Upload
independent
View
0
Download
0
Embed Size (px)
Citation preview
ARTICLE IN PRESS
1438-4221/$ - se
doi:10.1016/j.ijm
�Correspondof Skovde, SE-
fax: +46500 44
E-mail addr
International Journal of Medical Microbiology 299 (2009) 65–74
www.elsevier.de/ijmm
Mathematical modelling of the regulation of spa (protein A) transcription
in Staphylococcus aureus
Erik Gustafssona,b,�, Stefan Karlssona, Jan Oscarssonb,c, Peter Sog(arda,d,Patric Nilssona, Staffan Arvidsonb
aDepartment of Life Sciences, University of Skovde, SE-541 28 Skovde, SwedenbDepartment of Microbiology, Tumor and Cell Biology (MTC), Karolinska Institutet, SE-171 77 Stockholm, SwedencDepartment of Odontology, Oral Microbiology, Ume (a University, S-901 87 Ume (a, SwedendMolecular Medicine and Surgery (MMK), Karolinska Institutet, SE-171 77 Stockholm, Sweden
Received 15 October 2007; received in revised form 18 January 2008; accepted 26 May 2008
Abstract
In the present work a general systems biology approach has been used to study the complex regulatory networkcontrolling the transcription of the spa gene, encoding protein A, a major surface protein and an important virulencefactor of Staphylococcus aureus. A valid mathematical model could be formulated using parameter values, which werefitted to quantitative Northern blot data from various S. aureus regulatory mutants using a gradient search method.The model could correctly predict spa expression levels in 4 different regulatory mutants not included in the parametervalue search, and in 2 other S. aureus strains, SH1000 and UAMS-1. The mathematical model revealed that sarA andsarS seem to balance each other in a way that when the activating impact of sarS is small, e.g. in the wild-type, therepressive impact of sarA is small, while in an agr-deficient background, when the impact of sarS is maximal, therepressive impact of sarA is close to its maximum. Furthermore, the model revealed that Rot and SarS actsynergistically to stimulate spa expression, something that was not obvious from experimental data. We believe thatthis mathematical model can be used to evaluate the significance of other putative interactions in the regulatorynetwork governing spa transcription.r 2008 Elsevier GmbH. All rights reserved.
Keywords: Staphylococcus aureus; Protein A (spa); Regulation; Mathematical model; Parameter estimation
Introduction
Staphylococcus aureus is a common human patho-gen responsible for a variety of diseases ranging frommild cutaneous to deep systemic infections, such as
e front matter r 2008 Elsevier GmbH. All rights reserved.
m.2008.05.011
ing author at: Department of Life Sciences, University
541 28 Skovde, Sweden. Tel.: +46 500 448657;
8499.
ess: [email protected] (E. Gustafsson).
osteomyelitis, endocarditis, and bacteremia. The patho-genesis is very complex, and virulence depends on theproduction of more than 40 different virulence factors(Arvidson and Tegmark, 2001), which are coordinatelycontrolled by a number of global regulators, e.g. agr,arl, svr, srr, sae, mgrA, sarA, and several sarA-homologs, rot, sarR, sarS, sarT, sarU, sarV, and sarX
(Cheung et al., 1992; Fournier and Hooper, 2000;Garvis et al., 2002; Giraudo et al., 1994; Janzon et al.,1986; Luong et al., 2003; Manna and Cheung, 2001,
ARTICLE IN PRESSE. Gustafsson et al. / International Journal of Medical Microbiology 299 (2009) 65–7466
2003, 2006; Manna et al., 2004; McNamara et al., 2000;Recsei et al., 1986; Schmidt et al., 2001; Tegmark et al.,2000; Yarwood et al., 2001).
Staphylococcal protein A (Spa) is a major surfaceprotein found in essentially all strains of S. aureus
(Forsgren, 1969). Protein A binds to the Fc-region ofIgG thereby hypothetically preventing phagocytosis(Dossett et al., 1969; Forsgren, 1969) and, in addition,binds to von Willebrand factor (Hartleib et al., 2000).Recent studies have revealed that protein A inducesinflammatory responses in human airway and cornealepithelial cells (Gomez et al., 2004; Kumar et al., 2007)and also triggers T cell-independent B cell proliferation(Bekeredjian-Ding et al., 2007). The importance ofprotein A in infections has been demonstrated in amurine septic arthritis model (Palmqvist et al., 2002) andin subcutaneous infections in mice (Patel et al., 1987).Production of protein A in S. aureus is controlled byseveral global regulators including agr (RNAIII), sarA,sarS, sarT, rot, and mgrA, which together form acomplex regulatory network (Fig. 1) (Oscarsson et al.,2005).
SarS, which is an activator of spa transcription, isencoded directly upstream of spa and is a key regulatorin this network (Tegmark et al., 2000). Previous studieshave shown that expression of sarS is upregulated in agr
and sarA mutants (Tegmark et al., 2000), resulting inincreased spa expression. In an agr mutant, upregulationof sarS and spa required rot and sarT (Said-Salim et al.,2003; Schmidt et al., 2003), which seemed to counteractthe repressive effect of sarA (Oscarsson et al., 2005),
Fig. 1. Schematic illustration of the regulatory system
controlling spa transcription in S. aureus (arrows indicate
activation and bars repression). RNAIII represses sarT
transcription (Schmidt et al., 2001) and neutralizes Rot activity
(Geisinger et al., 2006; McNamara et al., 2000; Said-Salim
et al., 2003). Rot directly stimulates both sarS and spa
transcription but stimulation of sarS by rot requires sarT and
is only required in the presence of sarA (Oscarsson et al.,
2005). SarA is a direct repressor of sarS (Oscarsson et al.,
2005) and spa (Chien et al., 1999; Sterba et al., 2003), and also
suppresses sarT (Schmidt et al., 2001). SarS is a direct activator
of spa transcription (Cheung et al., 2001; Tegmark et al., 2000).
