How to Select and Interpret Molecular Strain Typing Methods for... Bacterial Infections: A Review for Healthcare Epidemiologists

How to Select and Interpret Molecular Strain Typing Methods for... Bacterial Infections: A Review for Healthcare Epidemiologists

How to Select and Interpret Molecular Strain Typing Methods for Epidemiological Studies of
Bacterial Infections: A Review for Healthcare Epidemiologists
Author(s): Fred C. Tenover, Robert D. Arbeit, Richard V. Goering
Reviewed work(s):
Source: Infection Control and Hospital Epidemiology, Vol. 18, No. 6 (Jun., 1997), pp. 426-439
Published by: The University of Chicago Press on behalf of The Society for Healthcare Epidemiology of
America
Stable URL: http://www.jstor.org/stable/30141252 .
Accessed: 02/11/2011 15:18
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].
The University of Chicago Press and The Society for Healthcare Epidemiology of America are collaborating
with JSTOR to digitize, preserve and extend access to Infection Control and Hospital Epidemiology.
http://www.jstor.org
S S SSI
ma
SHEA
How
to
Molecular
OS
Position
Strain
Interpret
Typing
Methods
of
Studies
Epidemiological
A
Review
for
Paper
and
Select
Healthcare
I
Bacterial
for
Infections:
Epidemiologists
V.Goering,PhD;
FredC.Tenover,PhD;RobertD. Arbeit,MD;Richard
the Molecular
ofAmerica
Groupof the SocietyforHealthcare
TypingWorking
Epidemiology
ABSTRACT
Straintypingis an integralpartof epidemiological
of nosocomialinfections.Methodsfor disinvestigations
tinguishingamongbacterialstrainshave improveddraoverthe last5 years,duemainlyto the introducmatically
tion of moleculartechnology.Althoughnot all molecular
techniquesare equallyeffectivefor typingall organisms,
is the techniquecurrently
pulsed-field
gel electrophoresis
favoredformostnosocomial
to aidepipathogens.Criteria
Outbreaksof nosocomialinfectionscontinueto
occur among patients in a variety of healthcaresettings.1,2Although fungi, viruses, and parasites can
cause nosocomialinfections,bacterialagents remain
the most commonlyrecognizedcause of outbreaksof
disease.' In hospitals,the use of indwellingcatheters,
ventilators,and a variety of other medical devices
often serve as sources and conduitsfor bacteriaand
make efforts to control outbreaksmore difficult.3In
addition,the decreasedeffectiveness of some antimicrobialagents due to the emergence of resistantbacteria4complicatesinfectioncontrolefforts.5
Investigationsof presumedoutbreaksof bacterial infectionsin hospitals often require strain typing
data to identifyoutbreak-related
strainsand to distin-
resultshavebeenpublished.
demiologistsin interpreting
Nucleicacidamplification-based
typingmethodsalso are
to manyorganismsandcanbe completedwithapplicable
in a single day,but interpretivecriteriastill are under
debate.Straintypingcannotbe used to replacea sound
epidemiologicalinvestigation,but serves as a useful
adjunct to such investigations (Infect Control Hosp
Epidemiol1997;18:426-439).
guish epidemic from endemic or sporadicisolates.6
Straintypingdata,particularlythose that are generated by using newer molecular methods, have been
helpfulfor investigatingoutbreakscaused by a wide
range of bacterialpathogens, including methicillinresistant strains of Staphylococcus aureus (MRSA),7
enterococci(VRE),8
Pseudomonas
vancomycin-resistant
and Klebsiellapneumoniae.10
The moleaeruginosa,-9
cular approaches, which include pulsed-field gel
electrophoresis (PFGE), arbitrarilyprimed polymerase chainreaction(AP-PCR)assays, and plasmid
fingerprinting (PF), have several advantages over
traditional typing methods, including higher discriminatorypower,broaderapplicationto a varietyof
bacterialspecies, and, at times, speed.6While many
From the HospitalInfectionsProgram (Dr Tenover),Centersfor Disease Controland Prevention,Atlanta, Georgia;the Boston
Veterans'Affairs Medical Center (Dr Arbeit), Boston, Massachusetts;the Department of Medical Microbiology(Dr Goering),
CreightonUniversity,Creighton,Nebraska;and the Universityof Texas (Barbara E. Murray,MD), Houston, Texas;Mayo Clinic
(David H. Persing, MD, PhD), Rochester,Minnesota;the Universityof Iowa (MichaelA. Pfaller,MD), Iowa City,Iowa; and Cook
CountyHospital (RobertA. Weinstein,MD), Chicago,Illinois.
Addressreprintrequeststo Fred C. Tenover,PhD, NosocomialPathogensLaboratoryBranch (G08), Centersfor Disease Control
and Prevention,1600 CliftonRd, Atlanta, GA 30333.
95-SR-206.TenoverFC, ArbeitRD, GoeringRV theMolecularTypingWorkingGroupof the Societyfor HealthcareEpidemiology
ofAmerica.How to selectand interpretmolecularstrain typingmethodsfor epidemiologicalstudiesof bacterialinfections:a reviewfor
healthcare
InfectControlHospEpidemiol1997;18:426-439.
epidemiologists.
Vol. 18 No. 6
SHEA POSITION
PAPER
of these newer methods now are utilized in clinical
and reference laboratoriesaround the world, there
oftenis disagreementon how to interpretthe results.
In most cases, the focus of discussion relates to the
degree of variabilityin the test results that can occur
before two isolates are determinedto belong to different strains.'1
The goal of this position paper is to provide
healthcareepidemiologistswith an overview of the
methods availablefor analyzingbacterialisolates suspected of causing nosocomialoutbreaksand to propose criteriafor interpretingthe results of those studies. The emphasiswillbe on those molecularmethods
whose utilityhas been validatedby studiesperformed
in multiplelaboratorieson large sets of isolates.More
traditionaltechniques,such as staphylococcalphage
typing,biotyping,and pyocintyping,12are not considered here. It is criticalthatepidemiologistsunderstand
and appreciatethatthe results of straintypingstudies
do not substitutefor a sound epidemiologicalinvestigation. Rather,the laboratorystudies and the epidemiologicalstudies of suspected outbreaksshould
be analyzedin paralleland should be viewed as complementarycomponentsof the overallinvestigation.
BACKGROUND
CONSIDERATIONS
The abilityof moleculartypingsystemsto distinguish amongepidemiologicallyunrelatedisolates is a
reflectionof the genetic variationseen in the chromosomal DNA of a bacterialspecies.'3Usually,this variabilityis high, and differentiationof unrelatedstrains
can be accomplishedusing any of a varietyof techniques.14However,the factorsthat enable bacteriato
cause infection often are not uniformlydistributed
withina species.Thus,the organismsmost commonly
associatedwithinfectionsoftenare a smallersubset of
the manystrainsthat constitutea species.1"As a consequence, this subset may exhibit relatively little
genetic diversity,and it can be difficultto differentiate
amongstrainseven with newermoleculartechniques.
Methicillin-resistantStaphylococcusaureus are an
example of this phenomenon.Because most MRSA
recoveredfrom patientsare derivedfrom a relatively
smallnumberof clones,16there are a limitednumber
of straintypes that can be differentiatedregardlessof
the typing method used. While phenotypic methods
often cluster isolates of MRSAinto a few broad groups,
most molecular methods are capable of differentiating
endemic from epidemic strains."7
CHARACTERISTICS
OF TYPING
METHODS
Typing methods fall into two broad categories:
phenotypicmethods and genotypicmethods. Phenotypic
427
methods are those that characterizethe productsof
gene expression in order to differentiate strains.
Properties such as biochemical profiles, bacteriophage types, antigens present on the cell's surface,
and antimicrobialsusceptibilityprofilesall are examples of phenotypicpropertiesthat can be determined
in the laboratory.Because they involvegene expression, these properties all have a tendency to vary,
based on changes in growth conditions, growth
phase, and spontaneousmutation.
Genotypicmethods are those that are based on
an analysis of the genetic structure of an organism
and include polymorphismsin DNA restrictionpatterns based on cleavage of the chromosome by
enzymes that cleave the DNA into hundredsof fragments (frequentcutters), or into 10 to 30 fragments
(infrequentcutters), and the presence or absence of
extrachromosomalDNA.Genotypicmethodsare less
subject to naturalvariation,although they can be
affected by insertions or deletions of DNA into the
chromosome,the gain or loss of extrachromosomal
DNA, or randommutationsthat may create or eliminate restrictionendonucleasesites.
All typing systems can be characterized in
terms of typeability,reproducibility,discriminatory
power,ease of performance,and ease of interpretation.6The characteristicsof a numberof typingmethods are presentedin Table 1. Typeabilityrefersto the
ability of a technique to assign an unambiguous
result (type) to each isolate. Although nontypeable
isolates are more commonwith phenotypicmethods,
they have been recognizedwith most methods. For
example,with PFGE,a techniquethat is almost uniformly applicable to bacteria," some strains of
Clostridiumdifficileremainnontypeablebecause the
chromosomal DNA is degraded, presumably by
endogenousnucleases,beforeit can be cleavedproperly by the restriction endonucleases used in the
PFGEprotocol.18
A reproduciblemethod is one that yields the
same results upon repeattesting of a bacterialstrain.
