MiniReview

Bacterial virulence: can we draw the line?

Trudy M. Wassenaar *, Wim GaastraDivision of Bacteriology, Department of Infectious Diseases and Immunology, School of Veterinary Medicine, University of Utrecht, P.O. Box 80.165,

3508 TD Utrecht, The Netherlands

Received 22 February 2001; received in revised form 14 May 2001; accepted 14 May 2001

First published online 14 June 2001

Abstract

The molecular approach to microbial pathogenesis has resulted in an impressive amount of data on bacterial virulence genes. Bacterialgenome sequences rapidly add candidate virulence genes to electronic databases. The interpretation of this overwhelming information isobscured because every gene involved in pathogenicity is called a virulence gene, regardless of its function in the complex process ofvirulence. This review summarizes the changing concept of bacterial virulence and the detection and identification strategies followed torecognize virulence genes. A refined definition of virulence genes is proposed in which the function of the gene in the virulence process isincorporated. We propose to include the life-style of bacteria in the assessment of their putative virulence genes. A universal nomenclature inanalogy to the EC enzyme numbering system is proposed. These recommendations would lead to a better insight into bacterial virulence anda more precise annotation of (putative) virulence genes, which would enable more efficient use of electronic databases. ß 2001 Federationof European Microbiological Societies. Published by Elsevier Science B.V. All rights reserved.

Keywords: Bacterial virulence; Virulence gene nomenclature; Standardization; Comparative genomics

1. The concept of bacterial virulence

In 1890 Robert Koch postulated guidelines to establisha standard for evidence of causation in infectious disease.His postulates became the gold standard to de¢ne micro-bial virulence for over 100 years, despite limitations totheir experimental applications for a number of microor-ganisms. Revisions of Koch's postulates were introducedto encompass those limitations in which immunologicaland/or epidemiological proof of causation was added(see [1] for a recent review). With the development ofmolecular biological techniques, it became possible toidentify the genes encoding those factors responsible forvirulence. This resulted in molecular microbiology, inwhich the role and function of speci¢c genes (and thefactors they encode) in (bacterial) virulence was the subjectof investigation.

The quest for virulence genes evolved together with thetechnical development of molecular biology and geneticmodi¢cation of microorganisms. In the beginning of mo-lecular microbiology, genes were identi¢ed that encodedvirulence factors of known reputation and these wereused as probes to ¢nd analogs in other organisms. Thefunction of individual genes and the factors they encodein virulence could be determined by random and targetedmutagenesis. Later, identi¢ed genes with unknown func-tion were tested for their role in virulence. At present thechallenge is to ¢lter out virulence genes from completebacterial genomes, which can now be sequenced fasterthan the time needed to establish the role of one singlegene in virulence. To give such evidence, a molecular formof Koch's postulates was de¢ned [2] : (i) the phenotype orproperty under investigation should be associated withpathogenic members of a genus or pathogenic strains ofa species; (ii) speci¢c inactivation of the gene(s) associatedwith the suspected virulence trait should lead to a measur-able loss in pathogenicity or virulence; and (iii) reversionor allelic replacement of the mutated gene should lead torestoration of pathogenicity. An alternative postulate wasadded in case genetic manipulation was not possible : (iv)the induction of speci¢c antibodies to a de¢ned gene prod-

0378-1097 / 01 / $20.00 ß 2001 Federation of European Microbiological Societies. Published by Elsevier Science B.V. All rights reserved.PII: S 0 3 7 8 - 1 0 9 7 ( 0 1 ) 0 0 2 4 1 - 5

* Corresponding author. Present address: Molecular Microbiology andGenomics Consultants, Tannenstrasse 7, D-55576 Zotzenheim, Germany.Tel. : +49 (6701) 8531; Fax: +49 (6701) 901803;E-mail : wassenaar_t@yahoo.co.ukE-mail : w.gaastra@vet.uu.nl

FEMSLE 9995 29-6-01

FEMS Microbiology Letters 201 (2001) 1^7

www.fems-microbiology.org

uct should neutralize pathogenicity. This addition is some-times taken alone in that when antibodies against a certainmolecule protect an animal from disease, this is acceptedas su¤cient to call such a factor a virulence factor.

