST213 and ST222 clinical strains. Likewise, 1 h after infection, four of the environmental strains infected macrophages more efficiently than did the WT laboratory strain; 35 to 58% of the inoculum was viable and cell associated for the ST1 and ST777 SG1 strains and for both ST461 SG6 strains. Compared to ST461 strains, ST367 SG6 strains infected macrophages less efficiently (P Ͻ 0.01), with the infective cells ranging from 16 to 30% of each inoculum, a rate similar to that of the WT laboratory control strain. The single Detroit ST367 isolate (ED20) infected macrophages at a frequency similar to that of the three ST367 strains from Flint (EF1, EF3, and EF16). Thus, as judged by both direct and viable-cell-count assays, the clinical and environmental strains were competent to infect cultured mouse macrophages.
SG1 and SG6 L. pneumophila isolates each replicate efficiently in macrophages. The capacity of the environmental and clinical isolates to replicate within macrophages was first evaluated by microscopy. Infected macrophages were incubated for 20 h, fixed, and labeled with serotype-specific antibodies, and their DNA was stained. The WT laboratory strain replicated profusely during the primary infection period, unlike the negative-control strain, an avirulent dotA mutant (Fig. 6A). Likewise, numerous L. pneumophila progeny were visible within macrophages infected for 20 h with either of two different SG1 ST1 isolates, whether obtained from a patient (strain C45) or Flint premise plumbing (strain EF10). A similar pattern was observed for two SG1 ST213 clinical strains and representative SG6 ST461 and ST367 environmental strains.
To quantify L. pneumophila intracellular replication, CFU were enumerated 1, 24, 48, and 72 h after infection of mouse macrophages. During the 72-h infection period, the yield of each of the SG1 clinical strains, whether of ST213, ST222, or ST1, increased 120to 850-fold (Fig. 6B). Some strain-to-strain variation was apparent; for example, the yield of the two ST1 isolates increased 150-and 120-fold, whereas that of the two ST213 isolates increased 650-and 850-fold. Similarly, the number of CFU increased by Ͼ100 for six of the eight environmental strains analyzed, a group that included strains of SG1 ST1 and -777 and SG6 ST461 and -367. A more modest~50-fold increase in the number of CFU was observed for two strains (EF14 [SG6 ST461] and EF3 [SG6 ST367]). It was curious that seven of the eight environmental L. pneumophila strains achieved their maximal yield in CFU at 48 h, whereas two of the five clinical isolates and the laboratory strain did so at 72 h. Whether alterations to the bacteria and/or macrophages contribute to this kinetic difference requires more detailed analysis. Overall, regardless of their SG, ST, or origin in Flint or Detroit, each of the clinical and the environmental L. pneumophila isolates from Southeast Michigan had the capacity to infect, survive in, and replicate within cultured mouse macrophages, laboratory indicators of the potential to cause disease.

DISCUSSION
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  isolates ("C" followed by a number) and environmental isolates from Flint ("EF" followed by a number) and Detroit ("ED" followed by a number) of the serogroup and sequence type shown and then fixed and stained with FITC-labeled and serotype-specific antibodies. PE-phase Lp01 L. pneumophila served as the positive control at both 1 and 20 h; the negative control was dotA mutant bacteria after 20 h. All bacteria and macrophages were visualized by staining their DNA with DAPI (not shown). Similar patterns were observed in two other independent experiments. (B) The capacity of representative clinical and environmental isolates from Flint and Detroit to replicate in mouse macrophages was quantified by enumerating total CFU per sample 1, 24, 48, and 72 h after infection. Shown are means calculated from duplicate or triplicate samples at each time indicated, divided by the mean cellassociated CFU 1 h after infection. Error bars indicate SD. One-way analysis of variance was calculated by comparing the maximum yield of each strain to that of the WT laboratory strain: **, P Ͻ 0.01; ****, P Ͻ 0.0001. Similar results were obtained in two additional independent experiments, with the exception of C53, which was analyzed twice. 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 Ͻ 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 Ϫ26°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, Ͻ5% 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 Ͼ90% 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 Ͼ10% 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.

