Use of the API 20E system to identify veterinary Enterobacteriaceae

0095-1 137/80/07-0010/05$02.00/0

Use

of the API

20E

System

to

Identify Veterinary

Enterobacteriaceae

EVRYLL C. SWANSON AND MICHAEL T. COLLINS*

DepartmentofMicrobiology, Colorado State University, Fort Collins, Colorado80523

A total of 503 veterinary enteric bacterial pathogens obtained from state veterinary diagnostic laboratories were tested on API 20E strips to determine

whether this rapid microidentification system coukl be utiized for veterinary

clinicalmicrobiology. The API 20E stripaccurately identified 96% of the

veteri-nary isolates and misidentified 3%. Identifications by the APIsystem and the diagnostic laboratorieswereinagreementin83% of the isolates, disagreementon 16% of the isolates, and 1% werenotidentified by the API strip. Differences in identification occurredprimarily in distinguishing betweenKlebsiellaand Enter-obacterand between Enterobacter and Escherichia coli. These disagreements were most often due to incorrect identifications by the diagnostic laboratory ratherthanby the APIsystem. Biotype differences between humari and veteri-nary isolateswere compared. Significant differences werenoted in several bio-chemical reactions. The main differences observed for E. coli isolates were in ornithine decarboxylase production and melibiose fermentation. The largest differences for Salmonella occurred in arginine dihydrolase production, citrate utilization, and inositol fermentation, whereas for Klebsiella pneumoniae the main differenceswere noted inureaseproduction and nitrate reduction. These biotype differences, however, did notaffect the accurateidentification of orga-nismsonthe API strip.

A large portion of the work in a diagnostic

laboratory consists of the identification of

bac-terial pathogens from the family

Enterobacte-riaceae. Veternarydiagnosticlaboratory

iden-tifications generally rely onconventional tube

media andlackthestandardizationprovided by

rapidmicroidentificationsystems.These micro-test systems are designed for easier use and

providemorerapid andaccurateidentifications

(by use of a computer-generated data base).

Rapid microtestsystems could benefit a

veteri-nary diagnostic laboratory, but these systems

are designed specificallyfor human clinical

iso-lates,andtheirapplicability for identification of

veterinary bacterialpathogens hasnotbeen

es-tablished. Many diagnosticians believe that there are biotype differences between human andveterinary bacterialpathogenswhichwould renderrapidmicrotestsystemslessaccuratefor veterinary clinicalmicrobiology.Thisstudywas undertaken to determine whether any

signifi-cant biotype differences exist between

veteri-nary enteric bacterial pathogens and human clinicalisolatesandwhetherthe APIsystemwas

capableofidentifyingthesepathogens.The API

20Estrip was chosen because of thelarge num-beroftestsavailable(whichwasbeneficial when

comparing the biotypes between human and

veterinarypathogens)and because of its proven accuracy

(1,

6, 8, 11, 12).

10

MATERIALS AND METHODS Cultures. Five-hundred threeveterinary Entero-bacteriaceae were collected from seven veterinary diagnostic laboratories across the country. Each cul-turewasassignedacode number andlyophilized. The number andanimal source of each organism tested are shown in Table1.

Before testingonthe APIstrip,lyophilizedcultures werereconstituted withsterileTrypticase soy broth (BBL Microbiology Systems) and plated on blood agar.Cultures thatweregrowing stronglywerethen tested without further passage. Cultures that grew weaklywerepassedoncemore on blood agarbefore testing.

