Rubella hemagglutination inhibition test: false positive reactions in sera contaminated with bacteria

(1)

Rubella

Hemagglutination-Inhibition

Test:

False-Positive

Reactions in Sera Contaminated with Bacteria

J. B. CAMPBELL,* MYROSLAVA ROMACH, AND MARY L. ELLINS

Department of Microbiology and Parasitology, Faculty of Medicine, University of Toronto, Toronto, Ontario,

Canada

M5SlAl

Receivedfor publication 23 June 1976

We have shown that bacterial contamination of sera can have a marked influence on the results of the rubellahemagglutination-inhibition (HI)test. In

additionto increasingthe levels of nonspecific agglutinins, a number of common bacterial species tested had a significant effect on the HI titers. When sera free

from rubella-specific antibodies were contaminated with Bacillus subtilis or Pseudomonas

fluorescens,

HI titers ranged from 16 to 128 (expressed as the reciprocal of the highest serum dilution completely inhibiting hemagglutina-tion) after treatment withheparin/MnCl2. Ourobservations demonstrate,

there-fore, that bacterial contamination can be one of the causes of false-positive reactions in this test.

Itis awidespread practice in diagnostic labo-ratories to collect blood samples into nonsterile tubes (such as vacuum blood-collecting tubes) and to store the separated sera or plasmas in

clean, butnonsterile, containers. It is not sur-prising, therefore, that a certain proportion of

these samples become contaminated with bac-teria. Yetlittle attention has been paid so far to thepotential effects of these bacterial contami-nants and their products on theconstituents of serumorplasma.

In a study of the lipid patterns of contami-nated sera, it was shown that in certain sam-ples there was a decrease in lipoproteins and

phospholipids and a concurrentincrease in se-rumtriglycerides. Phospholipase C-producing, gram-positive, aerobic, sporeforming bacilli were isolated from these samples (13). Treat-ment of serum withphospholipase Chas been

reportedtobe aneffectivewayof inactivating

the lipoproteinaceous nonspecific inhibitors (NSI) of rubella virus hemagglutination (HA)

(5, 8). It mightappear, therefore, that for the rubella hemagglutination-inhibition (HI) test,

bacterial contamination of this sort would be

advantageous rather than the opposite. How-ever, partial hydrolysis of the phospholipid

moietiesby thisenzyme has been found to in-terfere with the formation of insoluble com-plexes between 3-lipoproteins and heparin/

metal cation mixtures (15), and treatment of serum with heparin/MnCl2 is one of the most

widelyused methods for the removal of lipopro-tein

NSI

of rubellaHA.Itispossible,

therefore,

that contamination with phospholipase C-pro-ducing bacteria may result in the

incomplete

removal of NSI, thus producing false-positive reactions.

Two situations in whichbacterial contamina-tionproduced false positives in the rubella HI test have, in fact, been reported. White and Tinnion (17) observed that sera contaminated byanorganism closelyresembling

Pseudomo-nas

fluorescens

containedanonspecific

inhibi-tor of rubella HA that was not removed by treatmentwithheparin/MnCl2 (17). Biano and co-workers(3) have also describeda caseof mis-taken diagnosis of congenital rubella due to contamination ofserawith thisorganism. Bac-terial actionappearedtobe in the,8-lipoprotein fraction ofserum, rendering this resistant to precipitation withheparin/MnCl2. It isalso in-terestingto note that over30years ago it was

reported that contamination of blood with P.

fiuorescens,

P. pyocyanea, and Bacillus

sub-tiliscaused excessive increasesinthe antistrep-tolysintitration (9).

In addition to affecting the levels of NSI,

there is also thepossibility that certain bacte-rialcontaminants can reduce the concentration ofspecificimmunoglobulins. Forexample,most strainsofStaphylococcusaureus containacell wall protein, the so-called protein A, which complexeswiththeFcportion of immunoglobu-linG(IgG) molecules. Sufficiently large quan-tities of S.aureus mightcause anappreciable

reduction in rubella-specific IgG levels, and this method has, in fact, been used for the removal ofIgG inorderto permitdirect deter-minationofspecific IgMlevels (2).

Yet another way that bacterial contamina-tionmight affect the rubella HItestis

by

the

389

on February 7, 2020 by guest

http://jcm.asm.org/

(2)

production of nonspecific agglutinins (NSA). For example, three strains of enteric bacteria were found to agglutinate the blood cellsofa number of animal species (16). Since human sera do not normally contain agglutinins for trypsin-treated human cells, use ofthese cells asindicatorsinthe rubellaHItestobviatesthe need for prior absorption of the sera with packed erythrocytes (11, 12). Introduction of NSA by bacterial contamination, therefore,

might mask low specific antibody titers and would necessitate repeating the test after ab-sorptionoftheagglutinatingmaterial.