2. Detection and identi¢cation of virulence genes

The molecular approach to study bacterial virulence hasresulted in a number of techniques that are based on dif-ferent principles (Fig. 1). Firstly, genetic methods are usedto obtain phenotypic evidence for a role in virulence. Theapproaches of gene inactivation and gene complementa-tion are based on the second and third molecular postu-lates given above. Both principles are applicable to one-step mechanisms. For example, when the production of abacterial toxin results in cellular damage, mutants withinactivated toxin genes no longer cause damage, and ex-pression of these genes in another organism introducestoxinogenic properties. However, pathogenic bacteria canemploy such complex processes that the experimental re-sults of inactivation/complementation of these studies maybe hard to interpret. Even under simpli¢ed in vitro con-ditions, a presumably straightforward process such as bac-terial invasion is driven and regulated by multiple genesand gene loci, which work in concert or are complemen-tary. Inactivation of one link of the chain may eliminateinvasiveness, but complementation in a heterologous sys-tem may require several genetic loci. On the other hand,inactivation of a factor may be overcome by alternativefactors so that loss of virulence is not observed, but com-plementation in a di¡erent genetic environment may havestrong phenotypic e¡ects.

Virulence factors are often immunogenic, thus when ac-quired immunity protects against disease, protective anti-bodies are frequently directed against virulence gene prod-ucts (see Fig. 1). But the argument cannot be reversed: notall antigens are virulence factors. Some are structural com-

ponents of the bacterial cell, and although these can havevirulent properties (for instance lipopolysaccharide (LPS)),their function for the bacteria is primarily structural.Chaperonins can give rise to antibodies, but their functionin protein folding and stress repair is of greater signi¢-cance than their virulence properties. On the other hand,virulence factor candidates are sometimes discarded be-cause they are not immunogenic during infection. Thevalue of such reasoning remains disputable, for instancefor intracellular pathogens.

As indicated in Fig. 1, a third approach to identifynovel virulence genes comes from the observation thatmany virulence genes display antigenic polymorphisms,presumably to evade the selection pressure of the hostimmune system [4]. The correlation between polymor-phism and virulence is so strong that the presence ofmechanisms to produce polymorphic factors is indirectevidence for a role of that factor in virulence. With thehigh throughput of sequencing data, it becomes possible toidentify putative virulence properties for genes based onthe polymorphic nature of their predicted translationproducts [5]. Although most known examples of polymor-phisms in bacteria are virulence genes, it is likely that thismechanism is employed for di¡erent functions also, forinstance for adaptation to environmental conditions, asmay be the case for contingency genes of certain bacterialspecies.

In addition to these approaches, several techniques havebeen developed to identify and characterize bacterial genesthat are induced during in vivo infection and, potentially,may play a role in pathogenesis [6,7]. The (transcriptional)regulation of these genes may be a re£ection of the newenvironment that bacteria have to adapt to after entering ahost. Such gene products may not be involved in patho-genesis directly although they are a requirement for sur-vival in the host.

A relatively new approach is to deduce informationfrom comparative genetics [8]. The annotation of newlysequenced genes that are now rapidly generated by high-throughput genome sequencing is based on sequence sim-ilarity. To apply this to virulence genes is risky for tworeasons. First, an acceptable level of sequence conserva-tion is interpreted as conservation of function. However,genes may have a niche-adapted function in a particularorganism, and this may be re£ected in the role in viru-lence. Even a high degree of genetic conservation mustbe experimentally tested to demonstrate functional conser-vation. Second, sequence similarity searches have resultedin a new phenomenon for which the term `putativism'would be appropriate. Sequence similarity to an experi-mentally de¢ned virulence gene results in an entry of a`putative virulence gene', and every next signi¢cant simi-larity to that gene will pass on this term to new genes,which may have little identity with the virulence genethat initiated the linkage. In a more severe case, the orig-inal entry of a `virulence factor' in public databases may

Fig. 1. Di¡erent approaches to identify virulence genes and virulence-as-sociated genes.