In the contextof an epidemiologicalstudy,this means
that the same strainrecoveredfromepidemiologically linked patients will give the identical (or nearly
identical) typing result. Poor reproducibility may
reflect technical variation in the method or biologic
variationoccurringduringin vivo or in vitropassage
of the organismsto be examined. Over time (a few
weeks to years, dependingon the species), the typing
patternsproducedby DNA-basedmethods, such as
PFGE and AP-PCR,will show some minor, natural
variation.19,20Thus, when analyzing results, it is
important to consider the length of time over which
the bacterial isolates were collected.
428
INFECTIONCONTROLAND HOSPITALEPIDEMIOLOGY
Tune 1997
TABLE 1
CHARACTERISTICS
OFTYPINGSYSTEMS*
Typing
System
Proportionof
Strains
Typeable
All
Biotyping
Antimicrobial
susceptibilityAll
patterns
Most
Serotyping
Plasmidfingerprintingt Most
REAof chromosomal
DNA All
withconventional
electrophoresis
RFLPanalysiswith
All
DNAprobest
PFGE
All
All
AP-PCR.
Reproducibility
Discriminatory
Power
Ease of
Interpretation
Ease of
Performance
Poor
Good
Poor
Poor
Moderate
Easy
Easy
Easy
Good
Good
Good
Fair
Good
Good
Moderate
Moderate
Difficult
Moderate
Moderate
Moderate
Excellent
Moderate
to Excellent
Excellent
Good
Moderate
Difficult
Moderate
Moderate
Moderate
Moderate
Excellent
Good
Abbreviations:
REA,restrictionendonucleaseanalysis;RFLP,restrictionfragment-length
PFGE,pulsed-field
AP-PCR,
polymorphism;
gel electrophoresis;
arbitrarily
chainreaction.
primedpolymerase
* Adapted
fromreference6.
t ForisolatesofStaphylococcus
aureusandcoagulase-negative
Forthegram-negative
plasmidanalysiswithrestriction
staphylococci,
digestionoftenis required.
organisms,
suchas Klebsiella,
orSerratia,whole-plasmid
Enterobacter,
analysiswithoutrestriction
digestionoftenprovidessufficientdiscriminatory
power.
tAs appliedto IS6110analysisofMycobacterium
andribotyping.
tuberculosis
Notethatinterpretive
criteriaforAP-PCR
haveyet to be standardized.
difficile,andselectedEnterobacteriaceae.
. As appliedto S aureus,Clostridium
The discriminatory power of a technique refers
to its abilityto differentiateamong epidemiologically
unrelatedisolates, ideally assigning each to a different type. Traditionalphenotypic methods, such as
antibiogramtyping, serotyping, and biotyping, frequentlyshow lowerdiscriminatorypowerthannewer
molecularmethods.6,13
Ease of performancereflectsthe cost of specialized reagents and equipment,the technicalcomplexity of a method,andthe effort requiredto learnandto
implementthe techniquein the laboratory.Most molecularmethods requirepurchase of new equipment,
some of which is costly ($4,000-$20,000).However,
these methods are learned easily and are widely
applicableto a variety of species. Many traditional
methods also involveconsiderablecosts in laborand
materials,but are restrictedto a single or relatively
few species. For example, bacteriophage typing,
which is used primarilyfor S aureusand a few other
bacterialspecies, requiresthe maintenanceof bacteriophage stocks that constantlymust be replenished
and titered, a process that is both time-consuming
and labor-intensive.
Finally, ease of interpretationrefers to the
effort and experience requiredto obtainuseful, reliable typinginformationusing a particularmethod.At
present,the interpretationof the results of molecular
methods remains an area of active discussion.
However,this is in contrastto methods such as bac-
teriophagetyping and pyocin typing, which require
significant expertise to perform and interpret and
often still yield ambiguousresults.
INDICATOR
TECHNIQUES
(PHENOTYPIC
METHODS)
Biotyping
In the 1960s and early 1970s, identification of
bacterial species frequently was undertaken using
racks of tubes representing a variety of biochemical
tests, and the variability of certain tests, such as
indole, H2S,or pigmentproduction,served as markers for particular strains. Thus, biotyping emerged as
a useful tool for epidemiological investigations.
Today,identificationof bacterialspecies normallyis
accomplishedby using a combinationof biochemical
and immunologictests, many of which now are performed using commercialkits or automateddevices.
However,biotypingusing automatedmethods relies
on a variety of novel substrates,and some of these
tests, such as carbohydratefermentations,are highly
variableeven withinisolates of the same strain.Thus,
biotyping, like most phenotypic methods, has only
modest reproducibility,
because microorganismscan
alter unpredictablythe expression of many cellular
products.Moreover,contemporarybiotypingtypically
has poor discriminatory power and cannot differenti-
ate among some of the current nosocomialproblem
pathogens, such as enterococci, where biochemical
Vol.18 No.6
SHEA PosITION
PAPER
diversityis uncommon.Occasionally,outbreaksare
observedthatare causedby bacterialstrainsthatrepresent unusual species or unusual biotypes of common species, for example, H2S-producing
isolates of
Escherichiacoli. In such situations,additionaltyping
techniques may not be needed. However,even clusters of unusual isolates may not always indicate a
common-sourceoutbreak,as indicated by a recent
report in which four isolates of Leptotrichiabuccalis,
an unusual anaerobicgram-negativebacillus, recovered from blood culturesof four differentbone marrow transplantpatients,were found to be unrelated
The investiby PFGEandfatty-acidprofileanalysis.21
gation of the suspected outbreakrevealed that each
of the patientshad undergone dental manipulations
priorto developingbacteremia.In addition,all of the
patientshadbeen placedon prophylacticantimicrobial
agents to which the L buccalisisolateswere resistant.
Thus, each neutropenicpatientdevelopedbacteremia
with his own endogenous strainof L buccalis,which
served as an opportunisticpathogen.Nonetheless,it
shouldbe notedthatoutbreakscan, in some cases, be
causedby multiplepathogens.
429
the associationof certain Salmonellaserotypes with
foodbornedisease, and the associationbetween specific pneumococcalserotypes and invasive disease,
particularlyin children,serotypingcontinuesto be a
valuable typing technique. Nonetheless, PFGE has
been shown to resolve distinct clonal strains within
individualserotypes of both Salmonellaand pneumococci, thus indicatingthat it is a more discriminatory
typingtool.27,28
GENOTYPIC
METHODS
Overthe last severalyears, six moleculartechniques have emerged as the methods of choice for
typing bacterial isolates. They are PF; restriction
endonuclease analysis (REA)of plasmid DNA; REA
of chromosomal DNA using frequent cutting
enzymes and conventionalelectrophoresis;restriction fragment-lengthpolymorphism(RFLPtyping)
analysisusing DNA probes;PFGE;and AP-PCRand
other related nucleic acid amplification-based
typing
methods.In addition,PCR-DNAsequencingmethods
are just beginningto be utilized.
Plasmid Fingerprinting
Antimicrobial Susceptibility Patterns
Plasmid fingerprintingwas the first molecular
Antimicrobialsusceptibilitypatterns also have method to be used as a bacterial typing tool.29,30
relativelypoor discriminatorypower,because antimi- Plasmidsare extrachromosomalDNA elements that
crobial resistance is under tremendous selective are present in most clinicalisolates and can be idenandoften is asso- tifiedreadilyby simplecell lysis proceduresfollowed
pressurein healthcareinstitutions22
ciatedwith mobilegenetic elements (eg, transposons by agarosegel electrophoresisof the lysates (Figure
and plasmids).23Changes in antibiogramsalso may
1).31The numberand size of the plasmidspresent is
reflect spontaneous point mutations, such as seen
used as the basis for strainidentification.This strain
with fluoroquinolones.24
Thus, isolates that are epi- typing technique has been used successfully for
demiologically related and otherwise genetically analysis of outbreaksof nosocomialinfections29and
indistinguishablemay manifest different antimicro- community-acquired
infections32caused by a variety
bial susceptibilitiesdue to acquisitionof new genetic of species of gram-negativerods.
material over time25 or the loss of plasmids.26
Conversely, unrelated isolates may have indistin- REA of Plasmid DNA
Some strains of bacteriacontain only a single
guishable resistance profiles, which may represent
acquisitionof the same plasmidby multiplespecies
large plasmid, often in the size range of 100 to 150
kilobases (kb). Because it is difficultto differentiate
(a "plasmidoutbreak").
plasmidsin this size range,especiallythose thatvary
Serotyping
by only 10 kb to 15 kb, some investigators have
added a restriction endonuclease digestion step to
Serotyping, a nonmolecular method, uses a
series of antibodies to detect different antigenic try to increase the discriminatory power of agarose
determinantson the surface of the bacterial cell.