SarT directly stimulates sarS but the stimulation of sarS by
SarT requires rot, and is only required in the presence of sarA
(Oscarsson et al., 2005; Schmidt et al., 2003). For further
explanations see text.
whereas in a sarA mutant, derepression of sarS and spa
was independent of rot and sarT (Schmidt et al., 2003).In addition, different studies have shown mgrA to berequired for (Luong et al., 2003) or to repress (Luonget al., 2006; Oscarsson et al., 2005) sarS and spa
expression. However, as mgrA mutants also exhibited agrowth defect (Ingavale et al., 2003; Oscarsson et al.,2005; Truong-Bolduc et al., 2003) and variably reducedlevels of RNAIII (Ingavale et al., 2003, 2005; Oscarssonet al., 2005), the actual role of mgrA in sarS and spa
expression is difficult to assess. Because of this we havechosen to model the regulation of spa in an mgrA+
background. RNAIII, which is synthesized at the end ofthe exponential phase of growth (Janzon and Arvidson,1990; Novick et al., 1993) in response to accumulationof an auto-inducing peptide, AIP (Ji et al., 1995),represses spa transcription by inhibiting translation ofrot (Geisinger et al., 2006), which is required for sarS
and spa expression (Oscarsson et al., 2005; Said-Salimet al., 2003). RNAIII has also been shown to down-regulate spa expression by base-pairing with the ribo-some binding site of the spa mRNA, thereby recruitingendoribonuclease III, which subsequently degrades thespa messenger (Huntzinger et al., 2005).
The regulatory network controlling spa transcriptionthus consists of several feed-forward loops (Shen-Orret al., 2002). A feed-forward loop is a three-gene patternthat is composed of 2 input transcription factors, one ofwhich regulates the other, both jointly regulating atarget gene. The three-gene pattern composed of the 2transcription factors, sarA and sarS, is a coherent typeII basic-building block. This means that the sign(negative) of the direct regulatory route (from sarA tospa) is the same as the overall sign of the indirectregulatory path (from sarA via sarS to spa), as indicatedin Fig. 1 (Mangan and Alon, 2003). The basic-buildingblock composed of agr, rot, and spa belongs to the samecoherent type. However, the three-gene pattern com-posed of rot, sarS, and spa is a coherent type I basic-building block (Mangan and Alon, 2003), meaning thatthe sign of the direct and indirect regulatory paths ispositive. The dynamical functions for different classes ofsingle feed-forward loop circuits have been extensivelystudied (Mangan and Alon, 2003; Wall et al., 2005).Interestingly, there is no evidence of negative feedbackloops within the regulatory network. Positive feedbackwithin the system is represented by auto-activation ofagr (i.e. biosynthesis of RNAIII) as transcription of theagr operon is auto-activated by AIP (Ji et al., 1995).However, as the agr system acts as a bistable switch(Gustafsson et al., 2004), we have chosen to model a cellin which agr is fully activated (maximum levels ofRNAIII).
Our current view of the regulatory network control-ling spa transcription (Fig. 1) is mainly based on theanalysis of spa transcription in derivatives of strain
ARTICLE IN PRESSE. Gustafsson et al. / International Journal of Medical Microbiology 299 (2009) 65–74 67
NCTC8325-4 harboring different combinations ofregulatory mutations (Cheung et al., 2001; Chienet al., 1999; Geisinger et al., 2006; McNamara et al.,2000; Oscarsson et al., 2005; Said-Salim et al., 2003;Schmidt et al., 2001, 2003; Sterba et al., 2003; Tegmarket al., 2000). Because of the complexity of the system it isalmost impossible to know whether this network couldfunction in a way that could explain the quantitativechanges in spa expression in response to inactivation ofdifferent regulatory loci. Experimentally, it is alsodifficult to analyze the effect of small changes inconcentration of individual or combinations of regula-tors. To investigate this, a mathematical model of thenetwork was set up using ordinary differential equa-tions. To be able to make quantitative predictions withthe model, parameter values (e.g. transcription rates andbinding constants) were identified, which gave the bestcorrelation between spa expression data generated invitro (spa mRNA quantification) and in silico (outputdata from the mathematical model). A very goodcorrelation was achieved, showing that the proposednetwork structure can describe the in vitro data. Usingthe mathematical model, we could predict the relativeimpact of small/large changes in concentration of theindividual regulators on spa expression and found thatsarA and sarS seem to balance each other, whereasrot and sarS act synergistically.
Materials and methods
Bacterial strains and cultivation conditions
Bacterial strains used in this study are listed inTable 1. S. aureus strains were grown on Nutrient
Table 1. Bacterial strains used in this study
Strain Relative characteristics Reference
8325-4 Prototype S. aureus strain,
rsbU
Novick (1967)
RN6390 Laboratory isolate of 8325-4,
rsbU
Peng et al.
(1988)
SH1000 8325-4, rsbU+ Horsburgh et al.
(2002)
WA1029 RN6390, agr-null, rot::tet,
sarS::ermB (TcR, EmR)
Oscarsson et al.
(2006b)
WA1049 RN6390, agr-null, rot::tet,
sarA::km, sarS::ermB (TcR,
KmR, EmR)
Oscarsson et al.
(2006a)
WA1217 RN6390, agr-null, sarS::ermB
(EmR)
Oscarsson et al.
(2006b)
WA1428 RN6390, agr-null, sarA::km,
sarS::ermB (KmR, EmR)
Oscarsson et al.
(2006a)
WA1430 RN6390, agr-null, rot::tet,
sarA::km (TcR, KmR)
Oscarsson et al.