FEMSLE 9995 29-6-01

T.M. Wassenaar, W. Gaastra / FEMS Microbiology Letters 201 (2001) 1^72

not be backed up with published experimental evidence.For example, the database entry of Salmonella typhimuri-um `virulence factor MviN' has resulted in the identi¢ca-tion of homologs of this gene in many species, such asCampylobacter jejuni, Escherichia coli K12, Neisseria men-ingitidis, Streptomyces coelicolor, Deinococcus radians,Thermotoga maritima, Helicobacter pylori, Rickettsia pro-wazekii, Treponema pallidum, and many others. It is hardto imagine that this gene encodes a virulence factor whenit is present in organisms with such diverse life-styles,some of which are not even pathogens. In all cases theannotation of the homolog mentions `(putative) virulencegene/factor' although no published record could be iden-ti¢ed describing experimental evidence for a role of MviNin virulence in S. typhimurium. Mis-annotation based on`putativism' is quite common, and contaminates electronicdatabases with misinformation.

Two other pitfalls of comparative genetics are facingopposite directions. On the one hand di¡erent genes thatshare no sequence homology can have identical functions,as is demonstrated for actA of Listeria monocytogenes andicsA in Shigella £exneri whose gene products recruit hostcell actin (discussed in [3]). On the other hand sequencehomology does not always predict function, or functionaldomains may not be conserved, as illustrated by compar-ison of calmodulin genes in Saccharomyces cerevisiae andvertebrates [9].

Ideally, for the identi¢cation of virulence genes, severalapproaches should lead to the same gene or set of genes,and a virulence gene should have more than one of thecharacteristics listed in Fig. 1. Even then, a controversyremains whether a gene is interpreted as being a house-

keeping gene or a virulence gene. This situation prompts amore restricted de¢nition of virulence genes, and moreprecise terms than those currently used.

3. How to de¢ne virulence genes

The de¢nitions of bacterial virulence, and virulence fac-tors, that have been in use over time have been summa-rized elsewhere [10]. The number of genes nominated `vir-ulence genes' depends on the de¢nition used, as illustratedin Fig. 2. Most investigators draw a line somewhere be-tween circles 2 and 3 of Fig. 2, since genes involved inbasic cellular metabolism (`housekeeping genes') are notregarded as virulence genes. However, housekeeping genescan be screened as virulence genes, when their inactivationresults in attenuation of virulence. This will be recognizedfor those genes for which a function in cellular metabolismis known, such as aroA. When the function of a geneproduct is not known, attenuation after inactivation re-sults in the application of the term `(putative) virulencegene'. This problem is enlarged now that a total-genomemutagenesis approach is followed for pathogens whosecomplete genome sequence is available but whose viru-lence genes are still a mystery: every predicted (putative)open reading frame can now be mutated (providing thegenetic tools are available) and tested for attenuation. Inthe de¢nition of virulence genes one could include therequirement that the gene should be absent in non-patho-gens (or non-pathogenic strains). If we accept this, LPScannot be a virulence factor, since LPS genes are presentin both pathogens and non-pathogens.

Fig. 2. Depending on the de¢nition of virulence, more or fewer genes are called `virulence genes'. The number of virulence genes and virulence-associ-ated genes included in a given de¢nition are represented by concentric circles. In collection 1, only those virulence factors are included that are directlyinvolved in causing disease (`true virulence genes'). The addition of `virulence-associated genes' increases the number of identi¢ed virulence genes andthus the size of circle 2. The gene pool identi¢ed by inactivation and phenotypic characterization (see Fig. 1) includes all genes that lead to an attenu-ated phenotype as `virulence life-style genes' (circle 3). The remaining genes are other housekeeping genes, structural genes, and essential genes. The bor-der between collection 3 and the remainder cannot be exactly de¢ned.

FEMSLE 9995 29-6-01

T.M. Wassenaar, W. Gaastra / FEMS Microbiology Letters 201 (2001) 1^7 3

In order to exclude housekeeping genes from the set ofvirulence genes, the requisite is often added to Falkow'smolecular postulates that virulence genes should not beexpressed outside the host. This would again excludemany well-characterized and generally accepted virulencegenes, for which the genes encoding LPS-producing en-zymes are also an example: they are expressed undermost if not all circumstances. Moreover, the lack of ex-pression outside the host may be a re£ection of the appliedculture conditions. In conclusion, the border between vir-ulence-associated genes and housekeeping genes remainspoorly de¢ned.

