Molecular methods for serovar determination of Salmonella

17
http://informahealthcare.com/mby ISSN: 1040-841X (print), 1549-7828 (electronic) Crit Rev Microbiol, Early Online: 1–17 ! 2013 Informa Healthcare USA, Inc. DOI: 10.3109/1040841X.2013.837862 REVIEW Molecular methods for serovar determination of Salmonella Chunlei Shi 1,2 , Pranjal Singh 1 , Matthew Louis Ranieri 1 , Martin Wiedmann 1 , and Andrea Isabel Moreno Switt 1 1 Department of Food Science, Cornell University, Ithaca, NY, USA and 2 MOST-USDA Joint Research Center for Food Safety and Bor Luh Food Safety Center, Department of Food Science and Technology, Shanghai Jiao Tong University, Shanghai, China Abstract Salmonella is a diverse foodborne pathogen, which has more than 2600 recognized serovars. Classification of Salmonella isolates into serovars is essential for surveillance and epidemio- logical investigations; however, determination of Salmonella serovars, by traditional serotyping, has some important limitations (e.g. labor intensive, time consuming). To overcome these limitations, multiple methods have been investigated to develop molecular serotyping schemes. Currently, molecular methods to predict Salmonella serovars include (i) molecular subtyping methods (e.g. PFGE, MLST), (ii) classification using serovar-specific genomic markers and (iii) direct methods, which identify genes encoding antigens or biosynthesis of antigens used for serotyping. Here, we reviewed reported methodologies for Salmonella molecular serotyping and determined the ‘‘serovar-prediction accuracy’’, as the percentage of isolates for which the serovar was correctly classified by a given method. Serovar-prediction accuracy ranged from 0 to 100%, 51 to 100% and 33 to 100% for molecular subtyping, serovar-specific genomic markers and direct methods, respectively. Major limitations of available schemes are errors in predicting closely related serovars (e.g. Typhimurium and 4,5,12:i:-), and polyphyletic serovars (e.g. Newport, Saintpaul). The high diversity of Salmonella serovars represents a considerable challenge for molecular serotyping approaches. With the recent improvement in sequencing technologies, full genome sequencing could be developed into a promising molecular approach to serotype Salmonella. Keywords Molecular serotyping, salmonella serovar determination, Salmonella subtyping History Received 15 May 2013 Revised 1 August 2013 Accepted 21 August 2013 Published online 4 November 2013 Introduction Salmonella is one of the most prominent foodborne pathogens in the developing and developed world (Tauxe, 2002). In the USA alone, an estimated 1.03 million cases of domestically acquired foodborne salmonellosis occur each year (Scallan et al., 2011). Salmonella is transmitted by consumption of contaminated food or by contact with infected animals (Hoelzer et al., 2011). Importantly, a number of food products have been associated with foodborne salmonellosis, including eggs, dairy, vegetables and processed foods (Greig & Ravel, 2009). In addition, a number of animals (e.g. reptiles, chicken and other young birds) (Chiodini & Sundberg, 1981; Hoelzer et al., 2011; Wiedmann & Nightingale, 2009) have been found to carry this pathogen and a number outbreaks have been linked with exposure to these animals (Hoelzer et al., 2011; Loharikar et al., 2012). Salmonellosis symptoms vary from self-limiting diarrhea that usually lasts for 12 to 72 hours to long-lasting, disabling effects, such as reactive arthritis (Saarinen et al., 2002) and systemic disease, particularly in susceptible individuals (e.g. infants, immune-compromised individuals) (Feasey et al., 2012; Monack, 2012). While Salmonella is considered a public health concern, the food industry is impacted by salmonellosis as well, as recalls frequently result in major economic losses (GMA, 2010). As Salmonella transmission and routes of entry in the food chain can be complex and diverse (contamination can occur in one or more steps of food production with ingredients potentially sourced from through- out the world), tools that can differentiate Salmonella beyond the species level (e.g. to serovars) are essential to facilitate improved control of this pathogen (Olaimat & Holley, 2012). The Salmonella genus has two species named ‘‘enterica’’ and ‘‘bongori’’. The species S. enterica contains six subspe- cies. The six subspecies of S. enterica are enterica (subspe- cies I), salamae (subspecies II), arizonae (subspecies IIIa), diarizonae (subspecies IIIb), houtenae (subspecies IV) and indica (subspecies VI) (Brenner et al., 2000; Tindall et al., 2005). Recently, subspecies VII was recognized as a new group (Boyd et al., 1996; McQuiston et al., 2008; Porwollik et al., 2002). Importantly, S. enterica subspecies enterica is the most clinically significant group causing 99% of salmon- ellosis cases (Hadjinicolaou et al., 2009). Subspecies are further divided into serogroups and serovars (Grimont & Weill, 2007). The diversity of Salmonella is represented by more than 2600 serovars (e.g. Typhimurium, Montevideo, Agona) (Grimont & Weill, 2007; Guibourdenche et al., 2010). Address for correspondence: Andrea Isabel Moreno Switt, Department of Food Science, Cornell University, Ithaca 14853, USA. E-mail: [email protected] Critical Reviews in Microbiology Downloaded from informahealthcare.com by JHU John Hopkins University on 04/13/14 For personal use only.

Transcript of Molecular methods for serovar determination of Salmonella

Page 1: Molecular methods for serovar determination of Salmonella

http://informahealthcare.com/mbyISSN: 1040-841X (print), 1549-7828 (electronic)

Crit Rev Microbiol, Early Online: 1–17! 2013 Informa Healthcare USA, Inc. DOI: 10.3109/1040841X.2013.837862

REVIEW

Molecular methods for serovar determination of Salmonella

Chunlei Shi1,2, Pranjal Singh1, Matthew Louis Ranieri1, Martin Wiedmann1, and Andrea Isabel Moreno Switt1

1Department of Food Science, Cornell University, Ithaca, NY, USA and 2MOST-USDA Joint Research Center for Food Safety and Bor Luh Food Safety

Center, Department of Food Science and Technology, Shanghai Jiao Tong University, Shanghai, China

Abstract

Salmonella is a diverse foodborne pathogen, which has more than 2600 recognized serovars.Classification of Salmonella isolates into serovars is essential for surveillance and epidemio-logical investigations; however, determination of Salmonella serovars, by traditional serotyping,has some important limitations (e.g. labor intensive, time consuming). To overcome theselimitations, multiple methods have been investigated to develop molecular serotypingschemes. Currently, molecular methods to predict Salmonella serovars include (i) molecularsubtyping methods (e.g. PFGE, MLST), (ii) classification using serovar-specific genomic markersand (iii) direct methods, which identify genes encoding antigens or biosynthesis of antigensused for serotyping. Here, we reviewed reported methodologies for Salmonella molecularserotyping and determined the ‘‘serovar-prediction accuracy’’, as the percentage of isolates forwhich the serovar was correctly classified by a given method. Serovar-prediction accuracyranged from 0 to 100%, 51 to 100% and 33 to 100% for molecular subtyping, serovar-specificgenomic markers and direct methods, respectively. Major limitations of available schemes areerrors in predicting closely related serovars (e.g. Typhimurium and 4,5,12:i:-), and polyphyleticserovars (e.g. Newport, Saintpaul). The high diversity of Salmonella serovars represents aconsiderable challenge for molecular serotyping approaches. With the recent improvement insequencing technologies, full genome sequencing could be developed into a promisingmolecular approach to serotype Salmonella.

Keywords

Molecular serotyping, salmonella serovardetermination, Salmonella subtyping

History

Received 15 May 2013Revised 1 August 2013Accepted 21 August 2013Published online 4 November 2013

Introduction

Salmonella is one of the most prominent foodborne pathogens

in the developing and developed world (Tauxe, 2002). In the

USA alone, an estimated 1.03 million cases of domestically

acquired foodborne salmonellosis occur each year (Scallan

et al., 2011). Salmonella is transmitted by consumption of

contaminated food or by contact with infected animals

(Hoelzer et al., 2011). Importantly, a number of food products

have been associated with foodborne salmonellosis, including

eggs, dairy, vegetables and processed foods (Greig & Ravel,

2009). In addition, a number of animals (e.g. reptiles, chicken

and other young birds) (Chiodini & Sundberg, 1981; Hoelzer

et al., 2011; Wiedmann & Nightingale, 2009) have been found

to carry this pathogen and a number outbreaks have been

linked with exposure to these animals (Hoelzer et al., 2011;

Loharikar et al., 2012).

Salmonellosis symptoms vary from self-limiting diarrhea

that usually lasts for 12 to 72 hours to long-lasting, disabling

effects, such as reactive arthritis (Saarinen et al., 2002) and

systemic disease, particularly in susceptible individuals

(e.g. infants, immune-compromised individuals) (Feasey

et al., 2012; Monack, 2012). While Salmonella is considered

a public health concern, the food industry is impacted by

salmonellosis as well, as recalls frequently result in major

economic losses (GMA, 2010). As Salmonella transmission

and routes of entry in the food chain can be complex and

diverse (contamination can occur in one or more steps of food

production with ingredients potentially sourced from through-

out the world), tools that can differentiate Salmonella beyond

the species level (e.g. to serovars) are essential to facilitate

improved control of this pathogen (Olaimat & Holley, 2012).

The Salmonella genus has two species named ‘‘enterica’’

and ‘‘bongori’’. The species S. enterica contains six subspe-

cies. The six subspecies of S. enterica are enterica (subspe-

cies I), salamae (subspecies II), arizonae (subspecies IIIa),

diarizonae (subspecies IIIb), houtenae (subspecies IV) and

indica (subspecies VI) (Brenner et al., 2000; Tindall et al.,

2005). Recently, subspecies VII was recognized as a new

group (Boyd et al., 1996; McQuiston et al., 2008; Porwollik

et al., 2002). Importantly, S. enterica subspecies enterica is

the most clinically significant group causing 99% of salmon-

ellosis cases (Hadjinicolaou et al., 2009). Subspecies are

further divided into serogroups and serovars (Grimont &

Weill, 2007).

The diversity of Salmonella is represented by more than

2600 serovars (e.g. Typhimurium, Montevideo, Agona)

(Grimont & Weill, 2007; Guibourdenche et al., 2010).

Address for correspondence: Andrea Isabel Moreno Switt, Departmentof Food Science, Cornell University, Ithaca 14853, USA. E-mail:[email protected]

Cri

tical

Rev

iew

s in

Mic

robi

olog

y D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y JH

U J

ohn

Hop

kins

Uni

vers

ity o

n 04

/13/

14Fo

r pe

rson

al u

se o

nly.

Page 2: Molecular methods for serovar determination of Salmonella

Traditional serotyping of Salmonella has been used for

decades worldwide, and has been crucial for foodborne

disease surveillance and outbreak investigations (Wattiau

et al., 2011; Winokur, 2003). Unfortunately, traditional

serotyping has a number of limitations; it is laborious, slow

and can be imprecise (McQuiston et al., 2011). Consequently,

several molecular methods to classify Salmonella serovars

have been developed to complement or replace this method-

ology (Foley et al., 2007; Wattiau et al., 2011; Wiedmann &

Nightingale, 2009). To compare molecular methods that can

be used to predict Salmonella serovars, we identified here

the percentage of isolates for which the serovar was correctly

classified by a given method, which we defined as the

‘‘serovar-prediction accuracy’’. While this provides a basis

for comparison between methods and studies, the serovar-

prediction accuracy depends considerably on the isolates

included in a study and the values reported here need to be

treated with considerable caution.

Traditional Salmonella serotyping

The White-Kauffmann-Le minor scheme is the traditional

method used for designation of Salmonella serovars (Dera-

Tomaszewska, 2012; Guibourdenche et al., 2010). In this

phenotype-based approach, surface antigens are detected by

agglutination of bacterial cells using antisera (Schrader et al.,

2008). According to this scheme, a serovar is determined on

the basis of somatic (O), flagellar (H) and capsular (Vi)

antigens present on the surface of Salmonella (Brenner et al.,

2000). The somatic antigen, which is a polysaccharide present

in Salmonella lipopolysaccharide (LPS) (Reeves et al., 1996;

Schnaitman & Klena, 1993), specifies Salmonella serogroups

and together with the flagellar antigen reactions, is used to

determine a serovar (Schrader et al., 2008). The White-

Kauffmann-Le minor scheme recognizes 46 different somatic

serogroups and 114 different flagellar antigens that, in several

combinations, result in over 2600 different serovars

(Guibourdenche et al., 2010). There are two flagellar antigens

in salmonellae, designated phase I H-antigen (H1) and

phase II H-antigen (H2) (McQuiston et al., 2004, 2011).

The expression of flagellar antigens H1 and H2 is coordinated

via a phase variation mechanism (Bonifield & Hughes, 2003;

Silverman et al., 1979; Yamamoto & Kutsukake, 2006),

therefore, serovars which express two flagellin types are

called diphasic; those with only one flagellar antigen type are

designated monophasic (e.g. serovar 4,5,12:i:-) (McQuiston

et al., 2011). Three of the Salmonella enterica subspecies (i.e.

subspecies IIIa, IV, VII), in addition to S. bongori, lack the

H2-antigen, and thus are considered monophasic (McQuiston

et al., 2008). In rare cases, Salmonella are triphasic,

expressing a third flagellar antigen (e.g. serovars Rubislaw

and Salinatis) (Old et al., 1999; Smith & Selander, 1991). In

some serovars (e.g. Typhi, Dublin), a capsular antigen or Vi is

present acting as a virulence-associated factor (Jansen et al.,

2011).

For traditional serotyping of Salmonella, a laboratory

requires more than 250 different typing antisera as well as

350 different antigens for preparation and quality control of

the antisera (Fitzgerald et al., 2006; McQuiston et al., 2004).

In addition, for less common antigens, commercial antisera

are typically unavailable, or their quality is highly variable

(McQuiston et al., 2004). Traditional serotyping is labor

intensive, requires a minimum of 3 days for a single isolate,

and in some cases can take longer depending upon the

complexity of the serovar (Kim et al., 2006). Other limitations

of traditional serotyping include a possible loss of expression

of an antigen required for serotyping. For example, rough

strains do not express the O-antigen and do not react with

O-antisera. Additionally, mucoid strains produce a capsule

around the bacteria which blocks detection of O-antigens, and

non-motile strains do not express flagellar antigens

(Fitzgerald et al., 2007). In addition, traditional serotyping

is a time consuming process requiring trained technicians.

All the above has increased the scientific interest in develop-

ing a reliable molecular approach for serotyping or serovar

prediction of Salmonella, which could be correlated with the

White-Kauffmann-Le minor scheme. This scheme has been

used for about 70 years, and a considerable amount of data

continues to be generated using this universal serotyping

language.

Overview of molecular approaches to Salmonellaserotyping

A number of different molecular approaches for Salmonella

subtyping, especially DNA based methods, have been

developed (Wattiau et al., 2011; Wiedmann, 2002).

Molecular subtyping methods have many advantages over

traditional methods, such as an increased discriminatory

power, better standardization and better reproducibility

(Herrera-Leon et al., 2007). However, many of these

techniques are still novel and research is needed to improve,

optimize and validate them.

Here, we conceptually separated molecular serotyping

approaches into the following three categories: (i) methods

that could predict serovars based on molecular subtype

(e.g. PFGE, ribotyping, rep-PCR), (ii) methods based on

serovar-specific genomic markers and (iii) direct methods

that target genes encoding antigens. Most of the initial DNA

based subtyping methods for foodborne pathogens were based

on the generation of banding patterns, from either genomic

or plasmid DNA (Hartmann & West, 1997; Wachsmuth et al.,

1991). These patterns are generated after restriction digestion

or from PCR amplified DNA fragments (Nair et al., 2000;

Ribot et al., 2006). DNA sequencing-based subtyping

methods include, for example, multilocus sequence typing

(MLST), which classifies Salmonella subtypes according to

allelic profiles of selected housekeeping genes (Achtman

et al., 2012; Enright & Spratt, 1999).