(2006a)
agar-plates (Difco). S. aureus strains were preculturedovernight (16–18 h) in 30ml Tryptic Soy Broth (Difco).When required, 10 mgml–1 tetracycline, 50 mgml–1 kana-mycin, 5 mgml–1 erythromycin, or 5 mgml–1 lincomycinwas added to the culture media. Cells were collected bycentrifugation and used to inoculate 100ml of BrainHeart Infusion (BHI) broth (Difco) in 1-l baffled flasksto give an optical density at 600 nm (OD600) of 0.5 andincubated on a rotary shaker (180 rpm) at 37 1C.
Northern blot analysis
Total S. aureus RNA was prepared using the FASTRNA-blue kit (Bio 101) according to instructions fromthe manufacturer. The concentration of RNA wasdetermined by measuring the absorbance at 260 nm.Samples containing 10 mg of total RNA were analyzedby Northern blotting as described previously (Morfeldtet al., 1988). For Northern hybridization, internalfragments of 16S rRNA (nt 11–1022; GenBank acces-sion no. X68417) and spa (nt 190–620; V01287) wereamplified by PCR, radio-labeled with [a-32P]-dCTP(Amersham) using a random prime labelling kit (RocheMolecular Biochemicals) and used as probes. Radio-activity was detected by a radioisotope imaging system(phosphorimager 445SI; Molecular Dynamics) andquantified using the ImageQuant software.
Quantitative real-time PCR (qPCR) analysis
Quantitative real-time PCR (qPCR) was performedusing the 7300 Real-time PCR system (Applied Biosys-tems). Concentration of RNA was determined bymeasuring the absorbance at 260 nm on ND-1000(Nanodrop Technologies, Inc.). One mg of total RNAwas converted to cDNA using High Capacity cDNAArchive Kit (Applied Biosystems) according to instruc-tions from the manufacturer. Each qPCR reaction (finalvolume 25 ml) was run in triplicates, containing1�Power SYBR Green PCR Master Mix kit (AppliedBiosystems) with 0.25 mM of forward and reverseprimers, and the cDNA diluted 1000� . Oligonucleotideprimers (Table 3) were designed using Primer Express v.3.0 (Applied Biosystems). 16S rRNA was used asendogenous control.
Results and discussion
The mathematical model
The mathematical model describes spa transcriptionin response to cellular concentrations of RNAIII, Rot,SarA, SarT and SarS. The model is based on quantita-tive Northern blot data from various regulatory mutants
ARTICLE IN PRESSE. Gustafsson et al. / International Journal of Medical Microbiology 299 (2009) 65–7468
of strain NCTC8325-4 and describes an averagebacterial cell in a liquid culture. Spatial distributionsof regulators within the cell were handled as if allconcentrations were homogeneous, assuming that localconcentrations are proportional to mean cellular con-centrations. Except for the recently demonstrated effectof RNAIII on translation of rot mRNA (Geisingeret al., 2006), nothing is known about translationalcontrol of other factors involved in spa regulation. Inthe model, we have therefore assumed that eachregulator is produced in direct proportion to corre-sponding mRNA transcription.
The regulatory network (Fig. 1) was described by 6ordinary differential equations (see Supplementarymaterials). As the agr-system acts as a bistable switch(Gustafsson et al., 2004), meaning that RNAIII levelsare either very low or very high, the auto-activation ofagr was not included in the model. Eq. (1) describes thelevel of RNAIII when the agr system is activated. Eq. (2)describes the inhibitory effect of RNAIII on Rotactivity, i.e. prevention of rot mRNA translation(Geisinger et al., 2006). The stimulatory effect of sarA
on agr expression (Manna and Cheung, 2003; Schmidtet al., 2001) was neglected as sarA does not seem toaffect the maximal level of RNAIII but rather the timeat which the agr system is activated (Gustafsson et al.,2004; Tegmark et al., 2000). The expression of sarA isdescribed in Eq. (3). Repression of sarT by RNAIII andsarA together (Schmidt et al., 2001) is described inEq. (4). Eq. (5) describes the repressive effect of sarA on
Table 2. Relative transcription levels of sarS, sarT, and spa in diffe
data (in vitro) and generated with the model (in silico) after optimi
Target gene sarS sa
In vitro In silico In
Strain properties
Wild-type 1x 1x 1x
agr 5xa 5.0x 5x
sarA 5xa 5.0x 5x
sarS
agr sarA 5xa 5.0x 5x
agr sarS
sarA sarS
agr sarA sarS
rot 1x 0.9x
agr rot 0.5xb 0.9x
agr sarA rot 5xc 5.0x
sarT 1xd 0.9x
agr sarT 5xd 5.1x
sarA sarT 1xd 0.9x
The expression levels are normalized to the levels expressed by the wild-typeaTegmark et al. (2000).bSaid-Salim et al. (2003).cOscarsson et al. (2005).dSchmidt et al. (2003).eSchmidt et al. (2001).
sarS transcription (Tegmark et al., 2000). This repres-sion is counteracted by Rot and SarT in consort(Oscarsson et al., 2005). Finally, Eq. (6) describes theactivity of the spa promoter. This activity is positivelycontrolled by sarS and rot in a direct manner (Oscarssonet al., 2005; Tegmark et al., 2000). SarA suppresses spa
transcription by competing with SarS for binding (Gaoand Stewart, 2004; Oscarsson et al., 2005) and also in adirect way (Tegmark et al., 2000). Finally, RNAIII alsodestabilizes the spa mRNA by a direct interaction(Huntzinger et al., 2005).
Determination of parameter values generating an
optimal match between in silico and in vitro data
To be able to make quantitative predictions with themodel and to verify that the proposed network structurecan describe the in vitro data, we estimated theparameters of the 6 equations described above using agradient search method and quantitative Northern blotdata from a set of regulatory mutants defining theinteractions in the regulatory network. The parametersdescribe binding constants, transcription, and turn-overrates for each component in the system.