A solution to this problem is to distinguish genes di-rectly involved in the pathogenesis of an organism fromthose genes that are required for a pathogenic life-style.The properties of pathogenic bacteria that are required tosurvive, multiply, and cause damage to a host are: thecapacity to compete with other bacteria in the host ; togain a foothold within a speci¢c host ; to avoid normalhost defence mechanisms; to multiply once established;and in the course of this process to produce damage tothe host. This would be a de¢nition of a pathogenic life-style. Virulence life-style factors would be all factors thatare essential for this life-style; their genes could collec-tively be called virulence life-style genes. They make upa Pandora's box containing all genes related to virulenceof a particular pathogen (Fig. 3). A subset of virulencelife-style genes, in which virulence-associated genes are

excluded, are the true virulence genes. They must be ab-sent in non-pathogenic bacteria; their gene products mustbe involved in interactions with the host, and must bedirectly responsible for the pathological damage duringinfection.

The proposed de¢nition of true virulence genes excludesthose genes that are involved in survival and multiplica-tion in the host, and genes involved in expression, process-ing, or secretion of virulence factors; these genes could bede¢ned in subclasses of virulence-associated genes as indi-cated in Table 1. All virulence life-style genes can either bestructural genes (directly encoding the virulence life-stylefactors), or encode the enzymes to produce such factors.

Every pathogen possesses a diverse and unique set ofgenes to allow it to cause disease. Focusing on the indi-vidual role of each of these genes is essential to understandthe complex mechanisms behind disease. In addition therole of the host and his immune status in the outcome ofdisease need to be considered. Each pathogen has evolveda pathogenic strategy combining one or more mechanismsthat operate in a concerted manner. Understanding thelife-style of pathogenic and other bacteria can help toidentify the genes relevant to pathogenicity.

4. Bacterial life-styles

Our focus on pathogenicity and virulence factors some-

Fig. 3. The multi-compartment virulence Pandora's box. This box contains all virulence life-style genes. They consist of true virulent genes and viru-lence-associated genes. The true virulent genes are directly responsible for pathological damage and are absent in non-pathogens. Virulence-associatedgenes can be genes whose products process virulent factors (by post-translational modi¢cation, folding, secretion, etc.) ; genes encoding auxiliary viru-lence factors (that are needed for virulent factors to be active); genes that regulate expression of virulence genes; housekeeping genes that produce en-zymes required for those metabolic processes required for the pathogenic life-style of the organism. More classes are listed in Table 1. Virulence-associ-ated genes can be present in non-pathogens that live in association with a host (commensals, opportunists).

FEMSLE 9995 29-6-01

T.M. Wassenaar, W. Gaastra / FEMS Microbiology Letters 201 (2001) 1^74

Tab

le1

De¢

niti

ons

for

subc

lass

esof

viru

lenc

elif

e-st

yle

gene

s

Fir

stdi

git:

life-

styl

eof

orga

nism

Seco

nddi

git:

gene

clas

sD

e¢ni

tion

Exa

mpl

esof

this

clas

sE

vide

nce,

com

men

tsT

hird

and

furt

her

digi

ts:

subc

lass

es

PA

:vi

rule

nce

gene

sfr

omba

cter

iath

atar

eex

clus

ivel

ypa

thog

enic

1.T

rue

viru

lenc

ege

nes

The

irge

nepr

oduc

tsar

edi

rect

lyin

volv

edin

inte

ract

ions

wit

hth

eho

stan

dar

edi

rect

lyre

spon

sibl

efo

rth

epa

thol

ogic

alda

mag

e.T

hese

gene

sar

eex

clus

ivel

yex

pres

sed

inpa

thog

ens.

Cho

lera

toxi

n,an

thra

xto

xin,

botu

linto

xin,

shig

ato

xin,

Bor

dete

llaad

enyl

ate

cycl

ase

toxi

n,et

c.

