Sequence-Based Typing of Legionella pneumophila Serogroup 1 Offers the Potential for True Portability in Legionellosis Outbreak Investigation

Abstract

Seven gene loci of Legionella pneumophila serogroup 1 were analyzed as potential epidemiological typing markers to aid in the investigation of legionella outbreaks. The genes chosen included four likely to be selectively neutral (acn, groES, groEL, and recA) and three likely to be under selective pressure (flaA, mompS, and proA). Oligonucleotide primers were designed to amplify 279- to 763-bp fragments from each gene. Initial sequence analysis of the seven loci from 10 well-characterized isolates of L. pneumophila serogroup 1 gave excellent reproducibility (R) and epidemiological concordance (E) values (R = 1.00; E = 1.00). The three loci showing greatest discrimination and nucleotide variation, flaA, mompS, and proA, were chosen for further study. Indices of discrimination (D) were calculated using a panel of 79 unrelated isolates. Single loci gave D values ranging from 0.767 to 0.857, and a combination of all three loci resulted in a D value of 0.924. When all three loci were combined with monoclonal antibody subgrouping, the D value was 0.971. Sequence-based typing of L. pneumophila serogroup 1 using only three loci is epidemiologically concordant and highly discriminatory and has the potential to become the new “gold standard” for the epidemiological typing of L. pneumophila.

Legionellosis (infection by members of the genus Legionella) can range from mild respiratory illness to acute life-threatening pneumonia and is invariably acquired from an environmental source. Characterization of clinical and epidemiologically linked environmental isolates of Legionella is invaluable in locating the source and extent of infection, allowing implementation of corrective measures and treatment to prevent further infection. The majority of cases of legionellosis are caused by Legionella pneumophila, particularly serogroup (sg) 1 (20). Previous studies have demonstrated the utility of differentiating isolates belonging to this serogroup of L. pneumophila in order to confirm or refute epidemiological associations in outbreak investigations (5, 19, 22, 28, 29), but only in a local setting.

Isolates of L. pneumophila sg 1 can be rapidly subtyped by using monoclonal antibody (MAb) subgrouping with panels based on the international MAb subgrouping panel (13, 14). However, MAb subgrouping lacks discrimination, dividing L. pneumophila sg 1 into only 8 to 10 phenons, depending on which MAb panels are used (9). Consequently, in addition to this phenotypic typing method, a number of molecular methods have been successfully employed locally for epidemiological purposes. These include ribotyping, amplified fragment length polymorphism (AFLP) analysis, pulsed-field gel electrophoresis (PFGE), restriction fragment length polymorphism (RFLP) analysis, restriction endonuclease analysis (REA), and arbitrarily primed PCR (9). One of these methods, AFLP analysis (26), has been adopted as an international standard (10) and is now widely used by members of the European Working Group for Legionella Infections (EWGLI). Two recent interlaboratory studies (10, 11) have assessed the reliability and discriminatory power of AFLP analysis for epidemiological typing of L. pneumophila sg 1 isolates. While these studies have demonstrated that the method is robust and rapid and allows the exchange of data between laboratories without the need for exchange of isolates, a program of proficiency testing showed that a significant proportion of laboratories could not correctly identify isolates 100% of the time (10). Incorrect identification was typically a result of data analysis issues (e.g., difficulties in comparing the AFLP banding patterns) rather than of the method itself.

For this reason, we sought to develop a universal typing method, which would be simple, rapid, discriminatory, and truly “portable” and which might in the future be applied by all microbiology laboratories to aid in the investigation of legionellosis.

Multilocus sequence typing (MLST) was first described for Streptococcus pneumoniae by Enright and Spratt (8), who used seven housekeeping genes. Subsequently, several such schemes using nonselective housekeeping genes have been described (http://www.mlst.net). We prefer to reserve the use of the term MLST for schemes such as those described by Enright and Spratt (8) and others that use housekeeping genes (not under selective pressure) and to use “sequence-based typing” (SBT) as a more universal term to describe other schemes not using such targets.