Direct methods are based on PCR, sequencing or probes

that target the genes encoding the somatic (O) and flagellar

(H1 and H2) antigens (Braun et al., 2012). Finally, whole

genome sequencing of multiple serovars has allowed the

identification of serovar-specific genomic markers, which

have also been used to predict Salmonella serovars (Arrach

et al., 2008; Huehn & Malorny, 2009; Malorny et al., 2007).

While subtyping methods and genomic markers based

methods predict serovars based on molecular subtype, direct

methods allow for a direct comparison of obtained results

with traditional serotyping.

2 C. Shi et al. Crit Rev Microbiol, Early Online: 1–17

Cri

tical

Rev

iew

s in

Mic

robi

olog

y D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y JH

U J

ohn

Hop

kins

Uni

vers

ity o

n 04

/13/

14Fo

r pe

rson

al u

se o

nly.

Page 3: Molecular methods for serovar determination of Salmonella

Prediction of serovars with banding pattern-basedmolecular subtyping methods

Banding pattern based molecular subtyping methods target

the bacterial chromosome to classify bacteria into subtypes.

Cluster analysis can be used to identify groups of similar

subtypes that typically represent the same serovar (Wattiau

et al., 2011). Hence by analyzing the subtype of an individual

isolate, one can predict a serovar (Achtman et al., 2012; Gaul

et al., 2007).

Pulsed field gel electrophoresis

Pulsed field gel electrophoresis (PFGE) is a restriction

fragment length polymorphism based method. Enzyme-

restricted bacterial DNA is separated, with alternating electric

fields, into various fragments representing sizes up to 2000 kb

(Schwartz & Cantor, 1984; Singh et al., 2006). In the

past decade, several researchers have examined the potential

of PFGE to predict Salmonella serovars (Gaul et al., 2007;

Kerouanton et al., 2007; Nde et al., 2006; Weigel et al., 2004;

Zou et al., 2010) (see Table 1). These studies have analyzed

from 46 to 1128 isolates, representing from 6 to 40 different

Salmonella serovars.

Nde et al. (2006) compared PFGE patterns for 80

Salmonella isolates, representing a limited number (n¼ 6)

of serovars. A total of 79/80 isolates were clustered into

distinct clades representing a single serovar based on their

XbaI PFGE patterns. The exception was one S. Montevideo

isolate, which clustered together with three S. Senftenberg

isolates (Nde et al., 2006) (Suppl. Table 1). For this study we

calculated an overall serovar-prediction accuracy of 99%.

Even though the serovar-prediction accuracy was high, in

this study the number of serovars tested was too low to be

conclusive. In a separate study, PFGE patterns for 674

Salmonella isolates representing 12 serovars were clustered

and most of the serovars fell into distinct groups (573/674).

Exceptions included S. Typhimurium var. Copenhagen,

S. 4,[5],12:i:-, and S. Typhimurium, which were clustered in

the same group; and S. Putten and S. Agona, which also

clustered together (Gaul et al., 2007). The overall serovar-

prediction accuracy in this study was calculated as 85%

(Table 1). The study involving the greatest number of isolates

(n¼ 1128), representing 31 serovars, was conducted by

Kerouanton et al. (2007); in this study, the overall serovar

prediction accuracy was 97% (1088/1128). Incorrectly pre-

dicted serovars included S. Paratyphi B, S. Give, S. Saintpaul,

S. Agona, S. Montevideo, and S. Newport (Kerouanton et al.,

2007). Recently, 46 isolates representing 40 serovars were

analyzed by PFGE, yielding a serovar-prediction accuracy

of 75%. Serovars 4,5,12:i:-, Saintpaul and Typhimurium var.

Copenhagen matched with serovar Typhimurium; in addition,

for eight isolates PFGE patterns were found to be different

from all existing isolate patterns in the database (Ranieri

et al., 2013).

Overall, the serovar-prediction accuracy for PFGE ranged

from 75% to 99%; however, most studies did not include

adequate serovar diversity (e.g. isolates representing both

common and rare serovars from multiple sources and

locations) (Table 1). While PFGE has been a useful method

to subtype Salmonella and for outbreak investigations (e.g. Di

Giannatale et al., 2008; Jeoffreys et al., 2001; Ribot et al.,

2006), further characterization of diverse isolate sets is

necessary to identify the specific limitations encountered

when PFGE is used to predict Salmonella serovars.

In addition, a robust PFGE pattern database, of the most

prevalent Salmonella serovars, such as PulseNet, is essential

to interrogate isolates for Salmonella serotyping (Zou

et al., 2012).

Limitations with PFGE-based serovar prediction include:

(i) multiple serovars can have identical PFGE types, because

they recently emerged from a common ancestor, such as

S. Typhimurium versus S. 4,5,12:i:- (Guerra et al., 2000;

Hoelzer et al., 2010; Soyer et al., 2009; Wiedmann &

Nightingale, 2009) or (ii) a single serovar can show high level

of PFGE diversity, particularly serovars that are polyphyletic

(e.g. Newport, Saintpaul (Achtman et al., 2012; Alcaine

et al., 2006; Harbottle et al., 2006; Porwollik et al., 2004;

Sukhnanand et al., 2005)). Importantly, polyphyletic serovars

only represent a challenge if representatives of all phylogen-

etic groups representing a specific serovar are not included

in the database used for serovar prediction. Unfortunately,

the serovars mentioned above include some of the most

common Salmonella serovars associated with human and

animal cases (Hoelzer et al., 2011; CDC, 2009); hence errors

in serovar prediction could have important public health

implications.

Ribotyping

Ribotyping is another subtyping method based on restriction

fragment length polymorphism. It can categorize Salmonella

isolates into groups designed as ribotypes (Esteban et al.,

1993), which can correspond to specific serovars.

Conventional ribotyping involves cleavage of bacterial gen-

omic DNA with a restriction endonuclease, having conserved

cleavage sites inside and outside the rRNA operon; this is

followed by hybridization of electrophoretically separated

genomic DNA fragments with a ribosomal operon probe

(Bouchet et al., 2008). Importantly, high reproducibility

has been demonstrated with the Riboprinter Microbial

Characterization System (Dupont Qualicon, USA) (Ito

et al., 2003).

In seven studies (Bailey et al., 2002; Capita et al., 2007;

De Cesare et al., 2001; Esteban et al., 1993; Oscar, 1998;

Rodriguez et al., 2006; Ranieri et al., 2013) where ribotyping

was used to predict Salmonella serovars, the overall serovar-

prediction accuracy ranged from 39% to 100% (Table 1 and

Suppl. Table 1). The results varied depending on the serovar

diversity tested; in these studies from 2 to 40 serovars were

tested. In most cases, when limited serovars were involved,

the serovar-prediction accuracy of ribotyping was 100% (De

Cesare et al., 2001; Esteban et al., 1993). For example, EcoRI

and PvuII ribotyping of 112 Salmonella isolates, representing

only S. Enteritidis (n¼ 71) and S. Typhimurium (n¼ 41),

predicted the serovar of 100% of the isolates (De Cesare et al.,

2001). Similarly, another study predicted the serovars of all

isolates, which represented only three serovars (i.e. S.

Reading, S. Typhimurium and S. Senftenberg) (Esteban

et al., 1993). However, in studies with more serovar diversity

(i.e. more than nine serovars), the serovar-prediction accuracy

DOI: 10.3109/1040841X.2013.837862 Molecular methods for serovar determination of Salmonella 3

Cri

tical

Rev

iew

s in

Mic

robi

olog

y D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y JH

U J

ohn

Hop

kins

Uni

vers

ity o

n 04

/13/

14Fo

r pe

rson

al u

se o

nly.

Page 4: Molecular methods for serovar determination of Salmonella

Tab

le1

.C

om

par

iso

no

fd

iffe

ren

tm

ole

cula

rse

roty

pin

gm

eth

od

sfo

rS

alm

on

ella

spp

.

No

.o

fis

ola

tes

test

edN

o.

of

sero

var

ste

sted

Iso

late

sou

rces

Ser

ovar

-pre

dic

tio

nac

cura

cy5

Ref

eren

ces

Met

ho

ds

tha

tP

red

ict

Ser

ova

rsB

ase

do

nM

ole

cula

rS

ub

typ

esP

FG

E 80

6T

urk

eyp

roce

ssin

gp

lan

t9

9%

(79

/80

)(N

de

etal

.,2

00

6)

68

10

Sw

ine

farm

s8

4%

(57

/68

)(W

eigel

etal

.,2

00

4)

67

41

2S

win

e8

5%

(57

3/6

74

)(G

aul

etal

.,2

00

7)

86

68

Fo

od

anim

als,

pro

du

ctio

nfa

cili

ties

,an

dcl

inic

alsa

mp

les

96

%(8

32

/86

6)

(Zo

uet

al.,

20

10

)1

12

83

1F

oo

d,

anim

als,

hu

man

s,n

atu

ral

env

iro

nm

ent,

and

pro

cess

ing

pla

nts

97

%(1

08

8/1

12

8)

(Ker

ou

anto

net

al.,

20

07

)4

64

0H

um

anan

dca

ttle

75

%(3

5/4

6)

(Ran

ieri

etal

.,2

01

3)

Rib

oty

pin

g1

12

2(S

E,

ST

)1B

roil

eran

dtu

rkey

farm

s,an

dp

ou

ltry

mea

t-b

ased

foo

ds

10

0%

(11

2/1

12

)(D

eC

esar

eet

al.,

20

01

)1

21

3A

nim

als

10

0%

(12

1/1

21

)(E

steb

anet

al.,

19

93

)6

09

Ch

icken

carc

asse

s7

2%

(43

/60

)(C

apit

aet

al.,

20

07

)9

11

4B

eef

and

dai

ryca

ttle

op

erat

ion

s,sw

ine

pro

du

ctio

nfa

cili

ties

,an

dp

ou

ltry

farm

s7

5%

(68

/91

)(R

od

rig

uez

etal

.,2

00

6)

11

72

2B

roil

erch

icken

pro

cess

or

39

%(4

5/1

17

)(O

scar

,1

99

8)

25

93

2P

ou

ltry

fece

s,ca

rcas

sri

nse

s,sc

ald

wat

er,

dra

gsw

abs

73

%(1

88

/25

9)

(Bai

ley

etal

.,2

00

2)

46

40

Hu

man

and

catt

le7

4%

(34

/46

)(R

anie

riet

al.,

20

13

)R

AP

D-P

CR

42

2(S

E,

ST

)1H

um

ans

and

foo

d0

%(0

/42

)(R

izzi

etal

.,2

00

5)

11

22

(SE

,S

T)1

Bro

iler

,h

en,

and

turk

eyfa

rms

and

po

ult

rym

eat-

bas

edfo

od

s1

00

%(1

12

/11

2)

(De

Ces

are

etal

.,2

00

1)

23

51

2H

um

ansp

ora

dic

and

ou

tbre

akst

rain

s,fo

od

and

wat

eris

ola

tes

10

0%

(23

5/2

35

)(S

oto

etal

.,1

99

9)

85

22

Sew

age

effl

uen

tan

dsu

rfac

ew

ater

0(0

/85

)(B

urr

etal

.,1

99

8)

12

84

3P

atie

nts

and

lab

sto

ckco

llec

tio

ns

91

%(1

16

/12

8)

(Sh

ang

ku

an&

Lin

,1

99

8)

Rep

-PC

R7

02

(SE

,S

T)1

Dif

fere

nt

avia

nsp

ecie

ssa

mp

les

90

%(6

3/7

0)

(Mil

lem

ann

etal

.,1

99

6)

54

4T

urk

eyp

roce

ssin

gp

lan

ts1

00

%(5

4/5

4)

(An

der

son

etal

.,2

01

0)

68

10

Sw

ine

farm

s9

0%

(61

/68

)(W

eigel

etal

.,2

00

4)

70

15

Cli

nic

pat

ien

tsfo

rh

um

ans

96

%(6

7/7

0)

(Jo

hn

son

etal

.,2

00

1)

44

21

Po

ult

ry6

4%

(28

/44

)(W

ise

etal

.,2

00

9)

15

52

1P

ou

ltry

92

%(1

43

/15

5)

(Ch

enu

etal

.,2

01

2)

89

22

Sew

age

effl

uen

tan

dsu

rfac

ew

ater

0%

(0/8

9)

(Bu

rret

al.,

19

98

)6

54

9P

ou

ltry

mea

tan

dfe

ces

10

0%

(65

/65

)(V

anL

ith

&A

arts

,1

99

4)

13

38

02

Hu

man

fece

s,m

eat,

egg

s,fo

od

anim

als,

rep

tile

s,w

ater

,fa

rmen

vir

on

men

tan

deq

uip

men

t9

2%

(12

3/1

33

)(R

assc

hae

rtet

al.,

20

05

)4

64

0H

um

anan

dca

ttle

65

%(3

0/4

6)

(Ran

ieri

etal

.,2

01

3)

PC

R-R

FL

P3

07

Can

talo

up

ean

dch

ile

pep

per

farm

s1

00

%(3

0/3

0)

(Gal

leg

os-

Ro

ble

set

al.,

20

08

)4

11

0H

um

anfe

ces,

hu

man

,m

eat,

bee

f,sn

ake,

refe

ren

ceco

llec

tio

n2

%(1

/41

)(N

air

etal

.,2

00

2)

11

25

2P

ou

ltry

farm

s7

1%

(37

/52

)(H

on

get

al.,

20

03

)A

FL

P 30

15

Ref

eren

ceco

llec

tio

n1

00

%(3

0/3

0)

(Pet

ers

&T

hre

lfal

l,2

00

1)

11

02

5H

um

ano

rvet

erin

ary

sou

rce

96

%(1

06

/11

0)

(To

rpd

ahl

etal

.,2

00

5)

78

62

Po

ult

ry1

00

%(7

8/7

8)

(Aar

tset

al.,

19

98

)

4 C. Shi et al. Crit Rev Microbiol, Early Online: 1–17

Cri

tical

Rev

iew

s in

Mic

robi

olog

y D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y JH

U J

ohn

Hop

kins

Uni

vers

ity o

n 04

/13/

14Fo

r pe

rson

al u

se o

nly.