API system. The API system(AnalytabProducts, Inc., Plainview, N. Y.) is composed ofaplasticstrip with 20 microtubescontaining dehydrated biochemi-cals. It consists of thefollowingtests:

o-nitrophenyl-Bl-D-galactosidase

(ONPG), arginine dihydrolase, ly-sineandornithinedecarboxylase, citrate, H2S,urease, tryptophan deaminase, indole, Voges-Proskauer, ge-latinase, and fermentation of glucose, mannitol, inosi-tol,sorbitol, rhamnose,sucrose,melibiose, amygdalin, and arabinose. Thestripswereinoculated(witha 24-h culture from blood agar) and interpreted by the manufacturers recommended procedures. Each cul-turewasalsotestedonAPI motilityand O-Fmedia andstreakedonto a blood-MacConkey biplate. Posi-tivereactionswereconvertedto aseven-digitprofile number, and identificationsweremade with theAPI Profile Index (fourth edition)and werecomparedwith thediagnostic laboratories' identification. Code num-bersnotfoundinthe API index werecalledintothe

on February 7, 2020 by guest

http://jcm.asm.org/

API 20E FOR VETERINARY IDENTIFICATIONS TABLE 1. Organisms and their sources used inthis study

Organism| Total _vineB _cine-ineEquine _prine-Fene C eAviane_Pig Other _kow

un-Arizona 3 1 1 1

Citrobacterfreundii 13 5 3 2 1 1 1

Citrobacterdiversus-levinea 4 1 2 1

Enterobacter aerogenes 4 4

Enterobacter agglomerans 10 3 5 1 1

Enterobacter cloacae 25 8 1 8 1 1 4 1 1h

Escherichia coli 197 74 45 18 6 2 7 25 3 1 16

Edwardsiella tarda 2 2

Hafniaalvei 2 1 1

Klebsiella oxytoca 8 1 2 1 1 1 1 1

Klebsiella ozaenae 1 1

Klebsiellapneumoniae 45 6 4 15 1 1 1 14 3

Morganellamorganii 1 1

Proteusmirabilis 23 4 3 1 1 1 il Y 1

Proteusvulgaris 5 3 1 1

Providencia rettgeri 1 1

Providencia stuartii 3 2 1

Providenciaalcalifaciens 2 2

Salmonellasp. 117 17 7 14 1 1 9 12 12d 44

Salmonella cholerae-suis 28 16 4 8

SalmonellaparatyphiB 2 1 1

Salmonellapullorum 1 1

Serratiamarcescens 2 2

Serratialiquefaciens 1

Yersiniaenterocolitica 3 1 3

aAsidentifiedby the API 50E.

bRabbit.

cIguana.

dMonkey (nine), ferret(one), opossum(one),skunk(one).

API computer center for identification. When there was disagreement between API and the diagnostic laboratory, or when theprofile number was not listed in theindex, the culture was sent to the APIreference centerfor further testing on a 50-test strip to obtain a finalidentification. When identifications by API and thediagnostic laboratorywerein agreement,the iden-tification was considered correct and not tested fur-ther.

Biotyping.Apreliminaryexamination of the bio-type differencesbetween human and veterinary iso-lateswasconducted. For eachorganism,the percent-age of the isolatesthat were positive for each reaction wascalculated. This percent positivity for the veteri-naryisolates was compared with the API percentage chart, which is based onreactionsof clinicalisolates of human origin (and is based solely on reactions obtained on the API strip). The significance of the differenceinpercentages for each reaction was deter-minedbythechi-square test.

RESULTS

Comparison of API anddiagnostic

labo-ratoryidentifications. Bacterial identification agreements and disagreements between the API system and thediagnostic laboratorieswere re-viewed.Results were separated on the basis of whether the identification was listed in the API

Profile Index or whether it was necessary to

consultthe API computercenter.Thiswasdone

because anadvantage ofmost

rapid

identifica-tion systems is the

simplicity

and efficiency of

using anindex. Asshown inTable 2,we found

that 78% of theisolateswereidentified thesame

byAPIand the

diagnostic

laboratory when

using

the index, and 83% were identified the same

when computer identifications were included.

Fifteen and sixteenper centof the identifications

differed whenusing the index and the total data

base,

respectively.

Seven percentof the

profile

numbers were not in the API

index,

whereas

only 1% were not in the computer data base.