The advantages of sequence data are now well established and include portability between laboratories and the consequent ease with which global databases can be established (21). For this study, the genes chosen included four likely to be selectively neutral (acn, groES, groEL, and recA) and three likely to be under selective pressure (flaA, proA, and mompS). The groESL genes (also known as hsp10/60 and htpA/B) (17, 24) encode the 10-kDa (GroES) and 58-kDa (GroEL) common antigens in L. pneumophila. The 58-kDa common antigen was formerly designated the 60-kDa common antigen (24). The acn gene encodes a major iron-containing protein of L. pneumophila showing aconitase activity (23), and the recA gene encodes the RecA protein of L. pneumophila (30). The flaA gene encodes the flagellum subunit protein of L. pneumophila sg 1 (15), proA encodes a zinc metalloprotease found in L. pneumophila (3), and mompS encodes an outer membrane protein of ca. 29 kDa (6; E. Christoph and W. Ehret, unpublished data).

The aims of this study were (i) to determine the suitability of the various primer pairs and gene targets for amplification and sequencing and (ii) to evaluate an SBT approach, which could be used for the investigation of outbreaks of legionellosis caused by L. pneumophila.

MATERIALS AND METHODS

Bacterial strains.

Ninety-five clinical and environmental isolates of L. pneumophila sg 1 from 10 European countries comprising one epidemiologically “unrelated” panel of 79 isolates (panel 1), one epidemiologically “related” panel of 16 isolates (panel 2), and one stability panel of 5 isolates (panel 3) were analyzed (Table 1). Each of these isolates was obtained from the EWGLI Legionella culture collection and has a unique EUL number (http://www.ewgli.org). This collection of isolates was established by members of the EWGLI to facilitate epidemiological typing studies (9, 10, 11). For the unrelated panel (panel 1), 10 laboratories provided their 10 most recent (i.e., consecutive) endemic clinical isolates that conformed to the following criteria: each isolate (i) was obtained from a patient resident within that country during the 2 weeks before onset of symptoms and (ii) was not epidemiologically related (temporally or geographically) to any other isolate included in the panel. Not all laboratories were able to supply 10 isolates, and 11 were obtained from one laboratory, so 79 unrelated isolates were obtained. The epidemiologically related panel (panel 2) of 16 isolates comprised five sets of epidemiologically related isolates and two replicates of the same isolate. The stability panel (panel 3) comprised five variants of the same strain. Each isolate has been extensively characterized previously (1, 2, 9, 10, 11).

TABLE 1.