Page 5: Molecular methods for serovar determination of Salmonella

ML

ST 25

7C

hic

ken

s9

2%

(23

/25

)(L

iuet

al.,

20

10

)6

61

2C

attl

e,b

ird

s,h

ors

esan

do

ther

mam

mal

s9

9%

(65

/66

)(S

uk

hn

anan

det

al.,

20

05

)1

10

25

Hu

man

and

vet

erin

ary

sou

rce

98

%(1

08

/11

0)

(To

rpd

ahl

etal

.,2

00

5)

15

23

3R

efer

ence

coll

ecti

on

10

0%

(15

2/1

52

)(B

en-D

arif

etal

.,2

01

0)

42

57

55

4R

efer

ence

coll

ecti

on

88

%(3

75

3/4

25

7)

(Ach

tman

etal

.,2

01

2)

46

40

Hu

man

and

catt

le9

1%

(42

/46

)(R

anie

riet

al.,

20

13

)C

RIS

PR

typ

ing

17

11

0H

um

an,

env

iro

nm

ent

and

foo

d7

8%

(13

3/1

71

)(L

iuet

al.,

20

11

)7

44

13

0H

um

anan

dfo

od

pro

du

cts

98

%(7

30

/74

4)

(Fab

reet

al.,

20

12

)G

eno

mic

mar

ker

bas

edm

eth

od

s3

02

(SE

,S

T)1

Ref

eren

ceco

llec

tio

ns,

po

ult

ryfe

ces

and

foo

d1

00

%(3

0/3

0)

(Liu

etal

.,2

01

2)

63

2(S

E,

ST

)1P

igfe

ces

98

%(6

2/6

3)

(Lee

etal

.,2

00

9)

77

3V

eter

inar

y,en

vir

on

men

tal,

foo

d,

and

clin

ical

sou

rces

10

0%

(77

/77

)(A

lvar

ezet

al.,

20

04

)4

75

7P

refe

ctu

ral

lives

tock

hyg

ien

esa

mp

les

fro

md

iffe

ren

tfa

rms

and

geo

gra

ph

ical

reg

ion

sin

Jap

an1

00

%(4

75

/47

5)

(Ak

iba

etal

.,2

01

1)

41

23

Hu

man

s,p

igs

or

chic

ken

s9

0%

(37

/41

)(A

arts

etal

.,2

01

1)

27

33

0H

um

ans

99

%(2

71

/27

3)

(Kim

etal

.,2

00

6)

85

33

4E

gg

s,p

ou

ltry

and

mea

t7

3%

(62

/85

)(M

erte

set

al.,

20

10

)1

42

34

2C

lin

ical

sou

rces

,fe

edlo

tca

ttle

fece

s,tu

rkey

,sw

ine,

ho

rse,

and

rep

tile

s9

5%

(13

5/1

42

)(P

eter

son

etal

.,2

01

0)

75

45

8F

oo

d-p

rod

uci

ng

anim

als

95

%(7

14

/75

4)

(Wat

tiau

etal

.,2

00

8)

75

15

8H

um

ans

85

%(6

39

/75

1)

(Lea

der

etal

.,2

00

9)

97

37

Co

mm

on

inE

uro

pea

nU

nio

nfo

rh

um

ans

51

%(4

9/9

7)

(Raj

tak

etal

.,2

01

1)

60

26

Ch

icken

65

%(3

9/6

0)

(O’R

egan

etal

.,2

00

8)

24

12

No

tp

rov

ided

10

0%

(24

/24

)(A

rrac

het

al.,

20

08

)D

irec

tm

eth

od

sP

CR

and

PC

R-s

equen

cin

gb

ased

met

ho

ds

targ

etin

gO

and

Han

tigen

alle

les

45

21

Ref

eren

ceco

llec

tio

ns

33

%(1

5/4

5)

(Hir

ose

etal

.,2

00

2)

26

7B

lin

ded

3R

efer

ence

coll

ecti

on

s9

6%

(25

5/2

67

)(M

un

oz

etal

.,2

01

0)

40

0B

lin

ded

3R

efer

ence

coll

ecti

on

s9

7%

(38

6/4

00

)(C

ard

on

a-C

astr

oet

al.,

20

09

)5

00

30

Ref

eren

ceco

llec

tio

ns

85

%(4

23

/50

0)

(Her

rera

-Leo

net

al.,

20

07

)8

53

7R

efer

ence

coll

ecti

on

s5

8%

(49

/85

)(R

ajta

ket

al.,

20

11

)2

39

43

An

imal

san

dh

um

an5

4%

(12

9/2

39

)(H

on

get

al.,

20

08

)1

16

10

9H

um

anan

dca

ttle

88

%(9

6/1

09

)(R

anie

riet

al.,

20

13

)P

rob

eb

ased

met

ho

ds

targ

etin

gO

and

Han

tigen

alle

les

20

01

0S

G4

Ref

eren

ceco

llec

tio

ns

95

%(1

89

/20

0)

(Fit

zger

ald

etal

.,2

00

7)

50

01

00

CD

Cco

llec

tio

n8

0%

(40

2/5

00

)(M

cQu

isto

net

al.,

20

11

)1

61

6R

efer

ence

coll

ecti

on

s7

5%

(12

/16

)(Y

osh

ida

etal

.,2

00

7)

10

03

58

Ref

eren

ceco

llec

tio

ns

87

%(8

7/1

00

)(F

ran

kli

net

al.,

20

11

)1

05

34

3R

efer

ence

coll

ecti

on

s8

9%

(93

/10

5)

(Bra

un

etal

.,2

01

2)

1S

Sa

lmo

nel

laE

nte

riti

dis

,S

Sa

lmo

nel

laT

yp

him

uri

um

.2F

ive

un

typ

eab

leis

ola

tes

wer

en

ot

add

edin

the

calc

ula

tio

ns

of

sero

var

-pre

dic

tio

nac

cura

cy.

3O

nly

the

bli

nd

edse

to

fis

ola

tes

(iso

late

sero

var

dat

aw

ere

un

kn

ow

nfo

rth

eev

aluat

ion

)w

ere

incl

ud

edin

the

calc

ula

tio

ns

inth

ista

ble

.4S

sero

gro

up

s.In

this

stu

dy

on

lyse

rog

rou

pin

form

atio

nw

asav

aila

ble

.5P

erce

nt

of

iso

late

sfo

rw

hic

hth

ese

rovar

was

corr

ectl

ycl

assi

fied

wit

ha

giv

enm

eth

od

.

DOI: 10.3109/1040841X.2013.837862 Molecular methods for serovar determination of Salmonella 5

Cri

tical

Rev

iew

s in

Mic

robi

olog

y D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y JH

U J

ohn

Hop

kins

Uni

vers

ity o

n 04

/13/

14Fo

r pe

rson

al u

se o

nly.

Page 6: Molecular methods for serovar determination of Salmonella

was dramatically reduced (Table 1). Ribotyping of 60

Salmonella isolates, representing nine serovars, resulted in

an overall prediction-accuracy of 72% (43/60) (Capita et al.,

2007). In this particular study, the authors reported four

clusters composed of multiple serovars; specifically, (i) S.

Paratyphi B clustered with S. Enteritidis; (ii) S. Newport, S.

Enteritidis, S. Paratyphi B, and S. Infantis clustered together;

(iii) S. Newport clustered into an S. Enteritidis group; and (iv)

S. Infantis, S. Typhimurium, S. Enteritidis, S. Virchow, and S.

Derby clustered together. The lowest serovar-prediction

accuracy of only 39% (45/117) was obtained in one study

conducted with 117 Salmonella isolates, representing 22

serovars (Oscar, 1998). This low serovar-prediction accuracy

could be attributed to serovars with multiple ribotypes, such

as S. Worthington (4 ribotypes), S. Hadar (4 ribotypes),

S. Mbandaka (3 ribotypes), and S. Senftenberg (3 ribotypes).

A number of ribotypes also contained multiple serovars,

such as one ribotype composed of S. Typhimurium,

S. Brandenburg, S. Cerro, and S. Heidelberg (Oscar, 1998)

(Suppl. Table 1).

Ribotyping usually generates relatively few bands (typic-

ally 5–10 bands) (Bouchet et al., 2008), and while this can

facilitate pattern analysis (Bailey et al., 2002; Foley et al.,

2007), a reduced number of bands introduces some short-

comings, such as lower discrimination power and conse-

quently incorrect serovar prediction. This is more relevant

for those isolates representing closely related serovars (e.g.

serovars Typhimurium versus 4,5,12:i:- (Bailey et al., 2002;

Guerra et al., 2000). In addition, the PvuII Riboprinter pattern

database only contains 227 of the 2600 currently recognized

Salmonella serovars (Guibourdenche et al., 2010), limiting

the ability of serovar prediction when this database is used.

Random amplified polymorphic DNA-PCR

Subtyping with random amplified polymorphic DNA-PCR

(RAPD-PCR) is based on random DNA amplification of the

bacterial genome (Ellsworth et al., 1993; Holmberg & Feroze,

1996; Singh et al., 2006). RAPD-PCR uses arbitrary

primers of approximately 10 mers, which amplify the DNA

under flexible PCR conditions (Williams et al., 1990). The

discriminatory power of this method is greatly affected by

the choice of primers and PCR conditions, which also makes

this method difficult to reproduce (Smith et al., 2011).

There are only few studies that used RAPD-PCR to predict

Salmonella serovars; however, the results are rather contra-

dictory and serovar-prediction accuracy ranged from 0 to

100% (Burr et al., 1998; De Cesare et al., 2001; Rizzi et al.,

2005; Shangkuan & Lin, 1998; Soto et al., 1999) (Table 1),

which may reflect the typically poor reproducibility of

RAPD-PCR. In two of the studies (Burr et al., 1998; Rizzi

et al., 2005) we identified a 0% serovar-prediction accuracy;

in both studies RAPD-PCR produced identical patterns for

different serovars (Table 1). Conversely, in three other studies

RAPD-PCR was sensitive enough to predict Salmonella

serovars and the RAPD-PCR serovar-prediction accuracy

ranged from 91 to 100% (De Cesare et al., 2001; Shangkuan

& Lin, 1998; Soto et al., 1999) (Table 1). While in some

studies RAPD-PCR methods have shown high discriminatory

power, a major drawback is the lack of reproducibility.

Importantly, even slight changes in reagents or reaction

conditions may result in significant differences in the banding

patterns produced and hence potentially incorrect serovar

prediction (Ellsworth et al., 1993; Holmberg & Feroze, 1996;

Lin et al., 1996; Singh et al., 2006).

Repetitive element (Rep) PCR

Repetitive element PCR (Rep-PCR) amplifies genomic DNA

fragments using primers that target repetitive DNA elements

(Ridley, 1998; Versalovic et al., 1991). For Salmonella

subtyping, three sets of repetitive elements present in the

genome of Enterobacteriaceae have been used (i.e. 38-bp

repetitive extragenic palindromic (Rep) sequence, 126-bp

enterobacterial repetitive intergenic consensus (ERIC)

sequences, and 154-bp BOX sequence) (Gilson et al., 1990;

Hulton et al., 1991; Martin et al., 1992). These sequences allow

discrimination because they are genetically stable and vary

with respect to chromosomal location and copy number

between strains (Bennasar et al., 2000). Similarly to RAPD-

PCR, reproducibility is a problem in both Rep-PCR and ERIC-

PCR (Rasschaert et al., 2005). However, the DiversiLab�

system from bioMerieux Industry (Marcy-l’Etoile, France),

which is a commercial product for Rep-PCR, offers improved

reproducibility (Chenu et al., 2012; Healy et al., 2005).

In ten studies (Anderson et al., 2010; Burr et al., 1998;

Chenu et al., 2012; Johnson et al., 2001; Millemann et al.,

1996; Ranieri et al., 2013; Rasschaert et al., 2005; Van Lith &

Aarts, 1994; Weigel et al., 2004; Wise et al., 2009) that used

Rep-PCR to predict Salmonella serovars, the serovar-predic-

tion accuracy ranged from 64 to 100%, except for one study

conducted in 1998 (Burr et al., 1998), in which the serovar-

prediction accuracy was 0% (Table 1). The most comprehen-

sive study that used Rep-PCR to predict Salmonella serovars

included 133 isolates representing 80 serovars (Rasschaert

et al., 2005). The serovar-prediction accuracy in this study

was calculated as 92% (125/133); the exceptions included:

two isolates of S. Typhimurium var. Copenhagen that grouped

with S. Typhimurium, two isolates of S. Infantis that grouped

with S. 6,7:r:�, one isolate of serovar S. Enteritidis that

grouped with S. 9:�:�, and one S. Urbana that grouped with

S. Sundsvall (Suppl. Table 1). Similarly, in a recent study,

using the DiversiLab� kit, of 155 blinded Salmonella isolates

representing 21 serovars, the overall serovar-prediction

accuracy was 92% (143/155); specifically, ten isolates

resulted in ‘No Match’, one S. 4,12:-:- was incorrectly

matched to S. Typhimurium, and one S. Sofia isolate was also

incorrectly matched to S. Typhimurium (Chenu et al., 2012).

As with other subtyping methods that are based on banding

patterns, Rep-PCR encountered difficulty with prediction of

homologous and polyphyletic serovars (Weigel et al., 2004;

Wise et al., 2009). In addition, while reproducibility was

a limitation in some studies (Rasschaert et al., 2005), use of

a commercial standardized kit, such as the DiversiLab�

product, appears to considerably improve reproducibility.

PCR-restriction fragment length polymorphism(PCR-RFLP)

PCR-restriction fragment length polymorphism (PCR-RFLP)

is based on banding patterns that are obtained through

6 C. Shi et al. Crit Rev Microbiol, Early Online: 1–17

Cri

tical

Rev

iew

s in

Mic

robi

olog

y D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y JH

U J

ohn

Hop

kins

Uni

vers

ity o

n 04

/13/

14Fo

r pe

rson

al u

se o

nly.

Page 7: Molecular methods for serovar determination of Salmonella

restriction digestion of PCR-amplified DNA (Olive & Bean,

1999). In order to utilize PCR-RFLP as a subtyping method,

a target sequence containing polymorphisms that can differ-

entiate isolates up to the subspecies level must be identified

(Wassenaar & Newell, 2000). For Salmonella serovar predic-

tion, fliC and fljB, genes responsible for phase I and phase II

flagellar antigens, have been used as target for PCR-RFLP

(Dauga et al., 1998; Gallegos-Robles et al., 2008; Hong

et al., 2008; Kilger & Grimont, 1993; Shima et al., 2004).

Importantly, the discriminatory power of the method depends

also upon the type of restriction endonuclease chosen

(Wassenaar & Newell, 2000).

Three studies have used PCR-RFLP to predict Salmonella

serovars (Gallegos-Robles et al., 2008; Hong et al., 2003;

Naire et al., 2002). Two of these studies used H antigen

encoding genes (i.e. only fliC in one study (Gallegos-Robles

et al., 2008)) and fliC and fljB in the other (Hong et al., 2003),

the third study used groEL (Nair et al., 2002) (Table 1 and

Suppl. Table 1). As multiple targets, not restricted to

H-antigen encoding genes, can been used in a PCR-RFLP

scheme, studies that used this methodology are mentioned in

this section, instead of below in the section of this manuscript

that discusses methods that target antigen encoding genes.

Serovar-prediction accuracy among the three studies

ranged from 2% for the study that used groEL PCR-RFLP

to 100% for the study that used fliC PCR-RFLP. The study

which characterized the greatest number of isolates (n¼ 112),

representing the most diverse set of Salmonella serovars

(n¼ 52) used fliC and fljB PCR-RFLP (Hong et al.,

2003). This study showed a serovar-prediction accuracy

of 71% (37/52) (Table 1). Important non-typhoidal serovars

associated with human salmonellosis cases, including

S. Typhimurium, S. Dublin, and S. Enteritidis, could not be

differentiated with this method, because they shared the same

patterns (Hong et al., 2003).

While the PCR-RFLP method is technically simple to

perform, it relies on variation of a very small bacterial

genomic region for differentiation, which can adversely affect

the discriminatory power and can cause difficulty interpreting

the data produced (Olive & Bean, 1999). While fliC and fljB

PCR-RFLP may be able to differentiate homologous serovars

(e.g. S. 4,5,12:i:- versus S. Typhimurium), the discriminatory

power of PCR-RFLP is not high enough to differentiate

O-antigen negative variant serovars (e.g. S. Typhimurium var.