This 1%

represented

four

organisms (three

Esch-erichiacoli andoneEnterobacter

agglomerans)

whichreceivedan

"unacceptable

identification"

designation by the computer,

indicating

that

theirbiochemical testpattern didnotresemble any taxastoredin the data base.Cultureswith

profilenumbers not inthéindex but in the total

data base consistedofunusualor

possibly

vet-erinarybiotypesof E. coli (13isolates),

Salmo-nella (12isolates),E. agglomerans (4

isolates),

Proteus (3

isolates),

Klebsiella

(1

isolate),

and

Providencia stuartii

(1

isolate).

A common E.

coli

biotype

thatwas notlisted in the indexwas

5-144-113.

il

VOL. 12,1980

on February 7, 2020 by guest

http://jcm.asm.org/

The API system andthe diagnostic laborato-ries disagreed on 81 (16%) of the organisms. Theseconflicting identifications, alongwith the finalidentifications,areshowninTable3. Four-teenof the different identifications varied only inspecies, and three of these were isolates

iden-tifiedasKlebsiellaoxytocaby API,whereas the

diagnostic laboratories identified them as K.

pneumoniae (2 isolates) andK. ozaenae (1

iso-late). K. oxytoca is a new species designation

and isprobablynotin common usage in many

laboratories. Disagreementsinidentification

oc-curredmostcommonlywhendistinguishing

be-tween Klebsiella-Enterobacter identifications

andE.coli-Enterobacteridentifications. In most

cases, there was no common cause for these

differing identifications. Of the 81 conflicting

TABLE 2. Comparison of API 20E and diagnostic laboratoryidentifications

Total API data

Identifica- base

tions

No. % No. %

Same 394 78 419 83

Different 76 15 81 16

Nottisted 34 7 4 1

identifications, the APIsystemcorrectly identi-fied62ofthe isolates, the diagnostic laboratories correctly identified 15, and 4 organisms were misidentified by both.

Considering all 503 veterinary isolates, the API 20E system correctlyidentified 96%, misi-dentified3%, andwasunabletoidentify 1%. The diagnostic laboratories correctly identified 87% and incorrectlyidentified 13% of the organisms. Biotype differences. Biotype differences be-tweenhuman andveterinary isolateswere com-pared in fourtaxa(E.coli,Salmonella, K. pneu-moniae, and Enterobacter cloacae) where suf-ficient numbers of organisms were studied to permit statistical analysis. Table4lists the

per-centpositivity for each biochemical reaction for each organism. These percentages were com-pared with the APIpercentage chart, which is basedonhuman clinicalisolates.Significant dif-ferences(atthe1%level) for E.coli isolateswere noted inlysine decarboxylase (for which veteri-nary isolates were 91.2% positive and human isolates were 82.8% positive), ornithine decar-boxylase (veterinary 60.3%; human 75.7%), sor-bitol fermentation (veterinary 88.1%; human 94.4%), melibiose fermentation (veterinary 91.8%; human 67.1%), and arabinose

fermenta-TABLE 3. IdentificationsdifferingbetweenAPIand thediagnostic laboratory

APIidentification Diagnostic labora-_{tory identification} Final identification |_ItoryAPIidentification Diagnostic_{identification}

labora-

Finalidentification 18E.coli 1Edwardsiella 18E.coli 4E. cloacae 1 E.aerogenes 4 E. cloacae

4K.pneumoniae 2K.pneumoniae

5Kozaenae 1E.coli

4Enterobacter

2E.aerogenes 3E.aerogenes 2 K.pneumoniae 3 E.aerogenes

1Arizona 1 E. cloacae

1Y.

entero-colitica 7 C.freundii 3 E.coli 3 C.freundii

3Enterobacter 2C.freundii 14K.pneumoniae 3Kozaenae 14K.pneumoniae 1Salmonella8p. 1E.coli

6Enterobacter 1Salmonellasp.

1E.liquefaciens

2E. coli 1C. diversus- 1Klebseilla 1C. amalonaticus

1E.agglomerans levinia

1C.freundii

4P.mirabilis 3P.vulgaris 4P.mirabilis 3K. oxytoca 2K.pneumoniae 3K.oxytoca 1C.freundii

1K. ozaenae

1P.rettgeri 1 C.freundii 1P.rettgeri

2K. ozaenae 2K.pneumoniae 2K.pneumoniae

2P.vulgaris 2 P. mirabilis 2 P.vulgaris 7Salmonellasp. 4Citrobacter 7Salmonella sp.

