Journal of Infection (I988) I6, 47-54

The early diagnosis o f leg ionnaires ' disease in a Legionella pneumophila aeroso l - in fec ted guinea pig m o d e l ; c o m p a r i s o n o f a m e t h o d deve loped for detec t ing Legionella pneumophila

ant igens in ur ine and the d e m o n s t r a t i o n o f c ircu lat ing ant ibody by e n z y m e - l i n k e d i m m u n o s o r b e n t assay

Ann Williams and A. S. R. Featherstone

Experimental Pathology Laboratory, Public Health Laboratory Service Centre for Applied Microbiology and Research, Porton Down, Salisbury, Wiltshire

SP4 oJG, U.K.

Accepted for publication 8 August I987

S u m m a r y The appearance of Legionella pneumopkila antigen in the urine of guinea-pigs with experimental airborne legionnaires' disease was investigated and compared with that of emerging antibody by use of enzyme-linked immunosorbent assay. Models with high dose (acute) and low dose (chronic) infection were studied. Antigen was detected after 4o h in the high dose group but animals died (at 3 days) before an antibody response could be elicited. In the low dose group, antigen was detected 4 days after infection, well before serum antibody was detected (7-IO days). Small, but significant, amounts of antigen were detected up to I7 days after infection in surviving animals.

Although detection of L. pneumopkila antigen in urine has been proposed before, and achieved on an ad koc basis, the technique is not in routine, general use. This is due mainly to difficulties of evaluation in relation to other methods of early diagnosis in the human situation where infectious dose, time of infection, host uniformity and availability of samples present difficulties. The use, in this study, of a highly relevant aerosol-infected guinea-pig model of legionnaires' disease has avoided these uncertainties and hopefully proved the value of this technique for general routine use. The antigen detection test was shown to be rapid, sensitive and reliable, and allowed diagnosis of legionnaires' disease earlier than was possible by demonstrating antibody. In addition, the detection of antigen in urine is a convenient and non-invasive procedure.

Introduction

T h e laboratory diagnosis of legionnaires' disease (LD) is current ly made either by direct isolation of Legionella pneumopkila from clinical specimens x or more usually by demonstra t ing specific ant ibody by means of the indirect immunofluorescent ant ibody test ( IFAT) . 2 Attempts to isolate the causal organism, however, are often unsuccessful and seroconversion may take some weeks. Early and appropriate antibiotic therapy is essential for successful t rea tment of L D since the disease is acute and not all commonly used antibiotics are effective? It is therefore important that a non-invasive diagnostic method should be developed in order to allow earlier detection and hence greater specificity of antibiotic treatment.

oi63-4453/88/oioo47+o8 $02.00/0 (~) I988 The British Society for the Study of Infection

48 A. W I L L I A M S A N D A. S. R. F E A T H E R S T O N E

Several studies 4-8 have used immunological techniques to detect Legionella species antigens in body fluids. They have shown that antigen is easily demonstrated, particularly in urine, but to date it has proved difficult to quantitate and compare these antigen capture techniques in relation to other methods of early detection. As a result, detection of L. pneumophila antigen in urine, although apparently a useful and non-invasive technique offering a very useful screening capability, is not as yet generally in routine use. Investigation of LD antigens in body-fluids would be aided if specimens could be obtained from suspect cases at intervals during the infection up to the time of definite diagnosis (by culture and/or by demonstration of antibody) and on to convalescence. The low prevalence and sporadic nature of LD, together with uncertainties of current diagnosis, make such an investigation difficult. Valuable information may be obtained, however, from experimental infection of animals. A guinea-pig model of human LD was developed 9 in which animals are exposed to a small particle aerosol of L. pneumophila resulting in typical LD bronchopneumonia. This model system allows urine and serum samples to be taken at defined intervals following infection with various infective doses.

The object of this study was to evaluate a direct sandwich type enzyme- linked immunosorbent assay (ELISA) for detecting soluble L. pneumophila serogroup-specific antigen in samples of urine from aerosol-infected guinea- pigs and to establish its efficiency in relation to the more commonly used antibody detection technique.