Copenhagen versus S. Typhimurium); or serovars with highly

similar antigens (e.g. S. Enteritidis versus S. Dublin) (Dauga

et al., 1998; Masten & Joys, 1993). fliC and fljB PCR-RFLP

patterns of 264 Salmonella serovars were established in 1998

(Dauga et al., 1998); while 112 Salmonella isolates were

added in 2003 (Hong et al., 2003), no comprehensive efforts

have been executed to expand this database, which limits the

application of PCR-RFLP for Salmonella serotyping.

Amplified fragment length polymorphism (AFLP)

In AFLP, DNA is fragmented with two restriction enzymes,

one high cutting-frequency enzyme (e.g. MseI or TaqI) and

one average cutting-frequency enzyme (e.g. EcoRI, PstI).

Then, adapter sequences (double stranded oligonucleotides)

are linked to the ends of a group of restriction fragments.

This is followed by two PCRs with adapter-specific primers

under highly stringent conditions (Savelkoul et al., 1999;

Singh et al., 2006). Finally, PCR products are separated by gel

electrophoresis and the banding patterns (typically 40–200

bands) are obtained.

In three studies that used AFLP to predict Salmonella

serovars (Aarts et al., 1998; Peters & Threlfall, 2001;

Torpdahl et al., 2005), the serovar-prediction accuracy was

high (Table 1). Up to 100% of serovar-prediction accuracy

was identified in two of these studies (Aarts et al., 1998;

Peters & Threlfall, 2001), one analyzed 30 isolates repre-

senting 15 serovars and the other analyzed 78 isolates

representing 62 serovars (Table 1). According to the authors’

findings, isolates from the same serovar had an identical

AFLP profile. However, only a few isolates representing the

same serovar were actually interrogated (Aarts et al., 1998;

Peters & Threlfall, 2001; Torpdahl et al., 2005) (Suppl.

Table 1).

One of the advantages of AFLP is the fact that it can be

used to generate fingerprints from DNA of any origin

without prior sequence knowledge (Aarts et al., 1998).

AFLP generates results within 2 days; however, reproduci-

bility and band interpretation are major limitations (Torpdahl

et al., 2005). In addition, only limited efforts have been made

to optimize AFLP for Salmonella serovar prediction.

Limitations of serovar prediction by banding patternbased subtyping methods

Two main limitations for Salmonella serovar prediction

appear to be ubiquitous among all subtyping methods based

on banding patterns. These limitations include (i) serovar

prediction of highly homologous serovars, and (ii) serovar

prediction of polyphyletic serovars. In cases where multiple

serovars have identical banding patterns, these serovars are

typically highly similar or have a common ancestor. For

example, a mutation in the rfb cluster, which encodes the

O-antigen, may result in a change in O-antigen expression;

this is the case for S. Typhimurium (antigenic formula:

1,4,[5],12:i:1,2) and S. Typhimurium var. Copenhagen (anti-

genic formula: 1,4,12:i:1,2) (Hauser et al., 2011; Heisig et al.,

1995). Another example is a deletion or insertion in the fljAB

operon (Bonifield & Hughes, 2003), which may affect

H2-antigen expression, as is the case of S. Typhimurium

(1,4,[5],12:i:1,2) and S. 4,5,12:i:- (Laorden et al., 2010; Soyer

et al., 2009). Thus, banding pattern based serovar-prediction

of isolates belonging to these serovars may not be

reliable. For polyphyletic serovars, as serovars S. Newport,

S. Saintpaul and S. Kentucky (Alcaine et al., 2006; Harbottle

et al., 2006; Sangal et al., 2010; Sukhnanand et al., 2005),

multiple distinct patterns may represent the same serovar

(Didelot et al., 2011; Falush et al., 2006; Octavia & Lan,

2006; Sangal et al., 2010). For polyphyletic serovars,

clustering may lead to incorrect serovar prediction if isolates

in the database are not representative of all clades of a given

serovar. It is important to mention that the highly homologous

and polyphyletic serovars mentioned above are ranked among

the most common Salmonella serovars associated with human

and animal salmonellosis globally (CDC, 2009; Galanis et al.,

2006); therefore, an incorrect prediction could interfere with

DOI: 10.3109/1040841X.2013.837862 Molecular methods for serovar determination of Salmonella 7

Cri

tical

Rev

iew

s in

Mic

robi

olog

y D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y JH

U J

ohn

Hop

kins

Uni

vers

ity o

n 04

/13/

14Fo

r pe

rson

al u

se o

nly.

Page 8: Molecular methods for serovar determination of Salmonella

epidemiological investigations and surveillance of this

important foodborne pathogen.

Prediction of serovars with sequence-basedmolecular subtyping methods

Multilocus sequence typing (MLST)

MLST is a method based on the determination of nucleotide

sequences of internal regions of a series of typically

housekeeping genes (Achtman et al., 2012; Enright &

Spratt, 1999). Importantly, as a subtyping method, MLST

has two major advantages: (i) it is highly reproducible and

(ii) results can easily be exchanged between laboratories,

which make MLST a valuable tool for international and

national surveillance (Achtman et al., 2012; Torpdahl et al.,

2005). In addition, to facilitate the analysis, there are several

publicly accessible MLST databases for MLST allelic profiles

and allele sequences as well as software and web based data

analysis tools (e.g. http://pubmlst.org/, www.pasteur.fr/mlst/,

http://mlst.ucc.ie/mlst/dbs/, http://www.mlst.net/).

A number of studies have investigated the ability of MLST

to predict Salmonella serovars with different schemes,

including 3-gene, 4-gene or 7-gene MLST (Achtman et al.,

2012; Ben-Darif et al., 2010; Liu et al., 2010). Overall

serovar-prediction accuracy for MLST ranged from 88 to

100% (Table 1). Several 7-gene MLST studies using the same

housekeeping genes (i.e. aroC-dnaN-hemD-hisD-purE-sucA-

thrA) (Achtman et al., 2012; Kidgell et al., 2002; Ranieri

et al., 2013) have been reported, this facilitates comparison

among these studies. One study involving 25 isolates that

represented 7 serovars, obtained a serovar-prediction accuracy

of 92% (23/25) (Liu et al., 2010) (Suppl. Table 1), and

only two exceptions were found (i.e. S. Pullorum and

S. Heidelberg). In another study with 110 isolates, represent-

ing 25 serovars, a serovar-prediction accuracy of 98% was

obtained (108/110); exceptions in this study included

S. Goettingen and S.9,12:l,v:-, which shared the same

sequence type (ST) (Torpdahl et al., 2005). In a different

study, 52 strains representing 33 serovars were MLST typed,

with a serovar-prediction accuracy of 100% (Ben-Darif et al.,

2010). One recent publication that investigated 46 isolates

representing 40 serovars, found a serovar-prediction accuracy

of 91% (42/46) (Ranieri et al., 2013). Importantly, a very

comprehensive recent study with 4257 isolates representing

554 serovars obtained a serovar- prediction accuracy of 88%

(3753/4257). This study identified a number of serovars

that fell into multiple unrelated groups (e.g. S. Newport,

S. Paratyphi B and S. Oranienburg (Achtman et al., 2012)),

consistent with previous reports that identified at least some

of these serovars as polyphyletic.

Limitations of MLST as serovar predictor could arise in

serovars that shared a common ancestor (and hence show

identical MLST types for different serovars) and in polyphyl-

etic serovars. Advantages of MLST include: it is highly

reproducible, the MLST Salmonella databases allow

for sequence comparisons and serovar prediction, and

MLST can provide insight for phylogenetic inferences.

Whereas serotyping could misinterpreted phylogenetically

unrelated isolates as the same (same serovar, but different

evolutionary origin); MLST distinguishes evolutionary groups

(Achtman et al., 2012). Importantly, many of the observed

discrepancies between traditional serotyping results and

MLST-based serovar prediction are due to discrimination,

by MLST, of isolates with the same serovar into distinct

phylogenetic groups.

Clustered regularly interspaced short palindromicrepeats (CRISPRs) typing

Clustered regularly interspaced short palindromic repeat

(CRISPR) loci are found in many prokaryotes (Jansen et al.,

2002). In Salmonella two CRISPR loci, CRISPR1 and

CRISPR2, have been reported (Fricke et al., 2011; Touchon

& Rocha, 2010). The application of CRISPR typing is based

on the high degree of polymorphism of the spacers in these

loci, which can differentiate subtypes by the content of the

spacers (Fabre et al., 2012). For subtyping, these loci are

amplified by PCR, followed by sequencing of the PCR

products, and analysis of the CRISPR spacers (Fabre et al.,

2012; Liu et al., 2011).

Only few studies have investigated the ability of CRISPRs

typing to predict Salmonella serovars (Fabre et al., 2012;

Liu et al., 2011). Overall serovar-prediction accuracy for

CRISPRs typing ranged from 78 to 98% (Table 1). In one

study (Liu et al., 2011), 171 isolates representing 10 serovars

were characterized based on sequencing and clustering

analysis of fimH, sseL and CRISPR spacers, with the

overall serovar-prediction accuracy of 78% (133/171), excep-

tions were found with some of the isolates representing

serovars Saintpaul, Montevideo and Muenchen, which

occupied unique branches; and with S. Typhimurium,

S. Typhimurium var. Copenhagen, and S. 4,5,12:i:-, which

clustered together (Suppl. Table 1). In the other study (Fabre

et al., 2012), 744 isolates representing 130 serovars were

typed based on CRISPR spacer content comparison, with

the overall serovar-prediction accuracy of 98% (730/744).

Exceptions included S. Urbana, S. Johannesburg, S. Reading,

S. Pomona, S. Gueuletapee, S. Rubislaw, S. Goettingen and

S. Sandiego (Fabre et al., 2012), which had a few spacers

shared among isolates of the same serovars (Suppl. Table 1).

While CRISPR typing has been reported and optimized for

different organisms (Gomgnimbou et al., 2012), limited

studies have been conducted in Salmonella. Although one

of the studies involved a large number of isolates and

serovars, more studies are needed to evaluate the effectiveness

of CRISPR typing as Salmonella serovar predictor. Due to the

variability of CRISPR it is likely though that rapid diversi-

fication of these regions can lead to problems with regard to

CRISPR-based serovar classification.

Comparison of different banding pattern andsequence-based methods for serovar prediction

As part of this review, we compiled a comparison of the

serovar-prediction methods discussed above (Table 2). This

analysis included: analysis time of each step on a given

method, database availability, necessary equipment, cost per

isolate, and serovar-prediction accuracy. The biggest limita-

tion for many of these methods is the analysis time, which is

up to 43 h for AFLP, and up to 52 h for PFGE (Table 2).

Conversely, for Rep-PCR, MLST and CRISPRs typing the

8 C. Shi et al. Crit Rev Microbiol, Early Online: 1–17

Cri

tical

Rev

iew

s in

Mic

robi

olog

y D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y JH

U J

ohn

Hop

kins

Uni

vers

ity o

n 04

/13/

14Fo

r pe

rson

al u

se o

nly.

Page 9: Molecular methods for serovar determination of Salmonella

Tab

le2

.C

om

par

iso

nam

on

gsu

bty

pin

gm

eth

od

sth

atca

nb

eu

sed

top

red

ict

Sa

lmo

nel

lase

rov

ars.

PF

GE

Au

tom

ated

Rib

oty

pin

gR

AP

D-P

CR

Au

tom

ated

Rep

-PC

RP

CR

-RF

LP

AF

LP

ML

ST

9C

RIS

PR

typ

ing

9

An

alysi

sti

me1

26

–5

2h

10

h8

h5

h2

8h

21

–4

3h

8–

10

hP

roce

du

res

(tim

e)3

–1

6h

cell

lysi

s,3

–1

6h

rest

rict

ion

,2

0h

elec

tro

ph

ore

sis

2h

pre

par

atio

n,

8h

run

tim

e2

hD

NA

extr

acti

on

,3

hP

CR

,3

hel

ectr

op

ho

resi

s

2h

DN

Aex

trac

tio

n,

2h

PC

R,

1h

Ag

ilen

tB

ioan

aly

zer

21

00

2h

DN

Aex

trac

tio

n,

5h

PC

R,

3h

rest

rict

ion

,1

8h

elec

tro

ph

ore

sis

2h

DN

Aex

trac

tio

n,

3h

rest

rict

ion

,3

had

apto

rli

gat

ion

,2

hP

CR

,1

hel

ectr

op

ho

resi

s,2

hse

lect

ive

amp

lifi

cati

on

,4

to6

hp

oly

acry

l-am

ide

elec

tro

ph

ore

sis

2h

DN

Aex

trac

tio

n,

2h

PC

R,

1h

PC

Rp

uri

fica

tio

n3

–5

hse

qu

enci

ng

Nec

essa

ryeq

uip

men

t2P

FG

Eel

ectr

op

ho

r-es

issy

stem

,g

elim

agin

gsy

stem

and

anal

ysi

sso

ftw

are

Rib

oP

rin

ter�

Syst

eman

dan

aly

sis

soft

war

e3

PC

Rth

erm

alcy

cler

,el

ectr

op

ho

resi

su

nit

,an

dgel

ima-

gin

gsy

stem

Div

ersi

Lab

�sy

stem

3P

CR

ther

mal

cycl

er,

elec

tro

ph

ore

sis

un

it,

and

gel

ima-

gin

gsy

stem

PC

Rth

erm

alcy

cler

,au

tom

ated

sequ

ence

ro

rac

cess

tose

qu

en-

cin

gfa

cili

ty4

,5

PC

Rth

erm

alcy

cler

,el

ectr

op

ho

resi

su

nit

,gel

imag

ing

syst

eman

dac

cess

toa

sequ

en-

cin

gfa

cili

ty5

Co

stp

eris

ola

te6

11

51

70

28

(on

ere

acti

on

/is

ola

te)

10

54

0(o

ne

reac

tio

n/

iso

late

)N

A2

80

10

0(t

wo

loci

/is

ola

te)

Net

wo

rk/d

atab

ase

Pu

lseN

et(n

ot

pu

b-

licl

yav

aila

ble

)7A

vai

lab

lew

ith

Rib

oP

rin

ter7

No

tav

aila

ble

Avai

lab

lew

ith

Div

ersy

Lab

for

Div

ersi

Lab

7

No

tav

aila

ble

No

tav

aila

ble

Sev

eral

dat

abas

esp

ub

licl

yav

aila

ble

8

No

tav

aila

ble

Ser

ovar

pre

dic

tio

nac

cura

cy8

3.8

–9

8.8

%3

8.5

–1

00

%0

–1

00

%0

–1

00

%2

.4–

10

0%

96

.4–

10

0%

88

.0–

10

0%

77

.8–

98

.1%

1A

nal

ysi

sti

me

star

tin

gfr

om

asi

ng

leb

acte

rial

colo

ny

on

ap

late

,re

pre

sen

tin

ga

pu

recu

ltu

re.

2T

his

on

lyli

sts

spec

iali

zed

equ

ipm

ent

nec

essa

ry;

bas

iceq

uip

men

tty

pic

ally

avai

lab

lein

acl

inic

alm

icro

bio

log

yla

bo

rato

ry(e

.g.

incu

bat

or,

pip

ette

s,as

say

tub

es,

etc.

)is

no

tli

sted

.3T

ech

niq

ue

can

be

per

form

edw

ith

ou

tth

eco

mm

erci

alsy

stem

,b

ut

rep

rod

uci

bil

ity

isim

pro

ved

wit

hth

esy

stem

.4E

qu

ipm

ent

nee

ded

for

AF

LP

wit

hfl

uo

rop

ho

re-t

agg

edp

rim

ers.

5T

imes

pro

vid

edas

sum

ea

sequ

enci

ng

faci

lity

inth

esa

me

lab

ora

tory

or

nea

rby

(to

avo

idti

me

del

ays

asso

ciat

edw

ith

ship

pin

g).

