ESX-5-deficient Mycobacterium marinum ishypervirulent in adult zebrafish

Eveline M. Weerdenburg,1 Abdallah M. Abdallah,1,2†

Suman Mitra,3 Karin de Punder,2

Nicole N. van der Wel,2 Steve Bird,3

Ben J. Appelmelk,1 Wilbert Bitter1 andAstrid M. van der Sar1*1Department of Medical Microbiology and InfectionControl, VU University Medical Center, Amsterdam, theNetherlands.2Division of Cell Biology-B6, Netherlands CancerInstitute-Antoni van Leeuwenhoek Hospital (NKI-AVL),Amsterdam, the Netherlands.3Scottish Fish Immunology Research Centre, School ofBiological Sciences, University of Aberdeen, Aberdeen,UK.

Summary

ESX-5 is a mycobacterial type VII protein secretionsystem responsible for transport of numerous PEand PPE proteins. It is involved in the induction ofhost cell death and modulation of the cytokineresponse in vitro. In this work, we studied theeffects of ESX-5 in embryonic and adult zebrafishusing Mycobacterium marinum. We found thatESX-5-deficient M. marinum was slightly attenu-ated in zebrafish embryos. Surprisingly, the samemutant showed highly increased virulence in adultzebrafish, characterized by increased bacterialloads and early onset of granuloma formation withrapid development of necrotic centres. This earlyonset of granuloma formation was accompaniedby an increased expression of pro-inflammatorycytokines and tissue remodelling genes inzebrafish infected with the ESX-5 mutant. Experi-ments using RAG-1-deficient zebrafish showedthat the increased virulence of the ESX-5 mutantwas not dependent on the adaptive immunesystem. Mixed infection experiments with wild-type and ESX-5 mutant bacteria showed that thelatter had a specific advantage in adult zebrafish

and outcompeted wild-type bacteria. Together ourexperiments indicate that ESX-5-mediated proteinsecretion is used by M. marinum to establish amoderate and persistent infection.

Introduction

Mycobacterium tuberculosis is a very successful intracel-lular pathogen that is able to persist for decades within itshuman host. An important factor that determines patho-genicity of this infectious agent is secretion of virulencefactors. In order to transport proteins across their complexcell envelope, mycobacteria have a set of homologoussecretion systems which are also known as type VIIsecretion systems (T7SS) and are encoded by the ESXloci in their genomes (Abdallah et al., 2007; Bottai andBrosch, 2009). ESX-1 was the first T7SS to be identifiedand is the most studied one. It is required for secretion ofseveral virulence factors, such as ESAT-6 and CFP-10(Guinn et al., 2004). Pathogenic mycobacteria of theM. tuberculosis cluster that lack ESX-1, such as Myco-bacterium microti, the natural mutant Mycobacteriumbovis BCG and a deletion mutant of M. tuberculosis,display reduced virulence in vivo when compared with theESX-1 knock-in or wild-type M. tuberculosis strains (Pymet al., 2002; Lewis et al., 2003).

Recently, also ESX-5 was identified as an active proteinsecretion system (Abdallah et al., 2006; 2009). UnlikeESX-1, the ESX-5 gene cluster is restricted to slow-growing, mostly pathogenic mycobacteria such asM. tuberculosis and the fish pathogen Mycobacteriummarinum. In these species, ESX-5 has been shown to beresponsible for the transport of different members of thePE and PPE protein families (Abdallah et al., 2006; 2009;Daleke et al., 2011). For example, almost all membersof the PE_PGRS subfamily of PE proteins that areexpressed during in vitro growth of M. marinum, werefound to be secreted in an ESX-5-dependent manner(Abdallah et al., 2009). This indicates that this pathogenmay secrete more than 100 different proteins via ESX-5.PE and PPE proteins are specific for mycobacteria andpartially associated with the ESX loci. In several myco-bacterial species that contain ESX-5, PE and PPE genesare highly expanded throughout the whole genome (Geyvan Pittius et al., 2006). For instance, in M. tuberculosis,about 10% of the coding potential of the genome is

Received 26 August, 2011; revised 5 January, 2011; accepted 9January, 2012. *For correspondence. E-mail: a.vandersar@vumc.nl;Tel. (+31) 20 4448296; Fax (+31) 20 4448319.†Present address: Computational Biosciences Research Center, KingAbdullah University of Science and Technology, Thuwal, SaudiArabia.

Cellular Microbiology (2012) 14(5), 728–739 doi:10.1111/j.1462-5822.2012.01755.xFirst published online 15 February 2012

© 2012 Blackwell Publishing Ltd

cellular microbiology

devoted to PE and PPE proteins (Cole et al., 1998).Together, these characteristics suggest that these ESX-5-secreted proteins are important for virulence. However,their exact function is not clear.

Over the last years, different roles for PE and PPEproteins have been proposed. Several PE and PPE pro-teins were found to be located on the cell surface ofmycobacteria, suggesting interaction with macrophagesor other host cells (Brennan et al., 2001; Banu et al., 2002;Delogu et al., 2004). Other studies showed that individualPE and PPE proteins are involved in immune modulationof macrophages or virulence in vivo (Ramakrishnan et al.,2000; Li et al., 2005; Basu et al., 2007; Nair et al., 2009).In order to study the effects of a large group of PE and PPEproteins simultaneously, mycobacterial pathogens dis-turbed in ESX-5- mediated protein secretion can be used.With this approach, ESX-5 of M. marinum was found to beinvolved in the induction of cell death and modulation ofthe cytokine response of macrophages (Abdallah et al.,2008). However, in vivo experiments with ESX-5-deficientmycobacteria that can provide more insight into the role ofmultiple PE and PPE proteins during infection have notbeen described yet. Therefore, we made use of thezebrafish infection model to study the effect of all ESX-5effector proteins during M. marinum infection of a naturalhost. Surprisingly, our results show that loss of proteinsecretion by ESX-5 deficiency leads to hypervirulence inadult zebrafish. In contrast, disruption of ESX-5 does notaffect virulence in macrophages and even leads to attenu-ation in zebrafish embryos. Our data show that myco-bacterial ESX-5-mediated protein secretion plays animportant role during infection in vivo.