Materials and methods

Organisms

Legionella pneumophila serogroup I (Corby strain, supplied by Dr R.A. Swann, John Radcliffe Hospital, Oxford) was cultured on buffered charcoal yeast extract agar. 1°

Infection o f animals

The female Dunkin-Hart ley guinea-pigs used had been screened for anti- bodies to L. pneumophila and were seronegative. Two groups of I6 guinea- pigs weighing 300-350 g were exposed to a small particle (5/~m diameter) aerosol of L. pneumophila by use of a mobile Henderson apparatus 11 in conjunction with a three-jet Collison spray, as previously reported. 9' 12 Inhaled doses were calculated such that, in one group, individual animals received i x Io 4 viable organisms (approximately Io LD5o for this strain) i.e. an acute infection, while those in the other group received one-fiftieth of this dose, i.e. a non-lethal dose. Similar guinea-pigs in another group were given a low dose infection and used to provide samples of serum.

Collect ion o f samples

Urine samples were obtained by manual expression of the bladder and were taken before infection and then at regular intervals throughout the disease. Samples from animals in the group with acute infection were taken after I6 h

Detection of Legionella antigen in urine 49

and at 8 hourly intervals until death at approximately 72 h. Samples from the other group (low infective dose) were collected every 24 h for 4 days then at days 6, 7, 8, 9, I I , 15, 16, I7 and 88. All samples were stored at 4 °C. Sample collection depended on the presence of urine in the bladder. On occasions it was not possible to obtain urine from every animal. Guinea-pigs were bled by cardiac puncture after ether anaesthesia. The blood was allowed to clot and, after separation, the serum was stored at - 2 0 °C until assayed.

Antiserum and preparation of conjugate

An anti-L, pneumophila serogroup I antiserum was obtained by immunising rabbits with a saline extract of strain W74/8I. 13 The IgG fraction of this was obtained by affinity chromatography on a protein A sepharose column (Pharmacia Fine Chemicals). Anti-L. pneumophila serogroup I IgG was conjugated to horseradish peroxidase (HRP) by the method detailed by Nakane and Kawaoi. 14

ELISA for antigen

Microtitre plates (Nunc Immunoplate I) with 96 wells were used. Each well was coated with IOO#1 anti-L, pneumophila IgG at 3"5 #g/ml in 0"05 M carbonate buffer (pH 9"5). Plates were incubated overnight at room temperature. In separate plates, serial dilutions of urine were made in phosphate buffered saline (PBS) containing o-I ~o Tween 20 and IO% new- born calf serum. Twofold dilutions were made through eight wells. Each well of the coated plate was washed three times in PBS-Tween before ioo#l volumes were transferred from the dilution plates to corresponding wells in the coated plates. These were covered and incubated with shaking for 2 h at room temperature. Controls consisting of normal guinea-pig urine and a positive serogroup I specific antigen standard i.e. lipopolysaccharide (LPS), extracted from a serogroup I strain of L. pneumophila ~ were included with each plate.

After incubation, the plates were washed as previously and IOO #1 anti- L. pneumophila serogroup I -HRP conjugate were added to each well before incubating the plates for a further 2 h. After washing them again, bound peroxidase enzyme was determined by adding Ioo#l of the substrate tetramethylbenzidine (TMB). TMB was dissolved in dimethyl sulphoxide (IO mg/ml) and this solution was then diluted to 1% in acetate buffer (0"05 M; pH 6"0). Immediately before addition to wells, 30 % hydrogen peroxide was added ( 4 # l / I o m l buffer). The blue colour reaction was stopped after lO-15 min shaking by the addition of 1% sulphuric acid. The subsequent yellow colour was measured spectrophotometrically at 450 nm in a Titertek Multiscan plate reader.

Results were expressed as titres (reciprocals of the log 2 dilutions), obtained by interpolation, and equivalent to 50 ~o of the maximum absorbance reading of the standard above that given by a blank consisting of conjugate only.

A working standard antigen consisting of the serogroup I LPS, titrated three times within the assay, provided a dose-response curve with which sample titrations were compared.

