J Clin Pathol 1992;45:1079-1083

Phenotypic methods for speciating clinicalAeromonas isolates

M H Wilcox, A M Cook, K J Thickett, A Eley, R C Spencer

AbstractAims: To establish the suitability of cur-rently available phenotypic methods forspeciation of clinical Aeromonas isolatesin diagnostic microbiology laboratories.Methods: Using 62 Aeromonas spp, threeschemes based on biochemical reactionswere compared: a series of conventionaltests; a system based on the suicide phe-nomenon, comprising two tubes in total;and a commercially available test, API 20NE, augmented with a plate assay for ,Bhaemolysin production. The whole celland outer membrane protein (OMP) pro-files of strains were examined by sodiumdodecyl sulphate polyacrylamide gel elec-trophoresis (SDS PAGE), according to theresults ofthe above schemes, to determinethe intra-species homogeneity.Results: Ninety per cent of strains wereidentified satisfactorily according to con-ventional criteria. For these strains,agreement was obtained using the suicidephenomenon and API schemes in 93% and88% of cases, respectively. The threeschemes concurred for 82% of strains.Whole cell protein profiles were unsuit-able for comparing strains within a spe-cies. However, OMP patterns were similarfor 890% ofA caviae and 63% ofA hydro-phila.Conclusion: Phenospeciation of clinicalAeromonas isolates by the scheme basedon the suicide phenomenon is simple toperform and accurate, and suitable foruse in the diagnostic laboratory. OMPprofiles are potentially useful for confirm-ing the identity of A caviae and most Ahydrophila, but not A sobria.

parison, using biochemical tests and otherphenotypic markers at least seven pheno-species can be differentiated.3 5A classification based on genospeciation

therefore presents the problem that a numberof genospecies can only be identified usingDNA hybridisation techniques, which are notavailable in most diagnostic laboratories. Bycomparison, phenotyping is much simpler,usually requiring standard biochemicaltests.` Kuijper et al have also described aprotein phenotyping method for aeromonadswhich compared outer membrane proteins(OMPs).8 They found that this method couldbe used to identify strains of A hydrophila, Acaviae, and A sobria (A veronii biovar sobria).However, the protein traces obtained hadmany bands and were therefore difficult toanalyse, with strains of A hydrophila and Asobria showing considerable variation withinspecies.

Evidence is accumulating to suggest thatparticular Aeromonas spp cause infection incertain clinical situations. For example, Acaviae seems to be associated with gastro-enteritis in infants.9 " Simple and reliablespeciation techniques are required if diagnosticmicrobiology laboratories are to be able toidentify Aeromonas isolates, and so providefurther information on their prevalence andpathogenicity. In this study three biochemicalphenotypic methods for the speciation ofclinical Aeromonas isolates were compared:conventional biochemical typing6 7; speciationaccording to the suicide phenomenon'2; and acommercial method (API 20NE strips, API,Basingstoke) in conjunction with a simplehaemolysin plate assay. The results of thesebiochemical methods were correlated withboth whole cell and OMP profiles.

(3 Clin Pathol 1992;45:1079-1083)

Department ofExperimental andClinical Microbiology,University of SheffieldMedical School, BeechHill Road, SheffieldS10 2RXM H WilcoxA M CookK J ThickettE EleyR C SpencerCorrespondence to:Dr M H Wilcox

Accepted for publication8 July 1992

Members of the genus Aeromonas were firstrecognised 100 years ago. Since then thetaxonomy of these bacteria has undergoneconstant modification, resulting in consider-able confusion.1̀ Recently, in an effort toclarify the situation, DNA-DNA hybridisationtechniques have been used. (Farmer JJ, et al.

Abstract presented at the First InternationalWorkshop on Aeromonas and Plasiomonas,1986).4 However, the results of such workhave, in some instances, complicated the tax-onomy of Aeromonas spp still further. Cur-rently the genus can be split into 13 DNAhybridisation groups of genospecies. By com-

MethodsSixty two Aeromonas spp were studied, com-prising 27 and 29 faecal isolates from adultsand children, respectively, who had been inves-tigated for gastroenteritis in Sheffield. Sixreference strains, including three kindly sup-plied by T Donovan (PHLS Laboratory, Ash-ford, Kent), and American Type CultureCollection strains ATCC 15468, 7966, and43979 (A caviae, A hydrophila, and A sobria,respectively) were used. All strains were oxi-dase positive, fermentative Gram negativebacilli, resistant to both low (10 ,ug disc) andhigh (150 jig disc) concentrations ofthe vibrio-static agent 0129.