gel electrophoresis (Figure 2).33 While this can be
Serotypingis one of the classic strain typing tech- helpful, large plasmids produce many restriction
niques that has been used over the years for epi- fragments, which can make interpretation more difdemiologicalstudies of many species of bacteria.It ficult, especially when multiple large plasmids are
remains a key method for typing isolates of
present. Thus, for gram-negative rods, the REA step
Salmonella, Shigella, and pneumococci. However, no longer is performed in most laboratories.
maintaining stocks of typing sera (including the
However, for analysis of staphylococci, where the
>2,200antiserarequiredfor definitiveSalmonellatyp- plasmids typically are <50 kb, REA appears to
increase the discriminatory power of the analysis,
ing) is a majorlimitationof this method. Because of
INFECTIONCONTROLAND HOSPITALEPIDEMIOLOGY
430
A
B
CHROMOSOME
0
PLASMID1
PLASMID2
C
CLEAVEWITHBamHI
0
I
June 1997
5' G1
C 3'
GATe
SF
A
E
B
CHROMOSOMAL
FRAGMENTS -z
MOLECULAR
SIZE
STANDARDS
techFIGURE1. Schematicdrawingof the plasmidfingerprinting
niqueusingagarosegel electrophoresis.Theovalon the left signifies a typicalgram-negativerod,and the circleon the rightsignifies
a typicalgram-positive
coccus. Cellsare lysed usingdetergentsat
highpH,the chromosomalDNAis removed,and the plasmidDNA
is appliedto an agarosegel thatthen is stainedand photographed.
Some chromosomalfragmentsusuallyare visibleon the gel and
12-15
serve as an internalmolecularsize standard(approximately
kb).PlasmidDNAwithinthe bacteriais shownin circularform.Each
plasmidis numberedwithinthe cell, and its corresponding
position
in the agarosegel is indicatedbythe same number.
because the number of restrictionfragmentsgenerated usually is <20.33Digestion also makes the patterns of the restriction fragments produced from
staphylococcalplasmids easier to analyze than the
undigested profiles, which often show multiple
forms for plasmidsof less than 15 kb, because circular and linearforms of the plasmidmigrateat different rates than the covalently closed circularform.
Plasmid fingerprintingis technically simple to perform and requires relativelyinexpensive equipment
($1,500-$3,000).At this time, the method is used primarilyas an alternativetechniquefor staphylococcal
isolates, which frequently carry multiple plasmids,
and for selected species of Enterobacteriaceae,
which
often have large distinctiveplasmids.
When applyingthe plasmidfingerprintingtechnique, investigatorsmust be awareof two confounding factors. First, it is possible that plasmids can
spread to multiple species of bacteria, causing a
plasmid outbreak in which unusual antibiograms
are recognized in multiple species. This has been
recognizedboth in gram-negativerods and in staphylococci.25,34
Second, it is importantto appreciatethat
the structureof individualplasmidsand the plasmid
contentof a particularstrainmayvary overtime.This
variabilityreflects two factors: over time, plasmids
can be lost spontaneously or acquired from other
organisms, and plasmids frequently carry smaller
1
-HMOLECULAR SIZE
STANDARDS
2
FIGURE2. Schematicdrawingof restrictionendonucleasedigestion of two uniqueplasmids,followedby agarosegel electrophoresis. Different-sized
restrictionfragmentsare denoted by the different patterns.
mobile genetic elements (transposonsand insertion
sequences) that promote duplicationsand deletions
of DNA segments. Both plasmids and transposons
often include antimicrobialresistance determinants
and thus are subject to considerableselective pressure within hospitals due to antimicrobialagent
use.22In general, plasmidfingerprintingis most useful for epidemiologicalstudies that are limited both
temporallyand geographically.In selected instances,
plasmid fingerprintingmay complementother techniques, such as PFGEanalysis,by providinga basis
for differentiatingisolates that are related genotypically but are separatedepidemiologicallyby moderate time periods,such as severalmonths.35
Gel Electrophoresis Techniquesfor Analysis of
Chromosomal DNA
There are two methods of typing organisms
based on fragment patterns produced by cleaving
chromosomal DNA with restriction endonucleases.
The first method, often referred to as conventional
electrophoresis,uses a restrictionenzyme that cuts
the chromosome into hundreds of pieces (frequent
cutter), followed by standard agarose gel electrophoresis. Fragmentsthat are 25 kb to 0.5 kb are
resolved into a discernible banding pattern,
although a single band may contain fragments of
similarsize from several differentareas of the chromosome. Largerfragmentscoalesce at the top of the
gel or do not migrate into the gel. The second
method, PFGE,uses enzymes that cut chromosomal
DNA infrequently,generating from 10 to 30 bands,
Vol. 18 No. 6
SHEA PosITIONPAPER
2
z
A
OI
Sz
431
B
A
C
B
C
MEMBRANE TRANSFER
AND
PROBE HYBRIDIZATION
z
AGAROSEGEL ELECTROPHORESIS
BACTERIALCHROMOSOME
(with frequent restriction sites)
FIGURE
3. Schematicdrawing
of restriction
endonucleaseanalysisof
chromosomalDNAusing conventionalelectrophoresis.The box on
the rightrepresentsthe bandingpattemof hundredsof fragments
afterconventional
Eachbandmayconagarosegel electrophoresis.
taina numberof uniquechromosomal
fragmentsof similarsize.
followedby a novel form of electrophoresisthat can
separate fragments from 1 kb up to 1,000 kb (1
megabase).36Each method, and a variationof the
conventionalelectrophoresismethod,is describedin
greater detailbelow.
REAof ChromosomalDNA WithFrequent
Cutting Enzymes and Conventional
Electrophoresis
Eachrestrictionendonucleasecleaves DNAat a
particularsequenceof nucleotidesthatmaybe repeated numerous times around the chromosome. The
numberandsize of the restrictionfragmentsgenerated
by digesting a given piece of DNA reflects the frequencyand distributionof the restrictionsites. In conventionalREA,endonucleaseswith frequentlyoccurring sites in the bacterialgenome are used to digest
total DNA (plasmidand chromosome),thereby generatinghundredsof fragmentsrangingfromapproximately 0.5 to 50 kb in length (Figure 3). Such fragments can be separatedby size using agarose gel
electrophoresis,and the pattern can be detected by
stainingthe gel withethidiumbromide(orotherdyes)
and photographingunder ultravioletlight. Different
strains of the same bacterialspecies have different
REAprofiles(depictedas a series of bandson agarose
gels) because of variationsin their DNA sequences.
All isolates are typeableby REA;however,it can be
very difficultto interpretthe complexprofiles,which
consist of hundredsof bandsthatmaybe indistinctor
overlapping.Althoughthe approachhas been applied
to manyspecies, at this time, its primaryuse is as an
alternativetechniquefor analyzingC difficile.37
RFLPAnalysis Using DNA Probes
In this technique, chromosomal restriction
digestsproducedby frequentcuttingenzymesare sep-
(CHROMOSOMAL RESTRICTION FRAGMENTS)
FIGURE
4. Schematicdrawingof restrictionfragment-length
polymorphismanalysisusinga DNAor RNAprobe,such as IS6110 or
ribosomalRNA.The box on the left representsagarose gel electrophoresisof chromosomalDNAcleavedwitha restrictionendonuclease, and the boxon the rightrepresentsthe nylonfilterto which
the DNAhas been attachedand hybridized
with a specific probe.
Onlythe DNAfragmentson the nylonfilterthat bindthe probecan
be visualized.The organismsrepresentedin lanes A and B are
whereasthe isolaterepresentedin laneC is a difindistinguishable,
ferentstrain.
aratedby conventionalagarosegel electrophoresis,as
described above, and then the DNA fragments are
transferredonto a nitrocelluloseor nylon membrane
(Figure 4).8 The DNA on the membrane then is
hybridizedwith a specific chemicallyor radioactively labeled piece of DNA or RNA (a probe), which
binds to the relativelyfew fragments on the membrane that have complementary nucleic acid
sequences. Variationsin the numberand size of the
fragmentsdetected by hybridizationare referred to
as RFLPs.