Isolates of Legionella pneumophila serogroup 1 characterized by SBT

Panel and EUL no.	Sender designation	Country of origin	MAb subgroup	Related strain	Evidence of relatedness	Characteristics and sourcea
Epidemiologically unrelated panel (panel 1)
1	IBS-2	Switzerland	Philadelphia
2	IBS-25	Switzerland	Knoxville
3	IBS-026	Switzerland	Philadelphia
4	IBS-308	Switzerland	Allentown
6	IBS-319	Switzerland	Benidorm
7	IBS-320	Switzerland	Allentown
8	IBS-323	Switzerland	Allentown
13	83/41091	Scotland	Benidorm
14	84/11316	Scotland	Benidorm
16	84/51978	Scotland	Benidorm
17	93/8188	Scotland	Philadelphia
18	94/26760	Scotland	Bellingham
19	94/51727	Scotland	Knoxville
20	95/9654	Scotland	Benidorm
25	L3	France	Allentown
26	L12	France	OLDA
27	L13	France	Benidorm
28	L23	France	Allentown
29	L27	France	Knoxville
30	L48	France	France
31	L51	France	Allentown
32	L52	France	Benidorm
33	L215	France	France
36	1	Italy	Knoxville
37	2	Italy	Philadelphia
38	3	Italy	OLDA
39	4	Italy	Benidorm
40	5	Italy	Philadelphia
41	6	Italy	Allentown
42	7	Italy	Philadelphia
43	8	Italy	Philadelphia
48	006/96	Spain	Bellingham
49	16/96	Spain	Knoxville
50	60/96	Spain	Benidorm
51	16140/95	Spain	Benidorm
52	13195/95	Spain	Philadelphia
53	6332/95	Spain	OLDA
54	2691/94	Spain	France
55	2879/94	Spain	OLDA
60	001/92	Greece	Philadelphia
63	007/93	Greece	Knoxville
66	013/86	Greece	Knoxville
67	16/95	Greece	OLDA
68	LC3598	England and Wales	Benidorm
69	LC3720	England and Wales	Philadelphia
70	LC3759	England and Wales	Allentown
71	LC3868	England and Wales	Allentown
72	LC3771	England and Wales	Philadelphia
73	LC3832a	England and Wales	Philadelphia
74	LC3290	England and Wales	Philadelphia
75	LC3196	England and Wales	Benidorm
81	R25	Denmark	Bellingham
82	R86	Denmark	OLDA
83	R208	Denmark	Benidorm
84	R232	Denmark	OLDA
85	R239	Denmark	OLDA
86	R270	Denmark	Benidorm
87	R278	Denmark	Knoxville
88	R283	Denmark	OLDA
91	R293	Denmark	Bellingham
92	L279	Denmark	Bellingham
93	L437	Denmark	Oxford
97	LD91/94	Sweden	Knoxville
98	LD66/96	Sweden	Knoxville
99	LD35/95	Sweden	Bellingham
100	LD5/95	Sweden	Bellingham
101	LD10/94	Sweden	Benidorm
102	LD162/93	Sweden	Bellingham
103	LD127/93	Sweden	OLDA
104	LD320/92	Sweden	Oxford			/PICK>
105	LD689/91	Sweden	Benidorm
110	Un1-S 763	Germany	OLDA
111	Un2-Berlin1	Germany	Benidorm
112	Un3-Hannover1	Germany	OLDA
116	Un7-Wien60	Germany	Benidorm
117	Un8-Wien3	Germany	Benidorm
118	R1-P 281	Germany	Philadelphia
119	R2-Ulm2	Germany	OLDA
120	R4-Augsburg1	Germany	Benidorm
Epidemiologically related panel (panel 2)
120	R4-Augsburg1	Germany	Benidorm
121	R4-Augsburg1	Germany	Benidorm	EUL 120	Duplicate of EUL 120
73	LC3832a	England and Wales	Philadelphia
78	LC3832b	England and Wales	Philadelphia	EUL 73	Clinical isolate from same patientb
79	LC3832c	England and Wales	Philadelphia	EUL 73	Clinical isolate from same patient
71	LC3868	England and Wales	Allentown
76	LC3869	England and Wales	Allentown	EUL 71	Clinical isolate from same patientc
77	LC3870	England and Wales	Allentown	EUL 71	Clinical isolate from same patient
48	006/96	Spain	Bellingham
56	17/96	Spain	Bellingham	EUL 48	Clinical isolate from same patient
40	5	Italy	Philadelphia		Clinical isolate
47	12	Italy	Philadelphia	EUL 40	Related environmental isolate
140	1956X/96	Spain	Knoxville		Clinical isolate from a Madrid outbreak
141	2099X/96	Spain	Knoxville	EUL 140	Clinical isolate from same outbreak
142	208/96	Spain	Knoxville	EUL 140	Related environmental isolate
143	209/96	Spain	Knoxville	EUL 140	Related environmental isolate
Stability panel (panel 3)
135	Corby RA/LC 4404	England and Wales	Knoxville			Corby variant obtained after 7 passes in amoebae
136	Corby CA/LC 4405	England and Wales	Knoxville			Corby variant obtained after 100 passes on MH agar
137	Corby CAC/LC 4406	England and Wales	Knoxville			Corby variant obtained after 100 passes on BCYE agar
138	Corby Rif/LC 4407	England and Wales	Knoxville			Corby variant from a rifampin mutant
139	Corby 3/1-/LC 4408	England and Wales	Oxford			Corby variant from a MAb 3/1-mutant

^aMH, Mueller-Hinton; BCYE, buffered charcoal yeast extract.

^bEUL 73, 78, and 79 are each a single colony picked from the primary isolation plate.