1 079

on May 20, 2020 by guest. P

rotected by copyright.http://jcp.bm

j.com/

J Clin P

athol: first published as 10.1136/jcp.45.12.1079 on 1 Decem

ber 1992. Dow

nloaded from

Wilcox, Cook, Thickett, Eley, Spencer

CONVENTIONAL BIOCHEMICAL METHODSSpecification was achieved according to thecriteria of Popoff and Veron,6 modified byJanda et al.7 Briefly, the tests comprised:aesculin hydrolysis determined by inoculationof agar slants supplemented with aesculin(1 gl-') and ferric citrate (0 5 gl-') (30°C);Voges Proskauer reaction; acid production inarabinose and salicin peptone waters; gasproduction in glucose peptone water; gluco-nate oxidation; growth in the presence ofpotassium cyanide; and ,B haemolysin produc-tion determined by inoculation of Columbiaagar (Oxoid) plates containing 5% sheepblood. All tests were incubated in air at 37°C(unless otherwise stated) for 18-24 hours.

SUICIDE PHENOMENONThe fermentation of glucose, with or withoutgas production and pelleting of bacteria (sui-cide phenomenon) was determined by inocula-tion of vessels containing nutrient broth(Oxoid) with 0 5% w/v glucose, 0-0015% w/vbromcresol purple, and inverted tubes.2 Testswere read after incubation at 30°C for 18-24hours. Additionally, for this method of specia-tion aesculin hydrolysis, as determined above,was also tested. A hydrophila is non-suicidal,

Suicidephenomenon

//

/

A sobria.<,

>......Irophila

"..........fM H>z............

API 20 NE

A sobri\\;..\ l~~~~~~~~..........N;< ~"Doubtful

A hydrophida"A cav,ae ::::

®X) A hydrophila

PositiveAPI 20 NE

haemolysinb production

aerogenic and aesculin positive; A sobria issuicide variable, aerogenic and aesculin neg-ative; A caviae is suicidal, anaerogenic, andaesculin positive.'2

API 20NE AND HAEMOLYSIN PRODUCTIONIdentification strips were used according to themanufacturer's instructions and read afterincubation at 30°C for 48 hours. As thismethod cannot be used to differentiatebetween A hydrophila and A caviae strains,haemolysin production (positive for A hydro-phila and negative forA caviae) was also tested,as described above.7 '3SDS PAGE

Whole cell and OMP profiles were studied bysodium dodecyl sulphate polyacrylamide gelelectrophoresis (SDS PAGE).'4 Strains werecultured on a shaker in brain heart infusionbroth (Oxoid) at 30°C for 18 hours, and werethen centrifuged at 5000 x g for 20 minutes.Whole cell protein samples were prepared byresuspending the bacteria in distilled water toan optical density of 0.5 at 590 nm, and thencentrifuging 10 ml of this suspension at5000 x g for 20 minutes. Double strengthsample buffer (150,u1) (4 g SDS, 20 ml gly-cerol, 10 ml 2-mercaptoethanol, 12-5 ml 1MTRIS-buffer (pH 6 8), and 20 mg bromophe-nol blue per 100 ml distilled water) was addedto the deposit. After boiling for five minutes150,ul of distilled water were added andboiling continued for a further five minutes.The supernatant fluid collected after cen-trifugation at 10 000 x g for 10 minutes rep-resented the whole cell proteins.OMPs were prepared by resuspending bac-

terial cells, cultured as above, in 5 ml 50 mMTRIS-HCI buffer (pH 7.4) to an opticaldensity of 1-5 at 590 nm. Suspended bacteriawere lysed by sonication over ice for two oneminute periods at 12 ,um peak-to-peak ampli-tude (MSE Scientific Instruments, Crawley,England). After centrifugation at 5000 x g for10 minutes to remove cell debris OMPs wereseparated from the supernatant fluid by adding0-5 ml 20% w/v Sarkosyl (Sigma, Poole, Eng-land) and incubating at room temperature(20°C) for 30 minutes. OMPs were collectedby centrifugation at 50 000 x g for one hourat 4°C, and then resuspended in 50 ,ul 50 mMTRIS-HCI buffer. Single strength samplebuffer (150 ,ul) was added, and then boiled asdetailed for whole cell preparations.Whole cell and OMPs were separated on

12% SDS PAGE gels (Mini Protean II Biorad,Hemel Hempstead, England), loading 10 jgprotein per lane. Molecular size standards(Sigma) were run concurrently. Followingelectrophoresis, gels were fixed and stainedwith Coomassie brilliant blue R250.