To investigate these possibilities more closely, we initiated a study of the effect of commonbacterial contaminantsonthe rubella HItest. The preliminary results reported here

indicate that whereas some common bacterial species have no deleterious effects, others can

influencethe assay results to a marked degree andmayberesponsible forsomeof the reports of falsepositivesthat limit the usefulness of the assaymethod.

MATERIALS AND METHODS

Rubella HA and HI tests. Rubella HA and the HI testswerecarried out as in previous work (6), using themodifiedCenter for Disease Control (CDC) pro-tocols withtrypsin-treatedhuman 0 cells as indica-tors(6,10, 11).Nonspecific inhibitors were removed

bytreatmentwithheparin/MnCl2asdirected in the CDC procedure except that 0.75 M MnCl2 was used instead of the recommended 1.0 M solution. This

decreased the incidence of nonspecific agglutina-tion. The sera were notabsorbed with blood cells

beforeassay.

Human type0, Rh-negative blood was obtained from a singledonor (J.B.C.) and was stored in Al-seversolutionat4°C forup to 3 weeks before use.

Trypsinization oftheerythrocytes was carried out

by themethod of Quirin et al. (11), using 2 x

crystal-lizedtrypsin(Sigma ChemicalCo., St. Louis, Mo.). The trypsinized cells were stored in dextrose-gela-tin-veronal(DGV) buffer, pH 7.3, for amaximum of 7days beforeuse.

Rubella HA antigen, lot number 2020-1, was

pur-chased from Connaught Diagnostics, Willowdale,

Ontario. It had been prepared from the HPV-77 strain ofvirusbyTween-ether extraction, andwas

obtainedasalyophilized preparationwith a titer of 128 HA units. Sodium heparin (5,000 U/ml) and HEPES (N-2-hydroxyethylpiperazine-N'-2-ethane-sulfonate)-buffered diluent (HEPES-saline-albu-min-gelatin), pH6.2, were alsoobtainedfrom Con-naughtDiagnostics. Alseversolution and dextrose-gelatin-veronal buffer were made up as recom-mended in the CDC procedure (10). Reagent-grade chemicalswere used in allformulations.

Assays were carried out in rigid polystyrene V plates (Linbro Chemical Co., New Haven, Conn.),

usingMicrotiter droppers and diluters (Cooke Engi-neering Co., Alexandria, Va.). All titers were

ex-pressedasthereciprocalof thehighestserum dilu-tionthattotallyinhibitedhemagglutination.

Bacteria.Pure cultures ofbacteria were obtained

from theDepartmentofMicrobiologyand Parasitol-ogyteaching collections. Sera were tested for

bacte-rial contaminants by streaking loopfuls on blood

agarplatesandincubatingthese at 25°C for 48 h. A

generalidentification of isolated colonies was then madebyGramstaining,by colony morphology,and

bythepresenceorabsence ofhemolysison theblood

agarplate. Furtherclassificationwasmade in some casesbybiochemicaltests.

Seratobe contaminated with aparticular

bacte-rium wereinoculatedwithloopfulsofplate cultures (1loopful/0.6ml ofserum). Thesampleswerethen left at25°C for 48 h beforetheywereassayedfor HI

activity.

Sera. Most oftheserausedinthepresentstudy

wereclinicalspecimensdrawn in severaldiagnostic laboratories for serumchemistryorrubella HI

anti-body assays. All sera werefrombloodcollected by venipunctureintononsterileVenojectvacuumblood

collecting tubes, without additives (Jintan Terumo Co., Ltd., Tokyo, Japan),which wereallowedto clot atroomtemperaturefor 6 to 12 h beforeseparation. Furtherdetails of the storageconditions of thesera aredescribed in Results.

RESULTS

Bacterial contamination ofsera. On three separate occasions, batches of clinical serum

specimensweretestedforbacterial contamina-tion. All sera were collected as described in Materials andMethods.

The first batch of 98 sera had been stored in clean but nonsterile, unused glass test tubes takenstraightfrom the manufacturer's carton. The samples were left uncovered at room tem-peraturefor up to 6 h and then stored at40Cfor about a week in racks covered with Parafilm. When plated on blood agar, 3 of the 98 sera (3.1%) were found to contain bacteria. These appeared tobe a mixture ofcoliforms, diphthe-roids, and staphylococci, with the gram-nega-tive rodspredominating.