The

path

olog

ical

dam

age

isin

duce

dby

puri

¢ed

gene

prod

ucts

and

the

gene

isth

est

ruct

ural

gene

for

thes

epr

oduc

ts.

Subc

lass

esac

cord

ing

toge

nefa

mili

es,

for

inst

ance

,.1

:R

TX

toxi

ns,

.2:

ente

roto

xins

,et

c.

HS

:vi

rule

nce

gene

sfr

omba

cter

iadi

spla

ying

host

-de

pend

ent

path

ogen

icit

y

2.C

olon

izat

ion

gene

sT

heir

gene

prod

ucts

enab

leco

loni

zati

onof

aho

stan

dde

term

ine

the

loca

lizat

ion

ofth

ein

fect

ion.

Adh

esin

s,¢m

bria

e,in

tim

in,

inva

sins

.In

acti

vati

onw

illre

sult

inde

crea

sein

colo

niza

tion

pote

ntia

l.T

hefa

ctor

sm

ake

cont

act

atth

esi

teof

colo

niza

tion

.

.1:

adhe

sins

,.2

:in

tim

ins,

.3:

inva

sins

,.4

:ac

cess

ory

gene

sof

2.1^

2.3

(e.g

.¢m

bria

lsu

buni

ts)

3.D

efen

sesy

stem

evas

ion

gene

sT

heir

gene

prod

ucts

are

invo

lved

inev

asio

nof

the

host

imm

une

syst

em.

Imm

unog

lobu

lin-s

peci

¢cpr

otea

ses,

cyto

toxi

nsdi

rect

edag

ains

tim

mun

ece

lls,

surf

ace

laye

rs,

slim

epo

lysa

ccha

ride

.

The

role

ofth

ese

gene

sm

ust

bees

tabl

ishe

dfo

rea

chpa

thog

en.

Subc

lass

esac

cord

ing

toth

esp

eci¢

cfu

ncti

on

OP

:vi

rule

nce

gene

sfr

omop

port

unis

tic

path

ogen

s4.

Pro

cess

ing

viru

lenc

ege

nes

The

irge

nepr

oduc

tsar

ein

volv

edin

the

bios

ynth

esis

ofvi

rule

nce

life-

styl

efa

ctor

sby

enzy

mat

icpr

oces

sing

.

Spec

i¢c

prot

ease

s,m

ethy

lase

s,ch

aper

onin

s,gl

ycos

yltr

ansf

eras

es,

wit

hvi

rule

nce

life-

styl

ege

nes

asa

subs

trat

e.

The

enzy

mat

icac

tivi

tyof

the

gene

prod

uct

mus

tbe

prov

en.

Thi

sac

tivi

tym

ust

not

beso

lely

dire

cted

tow

ards

viru

lenc

elif

e-st

yle

fact

ors.

Subc

lass

esac

cord

ing

toth

ety

peof

proc

essi

ng,

e.g.

.1:

chap

eron

ins,

.2:

met

hyla

ses,

.3:

glyc

osyl

tran

sfer

ases

,et

c.5.

Secr

etor

yvi

rule

nce

gene

sT

heir

gene

prod

ucts

are

resp

onsi

ble

for

secr

etio

nof

viru

lenc

elif

e-st

yle

fact

ors.

Typ

eII

Ise

cret

ion

mac

hine

ry,

type

Ise

cret

ion

mac

hine

ry.

The

role

ofth

ege

nepr

oduc

tsin

secr

etio

nof

viru

lenc

elif

e-st

yle

fact

ors

mus

tbe

prov

en.

The

irac

tivi

tym

ayno

tbe

sole

lydi

rect

edto

war

dsth

ese

fact

ors.

Subc

lass

es:

.1:

type

Ise

cret

ion

mac

hine

ryge

nes,

.3:

type

III

secr

etio

nm

achi

nery

gene

s6.

Vir

ulen

ceho

usek

eepi

ngge

nes

The

irge

nepr

oduc

tspr

ovid

enu

trie

nts

duri

ngco

loni

zati

on,

impr

ove

com

peti

tion

wit

hot

her

mic

robe

s,or

prov

ide

the

prop

erm

icro

envi

ronm

ent.