6T

hes

ear

eco

stes

tim

ates

per

iso

late

bas

edo

nth

ecu

rren

tch

arges

(US

D),

aso

fJu

ly2

01

3,

by

the

Lab

ora

tory

for

Mo

lecu

lar

Ty

pin

g(L

MT

)at

Co

rnel

lU

niv

ersi

ty;

tru

eco

sts

may

var

yco

nsi

der

ably

bas

edo

nn

um

ber

of

iso

late

ste

sted

per

yea

r,la

bo

rco

sts

etc.

;N

no

tav

aila

ble

.7D

atab

ase

isth

eli

mit

edac

cess

.8D

atab

ases

acce

ssib

leat

:h

ttp

://p

ub

mls

t.o

rg/,

ww

w.p

aste

ur.

fr/m

lst/

,h

ttp

://m

lst.

ucc

.ie/

mls

t/d

bs/

,h

ttp

://w

ww

.mls

t.n

et/

9A

nal

ysi

sti

me,

pro

ced

ure

s,an

deq

uip

men

tar

esi

mil

arfo

rC

RIS

PR

typ

ing

and

ML

ST

.

DOI: 10.3109/1040841X.2013.837862 Molecular methods for serovar determination of Salmonella 9

Cri

tical

Rev

iew

s in

Mic

robi

olog

y D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y JH

U J

ohn

Hop

kins

Uni

vers

ity o

n 04

/13/

14Fo

r pe

rson

al u

se o

nly.

Page 10: Molecular methods for serovar determination of Salmonella

total procedures is just a few hours (5–8 h). In addition,

whereas some methods described here require equipment that

could be used in other application within a laboratory (e.g.

PCR thermal cycler, electrophoresis unit, and gel imaging

system), other methods require specialized equipment with

limited or no use in other applications (e.g. PFGE electro-

phoresis system). Among methods based on banding patterns,

PFGE and AFLP were the only methods showing similar

serovar-prediction accuracy among the studies. The large

range in serovar-prediction accuracy (ranging from 0 to

100%) for RAPD-PCR, Rep-PCR and PCR-RFLP suggests

some limitations of these methods. In contrast, sequence

based methods are typically reproducible; in addition, the

serovar-prediction accuracy of sequenced based serotyping

methods was relatively high (77.8% to 100%). Finally, the

database used to interrogate the isolates is an important

limitation; as described above, in a number of studies the

serovar could not been assigned because it did not match with

any pattern or ST in the database used. Overall, the presence

of optimized protocols, and available databases for compari-

sons make PFGE and MLST practical methods for serovar

prediction, even though the fact that the PulseNet PFGE

database is not publicly available reduces the attractiveness

of this approach for some users.

Genomic marker based methods not targetingO and H antigen alleles

With the development of genomics, numerous genome

sequences of Salmonella serovars have been completed.

Comparative genomics analyses of these data provided

opportunities to identify regions in the genome that are

unique for a given serovar, or ‘‘serovar-specific genomic

markers’’ (Arrach et al., 2008). A number of schemes,

including PCR (multiplex and real-time PCR), and probe-

based schemes (Akiba et al., 2011; Alvarez et al., 2004;

Kim et al., 2006; Lee et al., 2009; Liu et al., 2012; Peterson

et al., 2010) have been developed to target these regions for

serovar prediction (Table 1).

The overall serovar-prediction accuracy of genomic marker

based methods ranged from 51 to 100% (Table 1). Most of the

studies using genomic marker based methods only tested

serovars Enteritidis, Typhimurium (Lee et al., 2009; Liu et al.,

2012), and 4,5,12:i:- (Alvarez et al., 2004). In 2011, Akiba

et al. used serovar-specific genomic regions (SSGRs) based

multiplex PCR to determine seven serovars including

S. Typhimurium, S. Choleraesuis, S. Infantis, S. Hadar,

S. Enteritidis, S. Dublin, and S. Gallinarum, with 100%

serovar-prediction accuracy (Akiba et al., 2011). In order

to allow for the correct identification of more serovars by

multiplex PCR, Kim et al. (2006) developed a double 5-plex

PCR scheme to identify 30 common clinical serovars,

targeting the unique genomic regions in Typhimurium LT2

(STM) and Typhi CT18 (STY) (Porwollik et al., 2004). The

overall serovar-prediction accuracy of this assay was 99%

(271/273); in only two isolates (i.e. one S. Chester and one

S. Infantis), results were not consistent with traditional

serotyping (Kim et al., 2006) (Suppl. Table 1). Later,

Peterson et al. (2010) added an additional 5-plex PCR

targeting Salmonella Typing Virulence (STV) that determines

the presence or absence of the genes spvC, invA, sseL, PT4

and STM7 into Kim’s scheme; this modified approach

allowed for identification of 42 serovars with an overall

serovar-prediction accuracy of 95% (135/142); the exceptions

were found with isolates representing serovars Anatum

(n¼ 2), Kentucky (n¼ 1), Saintpaul (n¼ 1), Weltevreden

(n¼ 1) and Westhampton (n¼ 2) (Peterson et al., 2010)

(Suppl. Table 1). Real-time PCR has also been used to predict

Salmonella serovars (Arrach et al., 2008; O’Regan et al.,

2008; Rajtak et al., 2011). Rajtak et al. (2011) developed

an approach that included three real-time PCR reactions

targeting fliC, fljB, sdr, spv, and floR; with this approach,

a serovar prediction accuracy of 51% (49/97) was obtained.

While only 19/35 Salmonella serovars were correctly pre-

dicted; these 19 serovars represent serovars commonly found

in the European Union (Rajtak et al., 2011). In another study,

a real-time PCR targeting four genes (sefA, sdf, fliC, aceK)

was tested with 60 isolates, representing 26 serovars

(O’Regan et al., 2008). This approach allowed for a serovar-

prediction accuracy of 65% (39/60). This assay allowed for

correct serovar prediction of all isolates representing selected

Salmonella serovars associated with poultry (Enteritidis,

Gallinarum, Typhimurium, Kentucky) as well as serovar

Dublin (associated with cattle). An improved serovar predic-

tion accuracy of 100% (24/24), using real-time PCR, was

obtained by Arrach et al. (2008). Upon investigation of gene

content in 291 strains, representing 32 serovars, these authors

developed a 146-gene real-time PCR approach that correctly

predicted the 24 isolates tested; these isolates represented 12

serovars (e.g. Dublin, Infantis, Typhimurium, Enteritidis

(Table 1)) (Arrach et al., 2008).

Some high throughput genomic based methods for serovar

prediction have been also developed. In a study that identified

Salmonella serovars using a Ligation Detection Reaction

(LDR) microarray assay, Aarts et al. (2011) designed 62

probes targeting 44 genes (related to pathogenicity, fimbriae,

antibiotic resistance, and serovar-specific genes); this array

identified isolates representing 23 serovars with an overall

serovar-prediction accuracy of 90% (37/41); the exceptions

represented (i) S. Enteritidis and S. Moscow, as well as

(ii) S. Hadar and S. Istanbul; each of these respective pairs

produced identical hybridization pattern (Aarts et al., 2011).

Another study determined Salmonella serovars using the

Universal Probe Salmonella Serotyping (UPSS) nanoPCR

chip (Kim et al., 2006), the overall serovar-prediction

accuracy was 73% (62/85).

There are also some commercial assay products currently

available, e.g. PremiTest� (PT; DSM Nutritional Products,

Switzerland); in a comparison study between traditional

serotyping and PremiTest assay (Wattiau et al., 2008), 754

strains representing 58 serovars were tested, and the overall

serovar-prediction accuracy was 95% (714/754) (Table 1).

The exceptions occurred as three types: (i) yielding discrepant

results with traditional serotyping; (ii) not recognized as

defined serovars; and (iii) generating a numeric code corres-

ponding to two possible serovars (Suppl. Table 1).

The widely used genomic markers for serovar determin-

ation are mainly flagellar and somatic antigen-encoding

genes, virulence, phage-associated, and antibiotic resistance

genes (Huehn & Malorny, 2009; Malorny et al., 2007;

10 C. Shi et al. Crit Rev Microbiol, Early Online: 1–17

Cri

tical

Rev

iew

s in

Mic

robi

olog

y D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y JH

U J

ohn

Hop

kins

Uni

vers

ity o

n 04

/13/

14Fo

r pe

rson

al u

se o

nly.

Page 11: Molecular methods for serovar determination of Salmonella

Porwollik et al., 2004). Since the genomic marker based

methods do not require complex information on a variety of O

and H antigen alleles, the design and manipulation are

relatively easier, and the result evaluation is simply based

on the presence or absence of certain amplicon bands or

hybridization signals. On the other hand, most of these

systems only allow for reliable serovar identification of a

limited number of serovars, typically the most common

serovars and/or those serovars for which genome sequence

data was available when a given assay was designed. The rapid

improvement in sequencing technologies has increased the

availability to obtain whole genome sequences of Salmonella

serovars. These sequences can be used for the identification of

further serovar-specific genomic markers (Huehn & Malorny,

2009; Malorny et al., 2007; Porwollik et al., 2004), which

will likely further improve these methods in the near future.

Molecular serotyping based on direct identificationand characterization of the genes encodingO biosynthesis pathways and H-antigens

We used the term ‘‘direct methods’’ here to refer to molecular

methods that target (i) genes that encode the enzymes

involved in somatic antigen (O) synthesis (typically found

in the rfb cluster (Jiang et al., 1991)) as well as (ii) genes that

encode the flagellar antigens (H1 and H2) (Braun et al., 2012;

Franklin et al., 2011; Yoshida et al., 2007), i.e. fliC (Smith

& Selander, 1990) and fljB (Vanegas & Joys, 1995). For

serotyping, molecular methods such as PCR, microarray or

sequencing-based strategies can be utilized to identify and

characterize these targets.

Genes responsible for somatic (O) antigen synthesis are

located within the rfb cluster in the Salmonella chromosome,

typically located between galF and gnd (Figure 1) (Samuel &

Reeves, 2003). Three types of genes are found within the rfb

cluster: (i) genes encoding proteins that facilitate the synthesis

of nucleotide sugars (e.g. rmlBDAC, ddhDABC, tyv),

(ii) genes encoding sugar transferases (e.g. wbaVUN, wbaP,

wbaBCD), and (iii) genes encoding O antigen polymerase

and transporters (e.g. wzx, wzy, and wzz) (Fitzgerald et al.,

2003; Samuel & Reeves, 2003). Importantly, the antigenic

differences in the 46 Salmonella O serogroups are mainly due

to the genetic variation in the gene content (Fitzgerald et al.,

2007) and not due to individual gene sequence variation

(Verma et al., 1998). In rare instances, antigenic factors

(e.g. O:24 and O:25) are encoded by genes outside the rfb

cluster (Fitzgerald et al., 2003).

Currently, sequences are available for the rfb cluster of 28

of the 46 serogroups, representing the serogroups covering the

most common Salmonella serovars. Genes related to sugar

biosynthesis and sugar transferases present a high level of

similarity, within serogroups (Fitzgerald et al., 2007; Samuel

& Reeves, 2003). Importantly, among all genes, wzx is almost

ubiquitous in rfb clusters sequenced to date (Fitzgerald et al.,

2003). Targets to identify Salmonella O antigens of serovars

commonly associated with human infection have been

identified (Tennant et al., 2010). However, to allow for

identification of the complete diversity of Salmonella

serogroups, sequence data for all O-antigen rfb clusters

needs to be determined.

Salmonella has two flagellar antigens, designated as phase

I (H1) and phase II (H2), but only one of them is expressed at

a time (Yamamoto & Kutsukake, 2006), a phenomenon

known as phase variation (Silverman et al., 1979). Phase I and

phase II are encoded by fliC and fljB, respectively; these

genes are located on different locations on the chromosome,

and their expression is regulated by the fljBA operon

(Aldridge et al., 2006). fliC and fljB are highly conserved at

their 50 and 30 ends; an alignment of 280 fliC and fljB alleles,

showed only 6 and 8 nucleotide substitutions, respectively,

within the first 37 and the last 30 nucleotides (McQuiston

et al., 2004). The middle region, corresponding approximately

to amino acids 181 to 390, is quite variable. This variable

region encodes the exposed surface and variable portion of

the flagellar filament (Joys, 1985; Kanto et al., 1991). Some

H antigens are composed of multiple factors which are a

group of antigens; for example, flagellar antigen H2:e,n,x is

composed of three distinct factors: e, n and x (Echeita et al.,

2002; McQuiston et al., 2004). The 114 distinct flagellar

antigens are classified according to immunological relation

into complexes (McQuiston et al., 2004). For example,

whereas flagellar antigens containing antigenic factors ‘‘g’’,

‘‘m’’ or ‘‘t’’ are members of the g complex; antigens not

containing these antigenic factors are members of non-g

complex (Mortimer et al., 2004). Within each complex, amino

acid sequences in the conserved regions have more than 95%

identity, whereas between different complexes only 75 to 85%

amino acid identity is observed; this indicates that

Figure 1. Linear representation of rfb genecluster for nine Salmonella serogroups (i.e.A, B, C1, C2, D1, E1, F and G). Arrowsindicate genes within the clusters. Potentialserogroup-specific targets are circled(Fitzgerald et al., 2003, 2006, 2007).

DOI: 10.3109/1040841X.2013.837862 Molecular methods for serovar determination of Salmonella 11

Cri

tical

Rev

iew

s in

Mic

robi

olog

y D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y JH

U J

ohn

Hop

kins

Uni

vers

ity o

n 04

/13/

14Fo

r pe

rson

al u

se o

nly.

Page 12: Molecular methods for serovar determination of Salmonella

immunologically related antigens are encoded by closely

related alleles (McQuiston et al., 2004). Consequently, these

flagellar genes can be used as targets for molecular

determination of Salmonella serovars, as is described below.

Serotyping methods recognize 63 different Salmonella

phase I antigenic factors and 37 different phase II antigenic

factors (Mortimer et al., 2004). As of January 2013, the

National Center for Biotechnology Information (NCBI)

includes 994 and 607 complete or partial Salmonella fliC

and fljB allele sequences, respectively (Suppl. Table 2).

PCR based methods targeting O and H antigen alleles

Traditional Salmonella serotyping requires identification of

variable O and H antigens; with molecular techniques,

serotyping can be done effectively and rapidly by identifying

the unique gene sequences associated with these antigens

(Luk et al., 1993; Mortimer et al., 2004). Current molecular

schemes include an initial multiplex PCR, which is conducted

to identify the serogroup of an isolate (Fitzgerald et al., 2007;

Herrera-Leon et al., 2007). Primers targeting serovars com-

monly associated with human salmonellosis cases have been

developed. In addition, separate PCRs or PCR-sequencing

approaches have been designed to determine H1 and H2

antigens (Garaizar et al., 2002; Herrera-Leon et al., 2004;

Mortimer et al., 2004; Ranieri et al., 2013).

Some studies have tested the ability to determine

Salmonella serovars by multiplex PCR (Cardona-Castro

et al., 2009; Herrera-Leon et al., 2007; Hirose et al., 2002;

Hong et al., 2008; Munoz et al., 2010; Rajtak et al., 2011).