In another survey, of 95 sera collected in reused, detergent-washed Venoject tubes and stored at40C for 1 to 3 days, 15 (16%) showed significant levels ofcontamination. The pre-dominant organism among these (8/15) was a

diphtheroid (Corynebacterium species), with the remainder of the samples being contami-nated with Alcaligenes faecalis (4/15), a coli-form (1/15), and a staphylococcus (2/15).

Over a period of several months, all sera assayed forrubella-specific HI antibodiesthat showed the presence ofnonspecific agglutinins werestored at -20°C forfurther testing. Most of these samples had been stored in reused, detergent-washedVenoject tubes. Whentested on blood agar plates, 24 out of 68 (35%) were

http://jcm.asm.org/

(3)

found to be contaminated with bacteria, some

samples withmorethanonespecies.Coliforms

andother enterobacteriaceae predominated, al-though staphylococciwerealsoquiteprevalent.

Effectof various bacteriaonthe rubellaHI

test.Fromthe work described inthe introduc-tion, the possibilities existed that

contamina-tionofserumwith certain bacteria might result

in: (i)theproductionof NSA; (ii)adecrease in

the specific antibodytiters; and/or (iii) an

in-crease ordecrease inthe level of NSI.

To eliminate the possible involvementof nat-urally occurring agglutinins, serum was

col-lected asepticallyfrom thedonorwho provided the erythrocytes for the HItest. Samplesofthis

serum wereinoculated with thevarious

bacte-ria and incubated for 2 days at 25°C before being assayed for rubella-specificHIantibodies andtotalHI (i.e.,rubella-specific HI plus NSI). Serial twofold dilutions of the serum controls werealso preparedso astoprovideameasureof

the levels of NSA. Results are summarized in

Table 1.Itcanbeseenthatmanyof the

bacte-riatestedproduced NSA thatwerenotremoved bytreatmentwithheparin/MnCl2.

The uncontaminated serum was shown on

several occasionsto havea specific rubella HI

titerof256. Twobacteria(Escherichia coli and Streptococcus faecalis) reduced thistiter signif-icantly (a fourfold increase or decrease being

consideredsignificant). One bacterium, B. sub-tilis, increased theapparentantibodytiter sig-nificantly. That this apparentincrease in

spe-cific antibodytiter was actually dueto an in-crease intheactivity of the NSIwasshown in

the following experiment.

Three organisms were chosen for further

study: P. fluorescens, B. subtilis, and S. au-reus. Serum samples from eight different

pa-tients, shown previouslytobe free of

rubella-specific antibody, wereinoculated with one or

the other of these organisms. After 2 days at 25°C, theywereassayed for rubella-specific and

totalHI asdescribed above. Resultsareshown

inTable2.

The uncontaminated controls showed a low

level of NSAin oneof the eight samples, and

total HI titers ranged from 128to4,096. These

arenormalrangesinoursystem.Treatment of

theserawithheparin/MnCl2removed alltraces

of NSA and NSI. Contamination with S. au-reusappearedtoreduce the NSI levels, and the remainderwasremoved completely after

treat-mentwith heparin/MnCl2. With the other bac-teria, however, the situation was quite

differ-ent.P.fluorescens producedalow levelof NSA

inall theseraand increased the NSIinmostof

them. Aftertreatmentwithheparin/MnCl2, all except one serum showed residual NSI. This was even more evident in the sera

contami-nated with B. subtilis. This bacterium

pro-duced very high levels of NSA, and the end

points of the NSI were not reached at serum

dilutions of 1:16,384. AswiththeP. fluorescens-contaminatedsamples,heparin/MnCl2was

un-abletoremoveallnonspecific inhibitory

activ-ity.

DISCUSSION

One of thedrawbacks of therubellaHItestis that there isno completely satisfactorywayof

ensuring that what is being measured is

ac-tually (and solely) specific antibody. By the incorporation of suitable controls in theassay,

the presence orabsence of NSAcanbereadily

monitored. We routinely test our serum

con-trolsthroughadilution of 1:32topermitsome

degreeofquantitationof the levelof NSA. It is

amuchmore difficult tasktomonitor the effi-ciencyof the methods for removal of NSI.