Ure

ase,

cata

lase

,su

pero

xide

dism

utas

e,si

dero

phor

es,

prot

eina

sein

hibi

tors

.F

lage

llaco

uld

also

belo

ngto

this

clas

sal

thou

ghth

eyar

est

rict

lysp

eaki

ngst

ruct

ural

com

pone

nts

ofth

eor

gani

sm.

Inac

tiva

tion

will

resu

ltin

decr

ease

inco

loni

zati

onpo

tent

ial

alth

ough

adi

rect

role

inco

loni

zati

onor

imm

une

evas

ion

isab

sent

.T

hese

gene

sar

elik

ely

tobe

pres

ent

inno

n-pa

thog

ens

asw

ell.

Subc

lass

esac

cord

ing

tofu

ncti

on

7.R

egul

ator

yge

nes

The

irge

nepr

oduc

tsar

ein

volv

edin

regu

lati

onof

viru

lenc

elif

e-st

yle

gene

expr

essi

on.

Alt

erna

tive

sigm

afa

ctor

s,gl

obal

regu

lato

rs,

spec

i¢c

tran

scri

ptio

nac

tiva

tors

,re

gula

tors

ofph

ase

vari

atio

nby

gene

/pro

mot

erin

vers

ion.

The

role

ofth

ese

gene

sin

viru

lenc

em

ust

bees

tabl

ishe

dfo

rea

chpa

thog

en.

.1:

two-

com

pone

ntre

gula

tors

,.2

:gl

obal

regu

lato

rs,

.3:

alte

rnat

ive

sigm

afa

ctor

s,et

c.

Pro

pose

dcl

assi

¢cat

ion

and

num

beri

ngsy

stem

for

bact

eria

lge

nes

enco

ding

viru

lenc

efa

ctor

s.F

orsi

mpl

icit

yon

lyhu

man

path

ogen

sar

eco

nsid

ered

.T

he¢r

stdi

git

ofa

give

nnu

mbe

rw

ould

beP

A,

HS,

orO

P,

depe

ndin

gon

the

life-

styl

eof

the

orga

nism

.T

hese

cond

digi

tis

dete

rmin

edby

the

func

tion

ofth

ege

nein

viru

lenc

e.G

enes

belo

ngin

gto

clas

ses

1^3

are

true

viru

lenc

ege

nes;

thos

ebe

long

ing

tocl

asse

s4^

7ar

evi

rule

nce-

asso

ciat

edge

nes.

Thi

rdan

dfu

rthe

rdi

gits

re¢n

eth

esy

stem

.F

orin

stan

ce,

Vib

rio

chol

erae

ente

roto

xin

wou

ldbe

aP

A:1

.2fa

ctor

.U

ropa

thog

enic

E.

coli

min

or¢m

bria

lsu

buni

tw

ould

bean

OP

:2.4

fact

or.

The

num

beri

ngsy

stem

coul

dbe

re¢n

edw

ith

mor

edi

gits

topr

ovid

ea

shor

than

dfo

rea

chun

ique

viru

lenc

ege

nein

anal

ogy

toth

eE

Cen

zym

eno

men

clat

ure.

FEMSLE 9995 29-6-01

T.M. Wassenaar, W. Gaastra / FEMS Microbiology Letters 201 (2001) 1^7 5

times leads to the incorrect concept that pathogenic bac-teria exist to cause disease in their host. Like every organ-ism, pathogens have adapted to occupy an ecologicalniche. Their close association with a host causes damageto their host. Often this damage is `coincidental', but itmay even be bene¢cial to the survival or spreading of thepathogen (for example liberation of nutrients by cell dam-age, or enabling contagion of the next host by inducingcoughing or diarrhea). The degree of damage is dependenton the equilibrium that results from the interplay ofpathogen and host. The conditions that result in diseasecan vary between individuals, and between host species.Disease can be the result of the micro-organism beingthe `wrong' host, while it lives as a commensal in otherhosts. The distinction between `pathogen' and `non-patho-gen' is not sharp, and the border between `virulence genes'and all other genes is also fuzzy.