For the parameter value searches, the relative mRNAlevels of spa, sarS, and sarT in wild-type as compared toa large set of regulatory mutants were used (Table 2).Except for sarT mRNA measurements, which wereextracted from published data (Schmidt et al., 2001), all
rent S. aureus regulatory mutants, obtained from experimental
zation
rT spa
vitro In silico In vitro In silico
1x 1x 1xe 5.0x 10xa 9.6xe 5.0x 5xa 5.6x
0.2xa 0.2xe 5.0x 20xa 18.3x
1xa 1.0x
1xa 0.9x
1.5xa 1.5x
1x 0.7x
0.5xb 0.7x
5xc 4.4x
1xd 1.0x
3xd 3.4x
5xd 5.6x
(8325-4).
ARTICLE IN PRESS
Fig. 2. Levels of spa transcription during mid post-exponential
phase of growth relative to the wild-type, in vitro (black bars)
and in silico, after fitting the parameter values to spa
transcription data using a gradient search method (white bars).
E. Gustafsson et al. / International Journal of Medical Microbiology 299 (2009) 65–74 69
mRNA levels were based on quantification of severalNorthern blot experiments in our laboratory (Oscarssonet al., 2005; Tegmark et al., 2000; and unpublisheddata). All data were from bacteria in the post-exponential phase of growth when agr is fully activatedin wild-type cells (Gustafsson et al., 2004; Tegmarket al., 2000). As indicated in several experiments, thelevel of RNAIII, Rot, SarA, SarS, and protein Aremained relatively stable over several hours of post-exponential growth (Geisinger et al., 2006; Tegmark,2000; Tegmark et al., 2000; and data not shown),suggesting that the system is close to equilibrium duringthis period. In the mathematical model, we thereforeconsider the system to be at equilibrium.
We defined a measure (penalty function, f) of howmuch gene expression in the corresponding in silicoknockout mutants differed from the in vitro data bytaking the sum of squares of the differences, in alogarithmic scale:
f ¼X
a
ðlog in vitroa � log in silicoaÞ2,
where index a varies over all combinations of knockoutsand all measured spa, sarS, and sarT mRNA levels(Table 2). The penalty function f is thus a function of theparameters in the model. The aim was to find parametervalues that minimize f. This was carried out using anumerical quasi-Newton method implemented asthe routine fmincon in the Optimization toolbox ofMATLAB v. 7.0.1 (The MathWorks, Inc.). We searchediteratively for minima to f starting from randomlygenerated values of all the parameters. Using thisapproach, we identified one potential global minimum(the minimum having the lowest value of the penaltyfunction). Using the parameter values belonging tothis minimum resulted in a very good match between
Table 3. Sequences of oligonucleotides used in the qPCR
reactions
Primers Oligonucleotide sequence (50-30)
16S rRNA
forward
AAT CAG AAA GCC ACG GCT AAC T
16S rRNA
reverse
CGC TTG CCA CCT ACG TAT TAC C
RNAIII forward TGT TCA CTG TGT CGA TAA TCC
ATT T
RNAIII reverse GGA GTG ATT TCA ATG GCA CAA G
SarA forward GCA CAA CAA CGT AAA AAA ATC
GAA
SarA reverse TTC GTT GTT TGC TTC AGT GAT TC
SarS forward CCA CCA TAA ATA CCC TCA AAC
TGT T
SarS reverse TCA TCT TCA GTT GAG CGT TCT
TTT
in silico and experimental in vitro data (Table 2 andFig. 2).
Experimental evaluation of the mathematical model
We have set values to the model parameters usingrelative expression levels of spa, sarS, and sarT in cellsfrom mid post-exponential phase of growth (4 h). Toevaluate the model, we simulated a cell in the lateexponential phase of growth (2 h) where the level ofRNAIII is 10-fold lower and sarA mRNA levels are five-fold higher than at mid post-exponential phase (4 h)(Tegmark, 2000; Tegmark et al., 2000; and data notshown). Using these data in our mathematical model, wepredicted that the level of spa mRNA would be 50%lower in late exponential compared to mid post-exponential cells. This is in good agreement withexperimental data showing a 50% reduction in spa
expression (Novick et al., 1993; Tegmark et al., 2000)(and see strain 8325-4 in Fig. 6), although they conflictwith the general idea that spa would be produced mainlyduring the early exponential phase of growth beforeRNAIII starts to accumulate.
To further evaluate the model, we used it to predictspa transcription in a number of S. aureus regulatorymutants, which have not been previously analyzed andwere therefore not included in the parameter valuesearches. According to previous studies (Oscarssonet al., 2005), rot seems to enhance spa transcriptionpartly in a sarS-independent way. According to ourmathematical model, spa transcription would decreaseseven-fold in an agr rot sarS triple mutant relative to an
ARTICLE IN PRESS
Fig. 3. Northern blot analysis of spa and 16S rRNA in strains.
(A) WA1029 (agr rot sarS), WA1049 (agr rot sarA sarS) and
WA1217 (agr sarS), and (B) WA1029 (agr rot sarS), WA1430
(agr rot sarA) and WA1428 (agr sarA sarS). Samples for
mRNA isolation were taken at the indicated time points
(hours) during growth of a representative culture.
E. Gustafsson et al. / International Journal of Medical Microbiology 299 (2009) 65–7470
agr sarS double mutant. To confirm this, we comparedspa mRNA levels in WA1029 (agr rot sarS) andWA1217 (agr sarS). As shown in Fig. 3a, spa transcrip-tion was clearly lower in WA1029, which support themathematical model and further support a directstimulatory role of rot in spa transcription. Further-more, the mathematical model predicted that inactiva-tion of sarA in the agr rot sarS triple mutant wouldresult in a seven-fold upregulation in spa transcription.As expected, Northern blot analysis revealed a clearincrease in spa mRNA levels in an agr rot sarA sarS
quadruple (WA1049) compared to an agr rot sarS triple(WA1029) mutant. We also compared spa transcriptionin 3 different triple mutants. According to the model,spa transcription would be upregulated 10- and 30-fold,respectively, in an agr sarA sarS and agr rot sarA,compared to an agr rot sarS triple mutant. This wasconfirmed by Northern hybridization (Fig. 3b) showingclearly increased spa mRNA levels in strains WA1428(agr sarA sarS) and WA1430 (agr rot sarA) relative toWA1029 (agr rot sarS). Although these predictions didnot exactly agree with experimental data, we considerthe model accurate enough to be used for further studiesof this regulatory network. One reason for the lack ofexact agreement could be slight differences in growthrate between some mutants, meaning that mRNA levelswere not determined at exactly the same point in thegrowth curve.