Un-TABLE 1. Effect of differentbacteria onthe rubella HItest

Titers beforetreatmentwithheparin/ Titersaftertreatmentwith

Bacterialcontaminant MnCl2 herparinlMnCl2

NSA HI NSA HI

Staphylococcusaureus 8 a2,048 8 512

Streptococcus faecalis _a ¢2,048 - 64

Alcaligenes faecalis 8 2,048 - 256

Bacillus subtilis >32 ¢2,048 >32 1,024

B.cereus 8 32,048 - 64

Serratiamarcescens 8 32,048 - 128

Escherichiacoli - ¢2,048 8 64

Pseudomonas aeruginosa 8 >2,048 - 256

P.fluorescens 8 ¢2,048 - 256

Klebsiella aerogenes 8 -2,048 8 128

Salmonellagallinarum - ¢2,048 8 128

Control(sterile) - 2,048 - 256

a -,Negative (<8).

391

http://jcm.asm.org/

(4)

00 " CD000CX tCO c c9COcOm ' N e I-000000(=== "

,- -~", "~a, 1~

00

00 00 C4 C14 LO ~

W-- eC Lo

00

4 00 cN CS111C4C41C0 C c- C, CD 0COm,4CO

00

o

Cli

A\

0000000000000000

Cicoicoc coiclico co

=CDCO ADw wC CD GD

V-A V-A A A A A A A

000X co

cli e-q

_ U CQI CQ O N

A

IS DXJQ D 00

Iqto W1N CD

CDCS t 00 000000

o

t 00 r;c_ a c

C000000000000000

CD1 CD COCD D CCC

.z 00 >z

V-0

4) q-C.) CUCO

0

Z S. a)

E--less physical techniques such as density gra-dientorflotationultracentrifugation areused,

there isnowayofdetermining whether all the NSI of rubella HA have been separated from

specific antibody. Even these methods arenot infallible, since NSI have been found in frac-tionscontaining IgG (7) and IgM (1) after su-crose gradient ultracentrifugation, and flota-tion methods are based upon the assumption

that all inhibitors have the properties of

lipo-proteins. Inaddition, these methods wouldnot be suitable for routinediagnostic use.

Within the confines of the HI test itself, the mostsatisfactory method ofmonitoring the re-moval of NSI istoinclude among the controlsa negative serum that contains veryhigh levels of NSI. If this serum repeatedly registers as negative in the test assay, it islikely that equal orlesser amountsof NSI present in othersera will also beeffectively removed. This is proba-bly a reasonable assumption in the case of ster-ile sera. Our data clearly demonstrate, how-ever, that if the sera are contaminated with bacteria, false-positive reactions may be pro-duced.

Atpresent, we do not know whether the bac-teriamodify and enhance existing inhibitors or whether they produce different ones. For the purposes of routine diagnosis of rubella infec-tionby the HI test,however, the problem is not a pressing one. It can even be avoided com-pletely by ensuring the sterility of the sera beforetesting. In situations where total steril-ity is impracticable, the chances of bacterial contamination can be reducedbyallowing the blood to clot at4°C andmaintainingthe sera at this temperature until assayed. Allowing the bloodtoclot at this temperaturehasthe added advantage that cold agglutinins, if present, wouldprobablybe removed byadsorption onto theerythrocytes (14).

Ourdataindicate thatcontamination ofsera with severaldifferent bacteria can result in the production of low levels of nonspecific aggluti-nation (high levels in the case of B. subtilis) that may not be removed by treatment with

heparin/MnCl2. The extent of NSA production, ofcourse, would depend on a number of varia-bles affecting bacterial growth, including the storage conditions of the sera and the presence or absence of antibiotics. Not all instances of NSA occurrence are due to bacterial contami-nation, however: other factors almost certainly contribute. For example, Biano et al. (4) have shown that an excess ofmanganous chloride can agglutinate erythrocytes. If a serum sam-ple has only low levels of

heparin/MnCl2-precip-itablematerial, sufficientmanganous ions may

0 _e

C.)

o

0

Co

C4

Co Co to C)

.2

Co

-o .0

C)

.)I

0

C.) Co.

04

z

4 z

CoZ

00

_ z

m

z

Z

4. co

v

z

Cl)

z

m

z

N oZ

z

Co Co 0

-4

-co

._

Co .0

eC

Co

http://jcm.asm.org/

(5)

remain inthe treated supernatant to produce this effect. We have found that reduction of the MnCl2 concentration from the CDC-recom-mended 1.0to 0.75 M decreases the incidence of NSA without having any deleterious effect on the removal of NSI. The NSA produced by the bacteria we have studied can be removed by absorption of the sera before assay with packed erythrocytes. However, one of the advantages ofusing trypsin-treated human erythrocytes as indicator cells is that routine serum absorption isunnecessary. Reintroduction of this step due to bacterial contamination, therefore, elim-inates thisadvantage.