1E. aerogenes 1P.stuartii 1Proteus 1P.rettgeri

2Arizona

2P.alcalifaciens 2P.mirabilis 2P.vulgaris

1Arizona 1E. coli 1E.coli

2S.dysenteriae 1S.typhi-suis 1Salmonella

1S. choleraesuis 1Edwardsiella 1Salmonellasp. 1 E.agglomerans group C

1E.agglomerans 5E.agglomerans 3E. coli 1E.agglomerans

2K.pneumoniae 2E. coli 1H.alvei 1E.coli I H. alvei 2K.pneumoniae

1Entericgroup8 1E.coli 1E.coli

on February 7, 2020 by guest

http://jcm.asm.org/

API 20E FOR VETERINARY IDENTIFICATIONS TABLE 4. Percentageof veterinary strains positive

for each biochemical reaction % Positive

Biochemical test S Salmo- K E. cloa-E. coli nella

pneu-(194)' sp. caise20

(113) (41) (20)

ONPGb 97.4 0.9 100 95

Argininedihydrolase 2.6 91.2 O 100 Lysinedecarboxylase 91.8 100 90.2 0 Ornithinedecarboxylase 60.3 100 O 95

Citrate 0 15.0 92.7 95

H2S 3.6 97.3 0 0

Urease 2.1 0 87.8 0

Tryptophan deaminase O 0 0 O

Indole 97.4 0 0 0

Voges-Proskauer 0 0 97.6 100

Gelatinase 0 0 2.4 0

Glucose 99.5 100 100 100

Mannitol 99.5 100 100 100

Inositol 1.0 59.3 100 25

Sorbitol 88.1 99.1 100 95

Rhamnose 87.6 98.2 100 90

Sucrose 44.8 0 100 95

Melibiose 91.8 94.7 100 95

Amygdalin 4.6 0 100 100

Arabinose 99.5 99.1 100 100

Oxidase 0 0 0 0

Nitratetonitrite 99.5 99.1 51.2 90

Nitrate togas 0 0 46.3 5

GrowthonMacConkey 100 100 100 100 agar

Glucose,oxidative 100 100 100 100

Glucose,fermentative 100 100 100 100 aNumbers inparenthesesarenumberof strains.

bONPG,

o-nitrophenyl-B-D-galactopyranoside.

tion (veterinary 99.5%; human 90.8%).

Signifi-cant differences forSalmonellawere observed in arginine dihydrolase (veterinary 91.2%; hu-man 69.2%)7 citrate (veterinary 15.0%; human

75.4%), H2S (veterinary 97.3%; human 85.7%),

inositol fermentation (veterinary 59.3%;human 33.7%), and melibiose (veterinary94.7%; human

78.2%). Reaction differences forK.pneumoniae

biotypes were recorded for urease production (veterinary 87.8%; human 63.6%), indole produc-tion (veterinary 0%; human 19.2%), nitrate re-duction to nitrites (veterinary 51.2%; human 100%), and nitrate reduction to nitrogen gas (veterinary 46.3%; human 0%). No significant differenceswerenoted forE. cloacae.

When comparing veterinary K. pneumoniae

biotypes with the human clinical biotypes ob-tainedby DeSilvaandRubin(4),wefoundthat

the most common K.pneumoniae biotype for human andveterinary isolateswasthesame (5-215-773). For veterinary isolates, this biotype occurred at a frequency of 62%, whereas for human isolates it was observed in 51% of the isolates(4). Thisdifferencewassignificantatthe 5%level.

The mostcommonE.coli biotype frequencies

are shown in Table 5. These differ somewhat from thebiotype frequencies obtained by Davies (3) in his study of 574 E. colistrains, but the samepatternsarepresent.