Examining up to 500 strains and 43 serovars, the overall

serovar-determination accuracy ranged from 33% to

97% (Table 1). In one initial study, Hirose et al. (2002)

attempted to discriminate S. Typhi, S. Paratyphi A and

19 other Salmonella serovars (Suppl. Table 1). This study

showed the lowest serovar-determination accuracy (33%)

(Hirose et al., 2002). Later, with more sequences available for

the rfb cluster, as well as fliC and fljB genes for different

serovars (Fitzgerald et al., 2003; Garaizar et al., 2002;

Herrera-Leon et al., 2004), several multiplex PCRs were

designed to determine O, H1 and H2 antigens. The study that

involved the most variety of antigen combinations targeted,

through multiplex PCRs: (i) five O-antigens (O:4; O:7; O:8;

O:9; O:3,10), (ii) eight H1-antigens (i; r; l,v; e,h; z10; b; d;

g complex), and (iii) seven H2-antigens (1,2; 1,5; 1,6; 1,7;

l,w; e,n,x; e,n,z15) (Herrera-Leon et al., 2007). This com-

bination of multiplex PCRs was tested with 500 iso-

lates representing 30 serovars; for 423/500 isolates the

serovars identified matched traditional serotyping, yielding

a serovar-determination accuracy of 85% (Table 1 and Suppl.

Table 1).

While multiplex PCR is a rapid and cost effective

alternative to the traditional serotyping approach, this tech-

nique also has some disadvantages. For example, this method

does not differentiate serovar variants due to phage conver-

sion, which results in some antigenic alterations, or subtle

point mutation in H1/H2 antigenic genes responsible for loss

of flagellar expression (Hong et al., 2008). Recently, a new

method based on sequence variation was reported (Ranieri

et al., 2013). In this study a multiplex PCR was used to

identify the serogroup, and PCR and subsequent sequencing

of partial fliC and fljB, was used to identify H1 and H2

antigens. This scheme allows for identification of point

mutations in fliC and fljB (Table 1). However, sequence data

for the complete diversity of O and H antigens are not yet

available, making it virtually impossible to yet design a

scheme that could correctly identify the whole diversity of

Salmonella serovars.

Probe based methods targeting O and Hantigen alleles

To allow for identification of Salmonella serovars, probe

based serotyping methods targeting O and H antigen alleles

have also been developed, including a multiplex bead-based

suspension array (Bio-Plex array) (Fitzgerald et al., 2007),

a microsphere-based liquid array (McQuiston et al., 2011),

a DNA-based microarray (Yoshida et al., 2007), a Salmonella

genoserotyping array (SGSA) (Franklin et al., 2011) and a

‘‘fast DNA serotyping’’ microarray (Braun et al., 2012)

(Table 1).

In studies with these probe-based methods, the overall

serovar-determination accuracy ranged from 75% to 95%,

based on evaluation of isolates representing up to 100

serovars (see Table 1). Two studies used probe-based methods

to determine either O-antigen or H-antigens. Fitzgerald

et al. (2007) used array-based method to differentiate ten

O-antigens among 200 isolates, obtaining an antigen deter-

mination accuracy of 95%; and McQuiston et al. (2011) used

a probe-based method to differentiate 36 different H-antigens

among 500 isolates, obtaining an antigen determination

accuracy of 80%. Yoshida et al. (2007) developed a

serotyping microarray targeting four O-antigen and eight

H-antigen alleles; the serovar-determination accuracy with

this array was 75% (12/16) with 16 strains representing 16

serovars (Yoshida et al., 2007). Franklin et al. (2011) and

Braun et al. (2012) added more O- and H-antigen alleles

into the assay to increase the number of Salmonella serovars

that could be identified; these assays showed overall serovar-

determination accuracies of 87% and 89%, respectively

(Braun et al., 2012; Franklin et al., 2011). The array reported

by Franklin et al. (2011) differentiated 18 somatic serogroups

and 41 flagellar antigens, while the array reported by

Braun et al. (2012) discriminated 28 O-antigens and 86

H-antigens. Limitations of this array included cross-

reaction of highly similar serogroups (e.g. O:2 with O:9).

For example, identical microarray patterns were generated

for S. Enteritidis, S. Nitra and S. Blegdam (Braun et al., 2012)

(Suppl. Table 1).

With the methods targeting O and H antigen encoding

genes, as described above (PCR-based and probe-based),

problems and discrepancies with serovar identification of

isolates usually represent three issues: (i) rare serovars are

not covered in the assay, (ii) results are not congruent with

the traditional serotyping (e.g. cross-reaction of highly

similar serogroups), and (iii) monophasic, rough, mucoid,

and nonmotile samples are positively identified (Braun et al.,

2012; Fitzgerald et al., 2007; Franklin et al., 2011; McQuiston

et al., 2011). The first and second exceptions can be revised

through technical improvements, which include the design

12 C. Shi et al. Crit Rev Microbiol, Early Online: 1–17

Cri

tical

Rev

iew

s in

Mic

robi

olog

y D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y JH

U J

ohn

Hop

kins

Uni

vers

ity o

n 04

/13/

14Fo

r pe

rson

al u

se o

nly.

Page 13: Molecular methods for serovar determination of Salmonella

of new primers or probes as more sequences become public.

The third exception represents an advantage of methods

targeting O and H antigen encoding genes over traditional

serotyping, but makes it difficult to compare obtained results

with traditional serotyping (Fitzgerald et al., 2007). Overall,

PCR-based detection of O and H-antigens can be conducted in

any lab with basic equipment for molecular biology (e.g. PCR

thermocycler), but probe-based methods might require special

equipment.

Next generation sequencing technologies asa future tool for Salmonella serotyping

In the recent years, considerable improvements in DNA and

genome sequencing technologies have occurred (MacLean

et al., 2009), which has led to a point where we have

approached affordable prices for sequencing of whole

bacterial genomes (Carrico et al., 2013). A number of

Salmonella serovars have been fully sequenced (e.g. den

Bakker et al., 2011a; Fricke et al., 2011; Jacobsen et al.,

2011; Moreno Switt et al., 2012). Importantly, this not only

had led to the identification of serovar-specific genomic

markers (as detailed above), but also to the application of

whole genome sequencing in outbreak investigations, single

nucleotide polymorphism detection and mobile elements

identification (Carrico et al., 2013; den Bakker et al., 2011b;

Diaz-Sanchez et al., 2013; Moreno Switt et al., 2012).

Several genome sequencing platforms (e.g. Illumina, Solid,

454, Ion torrent (Diaz-Sanchez et al., 2013; MacLean et al.,

2009)) are now routinely used to sequence bacterial

genomes. In addition, available algorithms and pipelines

for sequence assembly and comparative genomics analyses

have improved considerably (Carrico et al., 2013; Darling

et al., 2004; Zerbino & Birney, 2008). At the moment, one

limitation for the use of whole genome sequencing to predict

Salmonella serovars is the lack of a complete database that

could be used to interrogate sequence data (Ranieri et al.,

2013). Specifically, a comprehensive database that contains

(i) the sequences data for the rfb clusters encoding the

pathways for synthesis and (ii) sequence data for fliC and

fljB (as well as the corresponding phase 1 and 2 antigens, as

determined by traditional serotyping), as well as possibly

serovar-specific genomic markers, would be extremely

valuable to facilitate rapid full genome sequencing based

serotyping. While the application of whole genome sequen-

cing for Salmonella serovar prediction has not yet been

comprehensively investigated, the scientific community has

recognized that in the near future (5–10 years), whole

genome sequencing could become a widely used tool for

Salmonella serovar prediction and subtyping (Carrico et al.,

2013; The National Food Institute, 2011). We may see, in

the future, a whole genome sequencing approach for

Salmonella serotyping that includes (i) genome sequencing,

(ii) genome assembly, (iii) extraction of the ‘‘genes of

interest’’ (e.g. serovar markers), and (iv) interrogation of

these genes against a comprehensive database to predict

serovars. In addition, along with serovar prediction, DNA

sequences could be used for epidemiological investigations

or to predict the antimicrobial resistance pattern of the

sequenced isolates.

Conclusions

A number of studies have attempted to use molecular methods

to predict or determine Salmonella serovars. However,

the complexity and diversity of Salmonella serovars makes

this a noteworthy challenge. Attempts to date have focused

on predicting the most common serovars associated with

human salmonellosis. Schemes described in this review will

typically have difficulties correctly identifying rare serovars,

which is crucial in the case of emerging serovars or an

outbreak caused by an uncommon serovar. Errors determining

the most common serovars (e.g. Typhimurium, 4,5,12:i:-,

Newport, etc.) were common for a number of methods; these

errors are not trivial and could seriously interfere with

epidemiological investigations. Importantly, many methods

described here provide better resolution than traditional

serotyping and are thus valuable tools for subtyping of

foodborne pathogens. Whereas subtyping methods described

in this review can improve their accuracy for serovar

prediction by increasing the serovar coverage of the

associated subtype databases, direct methods and serovar-

specific markers can improve serovar prediction accuracy as

genome sequences of the remaining serogroups become

available. With the increased use of whole genome sequen-

cing, more sequences will become available, facilitating

improved design of molecular methods for Salmonella

serotyping in the near future.

Acknowledgements

The authors would like to express their deepest gratitude

to L.D. Rodriguez-Rivera for her critical reading of the

manuscript.

Declaration of interest

This project was supported by USDA-National Integrated

Food safety initiative grant 2008-51110-04333 as well as

USDA-NIFA Special Research Grants 2009-34459-19750 and

2010-34459-20756. The National Natural Science Foundation

of China (NSFC 31000779) supported C. Shi.

References

Aarts HJ, Vos P, Larsson JT, et al. (2011). A multiplex ligation detectionassay for the characterization of Salmonella enterica strains. Int JFood Microbiol 145:S68–78.

Aarts HJ, Van Lith LA, Keijer J. (1998). High-resolution genotyping ofSalmonella strains by AFLP-fingerprinting. Lett Appl Microbiol 26:131–5.

Achtman M, Wain J, Weill FX, et al. (2012). Multilocus sequence typingas a replacement for serotyping in Salmonella enterica. PLoS Pathog8:e1002776.

Akiba M, Kusumoto M, Iwata T. (2011). Rapid identification ofSalmonella enterica serovars, Typhimurium, Choleraesuis, Infantis,Hadar, Enteritidis, Dublin and Gallinarum, by multiplex PCR.J Microbiol Meth 85:9–15.

Alcaine SD, Soyer Y, Warnick LD, et al. (2006). Multilocus sequencetyping supports the hypothesis that cow- and human-associatedSalmonella isolates represent distinct and overlapping populations.Appl Environ Microbiol 72:7575–85.

Aldridge PD, Wu C, Gnerer J, et al. (2006). Regulatory protein thatinhibits both synthesis and use of the target protein controls flagellarphase variation in Salmonella enterica. Proc Natl Acad Sci USA 103:11340–5.

DOI: 10.3109/1040841X.2013.837862 Molecular methods for serovar determination of Salmonella 13

Cri

tical

Rev

iew

s in

Mic

robi

olog

y D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y JH

U J

ohn

Hop

kins

Uni

vers

ity o

n 04

/13/

14Fo

r pe

rson

al u

se o

nly.

Page 14: Molecular methods for serovar determination of Salmonella

Alvarez J, Sota M, Vivanco AB, et al. (2004). Development of amultiplex PCR technique for detection and epidemiological typing ofSalmonella in human clinical samples. J Clin Microbiol 42:1734–8.

Anderson PN, Hume ME, Byrd JA, et al. (2010). Evaluation of repetitiveextragenic palindromic-polymerase chain reaction and denaturedgradient gel electrophoresis in identifying Salmonella serotypesisolated from processed turkeys. Poult Sci 89:1293–300.

Arrach N, Porwollik S, Cheng P, et al. (2008). Salmonella serovaridentification using PCR-based detection of gene presence andabsence. J Clin Microbiol 46:2581–9.

Bailey JS, Fedorka-Cray PJ, Stern NJ, et al. (2002). Serotyping andribotyping of Salmonella using restriction enzyme PvuII. J Food Prot65:1005–7.

Ben-Darif E, Jury F, De Pinna E, et al. (2010). Development of amultiplex primer extension assay for rapid detection of Salmonellaisolates of diverse serotypes. J Clin Microbiol 48:1055–60.

Bennasar A, De Luna G, Cabrer B, Lalucat J. (2000). Rapid identifi-cation of Salmonella Typhimurium, S. Enteritidis and S. Virchowisolates by polymerase chain reaction based fingerprinting methods.Int Microbiol 3:31–8.

Bonifield HR, Hughes KT. (2003). Flagellar phase variation inSalmonella enterica is mediated by a posttranscriptional controlmechanism. J Bacteriol 185:3567–74.

Bouchet V, Huot H, Goldstein R. (2008). Molecular genetic basis ofribotyping. Clin Microbiol Rev 21:262–73.

Boy EF, Wang FS, Whittam TS, Selander RK. (1996). Molecular geneticrelationships of the salmonellae. Appl Environ Microbiol 62:804–8.

Braun SD, Ziegler A, Methner U, et al. (2012). Fast DNA serotyping andantimicrobial resistance gene determination of Salmonella entericawith an oligonucleotide microarray-based assay. PloS one 7:e46489.

Brenner FW, Villar RG, Angulo FJ, et al. (2000). Salmonella nomen-clature. J Clin Microbiol 38:2465–7.

Burr MD, Josephson KL, Pepper IL. (1998). An evaluation of ERIC PCRand AP PCR fingerprinting for discriminating Salmonella serotypes.Lett Appl Microbiol 27:24–30.

Capita R, Alonso Calleja C, Prieto, M. (2007). Prevalence of Salmonellaenterica serovars and genovars from chicken carcasses in slaughter-houses in Spain. J Appl Microbiol 103:1366–75.

Cardona-Castro N, Sanchez-Jimenez M, Lavalett L, et al. (2009).Development and evaluation of a multiplex polymerase chain reactionassay to identify Salmonella serogroups and serotypes. DiagnMicrobiol Infect Dis 65:327–30.

Carrico JA, Sabat AJ, Friedrich AW, Ramirez M. (2013). Bioinformaticsin bacterial molecular epidemiology and public health: databases,tools and the next-generation sequencing revolution. Euro Surveill 18:20382.

Centers for Disease Control and Prevention. (2009). Salmonella surveil-lance: annual summary, 2009. U.S. Department of Health and HumanServices, Centers for Disease Control and Prevention, Atlanta, GA.Available from: http://www.cdc.gov/ncezid/dfwed/pdfs/salmonellaannualsummarytables2009.pdf [last accessed 17 Jul 2013].

Chenu JW, Cox JM, Pavic A. (2012). Classification of Salmonellaenterica serotypes from Australian poultry using repetitive sequence-based PCR. J Appl Microbiol 112:185–96.

Chiodini RJ, Sundberg JP. (1981). Salmonellosis in reptiles: a review.Am J Epidemiol 113:494–9.

Darling AC, Mau B, Blattner FR, Perna NT. (2004). Mauve: multiplealignment of conserved genomic sequence with rearrangements.Genome Res 14:1394–403.

Dauga C, Zabrovskaia A, Grimont PA. (1998). Restriction fragmentlength polymorphism analysis of some flagellin genes of Salmonellaenterica. J Clin Microbiol 36:2835–43.

De Cesare A, Manfreda G, Dambaugh TR, et al. (2001). Automatedribotyping and random amplified polymorphic DNA analysis formolecular typing of Salmonella enteritidis and Salmonella typhimur-ium strains isolated in Italy. J Appl Microbiol 91:780–5.

den Bakker HC, Moreno Switt AI, Cummings CA, et al. (2011a).A whole genome SNP based approach to trace and identify outbreakslinked to a common Salmonella enterica subsp. enterica serovarMontevideo Pulsed Field Gel Electrophoresis type. Appl EnvironMicrobiol 77: 8648–55.

den Bakker HC, Moreno Switt AI, Govoni G, et al. (2011b). Genomesequencing reveals diversification of virulence factor content andpossible host adaptation in distinct subpopulations of Salmonellaenterica. BMC Genomics 12:425.