Negative

( Acaviae

Figure 1 Phenotypic characterisation of aeromonads using three different methods ofspeciation. The figures inside circles represent the number of strains identified as aparticular phenotype. The figures outside circles show the numbers of strains changingidentity (dotted lines) from the conventional scheme results. Strains identified as eitherAhydrophila orA caviae by API 20NE were divided according to haemolysin production.

ResultsThe results of the three biochemical schemesfor speciating Aeromonas spp are compared infig 1. Of 62 strains studied, six (10%) gaveinconclusive results according to the conven-tional scheme and are not included in the

Conventionaischeme

1 080

on May 20, 2020 by guest. P

rotected by copyright.http://jcp.bm

j.com/

J Clin P

athol: first published as 10.1136/jcp.45.12.1079 on 1 Decem

ber 1992. Dow

nloaded from

Phenotypic methods for speciating clinical Aeromonas isolates

remaining analysis. Janda et al7 defined eightprimary and nine supplementary phenotypicproperties for identifying clinical isolates ofAeromonas spp. The six strains that could notbe satisfactorily identified gave inappropriateresults for at least two primary reactions.However, the six strains were morphologicallysimilar to the remainder of Aeromonas isolatesand satisfied the broad criteria detailed in themethods section. Fifty two out of 56 strains(93%) gave the same species identification inthe conventional and suicide phenomenonschemes; the corresponding figure for the API20NE plus haemolysin assays was 49 (88%)strains. The three schemes gave concordantresults for 82% of strains tested. Four and twostrains not identifiable by the conventionalscheme were speciated as A sobria by the APIand suicide schemes, respectively. All 27strains identified by the conventional schemeas A caviae were haemolysin negative, while 15out of 16 A hydrophila were haemolysin pos-itive; seven (77%) A sobria were also haemoly-sin positive.Gas production from the fermentation of

glucose was frequently not detected usingglucose peptone water. For 12 strains gasproduction was reproducibly negative in thelatter method, but positive in the suicide tubeswhich contained nutrient broth and glucose,and these were incubated at 30°C instead of37°C. A separate experiment showed that themost critical factor was the temperature ofincubation, gas production being more reliableat 30°C as opposed to 37°C.The whole cell protein profiles contained in

excess of 40 bands which made comparisonbetween species and strains difficult, andspecies specific patterns could not be identifiedwith confidence (data not shown). The OMPprofiles of 62 Aeromonas spp were groupedaccording to the identity of strains using theconventional biochemical scheme of specia-tion. For two of the three phenospecies,namelyA caviae (fig 2) andA hydrophila (fig 3),

striking similarities were noted between mostisolates. A characteristic four band OMP pro-file was seen in 24 out of 27 (89%) A caviae,consisting of proteins with mean molecularweights (range) of 54 (51-60), 44 (40-48), 31(27-35) and 29 (26-33) kilodaltons. A differ-ent, but again characteristic, four band proteinprofile was observed in 10 out of 16 (63%) Ahydrophila, consisting of 52 (50-55), 44(42-64), 33 (31-35) and 28 (27-30) kilo-dalton proteins. As seen in A caviae, the twohigh molecular weight OMPs were expressedmore strongly than the remainder. Two isolatesidentified by conventional biochemical criteriaas A caviae had OMP profiles typical of Ahydrophila (tracks 3 and 4 from the right, fig 2).Both strains were haemolysin negative andidentified as A caviae by the API 20NEscheme, and in the suicide phenomenon onewas characterised as A caviae and the other"unknown". A sobria isolates had stronglyheterogenous OMP profiles (fig 4), with one ortwo strongly expressed proteins within the highmolecular weight ranges noted for the othertwo phenospecies. Of the six isolates notidentified by conventional criteria, two hadOMP profiles similar to the A hydrophilapattern (tracks 1 and 2 from the left, fig 5),while the remainder were unrecognisable. Bothof the former isolates were identified as Ahydrophila by the suicide phenomenon and API20NE schemes.