One common typing method that uses chromosomal DNA preparationsand a ribosomal RNA
probe is ribotyping.39Because all bacterial isolates
have one or more chromosomalrRNAoperons distributed around the chromosome, and because
those sequences are highly conserved, essentially
all bacterial isolates can be typed using probes
directed to the DNA sequences that encode the
rRNA loci using a single rRNA probe.39However,
enthusiasmfor this system has diminished,because
the approachhas proven to be only moderatelydiscriminatory.4042
Restriction fragment-length polymorphism
analysis using the DNA insertion element IS6110
currentlyis the method of choice for typing isolates
of Mycobacterium
An insertionelement
tuberculosis.43
is a piece of DNAwith a definedstructurethat is able
to move independentlyand to insert in multiplelocations in plasmidsor chromosomallocations,but does
not containantimicrobialresistance genes or genes
involvedin pathogenesis.IS6110 is present in essenand,because insertiallyall isolates of M tuberculosis,
tion sequences are mobile,the numberand chromo-
INFECTIONCONTROLAND HOSPITALEPIDEMIOLOGY
432
DIFFERENT ISOLATES z
O
12 34
56
BACTERIALCHROMOSOME
(with rare restriction sites)
z
S0
PFGE ELECTROPHORESIS
CHAMBER
PFGE ANALYSIS
FIGURE
5. Schematicdrawingof pulsed-fieldgel electrophoresis,
in whichchromosomalDNAis cleaved witha rarecuttingenzyme
followedby electrophoresis,using a uniquechamberand current
switchingprotocol.The boxon the rightis the agarosegel showing
the very large DNA fragments derived from the unique electrophoresischamber.
somal locations of the insertions vary greatly from
strainto strain.The approachhas provenreliableand
discriminatoryand, in most cases, can distinguish
successfullysporadicand clusteredcases of tuberculosis.44However, for isolates with fewer than five
copies of IS6110, the method has relativelypoor discriminatorypower and must be supplementedby
studies using other probes.
PFGE
Pulsed-field gel electrophoresis was first
described in 1984 as a tool for examiningthe chromosomalDNA of eukaryoticorganisms (Figure5).36
Subsequently,PFGEhas provento be a highly effective moleculartyping technique for many different
bacterialspecies.6,13,14,19
In this method,the bacterial
genome, which typicallyis 2,000 to 5,000 kb pairs in
size, is digested with a restrictionenzyme that has
relativelyfew recognition sites and thus generates
approximately10 to 30 restrictionfragmentsranging
from 10 to 800 kb. Essentiallyall of these fragments
can be resolved as a pattern of distinct bands by
PFGE,using a speciallydesigned chamberthat positions the agarose gel between three sets of electrodes thatform a hexagon aroundthe gel. Insteadof
applying an electric current to the gel in a single
direction,as is done in conventionalelectrophoresis,
in PFGE,the currentis appliedfirst in one direction
from one set of electrodes, then shifts to the second
set of electrodes for a short period of time (a pulse),
and then shifts to the third set of electrodes. Thus,
the electric field that causes the DNA to migratein
the gel is providedin pulses that alternatefromthree
sets of electrodes. This causes the DNA to wiggle
through the gel, and the back-and-forthmovement
June 1997
results in the higherlevel of fragmentresolutionseen
with the technique.All species are typeableby PFGE,
althoughthe isolationof intactchromosomalDNA is
technicallydifficultfor some species. As noted above,
the chromosomalDNA of some strains of C difficile
spontaneouslydegrades during the cell lysis procedure, making this typing approach impractical.'8
PFGEhas been appliedsuccessfullyto a widerangeof
bacterialspecies,both gram-positive(eg, staphylococci, enterococci,and mycobacteria)and gram-negative
and pseudomon(eg, E coli, otherEnterobacteriaceae,
In
PFGE
is
one
of
the most reproads)."
general,
ducible and highly discriminatorytypingtechniques
availableandis currentlythe typingmethodof choice
for many species. The major difficulties associated
with PFGErelateto the technicaldemandsof the procedure and initialcost of the equipment.Preparation
of suitable genomic DNA requires 1 to 3 days,
depending on the organisms tested, and the equipment required(includingthe electrophoresisapparatus and transilluminator)costs between $10,000and
$20,000.However,once the method is operationalin
a laboratory,it can be appliedreadilyto a wide range
of species with onlyminimalmodifications.The interpretationof PFGEgels is relativelystraightforward,
and consensus guidelinesfor correlatingvariationsin
restrictionprofileswith epidemiologicalrelatedness
were publishedrecently."
Typing Methods Using PCR
Polymerase chain reaction, which has been
used for several years for the direct detection of
many types of infectiousagents in clinicalsamples,45
has been adaptedfor use as a typingtool.4648
The hallmarkof PCRis the abilityto produceliterallymillions
of copies of a particularDNA segment with high
fidelity within 3 to 4 hours' time. The procedure
requires template DNA (or RNA if a reverse transcriptasestep is used initially),which maybe present
in the sample in minute quantities; two oligonucleotide primers,which flank the sequences on the
template DNA to be amplified (thus defining the
starting points for DNA polymeraseactivity);and a
heat-stableDNA polymerase.Efficient amplification
is accomplished readily for templates of less than
2,000base pairs,althoughtemplatesas large as 35 kb
now can be amplifiedby using newer polymerases.A
typicalPCRassay requires approximately3 hours to
complete 30 cycles, where each cycle consists of a
heat denaturationphase, in which double-stranded
DNA is melted into single strands; an annealing
phase, in which the primers bind to the target
sequences on the single strands; and an extension
phase, in which DNA synthesis proceeds from the
Vol.18 No. 6
433
SHEA PosITIoN PAPER
primers along each strand of the template DNA,
thereby generatingtwo new double-strandedcopies
of the originaltemplate.After30 such cycles, a single
initial copy of template DNA theoretically can be
amplifiedto 1 billioncopies.
Arbitrarily Primed PCR
ArbitrarilyprimedPCR,also referredto as the
randomlyamplifiedpolymorphicDNA assay,is a variationof the PCRtechniqueemployinga single short
(typically10 base pairs)primerthat is not targetedto
amplify any specific bacterial DNA sequence.46
Rather,at low annealing temperatures,the primer
will hybridizeat multiplerandomchromosomallocations and initiateDNA synthesis. If one copy of the
primerbinds to one strandof DNA, and anothercopy
of the primerbinds on the opposite strand of DNA
but in proximityof the first primer,a DNA fragment
will be synthesizedand amplificationof thatfragment
will occur (Figure 6).46The resulting PCRproducts
will represent a variety of different-sizedDNA fragments that are visualized by agarose gel electrophoresis.This approachhas remarkablegeneral
applicabilityand has been appliedto typing eukaryotic species, as well as manybacterialspecies.48
and discriminatorypowerof
The reproducibility
AP-PCRis a subjectof activediscussionandinvestigation. Recent reports have described specific primers
and
andconditionsfor analyzingisolatesof C difficile49
S aureus,50and have providedformalassessments of
the procedure.These studies have indicated that,
althoughAP-PCRis appreciablyfasterthan othertyping systems (often 20 to 30 organismscan be completedin a single day),the methodis much more susceptible to technical variationthan is routine PCR
employingprimersdirected at known sequences. In
AP-PCRthere can be substantialvariationsin the efficiencywith which the primersinitiateDNA synthesis
at a particularsite, dependingon even slightvariations
in the pH or ionic strength of the buffers used, the
temperatureof the reaction,or the source of the DNA
Consequently, on independent
polymerase.48,50,51
of the same strain,the numberof copies
amplifications
generated from a particular locus can be appreciably
different,yieldingwide variationsin the intensitiesof
individualfragments.These factorscan make it difficult to obtain reproducible patterns and interpretations with AP-PCR, particularly when attempting to
compare isolates tested on different days. At this time,
the most reliable results are obtained when a set of
isolates is tested in a single amplification reaction and
analyzed on a single electrophoretic gel. The comparison of AP-PCRfragment patterns obtained by testing
the same isolates in different laboratories often is
z
PRIMER BINDING
CHROMOSOMEt
*
AGAROSE GEL
ELECTROPHORESIS
FIGURE6. Schematic drawingof arbitrarily
primedpolymerase
chainreaction(AP-PCR)
technique.Theboxon the rightrepresents
an agarose gel showingthe productsof amplificationusing PCR
with a randomprimer.Whenthe primerbinds in two places on
to one another,a fragmentis
oppositestrandsof DNAin proximity
generatedthat can be recognizedby agarosegel electrophoresis.
problematic,as was shown recentlyin a multicenter
study.- Furthermore,standardguidelines for interpretationof AP-PCRare notyet available,andthe generalprinciplesthatare emergingforthe interpretation
of othermoleculartypingdata,such as for PFGE,cannot be appliedreadily.
PCR-DNASequencing
Although infrequentlyused at this time, DNA
sequencing of hypervariablegene sequences has
While this techshown promise as a typing tool.52,53
most
for
not
feasible
is
laboratories,
nique currently
the availabilityof less expensive DNA sequencing
machines may make this techniquemore accessible
in the future.
WHEN
TO USE STRAIN TYPING
Bacterialstrain typing data are most effective
when they are collected, analyzed, and integrated
into the results of an epidemiologicalinvestigation.