Dera-Tomaszewska B. (2012). Salmonella serovars isolated for the firsttime in Poland, 1995–2007. Int J Occup Med Environ Health 25:294–303.

Di Giannatale E, Prencipe V, Acciarri VA, et al. (2008). Investigation ofan outbreak of Salmonella enterica subsp. enterica serovar Hadar foodillness in the Abruzzi region of Italy. Vet Ital 44:405–27.

Diaz-Sanchez S, Hanning I, Pendleton S, D’Souza D. (2013). Next-generation sequencing: the future of molecular genetics in poultryproduction and food safety. Poultry Sci 92:562–72.

Didelot X, Bowden R, Street T, et al. (2011). Recombination andPopulation Structure in Salmonella enterica. PLoS Genet 7:e1002191.

Echeita MA, Herrera S, Garaizar J, Usera MA. (2002). Multiplex PCR-based detection and identification of the most common Salmonellasecond-phase flagellar antigens. Res Microbiol 153:107–13.

Ellsworth DL, Rittenhouse KD, Honeycutt RL. (1993). Artifactualvariation in randomly amplified polymorphic DNA banding patterns.BioTechniques 14:214–17.

Enright MC, Spratt BG. (1999). Multilocus sequence typing. TrendsMicrobiol 7:482–7.

Esteban E, Snipes K, Hird D, et al. (1993). Use of ribotyping forcharacterization of Salmonella serotypes. J Clin Microbiol 31: 233–7.

Fabre L, Zhang J, Guigon G, et al. (2012). CRISPR typing and subtypingfor improved laboratory surveillance of Salmonella infections. PloSone 7:e36995.

Falush D, Torpdahl M, Didelot X, et al. (2006). Mismatch inducedspeciation in Salmonella: model and data. Phil Trans R Soc Lond BBiol Sci 361:2045–53.

Feasey NA, Dougan G, Kingsley RA, et al. (2012). Invasive non-typhoidal Salmonella disease: an emerging and neglected tropicaldisease in Africa. Lancet 379:2489–99.

Fitzgerald C, Collins M, Van Duyne S, et al. (2007). Multiplex, bead-based suspension array for molecular determination of commonSalmonella serogroups. J Clin Microbiol 45:3323–34.

Fitzgerald C, Gheesling L, Collins M, Fields PI. (2006). Sequenceanalysis of the rfb loci, encoding proteins involved in the bio-synthesis of the Salmonella enterica O17 and O18 antigens:serogroup-specific identification by PCR. Appl Environ Microbiol72:7949–53.

Fitzgerald C, Sherwood R, Gheesling LL, et al. (2003). Molecularanalysis of the rfb O antigen gene cluster of Salmonella entericaserogroup O:6,14 and development of a serogroup-specific PCR assay.Appl Environ Microbiol 69:6099–105.

Foley SL, Zhao S, Walker RD. (2007). Comparison of molecular typingmethods for the differentiation of Salmonella foodborne pathogens.Foodborne Pathog Dis 4:253–76.

Franklin K, Lingohr EJ, Yoshida C, et al. (2011). Rapid genoserotypingtool for classification of Salmonella serovars. J Clin Microbiol 49:2954–65.

Fricke WF, Mammel MK, McDermott PF, et al. (2011). Comparativegenomics of 28 Salmonella enterica isolates: evidence for CRISPR-mediated adaptive sublineage evolution. J Bacteriol 193:3556–68.

Galanis E, Lo Fo Wong DMA, Patrick ME, et al. (2006). Web-basedsurveillance and global Salmonella distribution, 2000-2002. EmerInfect Dis 12:381–8.

Gallegos-Robles MA, Morales-Loredo A, Alvarez-Ojeda G, et al.(2008). Identification of Salmonella serotypes isolated from canta-loupe and chile pepper production systems in Mexico by PCR-restriction fragment length polymorphism. J Food Prot 71:2217–22.

Garaizar J, Porwollik S, Echeita A, et al. (2002). DNA microarray-basedtyping of an atypical monophasic Salmonella enterica serovar. J ClinMicrobiol 40:2074–8.

Gaul SB, Wedel S, Erdman MM, et al. (2007). Use of pulsed-field gelelectrophoresis of conserved XbaI fragments for identification ofswine Salmonella serotypes. J Clin Microbiol 45:472–6.

Gilson E, Bachellier S, Perrin S, et al. (1990). Palindromic unit highlyrepetitive DNA sequences exhibit species specificity withinEnterobacteriaceae. Res Microbiol 141:1103–16.

Gomgnimbou MK, Abadia E, Zhang J, et al. (2012). ‘‘Spoligoriftyping,’’a dual-priming-oligonucleotide-based direct-hybridization assay fortuberculosis control with a multianalyte microbead-based hybridiza-tion system. J Clin Microbiol 50:3172–9.

Greig JD, Ravel A. (2009). Analysis of foodborne outbreak data reportedinternationally for source attribution. Int J Food Microbiol 130:77–87.

Grimont P, Weill FX. (2007). Antigenic formulae of the Salmonellaserovars, 9th ed. WHO Collaborating Centre for Reference and

14 C. Shi et al. Crit Rev Microbiol, Early Online: 1–17

Cri

tical

Rev

iew

s in

Mic

robi

olog

y D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y JH

U J

ohn

Hop

kins

Uni

vers

ity o

n 04

/13/

14Fo

r pe

rson

al u

se o

nly.

Page 15: Molecular methods for serovar determination of Salmonella

Research on Salmonella, Geneva, Switzerland. Available from:http://www.pasteur.fr/ip/portal/action/WebdriveActionEvent/oid/01s-000036-089 [last accessed 17 Jul 2013].

Grocery Manufacturers Associatio GMA. (2010). Capturing recall costs.Available at: http://www.ey.com/Publication/vwLUAssets/Capturing_Recall_Costs/$FILE/Capturing_recall_costs.pdf. Accessed on May 5,2013.

Guerra B, Laconcha I, Soto SM, et al. (2000). Molecular characterisationof emergent multiresistant Salmonella enterica serotype [4,5,12:i:-]organisms causing human salmonellosis. FEMS Microbiol Lett 190:341–7.

Guibourdenche M, Roggentin P, Mikoleit M, et al. (2010). Supplement2003-2007 (No. 47) to the White-Kauffmann-Le Minor scheme. ResMicrobiol 161: 26–9.

Hadjinicolaou AV, Demetriou VL, Emmanuel MA, et al. (2009).Molecular beacon-based real-time PCR detection of primary isolatesof Salmonella Typhimurium and Salmonella Enteritidis in environ-mental and clinical samples. BMC Microbiol 9:97.

Harbottle H, White DG, McDermott PF, et al. (2006). Comparison ofmultilocus sequence typing, pulsed-field gel electrophoresis, andantimicrobial susceptibility typing for characterization of Salmonellaenterica serotype Newport isolates. J Clin Microbiol 44: 2449–57.

Hartmann FA, West SE. (1997). Utilization of both phenotypic andmolecular analyses to investigate an outbreak of multidrug-resistantSalmonella Anatum in horses. Can J Vet Res 61:173–81.

Hauser E, Junker E, Helmuth R, Malorny B. (2011). Different mutationsin the oafA gene lead to loss of O5-antigen expression in Salmonellaenterica serovar Typhimurium. J Appl Microbiol 110:248–53.

Healy M, Huong J, Bittner T, et al. (2005). Microbial DNA typing byautomated repetitive-sequence-based PCR. J Clin Microbiol 43:199–207.

Heisig P, Kratz B, Halle E, et al. (1995). Identification of DNA gyraseA mutations in ciprofloxacin-resistant isolates of Salmonellatyphimurium from men and cattle in Germany. Microb Drug Resist1:211–8.

Herrera-Leon S, McQuiston JR, Usera MA, et al. (2004). Multiplex PCRfor distinguishing the most common phase-1 flagellar antigens ofSalmonella spp. J Clin Microbiol 42:2581–6.

Herrera-Leon S, Ramiro R, Arroyo M, et al. (2007). Blind comparison oftraditional serotyping with three multiplex PCRs for the identificationof Salmonella serotypes. Res Microbiol 158:122–7.

Hirose K, Itoh KI, Nakajima H, et al. (2002). Selective amplification oftyv (rfbE), prt (rfbS), viaB, and fliC genes by multiplex PCR foridentification of Salmonella enterica serovars Typhi and Paratyphi A.J Clin Microbiol 40:633–6.

Hoelzer K, Soyer Y, Rodriguez-Rivera LD, et al. (2010). The prevalenceof multidrug resistance is higher among bovine than humanSalmonella enterica serotype Newport, Typhimurium, and 4,5,12:i:-isolates in the United States but differs by serotype and geographicregion. Appl Environ Microbiol 76:5947–59.

Hoelzer Karin, Moreno Switt AI, Wiedmann M. (2011). Animal contactas a source of human non-typhoidal salmonellosis. Vet Res 42:34.

Holmberg K, Feroze F. (1996). Evaluation of an optimized system forrandom amplified polymorphic DNA (RAPD)-analysis for genotypicmapping of Candida albicans strains. J Clin Lab Anal 10:59–69.

Hong Y, Liu T, Hofacre C, et al. (2003). A restriction fragment lengthpolymorphism-based polymerase chain reaction as an alternative toserotyping for identifying Salmonella serotypes. Avian Dis 47:387–95.

Hong Y, Liu T, Lee MD, et al. (2008). Rapid screening of Salmonellaenterica serovars Enteritidis, Hadar, Heidelberg and Typhimuriumusing a serologically-correlative allelotyping PCR targeting the O andH antigen alleles. BMC Microbiol 8:178.

Huehn S, Malorny B. (2009). DNA microarray for molecular epidemi-ology of Salmonella. Methods Mol Biol 551:249–85.

Hulton CS, Higgins CF, Sharp PM. (1991). ERIC sequences: a novelfamily of repetitive elements in the genomes of Escherichia coli,Salmonella typhimurium and other enterobacteria. Mol Microbiol 5:825–34.

Ito Y, Iinuma Y, Baba H, et al. (2003). Evaluation of automatedribotyping system for characterization and identification of verocy-totoxin-producing Escherichia coli isolated in Japan. Jpn J Infect Dis56:200–4.

Jacobsen A, Hendriksen RS, Aaresturp FM, et al. (2011). TheSalmonella enterica Pan-genome. Microb Ecol 62:487–504.

Jansen AM, Hall LJ, Clare S, et al. (2011). A Salmonella typhimurium-Typhi genomic chimera: a model to study Vi polysaccharide capsulefunction in vivo. PLoS Pathog 7:e1002131.

Jansen R, Embden JD, Gaastra W, Schouls LM. (2002). Identification ofgenes that are associated with DNA repeats in prokaryotes. MolMicrobiol 43:1565–75.

Jeoffreys NJ, James GS, Chiew R, Gilbert GL. (2001). Practicalevaluation of molecular subtyping and phage typing in outbreaks ofinfection due to Salmonella enterica serotype typhimurium. Pathology33:66–72.

Jiang XM, Neal B, Santiago F, et al. (1991). Structure and sequence ofthe rfb (O antigen) gene cluster of Salmonella serovar Typhimurium(strain LT2). Mol Microbiol 5:695–713.

Johnson JR, Clabots C, Azar M, et al. (2001). Molecular analysis of ahospital cafeteria-associated salmonellosis outbreak using modifiedrepetitive element PCR fingerprinting. J Clin Microbiol 39:3452–60.

Joys TM. (1985). The covalent structure of the phase-1 flagellar filamentprotein of Salmonella Typhimurium and its comparison with otherflagellins. J Biol Chem 260:15758–61.

Kanto S, Okino H, Aizawa S, Yamaguchi S. (1991). Amino acidsresponsible for flagellar shape are distributed in terminal regions offlagellin. J Mol Biol 219: 471–80.

Kerouanton A, Marault M, Lailler R, et al. (2007). Pulsed-field gelelectrophoresis subtyping database for foodborne Salmonella entericaserotype discrimination. Foodborne Pathog Dis 4:293–303.

Kidgell C, Reichard U, Wain J, et al. (2002). Salmonella typhi, thecausative agent of typhoid fever, is approximately 50,000 years old.Infect Genet Evol 2: 39–45.

Kilger G, Grimont PA. (1993). Differentiation of Salmonella phase 1flagellar antigen types by restriction of the amplified fliC gene. J ClinMicrobiol 31:1108–10.

Kim S, Frye JG, Hu J, et al. (2006). Multiplex PCR-based method foridentification of common clinical serotypes of Salmonella entericasubsp. enterica. J Clin Microbiol 44:3608–15.

Laorden L, Herrera-Leon S, Martınez I, et al. (2010). Genetic evolutionof the Spanish multidrug-resistant Salmonella enterica 4,5,12:i:-monophasic variant. J Clin Microbiol 48:4563–6.

Leader BT, Frye JG, Hu J, et al. (2009). High-throughput moleculardetermination of Salmonella enterica serovars by use of multiplex PCRand capillary electrophoresis analysis. J Clin Microbiol 47:1290–9.

Lee SH, Jung BY, Rayamahji N, et al. (2009). A multiplex real-time PCRfor differential detection and quantification of Salmonella spp.,Salmonella enterica serovar Typhimurium and Enteritidis in meats.J Vet Sci 10:43–51.

Lin AW, Usera MA, Barrett TJ, Goldsby RA. (1996). Application ofrandom amplified polymorphic DNA analysis to differentiate strainsof Salmonella enteritidis. J Clin Microbiol 34:870–6.

Liu B, Zhou X, Zhang L, et al. (2012). Development of a novel multiplexPCR assay for the identification of Salmonella enterica Typhimuriumand Enteritidis. Food Control 27:87–93.

Liu F, Barrangou R, Gerner-Smidt P, et al. (2011). Novel virulence geneand clustered regularly interspaced short palindromic repeat(CRISPR) multilocus sequence typing scheme for subtyping of themajor serovars of Salmonella enterica subsp. enterica. Appl EnvironMicrobiol 77:1946–56.

Liu WB, Chen J, Huang YY, et al. (2010). Serotype, genotype, andantimicrobial susceptibility profiles of Salmonella from chicken farmsin Shanghai. J Food Prot 73:562–7.

Loharikar A, Briere E, Schwensohn C, et al. (2012). Four multistateoutbreaks of human Salmonella infections associated with livepoultry contact, United States, 2009. Zoonoses Public Health 59:347–54.

Luk JM, Kongmuang U, Reeves PR, Lindberg AA. (1993). Selectiveamplification of abequose and paratose synthase genes (rfb) bypolymerase chain reaction for identification of Salmonella majorserogroups (A, B, C2, and D). J Clin Microbiol 31:2118–23.

MacLean D, Jones JD, Studholme DJ. (2009). Application of ‘‘next-generation’’ sequencing technologies to microbial genetics. Nat RevMicrobiol 7:287–96.

Malorny B, Bunge C, Guerra B, et al. (2007). Molecular characterisationof Salmonella strains by an oligonucleotide multiprobe microarray.Mol Cel Probes 21:56–65.

Martin B, Humbert O, Camara M, et al. (1992). A highly conservedrepeated DNA element located in the chromosome of Streptococcuspneumoniae. Nucleic Acids Res 20:3479–83.