Quantitative analysis of gene expression by the
mathematical model revealed synergy of Rot and
SarS on spa promoter activity
Using the model, we analyzed the direct effect of eachregulator on transcription of its target gene/s in the wild-type strain during post-exponential phase of growth.Minimal and maximal levels of transcription for eachtarget gene in the system was simulated, i.e. transcrip-tion level in the absence or presence of possibleactivators and repressors (top and bottom lines of boxesin Fig. 4). The theoretical potential of each regulator toaffect transcription of its target genes (grey area withinthe boxes in Fig. 4) was calculated by allowing theconcentration of the regulator to increase from zero toinfinitely. This analysis revealed that in a wild-typebackground increasing rot or sarS infinitely could onlyenhance spa transcription in a direct way up to 15% and30% of its maximum (Fig. 4a), respectively, suggestingthat Rot and SarS might act synergistically to activatespa. To investigate this, we analyzed the parameters inthe model, i.e. the effect of Rot and SarS alone ortogether on spa promoter activity in the absence of anyother influences. Interestingly, Rot alone could increasespa promoter activity by 70%, while SarS alone couldstimulate the promoter activity six-fold, whereas Rotand SarS together enhanced spa promoter activity24-fold (see Supplementary materials), clearly showinga synergistic effect. As SarA and SarS seem to competefor the same binding sites within the spa promoterregion (Gao and Stewart, 2004; Oscarsson et al., 2005),the negative effect of sarA would be greater when bothRot and SarS are present. Our in silico data also suggestthat spa transcription is completely inhibited when SarAbinds to the spa promoter in the absence of SarS andRot. This is in good agreement with experimental datashowing barely detectable spa transcription in a sarS
mutant (Tegmark et al., 2000). We also found thatbinding of Rot to the spa promoter in the presence ofSarA enhanced spa promoter activity. As over-expres-sion of rot did not affect binding of SarA to the spa
promoter (Oscarsson et al., 2005), Rot seems tostimulate spa promoter activity in a direct way and notby removing SarA. This is also consistent with theobservation that rot stimulated spa transcription equallyin the presence or absence of sarA (Oscarsson et al.,2005) (Fig. 3).
Since agr mutations are common in clinical isolates ofS. aureus, we also investigated the potential of eachregulator to affect transcription of its target genes in anagr-deficient background. In an agr mutant, sarS
transcription is upregulated and the rot mRNA istranslated. Because of the increased levels of Rotprotein, the ability of sarS to stimulate spa transcriptionis prominently increased (Fig. 4b). However, in thisbackground the regulatory potential of sarA to repress
ARTICLE IN PRESS
Fig. 4. An illustration of the theoretical potential of each
regulator to affect transcription of its target gene/s. Minimal
and maximal rates of transcription for each target gene in the
system are illustrated as top and bottom lines of boxes. The
theoretical potential of each regulator to affect transcription of
its target genes is illustrated as grey areas within the boxes. The
arrow within the box indicates whether the regulator positively
(arrow pointing upwards) or negatively (arrow pointing
downwards) influences transcription of its target gene. The
interaction arrows cut the boxes at present promoter activity
of the target gene. (A) shows the wild-type (8325-4) and (B) an
agr-deficient background.
E. Gustafsson et al. / International Journal of Medical Microbiology 299 (2009) 65–74 71
spa transcription is also elevated. Furthermore,although sarT has no regulatory potential on sarS
transcription in an agr+ background, the regulatoryimpact of sarT on sarS is enhanced in an agr mutantbecause of increased levels of Rot protein.
Fig. 5. Dose–response curves, calculated with the mathemat-
ical model, showing relative steady-state levels of spa
transcription with different levels of the regulators, RNAIII
(solid line), Rot (dashed line), SarA (dotted line), SarT (solid-
dotted line), and SarS (solid-squared line). The upper panel
shows the wild-type (8325-4) and the lower panel the rsbU+
derivative of 8325-4 (strain SH1000).
Changes in spa mRNA levels in response to altered
levels of individual regulators
One reason for developing the mathematical modelwas to be able to analyze the impact of small changes in
the expression of individual regulators instead of study-ing the effect of knockout mutants. We thereforegenerated dose–response curves showing the level ofspa mRNA at different concentrations of each regulatorin a mid post-exponential phase cell. As illustrated inFig. 5 (upper panel), expression of spa is most sensitiveto changes in concentration of sarA (slope of thedose–response curve at wild-type concentration) fol-lowed by sarS 4agr and rot 4sarT, i.e. sensitivity ofspa expression to sarA was two times higher than that tosarS and five times higher than that to agr/rot (seeSupplementary materials). Interestingly, even largechanges in sarT expression had no impact on spa
mRNA levels. Fig. 5 (upper panel) also illustrates thatsarA has the greatest impact (20-fold range of regula-tion) on spa regulation, followed by sarS (10-fold rangeof regulation) and agr/rot (10-fold range of regulation).Calculating dose–response curves for an agr mutantrevealed that the capacity of sarA and sarS to regulatespa expression was essentially the same as in the wild-type (data not shown).