^cEUL 71 was isolated from sputum by direct culture; EUL 76 was isolated from sputum via amoebal culture; and EUL 77 was isolated from feces by direct culture.

Study design.

The choice of nonselective and selective gene targets was based on the availability of sequence data for one or more strains of L. pneumophila. The SBT method was evaluated as outlined in the Consensus Guidelines of the European Study Group on Epidemiological Markers (25) and in previous studies (9, 10, 11). Typeability (T) was calculated as the proportion of isolates assigned to a type by sequencing of each target gene. Reproducibility (R) was assessed by analysis of one or more loci from 10 of the isolates in both our laboratories using a number of different conditions (including reagent manufacturers) for DNA extraction, PCR, and DNA sequencing. Initially, sequence data were obtained for the seven genes acn, flaA, groES, groEL, mompS, proA, and recA from 10 of the 16 isolates in the epidemiologically related panel (panel 2, EUL no. 48, 56, 71, 73, 76 to 79, 120, and 121), comprising four sets. The three genes yielding the maximum number of possible sequence types in these four sets, i.e., four sequence types in four sets, were chosen for further analysis of the remaining 6 isolates in panel 2 (EUL no. 40, 47, and 140 to 143) and of the 79 isolates in the epidemiologically unrelated panel (panel 1). Epidemiologic concordance (E) was expressed as the number of epidemiologically related sets of strains found to be indistinguishable by the typing system, divided by the total number of sets in the panel. The index of discriminatory power (D) was determined by calculating Simpson's index of diversity, as described by Hunter and Gaston (18), by the equation

where N is the total number of strains in the sample population, S is the total number of types described, and n_j is the number of stains belonging to the jth type (18).

Individual and combined D values were calculated for SBT by using flaA, proA, and mompS, together with MAb subgrouping results from previous studies. The stability of these three genes, targets that are likely to be under selective pressure (flaA, proA, and mompS), was assessed by analysis of the five variants of the same strain (Table 1, panel 3).

DNA extraction and PCR amplification.

Genomic DNA was extracted using the Nucleon BACC2 DNA extraction kit (Amersham Pharmacia Biotech, Little Chalfont, Buckinghamshire, United Kingdom) or the InstaGene Purification Matrix (Bio-Rad, Hercules, Calif.). Oligonucleotide primers targeting regions of each of the genes were designed to amplify 279- to 763-bp products encompassing regions of variation (Table 2). Amplification using PCR was performed in a reaction volume of 50 μl. Each reaction mixture contained 2.5 mM MgCl₂, 10 mM Tris-HCl (pH 8.3), 50 mM KCl, 200 μM (each) deoxynucleotide, 10 pmol of each primer (MWG Biotech Ltd., Milton Keynes, United Kingdom, or Microsynth AG, Balgach, Switzerland), and 1.0 to 2.5 U of Taq polymerase (Life Technologies, Paisley, United Kingdom, or Qiagen, Basel, Switzerland). Template DNA (ca. 10 to 100 ng) was added, and reaction mixtures with no added DNA served as negative controls. Amplification was performed in a DNA Engine (MJ Research) or a GeneAmp PCR system (Perkin-Elmer Applied Biosystems, Foster City, Calif.) using the following conditions: 35 cycles of 30 s at 94°C, 30 s at 50°C, and 30 s at 72°C.

TABLE 2.

Oligonucleotide primers designed for this study

Primer name	Sequence (5′-3′)	Gene	Positionsa	GenBank accession no. of reference sequence	Reference
groES-F	TCGTCCTTTACACGATCG	groES	161-178	X57520	16
groES-R	ACACCAGCAAGCATTTGT	groES	523-506
groEL-F	TGCTGCAGTAGAAGAAGG	groEL	1419-1436	M91673	24
groEL-R	GTTGGATCAAGAATACC	groEL	1697-1681
recA-F	ATCGATGCGGAACATGC	recA	442-458	X55453	30
recA-R	ACGGCGAATATCCAAACG	recA	846-829
acn-F	CAGGGATTTGCCATCAGG	acn	1379-1396	L22081	23
acn-R	GGTTATCGCTGCAATGAC	acn	2142-2125
proA-F	GATCGCCAATGCAATTAG	proA	1090-1107	M31884	3
proA-R	ACCATAACATCAAAAGCC	proA	1570-1553
mompS-F	GACATCAATGTGAACTGG	mompS	492-509	AF078136	Christoph and Ehret, unpublished
mompS-R	CAGAAGCTGCGAAATCAG	mompS	1032-1015
flaA-F	TTTCTCTGGCGCAAGCTTCC	flaA	625-644	X83232	15
flaA-R	GCTGCTTTGGCATAGGCAG	flaA	870-852