DiscussionThe large increase over the past decade in thenumber of publications on the genus Aeromo-nas has heralded the acceptance of several newgenospecies and phenospecies.' Similarly,studies now point to pathogenic roles forAeromonas spp that were previously unrecog-nised. For example, we and others have pro-vided evidence that A caviae is implicated as acause of gastroenteritis in children.9 11 Ahydrophila and A sobria can cause skin and soft

. l..,.~~~-6I 1 E~~~~~4. . _ = .~~-3

I1 1 1: i=g6~-2*_ -14

Figure 2 Outer membrane protein profiles of 27 strains identified as A caviae by conventional criteria. A caviae typestrain ATCC 15468 is in track 1 (from the left). Molecular weights are indicated at the far right side.

1081

on May 20, 2020 by guest. P

rotected by copyright.http://jcp.bm

j.com/

J Clin P

athol: first published as 10.1136/jcp.45.12.1079 on 1 Decem

ber 1992. Dow

nloaded from

Wilcox, Cook, Thickett, Eley, Spencer

-97-66

-45

-31

-22

-14

Figure 3 Outer membrane protein profiles of 16 strains identified as A hydrophila byconventional criteria. A hydrophila type strain ATCC 7966 is in track 1 (from the left).Molecular weights are indicated at the far right side.

tissue infection and septicaemia in both immu-nocompromised and immunocompetent indi-viduals.' In vivo and in vitro studies indicatethat the latter phenospecies are inherentlymore virulent that A caviae. `5-7 Unless diag-nostic microbiology laboratories are encour-aged to isolate and speciate these potentialpathogens, further clarification of the behav-iour ofAeromonas spp will be impeded.

Initial taxonomic studies based on 59 bio-chemical characteristics, of which seven werefound to be of discriminatory value, describedtwo mesophilic Aeromonas spp.6 Later mod-ifications assigned clinical -isolates to one ofthree phenospecies A hydrophila, A caviae, andA sobria7; this terminology has been applied inthis study. Several new phenospecies such asAtrota and A jandaei have been proposed, butthese are rarely recovered from clinical speci-mens. 18 iQ It is now appreciated that within

Figure 4 Outer membrane protein profiles of 13 strainsidentified as A sobria by conventional criteria. A sobriatype strain ATCC 43979 is in track I (from the left).Molecular weights are indicated at the far right side.

-97-66

-45

31

-22

-14

Figure 5 Outer membrane protein profiles of six strainsnot satisfactorily identified by conventional criteria.Molecular weights are indicated at the far right side.

each phenospecies more than one DNAhybridisation group exists. Nevertheless, mostclinical isolates represent DNA groups 1, 4,and 8, which correspond to the phenospeciesoutlined above.3Three biochemical phenospeciation schemes

have been compared which are amenable foruse in the diagnostic laboratory. The goldstandard results were taken as those obtainedfrom the conventional scheme, although suchassumptions may be flawed. This scheme istime consuming and laborious to perform, anddoes not easily lend itself to the occasionalspeciation of small numbers, because of prob-lems with the limited shelf-life of media. Whenhandling large numbers of strains it becomesmost unwieldy. Furthermore, the reproducibil-ity and subjectivity of tests such as growth inthe presence of potassium cyanide (noted alsoin the present study) is apparent.20 The varia-ble behaviour of Aeromonas spp at 30°C com-pared with 37°C has been described elsewhere,and the choice of incubation temperaturevaries between studies.2i 22 Our observationsof the variable results of gas production fromthe fermentation of glucose confirm thesefindings.The API 20NE and haemolysin assays were,

by contrast, easy to perform. However, theresults of this approach to speciation con-curred with the conventional scheme for 88%of isolates, compared with a figure of 93% forthe suicide phenomenon scheme (p > 0 1). Itwas interesting to note that the percentageacceptability of the API profile number bore norelation to the correct speciation ofA caviaelAhydrophila isolates. Hence a strain that hadeither a high or low probability of being Ahydrophila could be speciated as A caviae, forexample. The production of fi haemolysin was,however, an accurate discriminator betweensuch isolates. The speciation scheme based onthe presence or absence ofthe suicide phenom-enon (the susceptibility ofA caviae and someA