The mathematical model was based on transcriptionaldata from strain 8325-4, which is sigma B-deficient dueto a mutation in rsbU (Kullik et al., 1998). As the rsbU
mutation influences the expression of RNAIII, sarA,and sarS (Horsburgh et al., 2002; Oscarsson et al.,2006a), it could be argued that the model would not berepresentative for S. aureus strains in general. Wetherefore analyzed the rsbU+ strain SH1000, derivedfrom 8325-4 with respect to spa, RNAIII, sarA, and sarS
by Northern blotting and quantitative real-time PCR. Inmid post-exponential phase cells of SH1000, theRNAIII level was three-fold lower than in 8325-4,
ARTICLE IN PRESSE. Gustafsson et al. / International Journal of Medical Microbiology 299 (2009) 65–7472
whereas the sarA and sarS mRNA levels were two-foldhigher, respectively, as determined by qPCR (seeSupplementary materials). Using these data in themathematical model, spa expression was predicted tobe 50% higher in SH1000 than in 8325-4. This isconsistent with mRNA analysis (70% higher) (Fig. 6),suggesting that the model is also valid for SH1000. Thiswas further strengthened by the dose–response analysis(Fig. 5, lower panel), demonstrating that the impact ofthe different regulators was essentially unaffected by thersbU mutation.
To further validate the model, published transcriptiondata from the clinical isolate UAMS-1 (rsbU+) wereapplied to the model. This strain expressed reducedlevels of RNAIII (25-fold) and increased levels of sarA
(two-fold) and sarS (100-fold), respectively, relative toRN6390 (laboratory isolate of 8325-4) (Cassat et al.,2006). Assuming that our estimated parameter valuescan be applied to strain UAMS-1, the mathematicalmodel predicted 15-fold higher spa mRNA levels inUAMS-1 compared to RN6390. This is in goodagreement with the experimental data reported byCassat et al. (2006) showing eight-fold higher spa
mRNA levels in UAMS-1 than in RN6390.In the present study, we developed a mathematical
model of the regulatory network controlling spa
transcription in S. aureus. We have shown thatexperimental determination of individual parametervalues (i.e. transcription rates and binding constants)was not required but instead could be estimated on thebasis of quantified Northern blot data from a set ofregulatory mutants. The validity of the model wasverified by simulating spa expression in 4 differentregulatory mutants not included in the parameter valuesearch (Fig. 3). Using experimental RNAIII, sarA, andsarS transcription data from two rsbU+ strains,SH1000, and UAMS-1, we could also accurately predictthe spa mRNA levels (Fig. 6, see above). Our resultsindicate that although agr has been described as themain regulator of spa expression, sarA and sarS appearto be more important. The mathematical model alsorevealed that sarA and sarS seem to balance each otherin a way that when the activating impact of sarS issmall, e.g. in the wild-type, the repressive impact of sarA
Fig. 6. Northern blot analysis of spa in the strains SH1000
(rsbU+) and 8325-4 (wild-type). Samples for mRNA isolation
were taken at the indicated time points (hours) during growth
of a representative culture.
is small, while in an agr-deficient background, when theimpact of sarS is maximal, the repressive impact of sarA
is close to its maximum (compare Fig. 4A with Fig. 4B).The network controlling spa transcription is consideredrelatively robust in the sense that the system perfor-mance of spa regulation is not altered when theparameter values are changed and a single equilibriumwas always obtained. Furthermore, the model alsorevealed that rot and sarS act synergistically to stimulatespa expression, a matter that could be tested experi-mentally to gain more insight into the regulation of spa
expression. We believe that this mathematical model canbe used to evaluate the significance of other putativeinteractions in the regulatory network governing spa
transcription.
Acknowledgements
We thank Agneta Wahlquist for skillful technicalassistance. This work was financially supported by theKnowledge Foundation through the industrial Ph.D.program in Medical Bioinformatics at the Strategy andDevelopment Office (SDO) at Karolinska Institutet, bySkaraborg Hospital, Swedish Foundation for StrategicResearch, Swedish Society for Medical Research(SSMF) and the Swedish Research Council (projectno. 4513).
Appendix A. Supplementary materials
Supplementary data associated with this article can befound in the online version at doi:10.1016/j.ijmm.2008.05.011.
References
Arvidson, S., Tegmark, K., 2001. Regulation of virulence
determinants in Staphylococcus aureus. Int. J. Med.
Microbiol. 291, 159–170.
Bekeredjian-Ding, I., Inamura, S., Giese, T., Moll, H., Endres,
S., Sing, A., Zahringer, U., Hartmann, G., 2007. Staphy-
lococcus aureus protein A triggers T cell-independent B cell
proliferation by sensitizing B cells for TLR2 ligands.
J. Immunol. 178, 2803–2812.
Cassat, J., Dunman, P.M., Murphy, E., Projan, S.J., Beenken,
K.E., Palm, K.J., Yang, S.J., Rice, K.C., Bayles, K.W.,
Smeltzer, M.S., 2006. Transcriptional profiling of a
Staphylococcus aureus clinical isolate and its isogenic agr
and sarA mutants reveals global differences in comparison
to the laboratory strain RN6390. Microbiology 152,
3075–3090.
Cheung, A.L., Koomey, J.M., Butler, C.A., Projan, S.J.,
Fischetti, V.A., 1992. Regulation of exoprotein expression
ARTICLE IN PRESSE. Gustafsson et al. / International Journal of Medical Microbiology 299 (2009) 65–74 73
in Staphylococcus aureus by a locus (sar) distinct from agr.
Proc. Natl. Acad. Sci. USA 89, 6462–6466.
Cheung, A.L., Schmidt, K., Bateman, B., Manna, A.C., 2001.
SarS, a SarA homolog repressible by agr, is an activator of
protein A synthesis in Staphylococcus aureus. Infect.
Immun. 69, 2448–2455.
Chien, Y., Manna, A.C., Projan, S.J., Cheung, A.L., 1999.
SarA, a global regulator of virulence determinants in
Staphylococcus aureus, binds to a conserved motif essential
for sar-dependent gene regulation. J. Biol. Chem. 274,
37169–37176.