5 0 A. W I L L I A M S A N D A . S . R . F E A T H E R S T O N E

150

120

I10

I00

9O

-~ 8O o • == 7o = 60

• ~ 50

40

m

m

m

m

m

)

(2) I_____1#.I / , / (20 days)

2O

io ~.~ (2) (2) (4)(4)

I 2 5 4 5 6 7 8 9 I0 II 15 16 17 88

Time (days)

Fig. I. Appearance of L. pneumophila antigen in guinea-pig urine after high and low infectious doses and its relation to the appearance of circulating antibody. O, Antibody titre (ELISA) reciprocal of guinea-pig serum dilution (mean titre of four samples) ; i , antigen after high infectious dose (mean titre of eight samples); A , antigen after low infectious dose (mean titre of the number of samples shown in parentheses). Antigen titre is the reciprocal of the urine dilution at the 5 ° % end point. Bars indicate the range of values.

Antibody assay

The ELISA method was as previously described 15 but included use of a horse- radish peroxidase conjugate of rabbit anti-guinea pig immunoglobulin.

Results

Antigen was readily detected in the urine of guinea-pigs with experimental legionnaires' disease. The figure shows the change in antigen titre with varying times of sample collection for both the high and low dose of infection models. Earliest detection of antigen was at 40 h in the high dose group, with titres increasing until death at 72 h. In the low dose group, earliest detection was at 96 h. Titres rose to a peak at 6 days then fell sharply with low concentrations of antigen being detectable up to I7 days after infection. Antigen was not found in the 88 day samples. In addition, the figure shows that circulating antibody did not appear before 7- to days after aerosol infection, at least 3 days after L. pneumophila antigen was first detected in the urine.

Tables I and II show the amounts of antigen detected in #g/ml urine. These were calculated by comparing titres with that of a known standard (LPS). The pattern of variation in amounts detected with progression of disease paralleled that of the titres.

The sensitivity of the assay was determined by titrating a known amount of antigen (LPS solution: I mg/ml) and the detection limit estimated as o-oIz Fg

Detection of Legionella antigen in urine 5I

Table I Concentrations of L. pneumophila antigen in the urine of guinea-pigs in the low infective dose (chronic) group

Time Mean concentration Number of Range of values (days) of antigen (/zg/ml) animals* (/zg/ml)

4 0"07 z (0'06-0'08) 6 I" 5 I t 7 0"54 I (0'37-0"72 ) 8 0.026 6 (o.oi-0.o5) 9 0'025 I (0.02-0-03)

11 0"045 r (0"04-0"05) 15 0"02, I 0'02 I6 0"0 3 2 (0"02-0"04) I7 0"026 2 (O"Ol-0"04)

* Two samples/animal. t One sample.

Table II Concentration of L. pneumophila antigen in urine of guinea-pigs in the high infective dose (acute) group

Time Mean concentration Range of values (h) of antigen (/zg/ml)* (/zg/ml)

40 0"024 (o-oi8-0"o38) 48 0"088 (0"04-0" 13) 56 o-I69 (0"07-0"38 ) 64 0"657 (o-27-I'45) 72 I-2I (o'64-I'56)

* Mean of eight assays from four animals.

ml. It should be noted that the low antigen titres observed from days 8 to x7 corresponded to calculated concentrations of antigen two or three times higher than the minimum concentration detectable.

Discuss ion

Legionella urinary antigens have been detected in both human beings and guinea-pigs by several methods including ELISA, radioimmunoassay and reverse passive agglutination assays. Nevertheless, methods for detecting Legionella urinary antigen are not in routine general use. This is surprising since they are non-invasive techniques readily adaptable for screening purposes. It is possible that drawbacks with the assays and difficulties associated with evaluation in the human situation are responsible. Mangiafico and colleagues 16 and Tang and colleagues 7 developed reverse passive agglutination assays to detect antigen in samples of human urine and reported them to be sensitive and rapid. Preparation of the agglutinating reagent and frequent standardisation of other reagents, however, renders these assays far

5 2 A. W I L L I A M S A N D A. S . R . F E A T H E R S T O N E

less controllable and reproducible than an ELISA. Radioimmunoassays have been used by Kohler and colleagues 17 and Sathapatayavongs and colleagues s on urine from patients with LD. The latter study compared radioimmunoassay with ELISA and found the two to be of equivalent sensitivity and specificity but that the ELISA was less expensive and more suitable for clinical laboratories unwilling to use radioisotopes.