Since bacterial contamination undoubtedly increasesthe incidence of nonspecific aggluti-nation, wefeel that use of the human cell sys-temprovides a diagnostic advantage over sys-temsrequiring routine serum absorption (such as those using day-old chick cells). In those cases where levels of NSA are produced by

bacterial action, routine absorption may ac-tually contribute to the problem of false posi-tives, since the NSA will be removed but the increased levels of NSI may not be, and may

thereforegoundetected. Our observations indi-catethatifnonsterileseraarebeing tested, any level of NSA warrants caution in the interpre-tation of results. Certainly, the cause of high levels of agglutination should be investigated

further.

LITERATURE CITED

1. Al-Nakib, W., J. M. Best, and J. E. Banatvala. 1974. Rubella-specific IgM anda newinhibitor.Br.Med.J. 3:579.

2. Ankerst,J.,P.Christensen, L.Kjellen, and G. Kron-vall. 1974. A routine diagnostic test for IgA and IgM antibodies torubella virus: adsorption of IgGwith Staphylococcusaureus. J.Infect.Dis. 130:268-273. 3. Biano, S., W.Cochran, K. L.Herrmann, A. D.Hall,

and T.-W. Chang. 1975. Rubella reinfectionduring pregnancy. A case ofmistakendiagnosis of congeni-tal rubella.Am.J. Dis. Child. 129:1353-1356. 4. Biano, S.A., T.W. Chang, andJ. B.Daniels. 1974.

Rubella hemagglutination inhibition: removal of

nonspecificagglutinationdueto manganouschloride. Appl. Microbiol. 28:992-994.

5. Blom, H. A., and G. Haukenes. 1974. Identification of non-specific serum inhibitors of rubella virus hemag-glutination. Med. Microbiol. Immunol. 159:271-277. 6. Campbell, J. B., T. Grunberger, and M. L. Ellins. 1975.Influence of albumin on rubella virus hemag-glutination and the hemagglutination-inhibition test. Can. J. Microbiol.21:1172-1177.

7. Haukenes,G. 1973. A rubella hemagglutination inhibi-tor simulating antibody. Acta Pathol. Microbiol. Scand.Sect. B 81:719-723.

8. Haukenes, G., and H. Blom. 1975. False positive ru-bella hemagglutination inhibition reactions:

occur-rence and disclosure. Med. Microbiol. Immunol. 161:99-106.

9. Lofgren, S. 1944. Bacterialcontamination of blood sam-ples as a source of error inantistreptolysin titration. Acta Pathol. Microbiol. Scand. 21:768-774. 10. Palmer, D.F., J. J. Cavallero, K. L. Herrmann, J. A.

Stewart, and K. W. Walls (ed). 1974. Themodified rubellahemagglutination-inhibitiontest. In A proce-dural guide to the serodiagnosis oftoxoplasmosis, rubella, cytomegalic inclusion disease, and herpes simplex. Immunology Ser. no. 5. Center for Disease Control, Atlanta, Ga.

11. Quirin, E. P., D. B. Nelson, and S. L. Inhorn. 1972. Use oftrypsin-modified human erythrocytesin ru-bellahemagglutination-inhibition testing. Appl. Mi-crobiol.24:353-357.

12. Schmidt, N. J.,and J. Dennis. 1972. Modified hemag-glutination-inhibitiontestforrubellaemploying hu-mangroup 0 erythrocytes. Appl.Microbiol. 23:471-475.

13. Schwetner, H. A.,andH.S.Friedman. 1973.Changes inlipidvaluesandlipoproteinpatterns of serum sam-plescontaminatedwith bacteria.Am. J. Clin. Pathol. 59:829-833.

14. Smith, J.A.,and A. C.Cummins. 1976. Evaluation of a rubellahemagglutination inhibitiontestsystem. J. Clin. Microbiol. 3:5-7.

15. Srinivasan,S.R.,B.Radhakrishnamurthy, ahd G.S. Berenson. 1975.Studies on the interaction ofheparin with serum lipoproteins in the presence of Ca2, Mg+2, and Mn+2. Arch. Biochem. Biophys. 170:334-340.

16. Tohda, H. 1974. Cellagglutinating activities ofsome

enteric bacteria and effects of various factors on bac-teria-tumor cell agglutination. Sci. Rep. Res. Inst. TohokuUniv. Ser. C 21:10-16.

17. White, G. B. B., and K. H. Tinnion. 1975. A

non-specificinhibitor of rubellahemagglutination. Lan-cet ii:664.

http://jcm.asm.org/