Results

Replication of ESX-5-deficient M. marinum is normal inhuman macrophages

Previously, disruption of M. marinum ESX-5 has beenshown to affect cell death and cytokine release of mac-rophages (Abdallah et al., 2008). These results suggestthat secreted PE and PPE proteins play an important roleduring the initial steps of infection. To determine whetherthe ESX-5 secretion system is also important for intracel-lular replication and spread, THP-1 macrophages wereinfected with M. marinum wild-type and ESX-5 mutantstrains and bacterial growth was assessed. Intracellularcolony-forming unit (cfu) counts up to 3 days after infectiondid not show a significant difference between both strains,indicating that ESX-5 substrates have no significant effecton intracellular replication of M. marinum (Fig. 1A). Fluo-rescence microscopy confirmed that also cell to cellspread was not affected, since a comparable increase inthe percentage of infected macrophages throughout the

course of infection with both strains was observed(Fig. 1B). These results seem to contradict a report pub-lished previously, in which ESX-5 was suggested to beinvolved in macrophage escape and spreading to othercells (Abdallah et al., 2006). However, those earlier resultswere obtained with a mutant that turned out to be defectivein both ESX-5 and ESX-1 (Abdallah et al., 2009), indicat-ing that the observed effects could be due to the absenceof ESX-1 or a combination of the two mutations.

ESX-5 is involved in bacterial growth at the early stageof zebrafish embryo infection

Although disruption of ESX-5 did not have an effect onintracellular replication of M. marinum, it induces the pro-duction of pro-inflammatory cytokines such as TNF-a,

Fig. 1. ESX-5 is not required for bacterial replication and spread inmacrophages.A. PMA-stimulated THP-1 cells were infected with wild-type andESX-5-deficient M. marinum at an moi of 0.5 and at several timepoints intracellular bacterial growth was determined by cfu analysis.Values represent mean � standard error of four differentexperiments.B. THP-1 cells were infected with wild-type and ESX-5-deficientM. marinum at an moi of 10 and after 2 and 5 days the percentageof infected cells was determined by fluorescence microscopy.Values represent mean � standard error of four differentexperiments.

Hypervirulence of M. marinum ESX-5 mutant 729

© 2012 Blackwell Publishing Ltd, Cellular Microbiology, 14, 728–739

IL-12p40 and IL-6 in human macrophages within 24 h(Abdallah et al., 2008) (Fig. S1). This increase in pro-inflammatory cytokine levels may lead to enhancedattraction of other immune cells, resulting in improvedclearance of the bacteria in vivo. Alternatively, it couldfacilitate bacterial spread to neighbouring cells. In orderto investigate whether ESX-5 affects the outcome ofinfection in vivo, we infected zebrafish embryos withM. marinum wild-type and ESX-5 mutant strains anddetermined bacterial growth by whole embryo plating. Upto 5 days after infection, bacterial counts in zebrafishembryos infected with the ESX-5 mutant were signifi-cantly reduced when compared with the M. marinumwild-type strain (Fig. 2A). This effect could be partlycomplemented by reintroduction of the disrupted gene(Mmar_2676) in the ESX-5 mutant strain. At later timepoints, however, equal amounts of bacteria were found inall groups. These results indicate that the effect of inacti-vation of ESX-5-mediated protein secretion on theimmune response of macrophages may lead to increasedbacterial clearance during the first few days of infection inzebrafish embryos.

Since zebrafish embryos are transparent, clustering ofcells infected with mCherry expressing M. marinum strains

– a process called early granuloma formation (Davis et al.,2002) – can be visualized and followed in time. It is knownthat M. marinum strains deficient for ESX-1-mediatedprotein secretion are impaired in the induction of thisprocess as well as growth in zebrafish embryos (Volkmanet al., 2004; Stoop et al., 2011). Since ESX-5-deficientM. marinum is also attenuated in growth during the first fewdays of infection, we examined whether ESX-5 plays a rolein early granuloma formation. Using fluorescence micros-copy, we found that ESX-5-deficient M. marinum was stillable to elicit early granuloma formation, although theamount of granulomas was considerably reduced as com-pared with the wild-type strain (Fig. 2B and C). No differ-ences in the size of early granulomas were observed(Fig. 2D). Our results indicate that ESX-5-mediated proteinsecretion by M. marinum plays a moderate role in the earlyphase of infection of the zebrafish embryo.

ESX-5 deficiency leads to increased bacterial growth inadult zebrafish which is independent of the adaptiveimmune system

Zebrafish embryos are well suited to study the earlysteps of infection with M. marinum in a context of innate

Fig. 2. ESX-5 is important at the initial stage of infection of zebrafish embryos.A. Twenty-eight hpf zebrafish embryos were infected with 102 cfu of wild-type, ESX-5-deficient or complemented ESX-5-deficient M. marinumand bacterial growth was determined by cfu analysis after whole embryo plating. Values represent mean � standard error of cfu from at least10 fish per time point.B. Zebrafish embryos were infected as indicated and after 5 days clustering of infected macrophages (early granuloma formation) wasvisualized by fluorescence microscopy. Scale bars are 500 mm.C. Early granuloma formation in zebrafish embryos infected with wild-type or ESX-5-deficient M. marinum was quantified (Stoop et al., 2011).Values represent mean � standard error of four different experiments.D. High magnification of a cluster of infected macrophages from zebrafish embryos infected for 5 days with wild-type or ESX-5-deficientM. marinum. Scale bars are 10 mm.

730 E. M. Weerdenburg et al.

© 2012 Blackwell Publishing Ltd, Cellular Microbiology, 14, 728–739

immunity. However, chronic mycobacterial infection is notestablished in embryos due to the absence of full immu-nity. Therefore, we infected 1-year-old adult zebrafish withwild-type and ESX-5-deficient M. marinum strains and fol-lowed these fish in time. Normally, zebrafish infected withthe M. marinum wild-type strain used in this study (E11)do not show clear symptoms of disease until 6–8 weeks ofinfection (van der Sar et al., 2004). To our surprise,zebrafish infected with the isogenic M. marinum ESX-5mutant became moribund within 2–3 weeks of infection(Fig. 3A), and had to be euthanized in accordance withthe rules of our ethical committee. In contrast, adultzebrafish that were infected with the wild-type strainappeared completely healthy at this time. In a secondexperiment, we again observed increased virulence,although this time the ESX-5 mutant induced terminalillness after 4 weeks of infection. Determination of bacte-rial loads in the organs of infected fish revealed thatbacterial numbers of both wild-type and ESX-5-deficientM. marinum were similar after 6 days of infection.However, at 11 days post infection the ESX-5 mutantshowed significantly increased bacterial numbers (Fig. 3B

and C). In spleens and livers of fish infected with theESX-5 mutant, bacterial loads continued to increase up to4 weeks of infection, whereas cfu levels of the wild-typestrain remained relatively constant during this period. Withthe exception of the latest time point in the spleen, infec-tion with the complemented mutant strain resulted in cfulevels comparable to wild-type infection indicating that theobserved increase in bacterial numbers was due to theabsence of a functional ESX-5 secretion system.

