How to decrease TB

BCG-Induced Immune Suppression CID 2010:51 (15 July) 177

V I E W P O I N T S

Decrease in the Effectiveness of Bacille Calmette-GuerinVaccine against Pulmonary Tuberculosis: A Consequenceof Increased Immune Suppression by Microbial Antioxidants,Not Overattenuation

Douglas S. KernodleDepartments of Medicine and of Microbiology and Immunology, Vanderbilt University School of Medicine, Nashville, Tennessee

Mutations that arose in bacille Calmette-Guerin (BCG) daughter strains during decades of in vitro cultivation have longbeen suspected of reducing the efficacy of the BCG vaccine against pulmonary tuberculosis. Although concern was raised 6decades ago that BCG had become overattenuated, preferential use of relatively virulent BCG vaccines has not restoredefficacy. The recent discovery that as BCG evolved its production of antioxidants increased as a consequence of genomicduplications and other mutations suggests the alternative hypothesis that BCGbecame better at suppressing oxidant-dependentimmune responses. This new model of BCG evolution is supported by evidence indicating that reducing BCG antioxidantsenhances immunogenicity. Furthermore, some previously unexplained aspects of the performance of the BCG vaccine inclinical trials now make sense in the context of the new model. Finally, the model suggests that the risk of developingpulmonary tuberculosis is influenced by the balance between host-generated oxidants and microbial antioxidants that activateand suppress, respectively, the antigen-presentation pathways that protect the lungs.

For almost 9 decades, the live vaccineMy-cobacterium bovis bacille Calmette-Guerin(BCG) has been used against tuberculosis.In early studies, BCG vaccine was highlyefficacious. In the 1920s, vaccination ofnursing and medical students with BCGreduced pulmonary tuberculosis by 80%[1]. In the 1930s, a trial in North Americafound similarly high protection in the first2 decades after vaccination, with contin-ued benefit for 6 decades [2, 3]. Yet insubsequent studies BCG vaccine exhibitedmarked variability against pulmonary tu-berculosis and occasionally even appeared

Received 21 December 2009; accepted 31 March 2010;electronically published 4 June 2010.

Reprints or correspondence: Dr Douglas S. Kernodle, Divof Infectious Diseases, A2200 Medical Center North,Vanderbilt University Medical Center, Nashville, TN 37232-2582 ([email protected]).

Clinical Infectious Diseases 2010;51(2):177184! 2010 by the Infectious Diseases Society of America. Allrights reserved.1058-4838/2010/5102-0009$15.00DOI: 10.1086/653533

to increase the risk of developing tuber-culosis [4, 5]. This variability is highly sig-nificant ( ) and is regarded as be-P ! .0001ing indicative of true biological differences[4]. Yet despite its flawed record againstpulmonary tuberculosis, BCG remains re-liably efficacious in preventing tubercu-losis meningitis and miliary tuberculosisin young children [6]. Because pulmonarytuberculosis is more common than dis-seminated tuberculosis, BCG has had aminimal effect on the global burden oftuberculosis, estimated at 9.4 million newactive cases and 1.8 million deaths an-nually [7].Several hypotheses attempt to explain

the variable effectiveness of BCG againstpulmonary tuberculosis. These hypothesesinclude differences betweenBCGdaughterstrains (substrains), an inadequate dosageof BCG in some trials, interference by en-vironmental mycobacteria, genetic differ-ences in human populations, and geo-

graphic differences in clinical isolates ofM. tuberculosis. These hypotheses havebeen summarized elsewhere [4, 5, 8].This Viewpoint offers a new perspective

on the old idea that the variable efficacyof the BCG vaccine against pulmonary tu-berculosis involves differences betweenBCG daughter strains. In the context ofemerging data that reveal that mycobac-terial antioxidants suppress host immu-nity [912] and that antioxidant produc-tion increased as BCG evolved [13, 14],the relevant literature is reexamined to re-veal an association between BCGdaughterstrains that produce large amounts of an-tioxidants and poor efficacy against pul-monary tuberculosis. On the basis of par-tial but not fully conclusive evidence, anew model is proposed in which BCG,instead of becoming overattenuated,evolved to become better at suppressingthe CD8+ T cell responses needed for pro-tection against pulmonary tuberculosis.

putri cantika reviera

178 CID 2010:51 (15 July) Kernodle

Figure 1. Relative expression of genes en-coding antioxidants and other microbial factorsimplicated in immune evasion. Figure panelswere prepared by analysis of data in Table S4of the article by Brosch et al [14]. First, theexpression value for 2 reference isolates ofMy-cobacterium bovis was averaged, and the meanvalue was set as 1 (black line). Second, ex-pression values for bacille Calmette-Guerin(BCG) Tokyo 172 and Pasteur 1173P2 were dis-played relative to the mean values for the ref-erence M. bovis isolates. A, Antioxidant andother immune evasion genes duplicated in Tokyo172, Danish 1331, or Pasteur 1173P2 (full namesare given in Table 1). B, Antioxidant genes out-side the BCG duplication units. These includesodA (iron cofactored superoxide dismutase),tpx (thiol peroxidase), ahpC and ahpD (alkylhy-droperoxide reductases C and D), rubA (rubre-doxin), the oxidoreductase Rv1774, and mem-bers of the whiB family of protein disulfidereductases.