The preferredtypingmethodsfor variousorganisms
are presentedin Table 2. The hospitalepidemiologist
should initiate strain typing studies in consultation
with the hospital infectioncontrol laboratoryor the
hospital microbiology staff when investigating a
potential outbreak of an infectious disease.73This
may be triggeredby a noticeableincreasein the rate
of isolationof a particularpathogen,a clusterof infections on a particularward, or the recognitionin the
clinical microbiologylaboratoryof multipleisolates
with an unusualbiotypeor antibiogram.Straintyping
data should supplement,and not replace,a carefully
conducted epidemiological investigation. In some
cases, typingdatacan effectivelyrule out an outbreak
and thus avoid the need for an extensive epidemiologicalinvestigation.In othercases, straintypingdata
may reveal the presence of outbreaks caused by
434
INFECTIONCONTROLAND HOSPITALEPIDEMIOLOGY
June 1997
TABLE 2
ANDCOMMUNITY-ACQUIRED
PREFERRED
STRAINTYPINGTECHNIQUES
PATHOGENS
FORNOSOCOMIAL
Species
Reference Method*
aureus
Staphylococcus
Coagulase-negative
staphylococci
Streptococcus
pneumoniae
Enterococci
Escherichia
coli,tCitrobacter,
Proteus,Providencia
Serratia
Klebsiella,
Enterobacter,
Salmonella,Shigella
Pseudomonas
aeruginosa
Acinetobacter
Burkholderia,
Stenotrophomonas,
Clostridium
difficile
tuberculosis
Mycobacterium
otherthantuberculosis
Mycobacteria
PFGE
PFGE
PFGE
PFGE
PFGE
PFGE
Serotyping
PFGE
PFGE
AP-PCR
IS6110RFLP
PFGE
Alternative Methods
References
PF
AP-PCR,
PF
Serotyping
AP-PCR
PFt
PFGE
REA,PFGEs
REP-PCR
-
7, 11, 17,33, 35, 50, 54
11,26, 55
28, 56, 57
8, 42, 55, 58
11, 19,20, 32, 59
10, 11,29
11,27, 30,60, 61
9, 11,62-66
11,67, 68
18,37, 41, 49
43, 44, 69, 70
11,71, 72
Abbreviations:PFGE,pulsed-fieldgel electrophoresis;AP-PCR,arbitrarilyprimedpolymerasechain reaction;PF,plasmidfingerprinting(withor withoutrestrictionanalysis); REA,restriction endonuclease analysis of chromosomal DNA using conventionalelectrophoresis;RFLP,restriction fragment-lengthpolymorphismtyping using
IS6110; REP-PCR,repetitive-elementpolymerasechain reaction.
* Recommendationsare based on
published studies and consensus opinions of experts.
tE coli 0157:H7 must be identifiedby serotyping.
For the gram-negativeorganisms,whole-plasmidanalysiswithoutrestrictiondigests often is sufficient.
. Many strains of C difficileare nontypeableby PFGEdue to DNA degradation.
more than one strain. However,undue reliance on
straintypingin the absence of epidemiologicaldatais
an inefficientuse of laboratoryresources.11,73
SETS OF ISOLATES MOST
FOR TYPING
APPROPRIATE
STUDIES
Sets of epidemiologicallyrelatedisolates representing a putativeoutbreakspanninga periodof 1 to
3 months are appropriateorganismsfor straintyping
studies. This implies that the infectioncontrol staff
and hospitalepidemiologistalreadyhave made a preliminaryassessment and have determinedthe need
for strain typing data. One should not test sporadic
isolates, except as epidemiologicallyunrelatedcontrols (see below). The results of testing sporadicisolates oftenare misleadingandfrequentlylead to wasted resources in both the microbiologylaboratoryand
the infectioncontrolservice. Most of the techniques
describedabove can be appliedeffectivelyto 10 to 30
isolates per day,especiallyin dedicatedinfectioncontrol laboratories.
Control Strains
In attemptingto detect single-strainoutbreaks
involvingspecies that may be endemic, eg, MRSAor
multiplyresistant gram-negativebacilli, it is important to include epidemiologicallyunrelatedisolates
or historical controls to help differentiateendemic
problemsfrom new outbreaks.Outbreakscaused by
more than one organism (which occur with consid-
erablylower frequencythan those caused by single
strains) are much more complex to analyze and
requirelarge numbersof additionalcontrolstrainsin
order to recognize the outbreak strains. Although
endemic strainsmay cause clusters of infections,the
inclusion of historical controls from previous outbreakscanhelp determinewhetherthe recentisolates
are likelyto representa new outbreak.Additionalcontrols includethe testing of well-characterizedstrains,
such as S aureus NCTC 8325, E coli MG1655,and
Enterococcusfaecalis OG1RF,11
to ensure that each
step of the typingmethodis workingproperly.
INTERPRETATION
Interpretationof straintypingresults is facilitated greatly by an appreciationof the molecularbasis
of genetic variabilityof bacteriaand the technicalfactors that can affectresults.Three assumptionsusually are made:isolatesrepresentingthe outbreakstrain
are the recent progenyof a single (or common) precursor; such isolates will have the same genotype;
and, epidemiologicallyunrelatedstrainswill have different genotypes. Ideally,straintypingwill providea
clear, objective basis for identifying the outbreak
strain and distinguishingit from epidemiologically
unrelatedisolates.
In practice,the interpretationof typing data is
complicatedby the fact that isolates from an ongoing
outbreakmay demonstratesome, albeittypicallylimitThe purposeof interpretivecried, genetic variability.
SHEA PosITIONPAPER
Vol.18 No. 6
435
TABLE 3
OF MOLECULARTYPINGANALYSES*
GENERALPRINCIPLESFOR THE INTERPRETATION
Microblologic
Interpretation
Based on
TypingResults
Indistinguishable
Closely related
Possibly related
Different
No. of Genetic
Differences
ComparedWith
OutbreakStrain
TypicalNo. of
FragmentDifferences
Comparedto
OutbreakPattern
0
1
2
3
0
2-3
4-6
>7
Epidemiological
Correlation
Isolate is part of the outbreak
Isolate probablyis part of the outbreak
Isolate possibly is part of the outbreak
Isolate is not part of the outbreak
fromreference11.
*Adapted
teriais to establisha guide for distinguishingtrue differences in strainsfrom the naturalgenetic variation
that occurs over time withina given strain.For illustration,assumethata set of up to 20 putativeoutbreak
isolates has been typed and that the analysis has
detecteda subset of isolateswith a common (modal)
type, which is presumed to represent the outbreak
strain.Typically,among the other isolates in the set,
some have similartypes (as represented,for example,
by a few bandchanges in a PFGEpattern),and some
are distinctlydifferenttypes (distinctivePFGE patterns). The interpretativecriteriashould provideconsistent,objectiveguidelinesfor correlatingthe level of
variationobserved between an individualisolate and
the putativeoutbreakstrainwith an estimate of the
likelihoodthat the isolate is, in fact, part of the outbreak (Table 3). To provide a generally applicable
approach,this correlationfocuses on the numberof
genetic events requiredto generatethe observedtyping variation,rather than on the types of specific
changesobservedin a particulartypingsystem.In the
examplecited,there is a groupof isolatesthatproduce
identicaltyping patterns,ie, the presumed outbreak
strain.Because only a smallportionof the organisms'
genetic complementis undergoinganalysis,isolates
that give identicalresults are classified as "indistinA more detailed analysis
guishable,"not "identical."
theoreticallycould uncoverdifferencesin the isolates
that appearedto give identicalpatternsbut that were
epidemiologicallyunrelated.However,when a set of
epidemiologicallylinked isolates are analyzed,this is
unlikelyto occur.
ClassifyingMinor TypingPattern Variations
The next considerationis how to classifythose
isolateswithminortypingpatternvariations.A number
of studiesusing PFGEand othertypingmethodsindicate that single genetic events (ie, randommutations
that may destroyor create a new restrictionendonuclease site, or deletionsor insertionsof new DNAsuch
as plasmids,bacteriophages,or insertionsequences)
occur unpredictablyeven within the time span of a
well-definedoutbreak(1-3months).Therefore,interpretivecriterianeed to accommodatesuch naturalvariation.The criteriafor interpretingstraintypingresults
are given in Table 3, with their epidemiologicalcorollaries. The changes are depictedin cartoonform in
Figure 7. When differencesare encounteredin the
straintypingresults,it is importantto attemptto ascertainwhattype of genetic eventmayhave led to the differences. For example,if a single band increases in
size, could an insertionof new DNA, such as a bacteriophageor transposon,accountforthe changein size?
Woulda randommutationin a restrictionendonuclease site account for the appearanceof two smaller
bandswhere a single largerbandhad been seen with
other epidemiologicallyrelated isolates?Such questions oftencan be answeredby carefulanalysisof the
patterns.If two genetic events have occurredand are
recognizedthrough the differencesin fragmentpatterns, the epidemiologicalinterpretationfalls into a
gray zone. The results may indicatethat the isolates
arerelated,especiallyif theywere collectedovera long
periodof time (3-6months),but there also is a possibilitythat the strainsare unrelated,and the similarity
was a resultof chance.In such cases, additionalinformation, such as the use of a second strain typing
method or supplementaryepidemiologicalanalysis,
should be sought. This often will help to clarifythe
relationshipamongthe isolatesexamined.Finally,isolates thatdifferby morethanthreegeneticeventsrepresent geneticallydifferentstrainsand shouldbe consideredunrelated.These generalprincipleshavebeen
applieddirectlyto the developmentof criteriafor interpretingPFGE(Table3; also see reference11).