Our experiments show that ESX-5-deficient M. mari-num becomes hypervirulent in adult zebrafish, but not inmacrophages or embryonic zebrafish. Since embryos relysolely on innate immunity for their defence against patho-gens, we reasoned that the adaptive immune systemmight mediate the increased virulence observed after 1week of infection in adult zebrafish. Therefore, we infectedadult ragI-/- zebrafish, which lack functional T and B lym-phocytes but possess macrophages and granulocytes(Petrie-Hanson et al., 2009). Determination of bacterialloads in livers and spleens 2 weeks post infectionrevealed that in ragI-/- zebrafish, ESX-5-deficient M. mari-num still had a growth advantage over the wild-type strain,

Fig. 3. In adult zebrafish bacterial growth of ESX-5-deficient M. marinum is highly increased.A. Kaplan-Meier survival curve of adult zebrafish following infection with 2 ¥ 104 cfu wild-type or ESX-5-deficient M. marinum. Fish wereterminated when signs of severe disease were displayed (i.e. skin ulcers, lethargy). At least 10 zebrafish per condition were used.Corresponding bacterial loads in organs of these fish after 16 days of infection can be found in Fig. S2.B and C. Adult zebrafish were infected with 2 ¥ 104 cfu wild-type, ESX-5-deficient, complemented ESX-5-deficient or 1 ¥ 104 cfu wild-typetogether with 1 ¥ 104 cfu ESX-5-deficient M. marinum and after 6, 11 and 28 days the bacterial loads in liver (B) and spleen (C) weredetermined by cfu analysis on 7H10 plates. Values represent mean � standard error of cfu counts obtained from six zebrafish per group.Statistical differences were determined by the Student’s t-test (*P < 0.05, **P < 0.01, ***P < 0.001).D. Adult isogenic rag1–/– and rag1+/+ zebrafish siblings were infected with 2 ¥ 104 cfu wild-type or ESX-5-deficient M. marinum and after 14days bacterial loads in livers and spleens were determined by cfu analysis on 7H10 plates. Values represent mean � SE of cfu countsobtained from at least three zebrafish per group.

Hypervirulence of M. marinum ESX-5 mutant 731

© 2012 Blackwell Publishing Ltd, Cellular Microbiology, 14, 728–739

indicating that its hypervirulence is not mediated by theadaptive immune system of the host (Fig. 3D). We founddecreased bacterial numbers in ragI-/- zebrafish ascompared with wild-type zebrafish, which contradictswith a previous report showing hypersusceptibility ofragI-/- zebrafish to M. marinum M-strain infection within 1week post infection (Swaim et al., 2006). Possibly, theincreased numbers of granulocytes that are present inragI-/- zebrafish (Petrie-Hanson et al., 2009) may lead toenhanced innate immunity and improved clearance of theM. marinum wild-type strain E11 that is used in this study.This isolate induces chronic progressive disease inzebrafish whereas the M-strain, which is used in mostother studies, causes acute disease (van der Sar et al.,2004). In addition, studies in ragI-deficient mice show thatincreased susceptibility to mycobacterial infection is notnecessarily reflected in the bacterial burden (Ehlers et al.,2001; Chackerian et al., 2002; Kursar et al., 2007).Overall, our results show that ESX-5-deficient M. mari-

num is hypervirulent in adult zebrafish due to mechanismsthat do not involve the adaptive immune system.

Absence of the ESX-5-mediated protein secretion leadsto rapid and increased granuloma formation inadult zebrafish

Histological analysis revealed a high amount of solidgranulomas in the spleens and pancreases of zebrafishinfected for 11 days with the M. marinum ESX-5 mutant,whereas only few granulomas were present in fishinfected with the wild-type strain or complemented mutant(Fig. 4A and B). Four weeks after infection, the highnumber of granulomas elicited by the ESX-5 mutantlargely destroyed the cellular organization of the pan-creas. About half of these lesions contained a caseousnecrotic centre and some of them were multifocal. At thistime point of 4 weeks after infection, granulomas in thepancreases of fish infected with the wild-type strain had

Fig. 4. ESX-5-deficient M. marinum causes increased granuloma formation in adult zebrafish.A. Adult zebrafish infected for 11 days with wild-type, ESX-5-deficient or complemented ESX-5-deficient bacteria were analysed for granulomaformation by HE staining of histological sections. Stars indicate (solid) granulomas. Scale bars are 20 mm.B. Quantification of the amount of granulomas in zebrafish infected with M. marinum wild-type or ESX-5 mutant strain. Values representmean � standard error of one (6 dpi), two (28 dpi) or three (11 dpi) fish per group.C. HE staining of a typical granuloma in the pancreas of wild-type or ESX-5 mutant infected zebrafish (28 dpi). Centre of granuloma inzebrafish infected with the ESX-5 mutant shows extensive necrosis. Star indicates solid granuloma. Scale bars are 50 mm.

732 E. M. Weerdenburg et al.

© 2012 Blackwell Publishing Ltd, Cellular Microbiology, 14, 728–739

developed to an extent comparable to the 11-day stage inESX-5 mutant infected fish (Fig. 4A and B) and were allnon-necrotic (Fig. 4B and C). Very few granulomas werepresent in the spleens. These data suggest that ESX-5has an effect on the timing of granuloma formation but noton the process itself.

Granuloma formation in zebrafish infected withESX-5-deficient M. marinum is accompanied byupregulation of pro-inflammatory genes

In macrophage infection experiments, ESX-5-deficientM. marinum was shown to specifically induce the produc-tion of pro-inflammatory cytokines (Abdallah et al., 2008)(Fig. S1). In order to get more insight in the immunologicalprocesses that occur during infection in vivo and whichmay underlie the observed hypervirulence, we measuredmRNA levels of several immunity-related genes in liversand spleens of adult zebrafish during infection withM. marinum wild-type and ESX-5-deficient strains andcompared them to uninfected fish. After 1 week of infectionwith either of the strains, we did not find clear up- ordownregulation for most of the genes tested (Fig. 5A),

except for an infection-dependent increase in expressionlevels of the matrix metalloproteinase gene mmp13. Twoweeks after infection, expression levels of this tissueremodelling gene, which is suggested be involved ingranuloma formation (van der Sar et al., 2009), were stillincreased in organs of fish infected with the ESX-5 mutantbut not in those infected with the wild-type strain (Fig. 5B).At this time point we also observed a specific upregulationof transcript levels of the pro-inflammatory cytokines tnf-a,ifn-g and il-1b in organs of zebrafish infected with ESX-5-deficient M. marinum. These gene expression patternscorrelate with the increased bacterial load and infection-associated pathology. Therefore, our results suggest thatthere are no major mycobacteria-induced modifications ofthe inflammatory response that precede specific outgrowthof ESX-5-deficient M. marinum, but that the upregulation ofpro-inflammatory cytokines at the infection site is a markerof disease progression. When mRNA levels of whole fishwere compared, specific effects of ESX-5-deficiency onhost gene transcription were, probably due to a dilution ofthe observed tissue effects, not observed (results notshown). This underlines the importance of using tissueswith high numbers of infected host cells.