Dossett, J.H., Kronvall, G., Williams Jr., R.C., Quie, P.G.,
1969. Antiphagocytic effects of staphylococcal protein A.
J. Immunol. 103, 1405–1410.
Forsgren, A., 1969. Protein A from Staphylococcus aureus. 8.
Production of protein A by bacterial and L-forms of S.
aureus. Acta Pathol. Microbiol. Scand. 75, 481–490.
Fournier, B., Hooper, D.C., 2000. A new two-component
regulatory system involved in adhesion, autolysis, and
extracellular proteolytic activity of Staphylococcus aureus.
J. Bacteriol. 182., 3955–3964.
Gao, J., Stewart, G.C., 2004. Regulatory elements of the
Staphylococcus aureus protein A (Spa) promoter.
J. Bacteriol. 186, 3738–3748.
Garvis, S., Mei, J.M., Ruiz-Albert, J., Holden, D.W., 2002.
Staphylococcus aureus svrA: a gene required for virulence
and expression of the agr locus. Microbiology 148,
3235–3243.
Geisinger, E., Adhikari, R.P., Jin, R., Ross, H.F., Novick,
R.P., 2006. Inhibition of rot translation by RNAIII,
a key feature of agr function. Mol. Microbiol. 61,
1038–1048.
Giraudo, A.T., Raspanti, C.G., Calzolari, A., Nagel, R., 1994.
Characterization of a Tn551-mutant of Staphylococcus
aureus defective in the production of several exoproteins.
Can J. Microbiol. 40, 677–681.
Gomez, M.I., Lee, A., Reddy, B., Muir, A., Soong, G., Pitt,
A., Cheung, A., Prince, A., 2004. Staphylococcus aureus
protein A induces airway epithelial inflammatory responses
by activating TNFR1. Nat. Med. 10, 842–848.
Gustafsson, E., Nilsson, P., Karlsson, S., Arvidson, S., 2004.
Characterizing the dynamics of the quorum-sensing system
in Staphylococcus aureus. J. Mol. Microbiol. Biotechnol. 8,
232–242.
Hartleib, J., Kohler, N., Dickinson, R.B., Chhatwal, G.S.,
Sixma, J.J., Hartford, O.M., Foster, T.J., Peters, G.,
Kehrel, B.E., Herrmann, M., 2000. Protein A is the von
Willebrand factor binding protein on Staphylococcus
aureus. Blood 96, 2149–2156.
Horsburgh, M.J., Aish, J.L., White, I.J., Shaw, L., Lithgow,
J.K., Foster, S.J., 2002. sigmaB modulates virulence
determinant expression and stress resistance: characteriza-
tion of a functional rsbU strain derived from Staphylococ-
cus aureus 8325-4. J. Bacteriol. 184, 5457–5467.
Huntzinger, E., Boisset, S., Saveanu, C., Benito, Y., Geiss-
mann, T., Namane, A., Lina, G., Etienne, J., Ehresmann,
B., Ehresmann, C., Jacquier, A., Vandenesch, F., Romby,
P., 2005. Staphylococcus aureus RNAIII and the endo-
ribonuclease III coordinately regulate spa gene expression.
EMBO J. 24, 824–835.
Ingavale, S.S., Van Wamel, W., Cheung, A.L., 2003.
Characterization of RAT, an autolysis regulator in
Staphylococcus aureus. Mol. Microbiol. 48, 1451–1466.
Ingavale, S., van Wamel, W., Luong, T.T., Lee, C.Y., Cheung,
A.L., 2005. Rat/MgrA, a regulator of autolysis, is a
regulator of virulence genes in Staphylococcus aureus.
Infect. Immun. 73, 1423–1431.
Janzon, L., Arvidson, S., 1990. The role of the delta-lysin gene
(hld) in the regulation of virulence genes by the accessory
gene regulator (agr) in Staphylococcus aureus. EMBO J. 9,
1391–1399.
Janzon, L., Lofdahl, S., Arvidson, S., 1986. Evidence for a
coordinate transcriptional control of alpha-toxin and
protein A synthesis in Staphylococcus aureus. FEMS
Microbiol. Lett. 33, 193–198.
Ji, G., Beavis, R.C., Novick, R.P., 1995. Cell density control of
staphylococcal virulence mediated by an octapeptide
pheromone. Proc. Natl. Acad. Sci. USA 92, 12055–12059.
Kullik, I., Giachino, P., Fuchs, T., 1998. Deletion of the
alternative sigma factor sigmaB in Staphylococcus aureus
reveals its function as a global regulator of virulence genes.
J. Bacteriol. 180, 4814–4820.
Kumar, A., Tassopoulos, A.M., Li, Q., Yu, F.S., 2007.
Staphylococcus aureus protein A induced inflammatory
response in human corneal epithelial cells. Biochem.
Biophys. Res. Commun. 354, 955–961.
Luong, T.T., Newell, S.W., Lee, C.Y., 2003. Mgr, a novel
global regulator in Staphylococcus aureus. J. Bacteriol. 185,
3703–3710.
Luong, T.T., Dunman, P.M., Murphy, E., Projan, S.J., Lee,
C.Y., 2006. Transcription profiling of the mgrA regulon in
Staphylococcus aureus. J. Bacteriol. 188, 1899–1910.
Mangan, S., Alon, U., 2003. Structure and function of the
feed-forward loop network motif. Proc. Natl. Acad. Sci.
USA 100, 11980–11985.
Manna, A., Cheung, A.L., 2001. Characterization of sarR, a
modulator of sar expression in Staphylococcus aureus.
Infect. Immun. 69, 885–896.
Manna, A.C., Cheung, A.L., 2003. sarU, a sarA homolog, is
repressed by SarT and regulates virulence genes in
Staphylococcus aureus. Infect. Immun. 71, 343–353.