Interpretation of AP-PCR,PF, and RFLP
Techniques
At this time, the interpretationof results generated by AP-PCR,plasmidfingerprinting,and the vari-
436
INFECTIONCONTROLAND HOSPITALEPIDEMIOLOGY
Changes to
400kb fragment
None
Gain of
Lose of
reetrlction site
400
250&150
Resulting
fragment(s)
A
Lane
Deletion
Insertlon
of 50 kb region
600
C
B
450
350
D
E
kb
600
*
500
PFGE
gel
400
200
)
)
..
"
)
,,
100
75
No. of
differences
0
3
3
2
2
FIGURE
7. Schematicdiagramshowingthe theoreticalchanges in
pulsed-fieldgel electrophoresisfragmentpatternsof an isolate as
a resultof variousgenetic events. LaneA, outbreakpattern;Lane
B, gain of a restrictionsite; LaneC, loss of a restrictionsite; Lane
D, insertionof DNAin an existingfragment;LaneE,deletionof DNA
froman existingfragment.Opencircle(o) indicatesfragmentspresent in outbreakpattemand missingin the test isolateaftergenetic event;asterisk(*) indicatesfragmentpresentaftergeneticevent
but absent fromoutbreakDattern.
ous nucleic acid probe-basedRFLPmethods, including IS6110, is empiric.Forthese techniques,the gray
zone between "indistinguishable"(identical typing
results) and "unrelated"(clearly different results)
often is large.51In part, this reflects the difficultyof
definingthe effects of single genetic events in some
of these typingsystems.
In DNA hybridization-based RFLP studies,
such as those using IS6110, two basic mechanisms
may alterthe patternsobserved. The simplest is the
generation or loss of a restriction site flankingthe
insertion sequence (IS) element and a subsequent
change in the size of the restrictionfragmentcarrying that element. Such changes often can be defined
by performing DNA hybridizationusing a second
DNA probe or by testing DNA cleaved with a different restrictionendonuclease.A more complex basis
for changes in the RFLPpatternis the loss, gain, or
movementof an entire IS element.The initialstudies
applyingIS6110 typingassumed that patternsdiffering by even a single band represented genetically
and epidemiologically different strains. However,
more recent data indicates that the IS6110 patterns
of isolates representinga definitecluster of M tuberculosiscases can differby single genetic events.
Interpretation of PF Results
When isolates from a suspected outbreakhave
three or more plasmidsin common (excludingopen
circularandlinearforms,which sometimesappearas
June 1997
additionallight bands for plasmids under 15 kb in
size31),as often is seen with strains of coagulasenegative staphylococci,K pneumoniae, and other
gram-negativebacilli,one usuallycan say with confidence that they are indistinguishable(genetic interpretation)and thereforelikely to be epidemiologically related (epidemiologicalinterpretation)without
furthertesting. With fewer plasmids,the discriminatory power of the test is decreased. Yet, experience
with this technique in the 1970s and 1980s, particularly when used to analyze suspected outbreaks of
gram-negativeorganisms causing nosocomial infections, was that, even when only a single plasmidwas
present, the techniquewas a very good indicatorof
strain identity.The problem that was recognized in
the 1980s was that of the plasmid outbreak,where
the same plasmid could be identified in multiple
strains of a species that could be differentiatedwith
other typing techniques.To a degree, with plasmids
smallerthan 50 kb, the discriminatorypowerof plasmid fingerprintingcan be increased with restriction
analysis of the plasmid DNA,36which often yields
fragment patterns that approach the numbers of
bands seen with PFGE, but this does not identify
potential plasmid outbreaks. Many enteric organisms, such as Serratiamarcescens,may have only a
single plasmid,andthese plasmidsoften are relatively conserved, showing similar fragment patterns
after restrictionendonucleasedigestion even in epidemiologicallyunrelatedstrains. If the plasmid fingerprintand restrictionendonucleasedata do not fit
the epidemiologicalpicture,it may be appropriateto
seek datageneratedby using a differenttypingtechnique for analysisof such isolates.
Interpretation of AP-PCRResults
ForAP-PCR,interpretation
of identicalfragment
patternsand of patternswith three or more fragment
differences (there are no restrictionenzyme sites to
consider in AP-PCR)is straightforward.However,
there are no hard-and-fast
criteriafor interpretinga
change in the size of a single band or the intensityof
several bands. For strains yielding results such as
these, it may be necessary to try alternate primers or
to vary the reactionconditions.There is no "universal
primer,"althoughprimerssuch as ERIC1,ERIC2,and
repetitiveelementprimershavebeen used to type several different bacteria and fungi.51,59,74
The reproducibility and discriminatory power of each primer
and amplificationprotocol need to be validated by analyzing sets of isolates that previously have been well
defined by epidemiological data or independent typing
studies. Such analyses are available for relatively few
species (eg, S aureus and C difficile). Because varia-
Vol.18 No.6
SHEA PosITIoNPAPER
tion in AP-PCRcannotbe tightlycoupledwith specific
genetic events, the principlesdefined previouslyfor
PFGEcannotbe readilyappliedto typicalAP-PCRpatterns.The developmentof generallyapplicablecriteria
for interpretationof AP-PCRdata remainsan area of
activeinvestigation.51
Nevertheless,a recentmulticenter study5.demonstratedthat, althoughparticipating
laboratoriesobtaineddifferentAP-PCRproducts,the
same epidemiologicalclusterswere identifiedsatisfactorily.Thus,wherevariabilityof fragmentsizes can be
demonstratedamongepidemiologicallyunrelatedisolates, those showingeitherno differencesor changes
only in bandintensitycan be consideredepidemiologicallyrelated.Changesin two bandsremaindifficultto
interpret.
A COMPARISON
437
A
B
1 23 4 56
7
C
1
2
3
p
L
4m.
5
6
7
1234567
]E
Mbd
40
an
[]
4
AID am
]
dWE
do
,
OF TECHNIQUES
Figure 8 shows the typing results obtainedby
using three common methods (PF, PFGE, and APPCR) to analyze six well-characterizedisolates of S
aureus, for which epidemiologicalinformationwas
available."As previouslyreportedby the CDC,17four
of the isolates (SB02, SB04, SB06, and SB11) were
associated with administrationof a contaminated
anesthetic. The first three isolates were recovered
fromthe same patient,andthe fourthfroma different
patient.IsolatesSB13andSB17were unrelatedto this
cluster,but manifestedthe same bacteriophagetype.
As shown in Figure 8, all three methods correctly
identifiedSB02, SB04, and SB06 (lanes, 2, 3, and 4,
respectively) as multipleisolates of a single strain.
However,only PFGE demonstratedthat SB11 (lane
B5) was a strain subtype related to the outbreak,
while SB13 and SB17 (lanes B6 and B7, respectively)
appearedsufficientlydifferentto representunrelated
strains.Conversely,plasmidanalysisdid not associate
isolate SB11 (laneA5) with the outbreakand did not
differentiatebetween SB11 and SB13 (lanes A5 and
A6, respectively). Arbitrarilyprimed-PCRalso was
unableto distinguishSB11 from SB13 (lanes C5 and
C6, respectively).This emphasizesthe importanceof
interpretingthe typing data in conjunctionwith epidemiologicaldata.
At this pointin time, we recommendthat PFGE
be consideredthe reference method for typingmost
nosocomialbacterialpathogens (Table2). Important
exceptions to this general recommendation are
serotypingfor Salmonellaand Shigella,AP-PCRfor C
difficile,and IS6110 RFLPtyping for M tuberculosis.
Alternatetechniquescan be used when the reference
technique is not availableor when further discriminationof a set of isolates is desirable.
DNA-based strain typing techniques are a
remarkablyuseful set of tools for complementingthe
FIGURE
8. Interrelationship
of StaphylococcusaureusstrainsSB02
(lane2), SB04 (lane3), SB06 (lane4), SB11 (lane5), SB13 (lane
6), and SB17 (lane 7). The isolates were comparedby: (A)EcoRI
patternsof plasmidDNAwiththe 1-kb ladder
restriction-fragment
LifeTechnologies,GrandIsland,NY)plus EcoRI-digest(Gibco-BRL
ed lambdaDNAas a molecularsize standard(lane 1); (B) SmaldigestedchromosomalDNAanalyzedby PFGEwith lambda(48.5kb) oligomersas a size standard(lane 1); and (C)AP-PCR
typing
3' witha Hindllldigest
utilizingthe ARB11primer5' CTAGGACCGC
of lambdaDNAas a molecularsize standard(lane 1).
epidemiological analysis of nosocomial outbreaks.