Fig. 5. Gene expression levels in spleens ofadult zebrafish during infection withM. marinum. RNA was extracted from spleensof infected or uninfected adult zebrafish after7 (A) and 14 (B) days of infection and mRNAlevels of ifn-g, tnf-a, il-1b, il-4, il-10, mmp-9and mmp-13 were measured. Geneexpression levels of zebrafish infected withM. marinum were compared with those ofmock-infected fish. Graphs represent relativeexpression levels in spleens of three fish pergroup.

Hypervirulence of M. marinum ESX-5 mutant 733

© 2012 Blackwell Publishing Ltd, Cellular Microbiology, 14, 728–739

M. marinum ESX-5 does not affect the hypoxicresponse in vitro

Since we found no indications for differential host geneexpression in response to infection or involvement of theadaptive immune system in the increased bacterialgrowth, rapid induction of granulomas and early death ofadult zebrafish infected with ESX-5-deficient M. marinum,we examined the possibility that these effects were notdue to an inappropriate host response but instead a resultof bacterial adaptations to the environment created by thehost. Studies on different clinical isolates of M. tuberculo-sis have shown that strains of the recently evolved Beijinglineage display increased virulence, probably as a resultof enhanced and constitutive expression of the DosRregulon. This regulon is a dormancy-associated transcrip-tional programme that is essential for mycobacteria tosurvive in hypoxic conditions and for a rapid resumption ofgrowth upon transition to more favourable conditions(Reed et al., 2007; Leistikow et al., 2010). Compared withmacrophages and zebrafish embryos, tissues of adultzebrafish are probably more prone to hypoxic conditions.For this reason increased growth of the M. marinumESX-5 mutant in adult zebrafish may be a result of anenhanced ability to adapt to anaerobic conditions. To testthis possibility we measured mRNA expression of theM. marinum orthologues of M. tuberculosis dosR and theDosR-regulated gene hspX in wild-type and ESX-5mutant strains under standard growth conditions in vitro.However, we did not find any differences in gene expres-sion indicating normal DosR-mediated gene regulation(Fig. 6A). In addition, bacterial growth in vitro undernormal and hypoxic conditions did not differ between thetwo strains (Fig. 6B and C), nor was there a difference inantibiotic sensitivity (not shown). Furthermore, ESX-5-deficient and wild-type M. marinum re-isolated frominfected adult zebrafish showed similar growth ratesduring re-infection of THP-1 cells, indicating that presencein the host does not lead to enduring changes in ESX-5-related growth characteristics (data not shown).

Wild-type mycobacteria cannot suppress hypervirulenceof the ESX-5 mutant strain

Previous in vitro experiments showed that the M. marinumwild-type strain is able to suppress the macrophage pro-inflammatory immune response that is elicited by ESX-5-deficient M. marinum (Abdallah et al., 2008). Therefore,we investigated whether the M. marinum wild-type strainis also able to suppress the detrimental effects caused bythe ESX-5 mutant strain in vivo. To this end we infectedadult zebrafish simultaneously with wild-type and ESX-5mutant bacteria. Total bacterial growth in this mixed infec-tion was comparable to wild-type-only infection up to 11

Fig. 6. Hypervirulence of ESX-5 mutant strain cannot be explainedwith in vitro growth characteristics.A. qRT-PCR analysis of M. marinum dosR [Mmar_3480 (1) andMmar_1516 (2)] and hspX_1 (Mmar_3484) mRNA levels. Geneexpression levels were normalized to sigA. Values represent theaverage � standard error of three experiments.B and C. Mycobacterial growth of M. marinum wild-type and ESX-5mutant strains in 7H9 liquid agar with normal (B) or low oxygenlevels (C) was determined by OD600 and cfu analysis at severaltime points. Values represent the average � standard error of threeexperiments.

734 E. M. Weerdenburg et al.

© 2012 Blackwell Publishing Ltd, Cellular Microbiology, 14, 728–739

days (Fig. 3B and C), suggesting that wild-type bacteriawere indeed able to suppress hypervirulence of theESX-5 mutant strain. However, 4 weeks after infection,bacterial loads in livers and spleens had increased tolevels observed in the mutant-only infection. Analysis ofresulting colonies showed that at this time point more than95% of all bacteria were ESX-5-deficient, whereas equalnumbers of wild-type and mutant bacteria had beeninjected. This indicates that the ESX-5 mutant had out-grown the M. marinum wild-type strain during infection. Inaddition, these data suggest that the wild-type strain is notable to suppress the detrimental effects caused by theESX-5 mutant and moreover, that wild-type bacteria werenot able to profit from the increased granuloma formationand putative immune modulation caused by the mutant.Together, our data indicate that ESX-5-deficient M. mari-num has a specific growth advantage in adult zebrafish.

Discussion

In this report, we show that inactivation of the ESX-5secretion system in M. marinum leads to increased viru-lence in a natural host. Hypervirulence of this M. marinumESX-5 mutant strain was characterized by highlyincreased bacterial loads and early death of adultzebrafish. In addition, well structured granulomas devel-oped much more rapidly and in higher numbers than inzebrafish infected by wild-type bacteria. Hypervirulenceas a consequence of gene deletion in mycobacteria hasbeen described before. For example, disruption of themce1 operon in M. tuberculosis leads to increasedbacterial growth, poorly organized granulomas and adiminished pro-inflammatory cytokine response in mice(Shimono et al., 2003). Interestingly, overexpression ofthe mce1 operon by disrupting the mce1 repressor proteinmce1R had a similar effect on virulence, although in thiscase an accelerated immune response and granulomaformation was observed (Uchida et al., 2007). The lattersituation is reminiscent of the results we obtained with theM. marinum ESX-5 mutant strain in zebrafish, opening theinteresting possibility of a link between ESX-5 and Mce1.

It is remarkable that the inability to secrete a substantialamount of proteins, which include all PE_PGRS proteinsthat are expressed in vitro and may add up to more than100 proteins, results in such a large increase in virulence.Given the evolutionary expansion of PE and PPE proteinfamily members in pathogenic mycobacterial species(Gey van Pittius et al., 2006), these proteins would beexpected to have an important function in virulence.One possible explanation for hypervirulence of ESX-5-deficient M. marinum would be that ESX-5-mediated PEand PPE protein secretion imposes a high metabolicburden on the bacteria explaining the increased bacterialgrowth rate of the mutant. However, our experiments

showed that this is not the case during growth in culturemedium or during infection of macrophages and zebrafishembryos. In addition, the ESX-5 mutant strain of M. mari-num did not display an increased growth rate in vitroduring hypoxia nor altered hspX/dosR gene expression,which are indicative for increased virulence in the M.tuberculosis Beijing family strains (Reed et al., 2007).Therefore our data suggest that the hypervirulent charac-teristics of ESX-5-deficient M. marinum are due to theinterplay between bacteria and host.

