Cases of pneumonia reported in Alcoi during the incubation period were considered suspected cases of legionellosis unless this diagnosis could be ruled out. A confirmed case of legionellosis was defined as a case of pneumonia with laboratory evidence of acute infection with L. pneumophila including 1) isolation of L. pneumophila serogroup 1 from respiratory secretions or lung tissue, 2) a >4-fold rise in antibody titers (from 128) against L. pneumophila serogroup1 by immunofluorescence in paired acute- and convalescent-phase serum specimens, or 3) detection of L. pneumophila serogroup 1 antigen in urine.

Cases that conformed to the case definition and/or had positive results in at least 1 of the following tests were considered suspected cases: 1) high (>256) antibody titer against L. pneumophila serogroup 1 in convalescent-phase serum, 2) seroconversion to L. pneumophila serogroup 1 by indirect inmunofluorescence in acute- and convalescent-phase serum, or 3) direct staining of bacteria in respiratory secretions or lung tissue by direct fluorescence using monoclonal or polyclonal antibodies against any Legionella spp. or serogroup, including serogroup 1.

A questionnaire was used to obtain personal information (age, gender, job, and free-time activities), clinical features (signs and symptoms, date of illness onset, date of hospitalization, previous hospitalizations, and medical history), predisposing and risk factors, place of residence, and recent urban mobility within Alcoi. Information about the patients’ addresses and usual itineraries (roads and places visited) in Alcoi was used to delimit an area of influence, or a buffer, for the location of the likely source of infection. A 500-m radius was considered around home and city itineraries for each patient by using vectorial cartography (scale 1:10,000), ortophotography (scale 1:5,000), and ArcView software (www.esri.com). The buffer for each patient was thus represented by a polygonal area; intersecting areas yielded a common area representing a likely location for the source of the outbreak.

Systematic environmental investigations are regularly performed in the municipal water distribution system, but when this outbreak was detected, an active search for L. pneumophila was made in patients’ homes (bulk water and biofilms from showerheads and taps) and the water distribution system (bulk water); results were negative. None of the other usual sources (e.g., public fountains, cooling towers, humidifiers) were found to pose a risk. The absence of usual risk sources led us to consider other possible sources of aerosols, including moving devices used in street cleaning and asphalt repaving; the latter had been observed in the risk area during the epidemiologic inspection.

For cultured isolates, the 7 loci in the European Working Group for Legionella Infections (EWGLI, www.ewgli.org) typing scheme were amplified as detailed by Gaia et al. (17) and Ratzow et al. (18). DNA from respiratory samples was amplified by using a seminested approach. For the first PCR we used the same primers and conditions as for cultured isolates. For the second PCR we used the internal primers described by Coscollá et al. (22) and used 2 μL of the first PCR products as template. Amplification conditions and profiles were the same as for the first PCR except for the annealing temperature (22). PCR products were purified by using a High Pure PCR Product Purification Kit (Roche Diagnostics, Mannheim, Germany). PCRs were conducted in a reaction volume of 50 μL containing 20 ng of genomic DNA, 1 U of Taq DNA polymerase, 200 μmol/L of each dNTP, 2.5 mmol/L of magnesium-free buffer, 2.5 mmol/L MgCl₂, and 0.2 μmol/L of each pair of primers. We adopted the gene notation used in the first L. pneumophila genome published (24); consequently, locus fliC corresponds to locus flaA used in the EWGLI typing scheme (17).

Purified DNA was directly sequenced by the dideoxy method by using a BigDye Terminator v3.0 Ready Reaction Cycle Sequencing Kit and was analyzed in an ABI PRISM 3700 sequencer (each from Applied Biosystems, Foster City, CA, USA). Sequencing of PCR products from respiratory samples differed only in the use of the same internal primers used in the second PCR of the seminested approach. Sequence chromatogram files were analyzed by using the Staden sequence analysis package (25).

Allelic profiles for the 7 sequenced genes were obtained from EWGLI and were aligned and compared with sequences derived in this study. Multiple sequence alignments were obtained by using ClustalX (26) and further refined by visual inspection.

A phylogenetic reconstruction was obtained with the concatenated alignment of sequences from the 7 loci analyzed. Models of nucleotide substitution were assessed by using the maximum-likelihood approach implemented in jModeltest (27). Maximum-likelihood phylogenetic trees were obtained with PHYML 3.0 (28) by using the previously derived models of nucleotide substitution for each locus. Support for the nodes was evaluated by bootstrapping with 1,000 pseudoreplicates.