Studies of guinea-pig urine assayed by ELISA 4'1s have detected antigen after intravenous or intraperitoneal injection with L. pneumophila. Such modes of inoculation are not as comparable to human LD, in which infection is via the respiratory tract, as the use of an aerosol used in this study.

ELISAs have also been used on samples of human urine 4' 5. s but results from these and other assays have shown variable success in detecting antigen. In all these human cases of LD, neither the time of infection nor the infective dose could have been known. Hence comparison of diagnosis by antigen detection with that of specific antibody titre or isolation of the organism in culture is very difficult to assess.

The above uncertainties have, to an extent, been avoided by the use of a relevant guinea-pig model in which dose, infecting strain of L. pneumophila, host and sample availability are standardised. This study has shown that, by use of an ELISA, antigen can be detected in the urine of guinea-pigs with experimental LD after 2 days and up to 17 days after infection. A quantitative variation in antigen excretion throughout the course of the infection has also been demonstrated.

Guinea-pigs infected with L. pneumophila by aerosol have been shown to develop a pyrexial, coalescing broncho-pneumonia closely resembling the disease in human beings. 9' 19 Lethally-infected guinea-pigs did not exhibit an antibody response before they died at approximately 3 days after infection. Antigen is detectable in the urine of such guinea-pigs after only 4o h post- infection. Guinea-pigs given a low dose chronic infection required between 7 and IO days to develop detectable antibodies and even longer to exhibit a four- fold rise in antibody titre. Non-lethally infected guinea-pigs in this study had detectable antigen as early as 4 days following infection. These observations indicate that diagnosis of experimental LD may be achieved earlier by the detection of urinary antigens than by demonstration of an emerging antibody response. The ELISA for detecting antigen thus has the potential for a rapid and reliable diagnostic test for LD, although further development and evaluation of the test may be required.

The ELISA system used in this study detects only L. pneumophila serogroup I antigen but the test could readily be modified for other serogroups by substitution with appropriate antibody as a reagent. Human infections by other serogroups of L. pneumophila do cause LD although that due to serogroup I is the most common. For routine diagnosis, therefore, the assay should include antibodies directed against the common disease-causing serogroups of L. pneumophila as well as those against other species.

Cross-reactions which may be found with other respiratory and urinary tract pathogens must be investigated thoroughly so as to minimise the risk of false positive results. Cross-reactions may theoretically arise in antibody detection tests but do not appear adversely to affect the specificity of the

Detection of Legionella antigen in urine 53

technique in practice. Workers using antigen detection tests on human urine have not found cross-reactions in samples from patients with other pulmon- ary and ur inary tract infections, 5-7, iv thereby suggesting a high specificity for the serogroup antigen assay. Cross-reactions between various strains of L. pneumophila would be useful for screening purposes. An ideal assay for diagnostic purposes would include an ant ibody as reagent which recognises a common determinant of all serogroups and species of Legionella.

Other body fluids such as serum may also be tested for L. pneumophila serogroup antigen, a l though concentrations are lower than those in urine and are sometimes undetectable. TM In the acute stages of infection, however, serum might be useful if urine is unobtainable, and may also be of value in retrospective studies. Ant igen detection assays require similar reagents and expertise to tests for ant ibody such as those currently in routine use for diagnosis of LD.

T h e E L I S A is a relatively simple test which does not require specialised equipment and may be performed in most laboratories. For accurate measurements a spectrophotometer is required but visual interpretat ion of results may often prove acceptable. T h e use of pre-coated microti tre plates and possible reduct ion of incubat ion times would improve the suitability of the assay for a routine hospital laboratory.

T h e E L I S A test described in this study for detecting L. pneumophila antigen in urine provides a rapid, non-invasive, reliable and sensitive test with the potential for improving t reatment of L D by identifying the causal agent in the ea r ly stages of infection, well before diagnosis by ant ibody detection is possible.

(This work was supported by an MRC project to Ann Williams. Both authors thank Dr R. B. Fitzgeorge for helpful advice and discussion as well as Dr J. W. Conlan for preparing some of the reagents.)