Previous experiments showed that ESX-5 substratessuppress the release of pro-inflammatory cytokines bymacrophages (Abdallah et al., 2008). Elevated pro-inflammatory cytokine levels may be the cause of thereduced bacterial numbers we found in zebrafish embryosduring the early phase of infection with the ESX-5 mutant.However, at later stages of infection attenuation is notobserved anymore in these embryos. In contrast, in adultzebrafish infected with the M. marinum ESX-5 mutantbacterial loads were already after 11 days of infection atleast 1 log higher as compared with the wild-type strain.One of the differences between embryonic and adultzebrafish is the presence of a functional adaptive immunesystem. In embryos, it takes several weeks to achieveimmunocompetence (Lam et al., 2004). Before this timepoint, bacterial infections are primarily dealt with by theinnate immune system. Adult zebrafish can also rely onthe adaptive immune system, which plays an importantrole in the defence against pathogens. This was illustratedby a study using adult ragI-/- zebrafish, which showed thedetrimental effects of the absence of adaptive immunityalready 1 week after mycobacterial infection (Swaimet al., 2006). As hypervirulence of the M. marinum ESX-5mutant was only observed in the immunocompetent adultzebrafish after more than 1 week of infection, we hypoth-esized that substrates of ESX-5 may interact with media-tors of the adaptive immune system leading to control ofinfection. However, infection of ragI-/- adult zebrafishshowed a growth advantage of ESX-5-deficient M. mari-num despite the absence of a functional adaptive immunesystem. Hypervirulence is therefore probably mediated byother factors that differ between embryonic and adultzebrafish. We showed that during infection of adultzebrafish with a mixture of both the M. marinum wild-typeand ESX-5 mutant strains, the mutant outcompeted thewild-type strain. The fact that wild-type bacteria were notable to profit from the mutant’s hypervirulence, suggeststhat the advantage of ESX-5-deficient M. marinum doesnot result from an altered extracellular environment orgeneral immune response but instead may be due tomore local and possibly intracellular effects.

Although many haematopoietic cell types are producedalready at early stages during embryonic development,these cells can be quite different from the ones that are

Hypervirulence of M. marinum ESX-5 mutant 735

© 2012 Blackwell Publishing Ltd, Cellular Microbiology, 14, 728–739

present in adult zebrafish. This is a result of several wavesof haematopoiesis that occur during development of theembryo (Herbomel et al., 1999). Therefore, there may bespecific markers of adult macrophages that are notpresent on or in primitive cells that interact with ESX-5substrates resulting in control of infection. It is also pos-sible that ESX-5-deficient M. marinum is able to spreadmore easily to other cell types. Due to the absence of cellsurface-localized PE_PGRS and other ESX-5-dependentproteins, the cell wall of ESX-5-deficient M. marinum isaltered (J. Bestebroer, unpublished), which may have aneffect on interaction with different types of host cells.The structural changes of the mycomembrane may alsohave an effect on the exposure of cell wall lipids, some ofwhich induce granuloma formation (Hamasaki et al.,2000). It is possible that altered exposure of these cellsurface lipids may result in dramatic differences in theoutcome of infection.

The granulomas that are formed in zebrafish duringinfection with both ESX-5 mutant and wild-type M. mari-num are structurally well organized, which is also charac-teristic for human M. tuberculosis granulomas. Asgranuloma development progresses, infected macroph-ages in the centre undergo necrosis, leading to the accu-mulation of caseum in which mycobacteria can multiply(Dannenberg, 1994; Swaim et al., 2006). Necrotizinggranulomas are usually formed at late stages of mycobac-terial infection. In organs of zebrafish, necrotic granulo-mas are present after 8 weeks of infection with wild-typeM. marinum (van der Sar et al., 2004). However, zebrafishinfected with ESX-5 mutant mycobacteria develop necro-tizing granulomas much more rapidly and in highamounts. The rapid necrosis, which facilitates mycobac-terial replication, may account for the increased cfunumbers we find in these fish. Possibly, the type of celldeath induced by the ESX-5 mutant strain during infectionis necrotic rather than apoptotic leading to specific out-growth. Recently, ESX-5 has been shown to play a role inhost cell death (Abdallah et al., 2011). As the induction ofapoptosis by PE proteins has also been reported (Balajiet al., 2007; Cadieux et al., 2011), it may be possible thatthe absence of these proteins may lead to a shift to amore necrotic type of cell death.

The rapid development of granulomas and eventualtissue destruction in zebrafish infected with the ESX-5-deficient M. marinum correlates with the prolonged induc-tion of the tissue remodelling genes mmp9 and mmp13.These genes encode matrix metalloproteinases thathave been suggested to be involved in granuloma forma-tion during mycobacterial infection (Taylor et al., 2006;Gonzalez-Avila et al., 2009; van der Sar et al., 2009). Thelarge increase in transcription of ifn-g, tnf-a and il-1b after2 weeks indicates high levels of inflammation induced bythe rapidly progressing infection, which probably leads to

tissue damage rather than effective bacterial clearance.At 1 week of infection, just before the induction ofgranulomas, we did not observe significantly alteredgene expression levels. Since an upregulation of pro-inflammatory cytokines has also been observed at theend-stage of E11 infection (6–8 weeks) (van der Sar et al.,2009), these data suggest that the host’s immuneresponse to the M. marinum ESX-5 mutant strain ispushed forward in time but not significantly different fromthe response to wild-type M. marinum infection.

In summary, we show here that the ESX-5 proteinsecretion system of M. marinum plays an important role inprogression of infection in a natural host. Our researchindicates that ESX-5-mediated protein secretion is impor-tant for the establishment of a moderate and persistentinfection. Consequently, PE and PPE proteins might bevery important for mycobacterial persistence. In thisrespect it is interesting to note that recently manymembers of these protein families were found to be abun-dantly expressed at the chronic state of M. tuberculosisinfection in guinea pigs (Kruh et al., 2010), supporting thishypothesis.

Experimental procedures

Bacterial strains and growth conditions

The mCherry-expressing M. marinum wild-type strain E11 (vander Sar et al., 2004), its isogenic ESX-5-deficient transposonmutant (Tn::ESX-5, transposon mutant of Mmar_2676) and thecomplemented ESX-5 mutant (Tn::ESX-5c) (Abdallah et al.,2009) used in this study were grown at 30°C on Middlebrook7H10 agar plates supplemented with 10% OADC (oleic acid–albumin–dextrose–catalase, BD Biosciences) or in shaking cul-tures in Middlebrook 7H9 liquid medium supplemented with 10%ADC (albumin–dextrose–catalase, BD Biosciences) and 0.05%Tween 80. Kanamycin, hygromycin and chloramphenicol wereadded when required at a concentration of 25, 50 and 30 mg ml-1

respectively. Anaerobic conditions were set up according to theWayne model (Wayne and Hayes, 1996).