In the discussion of virulence gene de¢nition, it is im-portant to take the life-style of microorganisms into ac-count, with emphasis on the probability for that organismto cause disease in a host. Bacterial life-styles can be or-dered with an increasing probability to cause disease, vary-ing from extremophiles (cryophile/thermophile, halophile,etc.), to non-colonizing bacteria (soil bacteria, marine bac-teria, etc.), to commensal colonizers, to opportunisticpathogens and to exclusive pathogens. Evolutionary stepssuch as horizontal transfer of genetic information wouldmore likely have an e¡ect on virulence (and more likely be¢xed in the population) when the transfer occurs betweenbacteria with a common life-style, or at the most with life-styles ordered near to each other on the scale of increasingprobability of causing disease. True virulence genes wouldbe found in the class of exclusive pathogens only, butother virulence life-style genes may be present in oppor-tunistic pathogens and in other colonizers, since thesegenes are required for a life-style in close associationwith a host. Virulence life-style genes present in non-col-onizing bacteria are a contradictio in terminis and implythat the gene has a di¡erent function in that particularorganism.

Pathogens are constantly evolving, because the bacterialand the host population, as well as the ecological condi-tions that provide the interplay of both, undergo constantchanges. Pathogens emerge and lose signi¢cance over time.Emerging infectious diseases are most likely caused byorganisms that are already opportunistic or true patho-gens and that have acquired additional DNA elementsencoding a `true virulence determinant', e.g. toxin-convert-ing bacteriophage encoding cholera toxin or the Shigatoxins. Thus, a shift towards pathogenicity can be causedby changes in the bacteria, or, alternatively by a change inthe susceptible hosts, or in the success of bacterial survivaland contamination routes ex vivo. Bacterial factors of op-portunists that are directly responsible for damage of sus-ceptible hosts could be de¢ned as `opportunistic virulencefactors' to di¡erentiate them from true virulence factors, a

term reserved for those organisms with an exclusivelypathogenic life-style.

5. Virulence genes ^ what's in a name?

How important is it which genes we call virulence genesand how we further subdivide or classify this group? Afterall, the potency of the gene lies in its function, as dictatedby its sequence, not in its name. For that reason genes aresubmitted to electronic databases with a description. Sup-pose one would like to get an overview of our currentknowledge on bacterial virulence genes and virulence fac-tors. By entering the key words `virulence factor ANDbacteria' in PubMed one would get over 1000 hits, andthis number increases with time. Searching the Proteindatabase of PubMed with these key words decreases thenumber of hits to 580, and searching the Nucleotide data-base gives `only' 370 hits. Are these all genes encodingvirulence factors? And can their function be learnedfrom the annotation? Here are some examples to illustratehow genes are currently annotated.

b a `virulence factor' homolog MviB is described in Aqui-fex aeolicus, a thermophile.

b gene XF2420 from Xylella fastidiosa is described as:product = `virulence factor' without further evidencewhy this is so.

b gene b1121 from E. coli K12 is described as: product =`homolog of virulence factor' and its function = `puta-tive factor; Not classi¢ed'.

b An `outer membrane virulence protein' of V. cholerae isinvolved in the early steps of iron uptake. Its expressionis iron-regulated, but does that su¤ce to name this avirulence gene?

b A gene encoding a cAMP binding protein from Pseudo-monas aeruginosa is a probable DNA binding regulatorthat is required for production of exotoxin A and pro-tease. Should we call this a virulence factor of an op-portunistic pathogen?

PubMed databases are not generally used in the waydescribed above. Suppose a more practical scenario. Acosmid library of a pathogen of which virulence genesare not yet characterized was screened in an in vitro modelfor virulence. Positive clones were sequenced and the iden-ti¢ed open reading frames were compared with entries ofthe public databases. Would the scienti¢c insight increasewith the hits mentioned above?

These examples illustrate why we need a better annota-tion than the general term `virulence factor'.