DISCUSSION

The differing identifications reached by the APIsystemand thediagnostic laboratorieswere

the result of misidentifications made either by

the APIsystemorby the diagnostic laboratory orby both.Ofthe 15 organisms that API

misi-dentified, 6 isolates were given a "good

like-lihood, lowselectivity" rating by thecomputer, meaning that several organisms could fit that profile number. In three of thesecases,the

cor-rect identification was given as the second or third choice in the index, and the additional tests listed in the index probably would have given thecorrectidentification.Also,whenusing thissystem,other factors suchas colonial mor-phology and serology should be considered be-forereachingafinalidentification. Our identifi-cationswerebasedsolelyonreactionsobserved onthe APIstrip.

The diagnostic laboratories misidentified 64 isolates. For economicreasons,mostveterinary diagnostic laboratories utilizeaminimal number ofbiochemical tests to identify each organism. Many isolates, particularly E. coli, are often identified solely by colonial morphology. In this instance, cultures resemblingE.coli morpholog-ically would bemisidentified. Inourstudy, 13% ofthedisagreementswereorganismsincorrectly identifiedasE.coliby the diagnostic laboratory. Forotheridentifications, often only triplesugar iron agar slants and motility indole ornithine deepsareutilized.Occasionally, additionaltests suchascitrate,urease,

o-nitrophenyl-,8-D-galac-topyranoside, ormalonatewererun,butuseof thesetestswashighly variable between diagnos-tic laboratories. Identifications often are only

madetogenusandnot tospecies.

Insomecases,itwasfoundthat thediagnostic

laboratorywasmisreadingormisinterpretingits

tests. Forexample,seven K.pneumoniaewere identifiedasEnterobacterbyonediagnostic

lab-TABLE 5. Mostcommonbiotypes among194strains ofE.coli

Biotype No. of strains %

5 144 572 54 27.8

5 044 552 27 13.9

5144552 25 12.9

5044572 13 6.7

1044552 9 4.6

5044542 6 3.1

5144113 5 2.6

42otherbiotypes 55 28.4

13 VOL. 12,1980

on February 7, 2020 by guest

http://jcm.asm.org/

oratory. Theirbiochemical results indicated that

the organism was motile and ornithine

decar-boxylase negative. The negative ornithine

reac-tion is correctfor Klebsiella, not Enterobacter,

and retesting revealed that the organism was

nomnotile instead ofmotile, thus also favoring

aKlebsiellaidentification.

Afewconflicting identifications probably also

were due to the wrong culture being tested.

Some cultures obtainedwere onprimary

isola-tionplates, anddifferent coloniesmayhave been

tested. Also, some cultures appeared to have

beenmislabeled.

Overall, the API identifications were more

accurate than thediagnostic laboratories'

iden-tifications. The API systemcorrectlyidentified

96% of theveterinary isolatesascompared with

87%correctly identified by the diagnostic

labo-ratories. This 96%comparesfavorablytofigures

previouslyreported for the APIsystem in

iden-tifying human clinical isolates (1,5, 6,10, 11, 12).

Also, the APIsystem gave morecomplete

iden-tifications (downtospecies) andwas more

con-sistent duetothe

Profile

Index.

Theapplicability of theAPIstrip for

biotyp-ing hasreceivedconflicting opinions (2-4,9,10).

Ithas been criticized because of its lack of

re-producibility

when testing the same organism

several times. We compared the biotypes ofa

population of

organisms

rather than individual

biotypes,thereby reducing the

variability

within

eachbiochemical reaction.

Also,

thepercentage

obtained for each biochemical reactionwas

gen-erated in the same manner as for the human

isolate percentages that are given in the API

comparison chart.

Although

significant

reaction

differences were observed (as noted in the

re-sults), they didnotaffect thecorrect

identifica-tion oftheorganisms by theAPIindex.

Since the APIsystem canaccommodatemany

different biotypes, it appears to be

capable

of

identifyingveterinary bacteriainthefamily

En-terobacteriaceae withahigh degree of accuracy.