Animals

Male 1-year-old Danio rerio wild-type and ragI-/- (hu1999)zebrafish were maintained at 28°C in 10 l tanks with aeratedfreshwater on a 14 h light/10 h dark cycle as described previously(van der Sar et al., 2004). Prior to infection experiments, wild-type and isogenic ragI-/- zebrafish were genotyped to confirmmutation of the ragI gene.

Cell culture and macrophage infection

The human monocytic cell line THP-1 was cultured at 37°C and5% CO2 in RPMI-1640 with Glutamax-1 (Gibco) supplementedwith 10% FBS, 100 mg ml-1 streptomycin and 100 U ml-1 penicil-lin. Monocytic differentiation into macrophage-like cells wasinduced by overnight incubation with 20 ng ml-1 PMA (Sigma).

736 E. M. Weerdenburg et al.

© 2012 Blackwell Publishing Ltd, Cellular Microbiology, 14, 728–739

For cell infection experiments, THP-1 cells were seeded in24-well plates (Costar) at a density of 4 ¥ 105 cells per well. Afterdifferentiation, cells were washed with infection medium(RPMI + 10% FBS) and infected with mycobacteria either at anmoi of 1 for 2 h (for mycobacterial replication analysis) or at anmoi of 20 for 1 h (for cytokine measurement) at 32°C and 5%CO2. Subsequently, cells were washed three times with infectionmedium and incubated at 32°C and 5% CO2. To analyse intrac-ellular mycobacterial replication, cells were collected at severaltime points, lysed with 1% Triton X-100 and plated in serialdilutions on 7H10 agar plates. For fluorescence microscopy,cells were seeded on glass coverslips, infected as describedabove and fixed for 2 h with 2% paraformaldehyde and 0.2%glutaraldehyde.

TNF-a secretion

Cell culture supernatants of THP-1 cells were harvested 24 hafter infection and assayed for TNF-a by ELISA kit (Biosource,Invitrogen) according to the manufacturer’s instructions.

Zebrafish infection

Mycobacteria were grown in 7H9 liquid medium to an OD600 of 1.0and washed in phosphate-buffered saline (PBS) prior to infection.Zebrafish embryos were infected 28 h post fertilization with 100cfu of M. marinum wild-type, Tn::ESX-5, Tn::ESX-5c or 1 nl ofPBS by micro-injection in the caudal vein, as described previ-ously (Stoop et al., 2011). Adult zebrafish were anaesthetized in0.02% MS-222 (Sigma) and injected intraperitoneally with 2 ¥ 104

cfu of M. marinum wild-type, Tn::ESX-5, Tn::ESX-5c or 1 ¥ 104

cfu wild-type together with 1 ¥ 104 cfu Tn::ESX-5. Uninfectedcontrol adult zebrafish were injected with 10 ml of PBS. By platingon 7H10 agar, the bacterial cfu in the injected inoculum wasconfirmed. All adult fish infection experiments were performeddouble-blind.

Bacterial quantification

At several time points after infection, at least 10 zebrafishembryos per group were homogenized in BBL Mycoprep (BDDiagnostics) and plated in 10-fold serial dilutions on 7H10 agarplates. Adult zebrafish were subjected to terminal anaesthesia inMS-222 at 6, 11, 14 and 28 days after infection. Subsequently,the livers and spleens of 6 (6, 11 and 28 dpi) or 3 (14 dpi)zebrafish per group were homogenized in BBL MycoPrep andplated in 10-fold serial dilutions on 7H10 agar plates to determinebacterial cfu.

Histopathological analysis

At 6, 11 and 28 days after infection, three adult zebrafish fish pergroup were fixed in 4% paraformaldehyde. Fixed animals wereembedded in paraffin and cut frontally in serial sections fromventral to dorsal. Sections were stained with haematoxylin andeosin (HE). A Zeiss Axioskop light microscope equipped with aLeica DC500 camera was used for imaging. Image J softwarewas used to adjust brightness and contrast.

RNA extraction and quantitative RT-PCR

At 7 and 14 days after infection, livers and spleens of three adultzebrafish from each group were snap frozen in liquid nitrogen andstored at -80°C. Zebrafish organs were homogenized in TRIzol(Invitrogen) using a drill and micropestles (Eppendorf). For myco-bacterial RNA isolation, pellets of 25 ml of bacterial cultures at anOD600 of 1.0 were resuspended in 1 ml of TRIzol. Bacterial cellswere disrupted by bead beating, followed by incubation at 60°Cfor 10 min. After centrifugation of homogenized zebrafish or lysedmycobacteria, supernatants were extracted with chloroform andRNA was precipitated with isopropanol. RNA pellets werewashed with 75% ethanol and taken up in RNase free water.Contaminating DNA was removed with RNase-free DNase(Fermentas) prior to RNA purification through a RNeasy MinEluteCleanup kit (Qiagen). SuperScript III reverse transcriptase (Invit-rogen) was used to generate cDNA, of which 1 mg of zebrafishRNA or 200 ng of mycobacterial RNA was used in the quantita-tive real-time PCR (qPCR) reactions. qPCR was performed usingthe SYBR GreenER qPCR kit (Invitrogen) and the LightCycler480 (Roche). Ct values were normalized to values obtained forb-actin, a constitutively expressed zebrafish gene, or the myco-bacterial housekeeping gene sigA.

Statistical analysis

Adult zebrafish infection experiments were performed at threeindependent occasions and similar differences between infectiongroups were found. Data shown here are derived from one rep-resentative experiment. Infection of ragI-/- zebrafish was per-formed twice. All statistical analyses were performed withGraphPad Prism software.

Acknowledgements

We thank Wim Schouten for fish care, Theo Verboom, Gunny vanden Brink and Esther Stoop for technical assistance, Ji-YingSong for help with histopathological analysis of the zebrafishsections and Peter Peters for providing the environment toperform microscopy experiments. This work has received fundingfrom the European Community’s Seventh Framework Pro-gramme ([FP7/2007–2013]) under grant agreement No. 201762.For providing the zebrafish knockout allele hu1999 we thank theHubrecht laboratory and the Sanger Institute Zebrafish MutationResource (ZF-MODELS Integrated Project; contract numberLSHG-CT-2003-503496; funded by the European Commission)also sponsored by the Wellcome Trust (grant number WT077047/Z/05/Z).