1082 on M

ay 20, 2020 by guest. Protected by copyright.

http://jcp.bmj.com

/J C

lin Pathol: first published as 10.1136/jcp.45.12.1079 on 1 D

ecember 1992. D

ownloaded from

Phenotypic methods for speciating clinical Aeromonas isolates

sobria to acidic conditions) was accurate, sim-ple, and inexpensive to perform. It requiresonly basic reagents and in total comprises onlytwo tubes per strain tested. When describingthis method of phenospeciation, Namdari andBottone tested 210 clinical and environmentalisolates and found that their results "paral-leled" those achieved by conventional cri-teria.'2 Whether strains that were onlysatisfactorily identified by the latter methodwere included in this analysis is unclear. Asimplified key for speciating Aeromonas whichcan distinguish eight phenotypes using seventests has recently been described.20 However,several of these species are rarely encounteredclinically and the key incorporates antibioticsusceptibility testing using the Kirby-Bauertechnique which is not favoured in Europe.We were interested to compare the results of

the above methods with the Aeromonas sppprotein profile phenotypes to determine theintra-species homogeneity. Several studieshave examined the protein profiles of Aeromo-nas spp, with widely differing patterns beingobtained.8 21 23-25 For example, the number ofwhole cell proteins clearly identified by SDSPAGE has varied from about six to 50.21 23-25Our results agree with the higher figure and,particularly when using mini-gel apparatus,this makes the interpretation of profiles diffi-cult. We were unable to identify species specificwhole cell protein profiles. An earlier reportobserved about 20 Aeromonas OMPs, whichgave somewhat "busy" protein profiles.8 Ourextraction methods identified about four to sixstrongly expressed OMPs, and distinct pat-terns were noted for 89% A caviae and 63% Ahydrophila, but not forA sobria. Clearly, differ-ing methods of protein extraction, includingcentrifugation speeds and cell lysis techniques,will influence the purity of OMP preparations,and hence, together with the sensitivity ofdetection methods, will determine the numberof protein bands observed by SDS PAGE. Twounidentifiable isolates and two that were speci-ated as A caviae by conventional criteria hadOMP patterns typical ofA hydrophila, and thetrue identity of these Aeromonas spp remainsuncertain.We have not attempted to redefine or sub-

divide the phenospecies A caviae, A hydrophila,and A sobria as this was not the aim of thestudy, and phenotypic criteria alone are insuffi-ciently discriminatory for this purpose. We arepresently investigating the potential of geno-typic methods such as ribotyping for groupingAeromonas spp, which may be able to clarifysuch problems. It seems clear that within thethree phenospecies heterogenous strains exist,although this fact does not detract from thebroad virulence properties ascribed to date toeach group.3 OMP profiles are, however, auseful means of confirming the phenospecia-tion ofA caviae and someA hydrophila isolates,as determined by simple biochemical schemessuch as that based on the suicide phenomenon.

The latter technique is particularly suitable foruse in the diagnostic microbiology laboratory.

1 Schubert RHW. Genus II Aeromonas, Kluyver and van Niel1936. In: Buchanan R E, Gibbons N E, eds. Bergey'smanual of determinative bacteriology. 8th edn. Baltimore:Wiliams and Wilkins Co, 1974:345-8.

2 Janda MJ, Duffey PS. Mesophilic aeromonads in humandisease: current taxonomy, laboratory identification, andinfectious disease spectrum. Rev Infect Dis 1988;10:980-97.

3 Janda MJ. Recent advances in the study of the taxonomy,pathogenicity and infectious syndromes associated withthe genus Aeromonas. Clin Microbiol Rev 1991 ;4:397-410.

4 Altwegg M, Steigerwalt AG, Altwegg-Bissig R, Luthy-Hottenstein J, Brenner DJ. Biochemical identification ofAeromonas genospecies isolated from humans. J ClinMicrobiol 1990;28:258-64.

5 Popoff M, Lallier R. Biochemical and serological character-istics of Aeromonas. In: Bergan T, ed. Methods inmicrobiology. Vol 16. London: Academic Press, 1984;127-45.

6 Popoff M, Veron M. A taxonomic study of the Aeromonashydrophila-Aeromonas punctata group. J Gen Microbiol1976;94:11-22.