6. Conclusions: where to draw the line

The handling of complex information requires simpli¢-

FEMSLE 9995 29-6-01

T.M. Wassenaar, W. Gaastra / FEMS Microbiology Letters 201 (2001) 1^76

cation by classi¢cation. Unfortunately in nature there isno black and white, only shades of gray. This applies tothe `pathogenicity' of microorganisms as well as to the`virulent' properties of their genes. If one can di¡erentiatewithin the collection of virulence genes, at least for elec-tronically stored annotation, the potential of database-generated research will increase. A re¢nement is requiredthat recognizes shades of gray. In this contribution it isproposed to de¢ne subclasses of virulence genes thatwould give weight to their function in pathogenicity. Ex-amples of such subclasses are given in Table 1. A nomen-clature could be developed in analogy to the EC enzymenumbering nomenclature, in which a code de¢nes the life-style of the organism, and the function of established vir-ulence genes. Such a code would simplify database entryand retrieval of information. Our current concept may notbe perfect, however a start must be made to reconsider thelabel `virulence' that is so eagerly attached to genes.

Comparative genomics should include multiple align-ment analysis of genes with signi¢cant similarity scores,to prove conservation of recognizable domains, before an-notation is accepted. A reference to an entry in the data-base of a homolog for which experimental evidence isavailable should be included. The challenge for genomesequence projects is to predict a function for as manygenes as possible. However, wrongly annotated genes areworse than `hypothetical protein' entries because annota-tion spreads through databases. When virulence genes aremore precisely annotated, this knowledge can be extrapo-lated and used for more exact entries of a newly sequencedgenes.

With the shades of gray in virulence and virulencegenes, the ¢nal conclusion would be that we cannotdraw a clear line between virulence genes and all others.The best we can do is to subdivide genes according to theirfunction. Until the role of a newly discovered `putativevirulence gene' is assessed for the organism (at the speciesor even subspecies level) predictions based on sequencesimilarity must be treated with a healthy amount of suspi-

cion. If those genes whose function has been experimen-tally investigated are annotated more precisely, we canmake optimal use of electronic databases and comparisonof genetic data.

Acknowledgements

This study was commissioned by the Netherlands Min-istry of Housing, Physical Planning and the Environment(VROM).

References

[1] Fredricks, D.N. and Relman, D.A. (1996) Sequence-based identi¢ca-tion of microbial pathogens: a reconsideration of Koch's postulates.Clin. Microbiol. Rev. 9, 18^33.

[2] Falkow, S. (1988) Molecular Koch's postulates applied to microbialpathogenicity. Rev. Infect. Dis. 10, S274^S276.

[3] Strauss, E.J. and Falkow, S. (1997) Microbial pathogenesis : genomicsand beyond. Science 276, 707^711.

[4] Deitsch, K.W., Moxon, E.R. and Wellems, T.E. (1997) Sharedthemes of antigenic variation and virulence in bacterial, protozoal,and fungal infections. Microbiol. Mol. Biol. Rev. 61, 281^293.

[5] Hood, D.W., Deadman, M.E., Jennings, M.P., Bisercic, M., Fleisch-mann, R.D., Venter, J.C. and Moxon, E.R. (1996) DNA repeatsidentify novel virulence gene in Haemophilus in£uenzae. Proc. Natl.Acad. Sci. USA 93, 11121^11125.

[6] Harb, O.S. and Abu Kwaik, Y. (1999) Probing the microenvironmentof intracellular bacterial pathogens. Microbes Infect. 1, 445^453.

[7] Chiang, S.L., Mekalanos, J.J. and Holden, D.W. (1999) In vivo ge-netic analysis of bacterial virulence. Annu. Rev. Microbiol. 53, 129^154.

[8] Frosch, M., Reidl, J. and Vogel, U. (1998) Genomics in infectiousdiseases: approaching the pathogens. Trends Microbiol. 6, 346^349.

[9] Lee, S.Y. and Klevit, R.E. (2000) The whole is not the simple sum ofits parts in calmodulin from S. cerevisiae. Biochemistry 39, 4225^4230.

[10] Casadevall, A. and Pirofski, L.A. (1999) Host-pathogen interactions:rede¢ning the basic concepts of virulence and pathogenicity. Infect.Immun. 67, 3703^3713.

FEMSLE 9995 29-6-01

T.M. Wassenaar, W. Gaastra / FEMS Microbiology Letters 201 (2001) 1^7 7