^aPositions of all primers are given according to the numbering of the reference sequence.

Purification and DNA sequencing.

Amplicons were purified using the Wizard PCR Preps purification system (Promega Corp., Madison, Wis.) or the Qiaquick PCR purification kit (Qiagen). Nucleotide sequences were determined by using the primers used for amplification together with either (i) the Dye Terminator Cycle Sequencing kit (Beckman Coulter), in which case the products were analyzed on a CEQ 2000 XL DNA analysis system or a CEQ 8000 genetic analysis system (Beckman Coulter), or (ii) the ABI PRISM BigDye Cycle Sequencing Ready Reaction kit with AmpliTaq DNA polymerase FS (Applied Biosystems), in which case the products were analyzed on a model 310 ABI DNA sequencer (Applied Biosystems). Sequencing and analysis were performed according to the instructions supplied by the manufacturers.

Sequence analysis.

Sequence analyses were performed using the Editseq and Megalign programs included in the Lasergene package (DNASTAR, Inc., Madison, Wis.); Kodon, version 1.0 (Applied Maths, Kortrijk, Belgium); or BioNumerics, version 3.0 (Applied Maths). For all analyses, data from forward and reverse sequencing primers were combined and aligned manually.

Data analysis.

On completion of the sequencing, the complete data set was analyzed in BioNumerics (Applied Maths).

Nomenclature.

For each gene of L. pneumophila analyzed, identical sequences were assigned the same allele number [e.g., flaA(1)] and different sequences were assigned distinct allele numbers [e.g., flaA(1), flaA(2), flaA(3), etc.]. For each isolate, the combination of alleles at each of the loci was defined as the sequence type or allelic profile (e.g., 1,2,3) by using a predetermined order, i.e., flaA, proA, mompS. Where sequence data were unavailable, a zero was entered into the sequence string.

Nucleotide sequence accession numbers.

The flaA, proA, and mompS sequences from the 95 isolates of L. pneumophila sg 1 are available from the authors or at http://www.ewgli.org.

RESULTS

Typeability.

All isolates included in the study, except two, yielded PCR products of the expected sizes and DNA sequences with primers from all genes tested. For two isolates (EUL no. 48 and 56), satisfactory amplification products and DNA sequences were not obtained with primers targeting the aconitase gene (acn).

Sequence variation.

The number of nucleotides included in the analysis from each of the gene targets ranged from 182 to 500 bp. All sequences used in the analyses were consensus sequences assembled from the results of the forward and reverse primers. Alleles were numbered consecutively even in the case of a single nucleotide difference.

The initial analysis of 10 of the 16 isolates in the epidemiologically related panel, including the number of polymorphic sites, the percentage of nucleotide substitutions, and numbers of silent and nonsilent mutations, is shown in Table 3, and the analysis of all 95 isolates with the three genes yielding the maximum number of possible sequence types in the initial analysis is shown in Table 4. The highest percentage of polymorphic sites was found with the mompS gene (9.09%), followed by proA (8.76%) and flaA (7.69%); these genes yielded a total of 16, 15, and 12 individual sequence types, respectively. Results of analyses of the 16 isolates in the epidemiologically related panel (panel 1) are shown in Table 5, and the allelic profiles of all 95 isolates with flaA, proA, and mompS are shown in Table 6.

TABLE 3.