7 Janda MJ, Reitano M, Bottone EJ. Biotyping ofAeromonasisolates as a correlate to delineating a species-associateddisease spectrum. J Clin Microbioll 984;19:44-7.

8 Kuijper EJ, van Alphen L, Leenders E, Zanen HC. Typingof Aeromonas strains by DNA restriction endonucleaseanalysis and polyacrylamide gel electrophoresis of cellenvelopes. J Clin Microbiol 1989;27:1280-5.

9 Namdari H, Bottone E J. Microbiologic and clinicalevidence supporting the role of Aeromonas caviae as apaediatric pathogen. J Clin Microbiol 1990;28:827-40.

10 Pazzaglia G, Sack RB, Salazar E, et al. High frequency ofcoinfecting enteropathogens in Aeromonas-associateddiarrhoea ofhospitalised Peruvian infants. J Clin Microbiol1991;29:1151-6.

11 Wilcox MH, Cook AM, Eley A, Spencer RC. Aeromonasspp. as a potential cause of diarrhoea in children. 7 ClinPathol 1992;45:959-63.

12 Namdari H, Bottone EJ. Suicide phenomenon in meso-philic aeromonads as a basis for species identification. JClin Microbiol 1989;27:788-9.

13 Kuijper EJ, Steigerwalt AG, Schoenmakers BSCIM, PeetersMF, Zanen HC, Brenner DJ. Phenotypic characterisationand DNA relatedness in human fecal isolates ofAeromo-nas spp. J Clin Microbiol 1989;27:132-8.

14 Lugtenberg B, Meijers J, Peters R, Van der Hoek P, VanAlphen L. Electrophoretic resolution of the major outermembrane protein of Escherichia coli K12 into fourbands. FEBS Len 1975;58:254-8.

15 Janda JM, Clark RB, Brenden R. Virulence of Aeromonasspecies as'assessed through mouse lethality studies. CurrMicrobiol 1985;12:163-8.

16 Millership SE, Barer MR, Tabaqchali S. Toxin productionby Aeromonas spp. from different sources. J Med Micro-biol 1986;22:311-4.

17 Watson IM, Robinson JO, Burke V, Gracey M. Invasivenessof Aeromonas spp. in relation to biotype, virulencefactors, and clinical features. Jf Clin Microbiol 1985;22:48-51.

18 Carnahan A, Fanning GR, Joseph SW. Aeromonas jandaei(formerly genospecies DNA group 9 A. sobria), a newsucrose - negative species isolated from clinical speci-mens. J7 Clin Microbiol 199 1;29:560-4.

19 Camahan AM, Chakraborty T, Fanning GR et al. Aeromo-nas trota sp. nov., an ampicillin susceptible speciesisolated from clinical specimens. J Clin Microbiol 1991;29:1206-10.

20 Carnahan AM, Behram S, Joseph SW. Aerokey II: a flexiblekey for identifying clinical Aeromonas species. J7 ClinMicrobiol 1991 ;29:2843-9.

21 Statner B, Jones JM, George LW. Effect of incubationtemperature on growth and soluble protein profiles ofmotile Aeromonas strains. J Clin Microbiol 1988;26:392-3.

22 Kuijper EJ, Steigerwalt AG, Schoenmakers BSCIM, PeetersMF, Zanen HC, Brenner DJ. Phenotypic characterisationand DNA relatedness in human fecal isolates ofAeromo-nas spp. J Clin Microbiol 1989;27:132-8.

23 Kokka RP, Vedros NA, Janda MJ. Electrophoretic analysisof the surface components of autoagglutinating surfacearray protein positive and surface array protein negativeAeromonas hydrophila and Aeromonas sobria. J ClinMicrobiol 1990;28:2240-7.

24 Stephenson JR, Millership SE, Tabaqchali S. Typing ofAeromonas species by polyacrylamide gel electrophoresisof radiolabelled cell proteins. J Med Microbiol 1987;24:113-8.

25 Millership SE, Want SV. Typing of Aeromonas species byprotein fingerprinting: comparison of radiolabelling andsilver staining for visualising proteins. Jf Med Microbiol1 989;29:29-32.

1 083

on May 20, 2020 by guest. P

rotected by copyright.http://jcp.bm

j.com/

J Clin P

athol: first published as 10.1136/jcp.45.12.1079 on 1 Decem

ber 1992. Dow

nloaded from