MATERIALS AND METHODS
Environmental surveillance. To evaluate the prevalence of legionellae in Flint premise plumbing, of the 187 total residences enrolled in the study, 130 were sampled and their water cultured according to the protocol described here. The majority (127) of homes were selected at random from Genesee County from tax parcel records; three other residences were selected from a list of potential self-identified participants, with preference given to residents based on perceived risk of acquiring Legionnaires' disease. Once residences were selected, institutional review board (IRB)-approved recruitment letters were mailed to each property. Within 2 weeks, study personnel knocked on residents' doors and attempted to obtain informed consent. If consent was obtained, a three-person team returned to collect environmental samples, complete an epidemiological survey focused on legionellosis risk factors, perform a psychosocial needs and resilience assessment, and provide referrals to social services if necessary.
From these 130 residences, a total of 357 water and biofilm samples were collected. At each location, when possible, 2-liter water samples were collected from the hot water tank drain valve, the primary shower used by residents, and the kitchen sink after 5 min of flushing (bypassing point-of-use filters when present). Each 2-liter water sample was collected into polyethylene terephthalate glycol (PETG) bottles containing 0.5 ml of 0.1 N sodium thiosulfate to remove any residual chlorine. After gentle mixing, Serogroup 6 L. pneumophila in Premise Plumbing ® shipped overnight to Pathogen Control Associates (Peachtree Corners, GA) for culturing. The remaining portion of the 2-liter sample was delivered to Henry Ford Hospital for further analysis (data not shown). To collect shower samples, the shower head was removed prior to the 2-liter first-draw water sample being collected, using hot water only. Once the shower water was collected, a biofilm sample was collected from the shower pipe neck by using a sterile 6-inch swab with a polyester tip, which was then placed in a 15-ml sterile Falcon tube with 5 ml of phosphate-buffered saline (PBS).
Legionella isolation from water samples. Water samples collected from the residential surveillance operation were cultured for legionellae by Pathogen Control Associates personnel using methods compatible with the relevant CDC laboratory protocol (59) and the International Organization for Standardization 11731:1998 standard (60). To identify suspect colonies to the genus level, colony characteristics were examined microscopically, and their requirement for L-cysteine was tested. To identify candidate isolates to the species and serogroup levels, monovalent and polyvalent direct fluorescent-antibody reagents were used in the direct fluorescence antibody and/or slide agglutination test as described previously (61,62). To eliminate the risk of experimental bias, premise plumbing samples were cultured by staff of Pathogen Control Associates Laboratories, who then transferred all legionella isolates to the Henry Ford Hospital System for coding, storage, and transfer to the University of Michigan for analysis (Table 1). From each culture-positive premise plumbing sample, only one L. pneumophila isolate was obtained and characterized. Therefore, one limitation of our study is that it cannot address how often residences are colonized by more than one strain of L. pneumophila, as recommended by ISO 11731:1998 (60).
Clinical isolates. Via a material transfer agreement, the Michigan Department of Health and Human Services Bureau of Laboratories provided 33 clinical isolates submitted from 2013 to 2016 by hospitals in Genesee, Wayne, and Oakland Counties in Michigan (Table 1). All specimens were deidentified prior to shipment to the University of Michigan Medical School.
Sequence-based typing. Genotypes were assigned using the sequence-based typing protocol developed by the European Working Group for Legionella Infections (65,66). Briefly, after extracting genomic DNA from an isolated colony, seven designated gene segments (flaA, pilE, asd, mip, mompS, proA, and neuA) were amplified by PCR. After purification, the DNA products were sequenced using the designated forward and reverse primers (65,66). To assign allele numbers, each DNA sequence was trimmed in silico by the Legionella SBT quality tool (http://www.hpa-bioinformatics.org.uk/cgi-bin/ legionella/sbt/seq_assemble_legionella1.cgi) and then used to query the L. pneumophila database (http://www.hpa-bioinformatics.org.uk/legionella/legionella_sbt/php/sbt_homepage.php). Together, the list of seven assigned allele numbers defines each strain's numerical sequence type (e.g., ST1). Novel alleles and STs were submitted to the ESGLI database. SBT analysis was performed on deidentified samples at the University of Michigan. Likewise, 33 clinical isolates were deidentified, coded, and shipped by the MDHHS to the University of Michigan for blind SBT analysis.