Sequence variation in seven different loci for 10 isolates of L. pneumophila comprising four sets of epidemiologically related isolates

Gene	No. of polymorphic nucleotide sites	No. of silent mutations	No. of nonsilent mutations	% of nucleotide substitutions	No. of sequence types/no. of related sets
flaA	13	10	3	7.14	4/4
mompS	20	12	5	5.67	4/4
proA	15	14	1	3.70	4/4
recA	13	13	0	3.67	3/4
acn	14	11	3	2.80	3/3a
groES	4	4	0	1.55	3/4
groEL	3	3	0	1.35	3/4

^aAmplification products were not obtained from one of the epidemiologically related pairs, EUL 48 and EUL 56.

TABLE 4.

Sequence variation in three different loci, for all 95 isolates of L. pneumophila tested

Gene	No. of polymorphic nucleotide sites	No. of silent mutations	No. of nonsilent mutations	% of nucleotide substitutions	No. of sequence types (alleles)
mompS	32	22	8	9.09	16
proA	31	29	2	8.76	15
flaA	13	10	3	7.69	12

TABLE 5.

Allelic profiles of L. pneumophila serogroup 1 from epidemiologically related sets

EUL no.	Sequence typea
	flaA	proA	mompS	recA	groES	groEL	acn
EUL 120	4	12	11	3	3	3	3
EUL 121	4	12	11	3	3	3	3
EUL 73	3	9	14	3	2	2	2
EUL 78	3	9	14	3	2	2	2
EUL 79	3	9	14	3	2	2	2
EUL 71	8	1	5	2	2	2	1
EUL 76	8	1	5	2	2	2	1
EUL 77	8	1	5	2	2	2	1
EUL 48	5	10	6	1	1	1	0
EUL 56	5	10	6	1	1	1	0
EUL 40	11	13	15	ND	ND	ND	ND
EUL 47	11	13	15	ND	ND	ND	ND
EUL 140	12	12	16	ND	ND	ND	ND
EUL 141	12	12	16	ND	ND	ND	ND
EUL 142	12	12	16	ND	ND	ND	ND
EUL 143	12	12	16	ND	ND	ND	ND

^a0, amplification products were not obtained. ND, not determined.

TABLE 6.

Allelic profiles of 95 isolates of L. pneumophila serogroup 1 from analysis of three loci showing MAb subgroup and country of origin

Sequence type (flaA, proA, mompS)	EUL no.	MAb subtype	Country of origin
1,1,1	1, 13, 14, 16, 17, 37, 38, 42, 43, 53, 60, 67, 82, 84, 85, 88, 117, 119	Benidorm, OLDA, Philadelphia	Denmark, Italy, Germany, Greece, Scotland, Spain, Switzerland
1,1,12	93	Oxford	Denmark
1,1,14	104, 110	OLDA, Oxford	Germany, Sweden
1,1,9	112	OLDA	Germany
1,4,9	72	Philadelphia	England and Wales
1,9,14	33, 74	France, Philadelphia	England and Wales, France
10,1,1	3	Philadelphia	Switzerland
2,1,2	4, 26, 28, 29, 32, 41, 49	Allentown, Benidorm, Knoxville, OLDA	France, Italy, Spain, Switzerland
2,1,4	8	Allentown	Switzerland
2,4,14	87	Knoxville	Denmark
2,5,2	7	Allentown	Switzerland
2,6,2	36	Knoxville	Italy
2,8,1	111	Benidorm	Germany
2,8,12	18	Bellingham	Scotland
3,4,7	81	Bellingham	Denmark
3,7,9	51	Benidorm	Spain
3,9,1	99	Bellingham	Sweden
3,9,11	98	Knoxville	Sweden
3,9,14	19, 20, 30, 52, 69, 73, 78, 79, 97, 118	Benidorm, France, Knoxville, Philadelphia	England and Wales, France, Germany, Scotland, Spain, Sweden
4,12,10	25	Allentown	France
4,12,11	6, 27, 39, 50, 75, 116, 120, 121	Benidorm	England and Wales, Germany, Switzerland, France, Italy, Spain
4,12,14	105	Benidorm	Sweden
5,2,6	31, 70	Allentown	England and Wales, France
5,10,6	48, 56, 68, 86, 103	Benidorm, Bellingham, OLDA	Denmark, England and Wales, Spain, Sweden
6,1,1	55	OLDA	Spain
6,1,2	66	Knoxville	Greece
6,3,2	63	Knoxville	Greece
6,4,7	92	Bellingham	Denmark
6,4,8	2	Knoxville	Switzerland
6,14,3	83	Benidorm	Denmark
6,9,14	135, 136, 137, 138, 139	Knoxville, Oxford	England and Wales
7,11,6	100	Bellingham	Sweden
7,11,13	101, 102	Bellingham, Benidorm	Sweden
8,1,5	54, 71, 76, 77	Allentown, France	England and Wales, Spain
9,15,11	91	Bellingham	Denmark
11,13,15	40, 47	Philadelphia	Italy
12,12,16	140, 141, 142, 143	Knoxville	Spain