Among patients with positive urine antigen test results, 11 cases of legionellosis were confirmed and L. pneumophila was isolated from 4. All patients required hospitalization, and all except 1 recovered. (The patient who did not recover had severe signs and symptoms and subsequently died.) The main signs and symptoms were fever (100% incidence), pneumonia (100%), headache (27.3%), myalgia (27.3%), diarrhea and/or vomiting (18.2%), and confusion (45.5%). The average age was 70 years, range 49–88 years. More men than women were affected (male:female ratio = 4.5).

Confirmed cases occurred from July 21 through September 17. According to the date of disease onset, the outbreak showed 3 epidemic waves: 2 cases in the second half of July, 8 cases in the first half of August, and 1 case in the second half of August (Figure 1).

No common indoor source of exposure was found, and the initial hypothesis was that the outbreak originated from environmental contamination of an unknown source capable of producing and dispersing large quantities of aerosols contaminated with L. pneumophila. The first 2 patients lived in the northern part of the city, which suggested that the source could be located in that area. The spatial distribution of patients’ buffers changed in August, thus indicating that the likely source of the outbreak had moved to the Santa Rosa quarter (Figure 2). The area of epidemic risk was modified accordingly, and the search for putative environmental sources focused on that neighborhood.

L. pneumophila was not isolated from samples derived from the municipal water supply, traditional risk facilities, and patients’ houses. The environmental investigation was extended to other potential sources, especially portable cleaning devices such as sweepers, hydrocleaners, and water tanks used to clean the streets, all of which used water from the municipal water supply. At that time, the Santa Rosa quarter was being repaved, and 1 of the machines used in the repaving process was a tank truck that carried water used by a large milling machine. The water in the tank was obtained from a natural spring untreated with chlorine or anything else. Because the average daytime temperature in Alcoi during July and August is 27°C, the water in the machine might have been warm enough for L. pneumophila growth. The milling machine had been working north of Alcoi around July 15 and in the Santa Rosa neighborhood from July 31 through August 20 (Figure 2). This activity fits spatially and temporally with the incubation period of confirmed cases. This machine was identified and removed from service on August 21, thus was able to cause the last infection detected on August 23 (Figure 2).

L. pneumophila serogroup 1 was isolated from the water in the tank and from the atomizers in the milling machine (Figure 3). L. pneumophila was also isolated from other machines used for street cleaning in the city, but these organisms were from serogroups other than serogroup1.

The milling machine and the water tank were immediately confined outside the city, sealed, cleaned, and disinfected. When the machine was put back into use, a new cleaning and maintenance protocol was implemented to prevent future contamination. The cleaning protocol consisted of treatment with chlorine (20 ppm) and removal of the atomizers and their replacement by gravity-based water distributors. Additionally, for operation in the city, use of a separate thermo-insulated water tank in which chlorine was continuously applied at 20 ppm was required, thus preventing any further growth of L. pneumophila. Additionally, water had to be obtained from the municipal system.

L. pneumophila sequence information was obtained from 7 uncultured sputum samples and from 4 cultures derived from the 11 patients studied. Complete allelic profiles, according to the EWGLI typing scheme, were obtained for 9 clinical samples in the study. All had the same profile, which corresponded to sequence type (ST) 578 in the EWGLI database (Table 1). Partial allelic profiles for the other 2 samples were consistent with ST578. The 9 L. pneumophila isolates derived from putative environmental sources, including the milling machine and the water tank (Table 2), were also characterized and corresponded to 4 STs: ST578 in 4 samples, ST1 in 3 samples, ST777 in 1 sample, and a new ST in sample 4159.

A phylogenetic tree was obtained from the concatenated alignment of the 2,984 bp (Figure 4). The phylogenetic analysis showed that 4 environmental samples were identical to the clinical samples (represented by ST578 C/E in Figure 4). Their comparison with the remaining environmental samples showed 14–50 nt differences to strains 4159 and 4143/4160/7970 (represented by ST1 in Figure 4), respectively. The comparison of these to reference strains showed that outbreak samples were more closely related to the Corby strain (29), a virulent human isolate from which it differed by only 6 nt.