References

I. Feeley JC. Diagnosis: culture. In: Katz SM, Ed. Legionellosis, Vol. II. Florida: CRC Press, I985: 3-9.

2. Wilkinson HZ. Diagnosis : serology. In : Katz SM, Ed. Legionellosis, Vol. II. Florida : CRC Press, I985: II-2O.

3. Fitzgeorge RB. Effect of antibiotics on the growth of L. pneumophila in guinea pig alveolar phagocytes infected in vivo by aerosol. J Infect I985 ; IO: I89-I93.

4. Berdal BP, Farshy CE, Feeley JC. Detection of Legionella pneumophila antigen in urine by enzyme-linked immunospecific assay. J Clin Microbiol I979; 9: 575-578.

5- Tilton RC. Legionnaires' disease antigen detected by enzyme-linked immunosorbent assay. Ann Intern Med I979; 90: 697-698.

6. Sathapatayavongs B, Kohler RB, Wheat LJ, White A, Winn WC, Girod JC, Edelstein PH. Rapid diagnosis of Legionnaires' disease by urinary antigen detection. Am J Med I98a; 72 : 576-582.

7- Tang PW, de Savigny D, Toma S. Detection of Legionella antigenuria by reverse passive agglutination. J Clin Microbiol I982; XS: 998-IOOO.

8. Bibb WF, Amow PM, Thacker L, McKinney RM. Detection of soluble Legionella pneumophila antigens in serum and urine specimens by enzyme-linked immunosorbent assay with monoclonal and polyclonal antibodies. J Clin Microbiol I984; 2o: 478-482.

54 A. W I L L I A M S AND A. S. R. FEATHERSTONE

9. Baskerville A, Fitzgeorge RB, Broster M, Hambleton P, Dennis PJ. Experimental transmission of Legionnaires' disease by exposure to aerosols of Legionella pneumophila. Lancet I98I ; i.i: i389-i39o.

io. Edelstein PH. Improved semiselective medium for isolation of LegioneUa pneumophila from contaminated clinical and environmental specimens. J Clin Microbiol 1981 ; I4: 298-3o3.

I I. Druett HA. A mobile form of the Henderson apparatus. J Hyg I969; 67: 437-448. I2. Baskerville A, Fitzgeorge RB, Broster M, Hambleton P. Histopathology of experimental

Legionnaires' disease in guinea pigs, rhesus monkeys and marmosets. J. Pathol I983; 139: 349-362.

I3. Conlan JW, Ashworth LAE. The relationship between the serogroup antigen and lipopolysaccharide of Legionella pneumophila. J Hyg I986; 96: 39-48.

I4. Nakane PK, Kawaoi A. Peroxidase-labelled antibody. A new method of conjugation. J Histochem Cytochem I974; 22: lO84-1o91.

I5. AshworthLAE, Fitzgeorge RB, Irons LI, Morgan CP, RobinsonA. Rabbitnasopharyngeal colonization by Bordetella pertussis : the effects of immunization on clearance and on serum nasal antibody levels. J Hyg I982; 88; 475-486.

i6. Mangiafico JA, Hedlund KW, Knott R. Rapid and sensitive method for quantitation of Legionella pneumophila serogroup I antigen from human urine. J Clin Microbiol I981 ; I3: 843-845.

17. Kohler RB, Wheat LJ, French MLV, Meenhorst PL, Winn WC, Edelstein PH. Cross reactive urinary antigens among patients infected with Legionella pneumophila serogroups I and 4 and the Leiden I strain. J Infect Dis 1985; 152: IOO7-IOI2.

i8. Lebrun L, Tram C, Lapierre F, Grangeot-Keros L, Pillot J. Detection of Legionella pneumophila antigens by ELISA in urine of experimentally infected guinea pigs. Ann Microbiol 1983; I34A: I55-I6I.

19. Fitzgeorge RB, Baskerville A, Broster M, Hambleton P, Dennis PJ. Aerosol infection of animals with strains of Legionella pneumophila of different virulence: comparison with intraperitoneal and intranasal routes of infection. J Hyg I983; 90: 81-89.