Thissystem is able to

identify

pathogens with

equal

orbetteraccuracy as

compared

to

conven-tionalmethods used bymostdiagnostic

labora-tories. Itis alsoadvantageousin that it is avery standardized system. Thestrip isalways

inocu-latedinthesamemanner,itcontains a standard

set of 20biochemical tests, and it isinterpreted withthe API Index. In this respect, results ob-tained within onelaboratoryand between differ-entlaboratories with API would be standardized and

easily

comparable. Although the API 20E strip ischeaperto usethanrunninga

compara-blesetofbiochemicaltests (7), theremaybea

slight increase in cost fordiagnosticlaboratories

that use onlya minimal battery oftests. This

cost would be offsetby the convenience of the

APIstrip, itsaccuracy,and thestandardization

ofresults obtained. Itisconcluded thatarapid

microtestsystemsuchasAPI 20E isfeasible for

identifying veterinary bacterial pathogens from

thefamily Enterobacteriaceae.

ACKNOWLEDGMENTS

This research was supported in part by Public Health Service BiomedicalResearchSupportGrant 5-507-RR-05458-15from the National Institutes of Health and Analytab Prod-ucts.

We thank thefollowing laboratories for providing us with veterinary enteric organisms: Animal Disease Laboratory,

Centraliae,Ill.;Wyoming State Veterinary Laboratory; and the VeterinaryDiagnostic Laboratories at University of Min-nesota,University ofGeorgia, Kansas State University, Iowa State University, and especially Colorado State University.

LITERATURE CITED

1. Brooks,K.A.,M.Jens,and T. M.Sodeman. 1974.A clinical evaluation of the API microtube system for identification of Enterobacteriaceae. Am. J. Med. Technol. 40:55-61.

2. Butler, D. A., C. M.Lobregat, and T. L. Gavan. 1975.

Reproducibilityof theAnalytab (API 20E) system. J. Clin. Microbiol. 2:322-326.

3.Davies, D. I. 1977. Biochemical typing of urinary Esche-richiacoli strains by means of the API 20E Enterobac-teriaceae system. J. Med.Microbiol.10:293-298. 4. DeSilva,M.I.,and S. J. Rubin.1977.Multiple biotypes

ofKlebsiellapneumoniae insingleclinical specimens. J.Clin.Microbiol.5:62-65.

5. Hayek,L.J.,and G. W.Wiilis.1976. Acomparison of twocommercialmethods for the identification of the Enterobacteriaceae-API 20E and the Enterotube-withconventionalmethods. J. Clin. Pathol. 29:158-161. 6. Nord, C. E., A. A.Lindberg,and A.Dahlback.1974. Evaluation of five test-kits-API, Auxotab, Enterotube, Pathotec andR/B-for identification of Enterobacte-riaceae. Med.Microbiol. Immunol. 159:211-220. 7. Robertson,E.A.,G. C.Macks, and J. D.MacLowry.

1976.Analysis of cost and accuracy of alternative

strat-egiesfor Enterobacteriaceaeidentification.Apple. Mi-crobiol.3:421-424.

8. Robertson, E.A.,and J. D.MacLowry.1974. Mathe-maticalanalysis of the API enteric 20profile register usinga computerdiagnostic model. Apple. Microbiol. 28:691-695.

9. Rubin,S.J.,S.Brock,M.Chamberland,andR. W.

Lyons. 1976. Combinedserotypingand biotypingof

Serratia marcescens. J.Clin. Microbiol.3:582-585.

10.Rutherford, I.,V.Moody, T. L. Gaven, L. W.Ayres,

andD.L.Taylor. 1977.Comparative studyof three methods of identification of Enterobacteriaceae. J. Clin. Microbiol.5:458-464.

11. Smith, P. B., K. M.Tomfohrde,D.LRhoden, andA. Balows.1972.APISystem:Amultitube micromethod foridentification of Enterobacteriaceae.Apple. Micro-biol.24:449-452.

12.Willis, G., and W. Y. Cook.1975. Acomparativestudy ofAPI, Encise,andconventionalmethods.Med. Tech-nologist 5:4.

on February 7, 2020 by guest

http://jcm.asm.org/