References

Abdallah, A.M., Verboom, T., Hannes, F., Safi, M., Strong, M.,Eisenberg, D., et al. (2006) A specific secretion systemmediates PPE41 transport in pathogenic mycobacteria.Mol Microbiol 62: 667–679.

Abdallah, A.M., Gey van Pittius, N.C., DiGiuseppeChampion, P.A., Cox, J., Luirink, J., Vandenbroucke-Grauls, C.M.J., et al. (2007) Type VII secretion – mycobac-teria show the way. Nat Rev Microbiol 5: 883–891.

Hypervirulence of M. marinum ESX-5 mutant 737

© 2012 Blackwell Publishing Ltd, Cellular Microbiology, 14, 728–739

Abdallah, A.M., Savage, N.D.L., van Zon, M., Wilson, L.,Vandenbroucke-Grauls, C.M.J.E., van der Wel, N.N., et al.(2008) The ESX-5 secretion system of Mycobacteriummarinum modulates the macrophage response. J Immunol181: 7166–7175.

Abdallah, A.M., Verboom, T., Weerdenburg, E.M., Gey vanPittius, N.C., Mahasha, P.W., Jimenez, C., et al. (2009)PPE and PE_PGRS proteins of Mycobacterium marinumare transported via the type VII secretion system ESX-5.Mol Microbiol 73: 329–340.

Abdallah, A.M., Bestebroer, J., Savage, N.D.L., de Punder,K., van Zon, M., Wilson, L., et al. (2011) Mycobacterialsecretion systems ESX-1 and ESX-5 play distinct roles inhost cell death and inflammasome activation. J Immunol187: 4744–4753.

Balaji, K.N., Goyal, G., Narayana, Y., Srinivas, M.,Chaturvedi, R., and Mohammad, S. (2007) Apoptosis trig-gered by Rv1818c, a PE family gene from Mycobacteriumtuberculosis is regulated by mitochondrial intermediates inT cells. Microbes Infect 9: 271–281.

Banu, S., Honore, N., Saint-Joanis, B., Philpott, D., Prevost,M.C., and Cole, S.T. (2002) Are the PE-PGRS proteins ofMycobacterium tuberculosis variable surface antigens?Mol Microbiol 44: 9–19.

Basu, S., Pathak, S.K., Banerjee, A., Pathak, S., Bhatta-charyya, A., Yang, Z., et al. (2007) Execution of macro-phage apoptosis by PE_PGRS33 of Mycobacteriumtuberculosis is mediated by Toll-like receptor 2-dependentrelease of tumor necrosis factor-a. J Biol Chem 282: 1039–1050.

Bottai, D., and Brosch, R. (2009) Mycobacterial PE, PPE andESX clusters: novel insights into the secretion of thesemost unusual protein families. Mol Microbiol 73: 325–328.

Brennan, M.J., Delogu, G., Chen, Y., Bardarov, S., Kriakov,J., Alavi, M., and Jacobs, W.R., Jr (2001) Evidence thatmycobacterial PE_PGRS proteins are cell surface constitu-ents that influence interactions with other cells. InfectImmun 69: 7326–7333.

Cadieux, N., Parra, M., Cohen, H., Maric, D., Morris, S.L.,and Brennan, M.J. (2011) Induction of cell death after local-ization to the host cell mitochondria by the Mycobacteriumtuberculosis PE_PGRS33 protein. Microbiology 157: 793–804.

Chackerian, A.A., Alt, J.M., Perera, T.V., Dascher, C.C., andBehar, S.M. (2002) Dissemination of Mycobacterium tuber-culosis is influenced by host factors and precedes theinitiation of T-cell immunity. Infect Immun 70: 4501–4509.

Cole, S.T., Brosch, R., Parkhill, J., Garnier, T., Churcher, C.,Harris, D., et al. (1998) Deciphering the biology ofMycobacterium tuberculosis from the complete genomesequence. Nature 393: 537–544.

Daleke, M.H., Cascioferro, A., de Punder, K., Ummels, R.,Abdallah, A.M., van der Wel, N., et al. (2011) ConservedPro–Glu (PE) and Pro–Pro–Glu (PPE) protein domainstarget LipY lipases of pathogenic mycobacteria to the cellsurface via the ESX-5 pathway. J Biol Chem 286: 19024–19034.

Dannenberg, A.M. (1994) Roles of cytotoxic delayed-typehypersensitivity and macrophage-activating cell-mediated-immunity in the pathogenesis of tuberculosis. Immuno-biology 191: 461–473.

Davis, J.M., Clay, H., Lewis, J.L., Ghori, N., Herbomel, P., andRamakrishnan, L. (2002) Real-time visualization ofMycobacterium–macrophage interactions leading toinitiation of granuloma formation in zebrafish embryos.Immunity 17: 693–702.

Delogu, G., Pusceddu, C., Bua, A., Fadda, G., Brennan, M.J.,and Zanetti, S. (2004) Rv1818c-encoded PE_PGRSprotein of Mycobacterium tuberculosis is surface exposedand influences bacterial cell structure. Mol Microbiol 52:725–733.

Ehlers, S., Benini, J., Held, H.D., Roeck, C., Alber, G., andUhlig, S. (2001) Alpha beta T cell receptor-positive cellsand interferon-gamma, but not inducible nitric oxide syn-thase, are critical for granuloma necrosis in a mouse modelof mycobacteria-induced pulmonary immunopathology.J Exp Med 194: 1847–1859.

Gey van Pittius, N.C., Sampson, S.L., Lee, H., Kim, Y., vanHelden, P.D., and Warren, R.M. (2006) Evolution andexpansion of the Mycobacterium tuberculosis PE and PPEmultigene families and their association with the duplica-tion of the ESAT-6 (esx) gene cluster regions. BMC EvolBiol 6: 95.

Gonzalez-Avila, G., Sandoval, C., Herrera, M.T., Ruiz, V.,Sommer, B., Sada, E., et al. (2009) Mycobacterium tuber-culosis effects on fibroblast collagen metabolism. Respira-tion 77: 195–202.

Guinn, K.M., Hickey, M.J., Mathur, S.K., Zakel, K.L., Grotzke,J.E., Lewinsohn, D.M., et al. (2004) Individual RD1-regiongenes are required for export of ESAT-6/CFP-10 and forvirulence of Mycobacterium tuberculosis. Mol Microbiol 51:359–370.

Hamasaki, N., Isowa, K.I., Kamada, K., Terano, Y.,Matsumoto, T., Arakawa, T., et al. (2000) In vivo adminis-tration of mycobacterial cord factor (trehalose 6,6′-dimycolate) can induce lung and liver granulomas andthymic atrophy in rabbits. Infect Immun 68: 3704–3709.

Herbomel, P., Thisse, B., and Thisse, C. (1999) Ontogenyand behaviour of early macrophages in the zebrafishembryo. Development 126: 3735–3745.