Serotyping. The serogroup designations for clinical and environmental isolates initially determined by clinical and commercial laboratories, respectively, were verified by immunolocalization of cellassociated bacteria 1 h after inoculation of mouse macrophage cultures using fluorescein isothiocyanate (FITC)-labeled serotype-specific antibodies. An L. pneumophila SG1-specific rabbit polyclonal antibody (Pierce PA1-73140) was diluted 1:1,000, and an L. pneumophila SG6-specific antibody (prepared by the CDC and a gift of Paul Edelstein of the University of Pennsylvania School of Medicine) was diluted 1:200. All bacteria and macrophages were visualized by staining their DNA with DAPI (4=6-diamidino-2phenylindole) (not shown).
Phylogenetic analysis. Phylogenetic relationships among the 18 environmental and 33 clinical L. pneumophila isolates were evaluated using the v3 eBURST online tool (http://eburst.mlst.net/). We defined a clonal complex as strains of those STs that share at least four of the seven alleles with at least one other member of the group.
Immunofluorescence microscopy. Microscopy was performed by culturing 2.5 ϫ 10 5 macrophages on 12-mm glass coverslips overnight prior to infection with PE-phase bacteria at a multiplicity of infection of 1. After the infection period, cultures were fixed and stained as described previously (41) using an L. pneumophila SG1-specific rabbit polyclonal antibody conjugated to FITC (Pierce; PA1-73140), which was diluted 1:1,000, and the FITC-conjugated L. pneumophila SG6-specific antibody (prepared by the CDC and a gift of Paul Edelstein), which was diluted 1:200. All bacteria and macrophages were identified by labeling them with the DNA stain DAPI, which was purchased as ProLong Gold antifade reagent (Molecular Probes) and included in the mounting medium.
Serum sensitivity. To quantify the sensitivity of L. pneumophila to human complement, human serum was purchased (Fisher Scientific; BP2657, lot 174248) and stored at Ϫ20°C. As a negative control, complement was heat inactivated by incubating an aliquot of serum at 56°C for 30 min just before use. PE-phase L. pneumophila cells (OD 600 , 3.5 to 4.1) were collected by centrifugation of 600-l culture aliquots at 5,000 rpm for 5 min, washed once with 1 ml of PBS, and resuspended in 0.5 ml of PBS. After dilution of duplicate 75-l cell aliquots into 5 ml of PBS, 50 l of cells was diluted into 450 l of nonimmune or heat-inactivated human serum to achieve a final concentration of approximately 5 ϫ 10 6 cells per ml of 90% serum. After a static incubation at 37°C for 90 to 120 min, cells were serially diluted in PBS and then plated in duplicate on CYE to enumerate CFU. In parallel, the CFU of each untreated inoculum was enumerated after serial dilution of duplicate aliquots. For each strain, the percent survival was calculated as (CFU in serum/CFU in inoculum) ϫ 100 from the mean of results from at least six aliquots per experiment in 2 to 4 independent experiments. Intracellular growth. The growth of L. pneumophila strains in bone marrow-derived macrophages from A/J mice (Jackson Laboratories) was assessed by enumerating CFU on CYE (T) medium essentially as described previously (67). PE-phase wild-type Lp01 and dotA mutant cells served as positive and negative controls, respectively. Shown are means Ϯ standard deviations (SD) calculated from triplicate cell preparations. The sequence type, serogroup, and origin of each strain are indicated in Table 1.