This article and the consensus guidelines recently
published by the European Study Group on
EpidemiologicalMarkers75
highlightthe strengthsof
molecular typing methods. Pulsed-field gel electrophoresis, in particular,has emerged as a technique that is broadlyapplicableto most nosocomial
pathogens. Interpretivecriteria for this technique
have been published.ArbitrarilyprimedPCRhas the
potential to provide a rapid, alternative typing
method,but has not been validatedfor manyspecies,
and standardizedinterpretativecriteria are not yet
available.Independentof the typing method used,
the applicationof strain typing to make decisions
affectingpatientcare must be made with an appreciation of the strengths and limitationsof molecular
methods and only in conjunctionwith the results of
an appropriateepidemiological investigation. The
results of straintyping studies producedin a laboratory withoutthe correspondingepidemiologicaldata
may be misleading.
REFERENCES
1. EmoriTG, GaynesRP.An overviewof nosocomialinfections,
including the role of the microbiology laboratory. Clin
MicrobiolRev 1993;6:428-442.
2. OliveKE,BerkSL.Infectionsin the nursinghome. ClinGeriatr
438
INFECTIONCONTROLAND HOSPITALEPIDEMIOLOGY
Med 1992;8:821-834.
3. MakiDG, RhameFS, MackelDC, BennettJV.Nationwideepidemicof septicemiacausedby contaminatedintravenousproducts, I: epidemiologic and clinical features. Am J Med
1976;60:471-485.
4. Tenover FC, Hughes JM. The challenges of emerginginfectious diseases: developmentand spread of multiplyresistant
bacterialpathogens.JAMA1996;275:300-304.
5. Hospital Infection Control Practices Advisory Committee,
Centersfor Disease Controland Prevention.Recommendations
for preventingthe spread of vancomycinresistance. Infect
ControlHospEpidemiol1995;16:105-113.
6. ArbeitRD.Laboratoryproceduresfor the epidemiologicanalysis of microorganisms.In: MurrayPR,BaronEJ,PfallerMA,
TenoverFC,YolkenRH,eds. Manualof ClinicalMicrobiology.
6th ed. Washington,DC: AmericanSociety for Microbiology;
1995:190-208.
7. El-Ahami
W,RobertsL,VickeryA, et al. Epidemiological
analysis of a methicillin-resistant
aureus outbreak
Staphylococcus
using restrictionfragmentlength polymorphismsof genomic
DNA.J GenMicrobiol1991;137:2713-2720.
8. MontecalvoMA, HorowitzH, Gedris C, et al. Outbreakof
vancomycin-, ampicillin-, and aminoglycoside-resistant
Enterococcus
faecium bacteremiain an adult oncology unit.
Antimicrob
1994;38:1363-1367.
AgentsChemother
9. SaderHS,PignatariAC,LemeIL,et al. Epidemiologictypingof
multiplydrug-resistantPseudomonasaeruginosaisolatedfrom
an outbreakin an intensivecareunit.DiagnMicrobiolInfectDis
1993;17:13-18.
10. GoubyA, NeuwirthC, BourgG, et al. Epidemiological
studyby
pulsed-fieldgel electrophoresisof an outbreakof extendedKlebsiellapneumoniaein a
spectrum P-lactamase-producing
geriatrichospital.J ClinMicrobiol1994;32:301-305.
11.TenoverFC, ArbeitRD, GoeringRV,et al. Interpretingchromosomal DNA restrictionpatterns producedby pulsed-field
gel electrophoresis:criteriafor bacterialstraintyping.J Clin
Microbiol1995;33:2233-2239.
12.Aber RC, Macke DC. Epidemiologictyping of nosocomial
AmJ Med 1981;70:899-905.
microorganisms.
13.Swaminathan
B, MatarGM.Moleculartypingmethods:definition, applications,and advantages.In: Persing DH, SmithTF,
TenoverFC,WhiteTJ,eds. DiagnosticMolecularMicrobiology:
Washington,DC:AmericanSociety
PrinciplesandApplications.
for Microbiology;1993:26-50.
14. GoeringRV.Molecularepidemiologyof nosocomialinfection:
analysis of chromosomal restriction fragment patterns by
pulsed-fieldgel electrophoresis.InfectControlHospEpidemiol
1993;14:595-600.
15. MusserJM.Molecularpopulationgenetic analysisof emerging
bacterial pathogens: selected insights. Emerg Infect Dis
1996;2:1-17.
16. KreiswirthB, KornblumJ, ArbeitRD,et al. Evidencefora clonal origin of methicillinresistance in Staphylococcus
aureus.
Science1993;259:227-230.
17.TenoverFC,ArbeitR,ArcherG, et al. Comparisonof traditional and molecularmethods of typing isolates of Staphylococcus
aureus.J ClinMicrobiol1994;32:407-415.
18. Kato H, Kato N, WatanabeK, et al. Applicationof typing by
pulsed-fieldgel electrophoresisto the studyof Clostridium
difficile in a neonatal intensive care unit. J Clin Microbiol
1994;32:2067-2070.
19. MaslowJN, SlutskyAM, ArbeitRD.The applicationof pulsedfieldgel electrophoresisto molecularepidemiology.In:Persing
DH, SmithTF,TenoverFC,WhiteTJ,eds. DiagnosticMolecular
Microbiology:Principlesand Applications.Washington,DC:
AmericanSocietyof Microbiology;1993:563-572.
20. van Belkum A, Meis J. Polymerasechain reaction-mediated
genotyping in microbial epidemiology. Clin Infect Dis
1994;18:1018-1019.
21. SchwartzDN, SchableB, TenoverFC, et al. Leptotrichiabuccalis bacteremiain patientstreatedin a single bone marrow
transplantunit. ClinInfectDis 1995;20:762-767.
June 1997
22. TenoverFC, McGowanJE Jr. Reasons for the emergence of
antibioticresistance.AmJ MedSci 1996;311:9-16.
23. Davies J. Inactivationof antibioticsand the disseminationof
resistancegenes. Science1994;264:375-382.
24. NakamuraS, NakamuraM, KojimaT, et al.gyrAandgyrBmutations in quinolone resistant strains of Escherichia coli.
AntimicrobAgentsChemother
1989;33:254-255.
25. LocksleyRM,CohenML, QuinnTC, et al. MultiplyantibioticresistantStaphylococcus
aureus:introduction,transmission,and
evolution of nosocomialinfection.Ann Intern Med 1982;97:
317-324.
26. MickelsenPA,PlordeJJ,GordonKP,et al. Instabilityof antibiotic resistancein a strainof Staphylococcus
epidermidisisolated
from an outbreakof prostheticvalve endocarditis.J InfectDis
1985;152:50-58.
27. Olsen JE, Skov MN, Threlfall EJ, et al. Clonal lines of
Salmonellaentericaserotypeenteritidisdocumentedby IS200-,
ribo-,pulsed-fieldgel electrophoresisand RFLPtyping.J Med
Microbiol1994;40:15-22.
28. McDougal LK, Rasheed JK, Biddle JW, Tenover FC.
Identification of multiple clones of extended-spectrum
cephalosporin-resistant
Streptococcus
pneumoniaein the United
States.AntimicrobAgentsChemother
1996;39:2282-2288.
29. SchabergDR,TompkinsLS,FalkowS. Use of agarosegel electrophoresis of plasmid deoxyribonucleicacid to fingerprint
gram-negativebacilli.J ClinMicrobiol1981;13:1105-1110.
30. TaylorDN, WachsmuthIK,ShangkuanY, et al. Salmonellosis
associated with marijuana-a multistateoutbreaktraced by
N EnglJ Med 1982;306:1249-1253.
plasmidfingerprinting.
31.TenoverFC. Plasmidfingerprinting:
a tool for bacterialstrain
identificationand surveillanceof nosocomialand communityacquiredinfections.ClinLabMed1985;5:413-436.
32. FornasiniM, Reeves RR, MurrayBE, et al. TrimethoprimresistantEscherichiacoli in households of childrenattending
daycare centers.J InfectDis 1992;166:326-330.
33. PfallerMA,WakefieldDS, HollisR, et al. The clinicalmicrobiology laboratoryas an aid in infectioncontrol.The application
of moleculartechniquesin epidemiologicstudiesof methicillinresistant Staphylococcus
aureus. Diagn MicrobiolInfect Dis
1991;14:209-214.
34.TompkinsLS, PlordeJJ, FalkowS. Molecularanalysis of Rfactors from multiresistantnosocomial isolates.J InfectDis
1980;141:625-636.
35. Carles-NuitMJ,ChristophileB, BrocheS, et al. DNApolymorphisms in methicillin-susceptibleand maethicillin-resistant
strains of Staphylococcus
aureus. J Clin Microbiol1992;30:
2092-2096.
36. SchwartzDC, CantorCR. Separationof-yeast chromosomesized DNAs by pulsed-fieldgel electrophoresis.Cell 1984;37:
67-75.
37. Clabots CR, Johnson S, Olson MM, et al. Acquisition of
Clostridiumdifficileby hospitalizedpatients:evidencefor colonized new admissionsas a source of infection.J InfectDis
1992;166:561-567.
38. SambrookJ, Fritsch EF, ManiatisT. MolecularCloning-A
LaboratoryManual. 2nd ed. Cold Spring Harbor,NY: Cold
SpringHarborPress; 1989.