Reproducibility.

Data from consensus sequences from loci from all 10 isolates tested by both laboratories using different methods of DNA extraction, PCR cycling conditions, and DNA sequencing platforms were in complete agreement (R = 1.00).

Epidemiologic concordance.

All six of the sets of isolates included in the epidemiologically related panel showed compelling evidence of epidemiological relatedness (9, 11). Previous analysis revealed concordant MAb subgrouping, RFLP analysis, REA, and AFLP analysis results (9, 11). Seven genes were sequenced for the 10 isolates representing three sets of epidemiologically related strains and two replicates of the same strain. Epidemiologic concordance (E) was calculated for each of the genes analyzed by using these 10 isolates, and for each locus there was complete concordance (E = 1.00). The aconitase gene (acn), the two common-antigen genes (groES and groEL), and recA could differentiate only two or three sets of strains out of the four sets, whereas proA, mompS, and flaA could distinguish among all four sets of isolates (Tables 3 and 5). The most variable gene was flaA, with 7.14% substitutions over a 182-bp region. Results from the additional two epidemiologically related sets (EUL no. 40, 47, and 140 to 143) comprising clinical and environmental isolates were also epidemiologically concordant (E = 1.00) and gave unique profiles for each set.

Discriminatory power.

Individual and combined indices of discrimination (D) were calculated for the three genes flaA, proA, and mompS by using the panel of 79 unrelated isolates (Table 7). Table 7 also shows the increase in the D value when SBT data are combined with MAb subgrouping data.

TABLE 7.

Indices of discrimination (D) calculated from the 79 unrelated isolates of L. pneumophila serogroup 1 by using one, two, or three loci with and without MAb subgrouping results

Gene(s)	D
proA	0.76
flaA	0.82
mompS	0.85
flaA + proA	0.88
proA + mompS	0.90
flaA + mompS	0.91
flaA + proA + mompS	0.92
flaA + proA + mompS + MAb	0.97

Stability.

Analysis of all five variants (EUL 135 to 139) in the stability panel (Table 1, panel 3) resulted in identical sequence profiles: flaA(6), proA(9), mompS(14). This was also a novel sequence profile not previously seen in the other isolates tested (Table 6).

DISCUSSION

Many phenotypic and genotypic methods for the epidemiologic typing of L. pneumophila sg 1 have been described (4, 12, 19, 27). Previous work by members of the EWGLI has sought to identify the best genotypic techniques for standardization (9). By using a well-defined panel of isolates (the same panel of 79 unrelated isolates used in this study), those methods showing good reproducibility (E values of 1.00) and D values of >0.89 were identified. Four methods, AFLP analysis (D = 0.891), RFLP analysis (D = 0.896), PFGE using NotI (D = 0.941), and REA using HhaI and/or HaeIII (D = 0.960 to 0.980), met these criteria; one of these, AFLP analysis, was subsequently adopted as the EWGLI standard method (11). However, the major disadvantage of all of these methods is that they require the interpretation of gel images. Data from these studies show that the comparison of the patterns obtained in different laboratories, even when the highly standardized EWGLI AFLP analysis protocol is followed, is not always satisfactory. The SBT approach exemplified by MLST (21) appears to offer a way to circumvent these problems.