Manna, A.C., Cheung, A.L., 2006. Expression of SarX, a
negative regulator of agr and exoprotein synthesis, is
activated by MgrA in Staphylococcus aureus. J. Bacteriol.
188, 4288–4299.
Manna, A.C., Ingavale, S.S., Maloney, M., van Wamel, W.,
Cheung, A.L., 2004. Identification of sarV (SA2062), a
new transcriptional regulator, is repressed by SarA
and MgrA (SA0641) and involved in the regulation of
autolysis in Staphylococcus aureus. J. Bacteriol. 186,
5267–5280.
McNamara, P.J., Milligan-Monroe, K.C., Khalili, S., Proctor,
R.A., 2000. Identification, cloning, and initial characteriza-
tion of rot, a locus encoding a regulator of virulence factor
expression in Staphylococcus aureus. J. Bacteriol. 182,
3197–3203.
Morfeldt, E., Janzon, L., Arvidson, S., Lofdahl, S., 1988.
Cloning of a chromosomal locus (exp) which regulates the
expression of several exoprotein genes in Staphylococcus
aureus. Mol. Gen. Genet. 211, 435–440.
ARTICLE IN PRESSE. Gustafsson et al. / International Journal of Medical Microbiology 299 (2009) 65–7474
Novick, R., 1967. Properties of a cryptic high-frequency
transducing phage in Staphylococcus aureus. Virology 33,
155–166.
Novick, R.P., Ross, H.F., Projan, S.J., Kornblum, J.,
Kreiswirth, B., Moghazeh, S., 1993. Synthesis of staphylo-
coccal virulence factors is controlled by a regulatory RNA
molecule. EMBO J. 12, 3967–3975.
Oscarsson, J., Harlos, C., Arvidson, S., 2005. Regulatory role
of proteins binding to the spa (protein A) and sarS
(staphylococcal accessory regulator) promoter regions in
Staphylococcus aureus NTCC 8325-4. Int. J. Med. Micro-
biol. 295, 253–266.
Oscarsson, J., Kanth, A., Tegmark-Wisell, K., Arvidson, S.,
2006a. SarA is a repressor of hla ((alpha)-Hemolysin)
transcription in Staphylococcus aureus: its apparent role as
an activator of hla in the prototype strain NCTC 8325
depends on reduced expression of sarS. J. Bacteriol. 188,
8526–8533.
Oscarsson, J., Tegmark-Wisell, K., Arvidson, S., 2006b.
Coordinated and differential control of aureolysin (aur)
and serine protease (sspA) transcription in Staphylococcus
aureus by sarA, rot and agr (RNAIII). Int. J. Med.
Microbiol. 296, 365–380.
Palmqvist, N., Foster, T., Tarkowski, A., Josefsson, E., 2002.
Protein A is a virulence factor in Staphylococcus aureus
arthritis and septic death. Microb. Pathog. 33, 239–249.
Patel, A.H., Nowlan, P., Weavers, E.D., Foster, T., 1987.
Virulence of protein A-deficient and alpha-toxin-deficient
mutants of Staphylococcus aureus isolated by allele
replacement. Infect. Immun. 55, 3103–3110.
Peng, H.L., Novick, R.P., Kreiswirth, B., Kornblum, J.,
Schlievert, P., 1988. Cloning, characterization, and sequenc-
ing of an accessory gene regulator (agr) in Staphylococcus
aureus. J. Bacteriol. 170, 4365–4372.
Recsei, P., Kreiswirth, B., O’Reilly, M., Schlievert, P., Gruss,
A., Novick, R.P., 1986. Regulation of exoprotein gene
expression in Staphylococcus aureus by agr. Mol. Gen.
Genet. 202, 58–61.
Said-Salim, B., Dunman, P.M., McAleese, F.M., Macapagal,
D., Murphy, E., McNamara, P.J., Arvidson, S., Foster,
T.J., Projan, S.J., Kreiswirth, B.N., 2003. Global regulation
of Staphylococcus aureus genes by Rot. J. Bacteriol. 185,
610–619.
Schmidt, K.A., Manna, A.C., Gill, S., Cheung, A.L., 2001.
SarT, a repressor of alpha-hemolysin in Staphylococcus
aureus. Infect. Immun. 69, 4749–4758.
Schmidt, K.A., Manna, A.C., Cheung, A.L., 2003. SarT
influences sarS expression in Staphylococcus aureus. Infect.
Immun. 71, 5139–5148.
Shen-Orr, S.S., Milo, R., Mangan, S., Alon, U., 2002.
Network motifs in the transcriptional regulation network
of Escherichia coli. Nat. Genet. 31, 64–68.
Sterba, K.M., Mackintosh, S.G., Blevins, J.S., Hurlburt, B.K.,
Smeltzer, M.S., 2003. Characterization of Staphylococcus
aureus SarA binding sites. J. Bacteriol. 185, 4410–4417.
Tegmark, K., 2000. Regulation of Virulence Gene Expression
in Staphylococcus aureus. Karolinska Institutet, Stockholm.
Tegmark, K., Karlsson, A., Arvidson, S., 2000. Identification
and characterization of SarH1, a new global regulator of
virulence gene expression in Staphylococcus aureus. Mol.
Microbiol. 37, 398–409.
Truong-Bolduc, Q.C., Zhang, X., Hooper, D.C., 2003.
Characterization of NorR protein, a multifunctional
regulator of norA expression in Staphylococcus aureus.
J. Bacteriol. 185, 3127–3138.
Wall, M.E., Dunlop, M.J., Hlavacek, W.S., 2005. Multiple
functions of a feed-forward-loop gene circuit. J. Mol. Biol.
349, 501–514.
Yarwood, J.M., McCormick, J.K., Schlievert, P.M., 2001.
Identification of a novel two-component regulatory system
that acts in global regulation of virulence factors of
Staphylococcus aureus. J. Bacteriol. 183, 1113–1123.