DOI: 10.3109/1040841X.2013.837862 Molecular methods for serovar determination of Salmonella 15

Cri

tical

Rev

iew

s in

Mic

robi

olog

y D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y JH

U J

ohn

Hop

kins

Uni

vers

ity o

n 04

/13/

14Fo

r pe

rson

al u

se o

nly.

Page 16: Molecular methods for serovar determination of Salmonella

Masten BJ, Joys TM. (1993). Molecular analyses of the Salmonellag . . . flagellar antigen complex. J Bacteriol 175:5359–65.

McQuiston JR, Herrera-Leon S, Wertheim BC, et al. (2008). Molecularphylogeny of the salmonellae: relationships among Salmonella speciesand subspecies determined from four housekeeping genes andevidence of lateral gene transfer events. J Bacteriol 190:7060–7.

McQuiston JR, Parrenas R, Ortiz-Rivera M, et al. (2004). Sequencingand comparative analysis of flagellin genes fliC, fljB, and flpA fromSalmonella. J Clin Microbiol 42:1923–32.

McQuiston John R, Waters RJ, Dinsmore BA, et al. (2011). Moleculardetermination of H antigens of Salmonella by use of a microsphere-based liquid array. J Clin Microbiol 49:565–73.

Mertes F, Biens K, Lehrach H, et al. (2010). High-throughput universalprobe Salmonella serotyping (UPSS) by nanoPCR. J MicrobiolMethods 83:217–23.

Millemann Y, Lesage-Descauses, MC, Lafont JP, Chaslus-Dancla E.(1996). Comparison of random amplified polymorphic DNA analysisand enterobacterial repetitive intergenic consensus-PCR for epidemio-logical studies of Salmonella. FEMS Immunol Med Microbiol 14:129–34.

Monack DM. (2012). Salmonella persistence and transmission strategies.Curr Opin Microbiol 15:100–7.

Moreno Switt AI, den Bakker HC, Cummings CA, et al. (2012).Identification and characterization of novel Salmonella mobileelements involved in the dissemination of genes linked to virulenceand transmission. PloS one 7:e41247.

Mortimer CK, Peters TM, Gharbia SE, et al. (2004). Towards thedevelopment of a DNA-sequence based approach to serotyping ofSalmonella enterica. BMC Microbiol 4:31.

Munoz N, Diaz-Osorio M, Moreno J, et al. (2010). Development andevaluation of a multiplex real-time polymerase chain reactionprocedure to clinically type prevalent Salmonella enterica serovars.J Mol Diagn 12:220–5.

Nair S, Schreiber E, Thong KL, et al. (2000). Genotypic characterizationof Salmonella typhi by amplified fragment length polymorphismfingerprinting provides increased discrimination as compared topulsed-field gel electrophoresis and ribotyping. J Microbiol Meth41:35–43.

Nair S, Lin TK, Pang T, Altwegg M. (2002). Characterization ofSalmonella serovars by PCR-single-strand conformation polymorph-ism analysis. J Clin Microbiol 40:2346–51.

Nde CW, Sherwood JS, Doetkott C, Logue CM. (2006). Prevalence andmolecular profiles of Salmonella collected at a commercial turkeyprocessing plant. J Food Prot 69:1794–801.

O’Regan E, McCabe E, Burgess C, et al. (2008). Developmentof a real-time multiplex PCR assay for the detection ofmultiple Salmonella serotypes in chicken samples. BMC Microbiol8:156.

Octavia S, Lan R. (2006). Frequent recombination and low level ofclonality within Salmonella enterica subspecies I. Microbiology 152:1099–108.

Olaimat AN, Holley RA. (2012). Factors influencing the microbial safetyof fresh produce: a review. Food Microbiol 32:1–19.

Old DC, Rankin SC, Crichton PB. (1999). Assessment of strainrelatedness among Salmonella serotypes Salinatis, Duisburg, andSandiego by biotyping, ribotyping, IS200 fingerprinting, andpulsed-field gel electrophoresis. J Clin Microbiol 37:1687–92.

Olive DM, Bean P. (1999). Principles and applications of methods forDNA-based typing of microbial organisms. J Clin Microbiol 37:1661–9.

Oscar TP. (1998). Identification and characterization of Salmonellaisolates by automated ribotyping. J Food Prot 61:519–24.

Peters TM, Threlfall EJ. (2001). Single-enzyme amplified fragmentlength polymorphism and its applicability for Salmonella epidemi-ology. Syst Appl Microbiol 24:400–4.

Peterson G, Gerdes B, Berges J, et al. (2010). Development ofmicroarray and multiplex polymerase chain reaction assays foridentification of serovars and virulence genes in Salmonella entericaof human or animal origin. J Vet Diagn Invest 22:559–69.

Porwollik S, Boyd EF, Choy C, et al. (2004). Characterization ofSalmonella enterica subspecies I genovars by use of microarrays.J Bacteriol 186:5883–98.

Porwollik S, Wong RM, McClelland M. (2002). Evolutionary genomicsof Salmonella: gene acquisitions revealed by microarray analysis. ProcNatl Acad Sci USA 99:8956–61.

Rajtak U, Leonard N, Bolton D, Fanning S. (2011). A real-timemultiplex SYBR Green I polymerase chain reaction assay for rapidscreening of Salmonella serotypes prevalent in the European Union.Foodborne Pathog Dis 8:769–80.

Ranieri ML, Shi C, Moreno Switt AI, et al. (2013). Salmonellaserovar prediction: comparison of typing methods with a newprocedure based on sequence characterization. J Clin Microbiol 51:1786-97.

Rasschaert G, Houf K, Imberechts H, et al. (2005). Comparison offive repetitive-sequence-based PCR typing methods for moleculardiscrimination of Salmonella enterica isolates. J Clin Microbiol 43:3615–23.

Reeves PR, Hobbs M, Valvano MA, et al. (1996). Bacterial polysac-charide synthesis and gene nomenclature. Trends Microbiol 4:495–503.

Ribot EM, Fair MA, Gautom R, et al. (2006). Standardization of pulsed-field gel electrophoresis protocols for the subtyping of Escherichiacoli O157:H7, Salmonella, and Shigella for PulseNet. FoodbornePathog Dis 3:59–67.

Ridley AM. (1998). Genomic fingerprinting by application of rep-PCR.Methods Mol Med 15:103–15.

Rizzi V, Migliorati G, Acciari V, et al. (2005). Surveillance system andrapid tracing of primary sources in food-borne outbreaks bySalmonella spp. Part II: molecular characterisation of some strainsof Salmonella enterica serovars Enteritidis and Typhimurium. Vet Ital41:265–79.

Rodriguez A, Pangloli P, Richards HA, et al. (2006). Prevalence ofSalmonella in diverse environmental farm samples. J Food Prot 69:2576–80.

Saarinen M, Ekman P, Ikeda M, et al. (2002). Invasion of Salmonellainto human intestinal epithelial cells is modulated by HLA-B27.Rheumatology 41:651–7.

Samuel G, Reeves P. (2003). Biosynthesis of O-antigens: genes andpathways involved in nucleotide sugar precursor synthesis andO-antigen assembly. Carbohydr Res 338:2503–19.

Sangal V, Harbottle H, Mazzoni CJ, et al. (2010). Evolution andpopulation structure of Salmonella enterica serovar Newport.J Bacteriol 192:6465–76.

Savelkoul P H, Aarts HJ, De Haas J, et al. (1999). Amplified-fragmentlength polymorphism analysis: the state of an art. J Clin Microbiol 37:3083–91.

Scallan E, Hoekstra RM, Angulo FJ, et al. (2011). Foodborne illnessacquired in the United States–major pathogens. Emer Infect Dis 17:7–15.

Schnaitman C A, Klena JD. (1993). Genetics of lipopolysaccharidebiosynthesis in enteric bacteria. Microbiol Rev 57:655–82.

Schrader KN, Fernandez-Castro A, Cheung WK, et al. (2008).Evaluation of commercial antisera for Salmonella serotyping. J ClinMicrobiol 46:685–8.

Schwartz DC, Cantor CR. (1984). Separation of yeast chromosome-sizedDNAs by pulsed field gradient gel electrophoresis. Cell 37:67–75.

Shangkuan YH, Lin HC. (1998). Application of random amplifiedpolymorphic DNA analysis to differentiate strains of SalmonellaTyphi and other Salmonella species. J Appl Microbiol 85:693–702.

Shima K, Terajima J, Sato T, et al. (2004). Development of a PCR-restriction fragment length polymorphism assay for the epidemio-logical analysis of Shiga toxin-producing Escherichia coli. J ClinMicrobiol 42:5205–13.

Silverman M, Zieg J, Hilmen M, Simon M. (1979). Phase variation inSalmonella: genetic analysis of a recombinational switch. Proc NatlAcad Sci USA 76:391–5.

Singh A, Goering RV, Simjee S, et al. (2006). Application of moleculartechniques to the study of hospital infection. Clin Microbiol Rev 19:512–30.

Smith NH, Selander RK. (1990). Sequence invariance of the antigen-coding central region of the phase 1 flagellar filament gene (fliC)among strains of Salmonella typhimurium. J Bacteriol 172:603–9.

Smith NH, Selander RK. (1991). Molecular genetic basis for complexflagellar antigen expression in a triphasic serovar of Salmonella. ProcNatl Acad Sci USA 88:956–60.

Smith SI, Fowora MA, Goodluck HA, et al. (2011). Molecular typing ofSalmonella spp isolated from food handlers and animals in Nigeria.Int J Mol Epidemiol Genet 2:73–7.

Soto SM, Guerra B, Gonzalez-Hevia MA, Mendoza MC. (1999).Potential of three-way randomly amplified polymorphic DNA analysis

16 C. Shi et al. Crit Rev Microbiol, Early Online: 1–17

Cri

tical

Rev

iew

s in

Mic

robi

olog

y D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y JH

U J

ohn

Hop

kins

Uni

vers

ity o

n 04

/13/

14Fo

r pe

rson

al u

se o

nly.

Page 17: Molecular methods for serovar determination of Salmonella

as a typing method for twelve Salmonella serotypes. Appl EnvironMicrobiol 65:4830–6.

Soyer Y, Moreno Switt A, Davis MA, et al. (2009). Salmonella entericaserotype 4,5,12:i:-, an emerging Salmonella serotype that representsmultiple distinct clones. J Clin Microbiol 47:3546–56.

Sukhnanand S, Alcaine S, Warnick LD, et al. (2005). DNA sequence-based subtyping and evolutionary analysis of selected Salmonellaenterica serotypes. J Clin Microbiol 43:3688–98.

Tauxe RV. (2002). Emerging foodborne pathogens. Int J Food Microbiol78:31–41.

Tennant SM, Diallo S, Levy H, et al. (2010). Identification by PCR ofnon-typhoidal Salmonella enterica serovars associated with invasiveinfections among febrile patients in Mali. PLoS Negl Trop Dis 4:e621.

The National Food Institute, D. T. U. (2011). Perspectives of a global,real – time microbiological genomic identification system -implica-tions for national and global detection and control of infectiousdisease. Available from: http://www.food.dtu.dk/�/media/Institutter/Foedevareinstituttet/Publikationer/Pub-2011/consensus report per-spectives of a global, real-time.ashx [last accessed 17 Jul 2013].

Tindall BJ, Grimont PA, Garrity GM, Euzeby JP. (2005). Nomenclatureand taxonomy of the genus Salmonella. Int J Syst Evol Microbiol 55:521–4.

Torpdahl M, Skov MN, Sandvang D, Baggesen DL. (2005). Genotypiccharacterization of Salmonella by multilocus sequence typing, pulsed-field gel electrophoresis and amplified fragment length polymorph-ism. J Microbiol Methods 63:173–84.

Touchon M, Rocha EP. (2010). The small, slow and specialized CRISPRand anti-CRISPR of Escherichia and Salmonella. PloS one 5:e11126.

Van Lith LA, Aarts HJ. (1994). Polymerase chain reaction identificationof Salmonella serotypes. Lett Appl Microbiol 19:273–6.

Vanegas RA, Joys TM. (1995). Molecular analyses of the phase-2antigen complex 1,2,.. of Salmonella spp. J Bacteriol 177:3863–4.

Verma NK, Quigley NB, Reeves PR. (1988). O-antigen variation inSalmonella spp.: rfb gene clusters of three strains. J Bacteriol 170:103–7.

Versalovic J, Koeuth T, Lupski JR. (1991). Distribution of repetitiveDNA sequences in eubacteria and application to fingerprinting ofbacterial genomes. Nucleic Acids Res 19:6823–31.

Wachsmuth IK, Kiehlbauch JA, Bopp CA, et al. (1991). The use ofplasmid profiles and nucleic acid probes in epidemiologic investiga-tions of foodborne, diarrheal diseases. Int J Food Microbiol 12:77–89.

Wassenaar TM, Newell DG. (2000). Genotyping of Campylobacter spp.Appl Environ Microbiol 66:1–9.

Wattiau P, Boland C, Bertrand S. (2011). Methodologies for Salmonellaenterica subsp. enterica subtyping: gold standards and alternatives.Appl Environ Microbiol 77:7877–85.

Wattiau P, Van Hessche M, Schlicker C, et al. (2008). Comparison ofclassical serotyping and PremiTest assay for routine identificationof common Salmonella enterica serovars. J Clin Microbiol 46:4037–40.

Weigel RM, Qiao B, Teferedegne B, et al. (2004). Comparison of pulsedfield gel electrophoresis and repetitive sequence polymerase chainreaction as genotyping methods for detection of genetic diversity andinferring transmission of Salmonella. Vet Microbiol 100:205–17.

Wiedmann M, Nightingale K. (2009). DNA-based subtypingmethods facilitate identification of foodborne pathogens. Food Tech63:44–9.

Wiedmann M. (2002). Subtyping of bacterial foodborne pathogens.Nutrition Rev 60:201–8.

William JG, Kubelik AR, Livak KJ, et al. (1990). DNA polymorphismsamplified by arbitrary primers are useful as genetic markers. NucleicAcids Res 18:6531–5.

Winokur PL. (2003). Molecular epidemiological techniques forSalmonella strain discrimination. Front Biosci 8:c14–24.

Wise MG, Siragusa GR, Plumblee J, et al. (2009). Predicting Salmonellaenterica serotypes by repetitive sequence-based PCR. J MicrobiolMethods 76:18–24.

Yamamoto S, Kutsukake K. (2006). FljA-mediated posttranscriptionalcontrol of phase 1 flagellin expression in flagellar phasevariation of Salmonella enterica serovar Typhimurium. J Bacteriol188:958–67.

Yoshida C, Franklin K, Konczy P, et al. (2007). Methodologies towardsthe development of an oligonucleotide microarray for determination ofSalmonella serotypes. J Microbiol Methods 70:261–71.

Zerbino DR, Birney E. (2008). Velvet: algorithms for de novo short readassembly using de Bruijn graphs. Genome Res 18:821–9.

Zou W, Lin WJ, Foley SL, et al. (2010). Evaluation of pulsed-field gelelectrophoresis profiles for identification of Salmonella serotypes.J Clin Microbiol 48:3122–6.

Zou W, Lin WJ, Hise KB, et al. (2012). Prediction system forrapid identification of Salmonella serotypes based on pulsed-field gel electrophoresis fingerprints. J Clin Microbiol 50:1524–32.

Supplementary material available online

Supplementary Tables 1 and 2

DOI: 10.3109/1040841X.2013.837862 Molecular methods for serovar determination of Salmonella 17

Cri

tical

Rev

iew

s in

Mic

robi

olog

y D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y JH

U J

ohn

Hop

kins

Uni

vers

ity o

n 04

/13/

14Fo

r pe

rson

al u

se o

nly.