MLST using bacterial housekeeping genes (21) is an adaptation of multilocus enzyme electrophoresis, revealing the slow accumulation of variation within the gene or the protein for which it codes. The choice of genes was based on those that would be likely to be selectively neutral (21). Sequence analysis of a number of loci (6 to 11 loci) can be used to address the population and evolutionary biology of a particular species (7, 21, 31; http://www.mlst.net). It is acknowledged, however, that other techniques may be more useful in studying individual outbreaks, because identification of microvariation is required in order to distinguish between circulating strains within a geographical area (21).

In this study we developed and evaluated an SBT approach for implementation in the investigation of outbreaks of legionellosis caused by Legionella pneumophila sg 1. Of the seven genes investigated, the four selectively neutral genes all showed less variation than the three genes under probable selective pressure (flaA, proA, and mompS). Furthermore, the aconitase primers failed to give satisfactory amplification products and DNA sequences from two epidemiologically related isolates, EUL 48 and EUL 56 (possibly due to sequence variation at the primer binding sites). Given their low overall variability in genetic sequence and the fact that we were seeking to develop a typing method rather than a phylogenetic tool, it was decided not to consider the nonselective genes further as epidemiological typing targets.

None of the three loci flaA, proA, and mompS gave a sufficiently high index of discrimination (range, 0.767 to 0.857) to be used on its own. However, use of a combination of mompS and proA gave a D value of 0.903, and use of all three loci combined gave a D value of 0.924. Although this is below the value of 0.95 considered to be ideal for a typing system used as a single typing method (25), it is still above the value of 0.90 considered to be desirable if the typing results are to be interpreted with confidence (18). Moreover, the D value of 0.903 achieved with three loci is above the value of 0.89 achieved with the AFLP analysis method (9) currently employed in L. pneumophila sg 1 outbreak investigation (9, 10, 11). When use of all three loci was combined with MAb subgrouping, as is usual in legionellosis outbreak investigations, this value rose to 0.971, well in excess of 0.95 (25). The isolates used in this study were chosen because they were well characterized and would allow ready comparison to previous studies using other methods. However, because the unrelated panel consists entirely of clinical isolates, the D values calculated here may well underestimate the true discrimination achieved when the SBT method is applied more widely. For these three genes (flaA, proA, and mompS) with the alleles identified from analysis of the panel 1 isolates, there is the potential to unambiguously resolve ca. 2,500 sequence types. Examination of other strains is likely to identify many more alleles, increasing the number of potential types still further.

While we believe that this study has demonstrated the validity of the SBT approach for L. pneumophila sg 1, there are a number of issues to be resolved before such a method can be used routinely. These include the following. What are the minimum standards required for the generation of suitable data? e.g., for SBT to be sufficiently robust, it is essential that both strands of the amplicon be sequenced. What size of PCR product should be sequenced? Sequencing of smaller (200- to 250-bp) regions showing sufficient variation, possibly by using internal primers, would allow a faster turnaround and be more economical. What are the optimal targets that should be examined? Together with our colleagues and members of the EWGLI, we are investigating additional gene targets (P. C. Lück and J. Etienne, personal communication).

By direct comparison of results from different loci (or distinct areas on the same loci) using the same panel of strains, a consensus should soon be reached, with the aim of using the best combination of gene targets. In an outbreak situation, depending on the numbers of isolates involved, it may still be desirable to screen isolates first by MAb subgrouping and/or AFLP analysis. It is anticipated that following agreement by members of the EWGLI on the most suitable gene targets for epidemiological typing of Legionella pneumophila sg 1, a sequence database, accessed via the EWGLI website, will be established. Following the sequencing of these gene targets from clinical or environmental isolates of L. pneumophila sg 1, users will be able to query the database, identify preexisting allelic profiles, and submit novel profiles.

In conclusion, we have demonstrated the advantages of SBT of L. pneumophila sg 1 over other current methods. By using only three loci, this methodology has the potential to become the new “gold standard” for the epidemiological typing of L. pneumophila.