Kruh, N.A., Troudt, J., Izzo, A., Prenni, J., and Dobos, K.M.(2010) Portrait of a pathogen: the Mycobacterium tubercu-losis proteome in vivo. PLoS ONE 5: e13938.

Kursar, M., Koch, M., Mittruecker, H.W., Nouailles, G.,Bonhagen, K., Kamradt, T., and Kaufmann, S.H.E. (2007)Cutting edge: regulatory T cells prevent efficient clearanceof Mycobacterium tuberculosis. J Immunol 178: 2661–2665.

Lam, S.H., Chua, H.L., Gong, Z., Lam, T.J., and Sin, Y.M.(2004) Development and maturation of the immune systemin zebrafish, Danio rerio: a gene expression profiling, insitu hybridization and immunological study. Dev CompImmunol 28: 9–28.

Leistikow, R.L., Morton, R.A., Bartek, I.L., Frimpong, I.,Wagner, K., and Voskuil, M.I. (2010) The Mycobacteriumtuberculosis DosR regulon assists in metabolic homeosta-sis and enables rapid recovery from nonrespiring dor-mancy. J Bacteriol 192: 1662–1670.

Lewis, K.N., Liao, R., Guinn, K.M., Hickey, M.J., Smith, S.,Behr, M.A., and Sherman, D.R. (2003) Deletion of RD1from Mycobacterium tuberculosis mimics bacille Calmette-Guérin attenuation. J Infect Dis 187: 117–123.

738 E. M. Weerdenburg et al.

© 2012 Blackwell Publishing Ltd, Cellular Microbiology, 14, 728–739

Li, Y., Miltner, E., Wu, M., Petrofsky, M., and Bermudez, L.E.(2005) A Mycobacterium avium PPE gene is associatedwith the ability of the bacterium to grow in macrophagesand virulence in mice. Cell Microbiol 7: 539–548.

Nair, S., Ramaswamy, P.A., Ghosh, S., Joshi, D.C., Pathak,N., Siddiqui, I., et al. (2009) The PPE18 of Mycobacteriumtuberculosis interacts with TLR2 and activates IL-10 induc-tion in macrophage. J Immunol 183: 6269–6281.

Petrie-Hanson, L., Hohn, C., and Hanson, L. (2009) Charac-terization of ragI mutant zebrafish leukocytes. BMCImmunol 10: 8.

Pym, A.S., Brodin, P., Brosch, R., Huerre, M., and Cole, S.T.(2002) Loss of RD1 contributed to the attenuation ofthe live tuberculosis vaccines Mycobacterium bovisBCG and Mycobacterium microti. Mol Microbiol 46: 709–717.

Ramakrishnan, L., Federspiel, N.A., and Falkow, S. (2000)Granuloma-specific expression of Mycobacterium viru-lence proteins from the glycine-rich PE-PGRS family.Science 288: 1436–1439.

Reed, M.B., Gagneux, S., DeRiemer, K., Small, P.M., andBarry, C.E., III (2007) The W-Beijing lineage of Mycobac-terium tuberculosis overproduces triglycerides and hasthe DosR dormancy regulon constitutively upregulated.J Bacteriol 189: 2583–2589.

van der Sar, A.M., Abdallah, A.M., Sparrius, M., Reinders, E.,Vandenbroucke-Grauls, C.M.J., and Bitter, W. (2004)Mycobacterium marinum strains can be divided into twodistinct types based on genetic diversity and virulence.Infect Immun 72: 6306–6312.

van der Sar, A.M., Spaink, H.P., Zakrzewska, A., Bitter, W.,and Meijer, A.H. (2009) Specificity of the zebrafish hosttranscriptome response to acute and chronic mycobacte-rial infection and the role of innate and adaptive immunecomponents. Mol Immunol 46: 2317–2332.

Shimono, N., Morici, L., Casali, N., Cantrell, S., Sidders, B.,Ehrt, S., and Riley, L.W. (2003) Hypervirulent mutant ofMycobacterium tuberculosis resulting from disruption ofthe mce1 operon. Proc Natl Acad Sci USA 100: 15918–15923.

Stoop, E.J.M., Schipper, T., Rosendahl Huber, S.K., Nezhin-sky, A.E., Verbeek, F.J., Gurcha, S.S., et al. (2011)Zebrafish embryo screen for mycobacterial genes involved

in granuloma formation reveals a novel ESX-1 component.Dis Models Mech 4: 526–536.

Swaim, L.E., Connolly, L.E., Volkman, H.E., Humbert, O.,Born, D.E., and Ramakrishnan, L. (2006) Mycobacteriummarinum infection of adult zebrafish causes caseatinggranulomatous tuberculosis and is moderated by adaptiveimmunity. Infect Immun 74: 6108–6117.

Taylor, J.L., Hattle, J.M., Dreitz, S.A., Troudt, J.M., Izzo, L.S.,Basaraba, R.J., et al. (2006) Role for matrix metalloprotein-ase 9 in granuloma formation during pulmonary Mycobac-terium tuberculosis infection. Infect Immun 74: 6135–6144.

Uchida, Y., Casali, N., White, A., Morici, L., Kendall, L.V.,and Riley, L.W. (2007) Accelerated immunopathologicalresponse of mice infected with Mycobacterium tuberculosisdisrupted in the mce1 operon negative transcriptional regu-lator. Cell Microbiol 9: 1275–1283.

Volkman, H.E., Clay, H., Beery, D., Chang, J.C.W., Sherman,D.R., and Ramakrishnan, L. (2004) Tuberculous granu-loma formation is enhanced by a Mycobacterium virulencedeterminant. PLoS Biol 2: e367.

Wayne, L.G., and Hayes, L.G. (1996) An in vitro model forsequential study of shiftdown of Mycobacterium tuberculo-sis through two stages of nonreplicating persistence. InfectImmun 64: 2062–2069.

Supporting information

Additional Supporting Information may be found in the onlineversion of this article:

Fig. S1. THP-1 cells were infected with wild-type, ESX-5-deficient or complemented ESX-5-deficient M. marinum at an moiof 20 and after 24 h cell culture supernatant was assayed forTNF-a levels by ELISA.Fig. S2. Bacterial loads (cfu) in organs of fish infected with theM. marinum ESX-5 mutant and wild-type strain after 16 days afterinfection. Statistical differences were determined by the Stu-dent’s t-test (*P < 0.05, **P < 0.01).

Please note: Wiley-Blackwell are not responsible for the contentor functionality of any supporting materials supplied by theauthors. Any queries (other than missing material) should bedirected to the corresponding author for the article.

Hypervirulence of M. marinum ESX-5 mutant 739

© 2012 Blackwell Publishing Ltd, Cellular Microbiology, 14, 728–739