39. StullTL, LiPumaJJ, EdlindTD. A broad-spectrumprobe for
molecularepidemiologyof bacteria:ribosomalRNA.J Infect
Dis 1988;157:280-286.
40. ArthurM, ArbeitRD,KimC, et al. Restrictionfragmentlength
polymorphismsamonguropathogenicEscherichiacoli isolates:
pap-relatedsequences compared with rrn operons. Infect
Immun1990;58:471-479.
41. KristjanssonM, SamoreMH,GerdingDN, et al. Comparisonof
restrictionendonucleaseanalysis,ribotyping,and pulsed-field
gel electrophoresisfor moleculardifferentiationof Clostridium
difficilestrains.J ClinMicrobiol1994;32:1963-1969.
42. GordilloME, SinghKV,MurrayBE. Comparisonof ribotyping
and pulsed-fieldgel electrophoresisfor subspecies differentiation of strains of Enterococcusfaecalis. J Clin Microbiol
1993;31:1570-1574.
Vol. 18 No. 6
PAPER
SHEA PosrrITION
43.van EmbdenJDA,CaveMD, CrawfordJT,et al. Strainidentifituberculosisby DNA fingerprinting:
cation of Mycobacterium
recommendationsfor a standardizedmethodology.J Clin
Microbiol1993;31:406-409.
44. SmallPM, HopewellPC, Singh SP,et al. The epidemiologyof
tuberculosisin San Francisco:a population-based
studyusing
conventionaland molecularmethods. N Engl J Med 1994;
330:1703-1709.
45. PersingDH. In vitronucleicacid amplification
techniques.In:
Persing DH, SmithT, TenoverFC, White T, eds. Diagnostic
andApplications.
Molecular
Washington,
Principles
Microbiology:
1993:51-87.
DC:AmericanSocietyforMicrobiology;
46. WelshJ, McClellandM. Fingerprintinggenomes using PCR
with arbitraryprimers.NucleicAcidsRes 1990;18:7213-7218.
47.WilliamsJGK, KubelikAR, Livak KJ, et al. DNA polymorphisms amplifiedby arbitraryprimersare useful as genetic
markers.NucleicAcidsRes1990;18:6531-6535.
48.van Belkum A. DNA fingerprintingof medicallyimportant
microorganismsby use of PCR Clin MicrobiolRev 1994;7:
174-184.
49. KillgoreGE, Kato H. Use of arbitraryprimed PCR to type
Clostridiumdifficileand comparisonof results with those by
immunoblottyping.J ClinMicrobiol1994;32:1591-1593.
50.van BelkumA, KluytmansJ, van LeeuwenW,et al. Multicenter
evaluationof arbitrarily
primedPCRfortypingof Staphylococcus
aureusstrains.J ClinMicrobiol1995;33:1537-1547.
51.Tyler KD,WangG, Tyler SD, JohnsonWM.Factorsaffecting
DNA finreliabilityand reproducibilityof amplification-based
gerprinting of representative bacterial pathogens. J Clin
Microbiol1997;35:339-346.
52. DuBose RF, DykhuiszenDE, Hartl DL. Genetic exchange
among naturalisolates of bacteria:recombinationwithinthe
phoA gene of Escherichiacoli. Proc Natl Acad Sci USA
1988;85:7036-7040.
53. DeanD, PattonM, StephensRS.Directsequenceevaluationof
the majorouter membraneprotein gene variantregions of
ChlamydiatrachomatissubtypesD', I', and L2'.InfectImmun
1991;59:1579-1582.
54.Winters MA, Goering RV, Boon SE, et al. Epidemiological
aureuscomparStaphylococcus
analysisof methicillin-resistant
ing plasmidtypingwith chromosomalanalysisby field inverLetters1993;2:
sion gel electrophoresis.MedicalMicrobiology
33-41.
55. GoeringRV,WintersMA.A rapidmethodforthe evaluationof
chromosomalDNAfromgram-positivecocci by field-inversion
gel electrophoresis.
J ClinMicrobiol1992;30:577-580.
56. LefevreJC,FauconG, SicardAM, GascAM. DNAfingerprintpneumoniaestrainsby pulsed-fieldgel elecing of Streptococcus
trophoresis.J ClinMicrobiol1993;31:2724-2728.
57. LundE, HenrichsenJ. Laboratorydiagnosis,serologyandepipneumoniae.MethodsMicrobiol
demiology of Streptococcus
1978;12:241-262.
58. MurrayBE, SinghKV,HeathJD, et al. Comparisonof genomic DNAs of differententerococcalisolates using restriction
endonucleases with infrequent recognition sites. J Clin
Microbiol1990;28:2059-2063.
59.VersalovicJ, KoeuthT, LupskiJR Distributionof repetitive
DNAsequencesin eubacteriaandapplicationto fingerprinting
of bacterialgenomes.NucleicAcidsRes1991;19:6823-6831.
60. BeltranP, MusserJM, HelmuthR, et al. Towarda population
439
genetic analysisof Salmonella:genetic diversityand relationships among strains of serotypes S choleraesuis,S derby,S
dublin,S enteritidis,S heidelberg,S infantis,S newport,and S
ProcNatlAcadSci 1988;85:7753-7757.
typhimurium.
61. LitwinCM, StormAL, ChipowskyS, RyanKJ.Molecularepidemiologyof Shigellainfections:plasmidprofiles,serotypecorrelation,andrestrictionendonucleaseanalysis.J ClinMicrobiol
1991;29:104-108.
62. Struelens MJ, Schwam V, Deplano A, Baran D. Genome
macrorestriction analysis of diversity and variability of
Pseudomonas aeruginosa strains infecting cystic fibrosis
patients.J ClinMicrobiol1993;31:2320-2326.
63.Allardet-Servent
A, BouzigesN, Carles-NuritM-J,et al. Use of
restriction endonucleases for DNA
low-frequency-cleavage
analysisin epidemiologicalinvestigationsof nosocomialbacterialinfections.J ClinMicrobiol1989;27:2057-2061.
64. GrundmannH, SchneiderC, HartungD, et al. Discriminatory
powerof three DNA-basedtypingtechniquesforPseudomonas
aeruginosa.J ClinMicrobiol1995;33:528-534.
65. Pradella S, Pletschette M, Mantey-StiersF, Bautsch W.
Macrorestriction
analysisof Pseudomonasaeruginosain colonized burn patients.EurJ Clin MicrobiolInfectDis 1994;13:
122-128.
66. R6mlingU, FiedlerB, BosshammerJ, et al. Epidemiologyof
chronicPseudomonas
aeruginosainfectionsin cystic fibrosis.J
InfectDis 1994;170:1616-1621.
67. GoubyA, Carles-NuritM-J,Bouziges N, et al. Use of pulsedfieldgel electrophoresisforinvestigationof hospitaloutbreaks
baumanii.J ClinMicrobiol1992;30:1588-1591.
of Acinetobacter
of differentPCR
68.Yuk-FongLiuP,ShiZ, LauY,et al. Comparison
of Burkholderia
(Pseudomonas)
approachesfor characterization
cepaciaisolates.J ClinMicrobiol1995;33:3304-3307.
69. CaveMD, EisenachKD,McDermottPF,et al. IS6110:consertuberculosis
vationof sequence in the Mycobacterium
complex
and its utilizationin DNA fingerprinting.Mol Cell Probes
1991;5:73-80.
70. FriedmanCR,StoeckleMY,JohnsonWD, Riley LW.Doublerepetitive-elementPCRmethod for subtypingMycobacterium
clinicalisolates.J ClinMicrobiol1995;33:1383-1384.
tuberculosis
71.Arbeit RD, Slutsky A, BarberTW, et al. Genetic diversity
avium causing monoclonal
among strains of Mycobacterium
and polyclonalbacteremiain patientswith AIDS.J InfectDis
1993;167:1384-1390.
72.WallaceRJJr,ZhangY,BrownBA,et al. DNAlarge restriction
fragmentpatternsof sporadicandepidemicnosocomialstrains
abscessus.
chelonaeandMycobacterium
of Mycobacterium
J Clin
Microbiol1993;31:2697-2701.
73. McGowanJE Jr, MetchockB. Infectioncontrolepidemiology
In:MurrayPR,BaronEJ,PfallerMA,
andclinicalmicrobiology.
TenoverFC,YolkenRH,eds. Manualof ClinicalMicrobiology.
6th ed. Washington,DC:AmericanSocietyfor Microbiology;
1995:182-189.
74.vanBelkumA, MelchersW,de PauwBE,et al. Genotypiccharacterizationof sequential Candida albicans isolates from
fluconazole-treated
neutropenicpatients.J InfectDis 1994;169:
1062-1070.
75. Struelens MJ, Members of the EuropeanStudy Group on
EpidemiologicalMarkers.Consensusguidelinesfor appropriate use and evaluationof microbialepidemiologictyping systems. ClinMicrobiolInfect1996;2:2-11.