The new model further suggests that invitro evolution enabled extensively culti-vated BCG substrains to survive longer invivo, which made them more effective atpreventing disseminated tuberculosis. Al-though overattenuation and increased im-mune suppression are mutually exclusivemodels of the evolution of BCG, the newhypothesis is not fully sufficient. Environ-mental and genetic factors are also crucial,especially in the context of their effect onhost oxidant-generating capacity duringvaccination.

EARLY CONCERNABOUT OVERATTENUATIONAND IDENTIFICATIONOF PHENOTYPIC DIFFERENCESAMONG BCG SUBSTRAINS

An early theory regarding the apparent de-cline in efficacy of the BCG vaccine againstpulmonary tuberculosis was that the vac-cine had changed over time. In 1949, Ir-vine expressed concern that the problemhas now become one of over-attenuation.Separated from its natural habitat for 42years, may not attenuation still be slow-ly progressing? [15, p 25]. By the mid-1950s, it had been shown that BCG sub-strains differ in characteristics, includinggrowth rate, their ability to persist in vivo,and their ability to protect mice againstM. tuberculosis infection [1619]. Daugh-ter strains are descendents of BCG; beforemodern technologies for preserving bac-teria became available, laboratories main-tained BCG by serial passage (ie, by trans-ferring part of an aging culture into freshmedia). Some strains were passaged11000 times before seed lots were pre-pared (eg, Pasteur 1173P2 and Danish1331), whereas in Japan BCG was pas-saged only 172 times (Tokyo 172) [14,20]. Serial passage at multiple sites causeddivergent evolution. Thus, substrains dif-fer from each other and from the originalvaccine, which no longer exists.

SUPERIORITY OF RELATIVELYVIRULENT BCG VACCINESIN ANIMAL MODELS

In an attempt to identify BCG daughterstrains still effective in humans, investi-gators turned to animal models andlearned that protection correlated with theinvasiveness and persistence of the vaccinestrain. Dubos and Pierce summarizedtheir findings in mice as follows: Sincethe most invasive substrains of BCG arethe most likely to elicit dependable andlasting immunity, it would appear at firstsight that they are best suited to the prac-tice of human vaccination.On the otherhand greater invasiveness increases the

incidence of adenopathies. The choiceof the optimal substrain of BCG involvesa compromise between the requirementsof the immunologist for dependable andlasting immunity and the concern of allfor innocuousness [19, p 713]. On thebasis of studies in golden hamsters, bankvoles, and guinea pigs, Bunch-Christensenet al also favored virulent BCG substrains,concluding that the lower virulence [ofsome substrains] is a sign of genetic mu-tation [and] consistent with the generalbiological experience that virulence is of-ten lost in vitro but that it practically neverincreases [21, p 65]. Thus, virulent vac-cines were assumed to be most like theoriginal BCG, which was 80% efficaciousin humans. This assumption influencedthe selection of 2 virulent substrains, Pas-teur 1173P2 and Danish 1331, for a vac-cination trial in the Chingleput region ofIndia [5, 22, 23]. However, despite theirhigh rank (second and third, respectively)among the dozen vaccines evaluated inanimals [22], neither vaccine protectedhumans against pulmonary tuberculo-sis [23].

SUPERIORITY OF BCG TOKYO172 IN PREVENTINGTUBERCULOSIS IN HUMANS

The failure to restore vaccination effec-tiveness in Chingleput with virulent sub-strains that worked best in animal modelsraised questions about whether overatten-uation is the reason for the decline in BCGeffectiveness and also about how well re-sults in animal models correlate with re-sults in humans. Concerned by the in-consistent ranking of vaccines by differentlaboratories [24], Comstock instead fo-cused on the details of clinical trials toidentify an effective BCG vaccine. He ar-gued that case-control studies conductedover a time span in which one BCGdaughter strain replaces another provideinsight into their relative effectiveness [2527]. Two studies, one in Indonesia in-volving disseminated and pulmonary tu-berculosis (16 and 88 cases, respectively)and another in Colombia involving pul-

BCG-Induced Immune Suppression CID 2010:51 (15 July) 179

Table 1. Antioxidant and Other Immune Evasion Genes in Bacille Calmette-Guerin (BCG)Duplication Units

Protein function (gene name)Duplication

unitTokyo172

Danish1331

Pasteur1173P2

Thioredoxin reductase (trxB2) DU1 +Thioredoxin (trxC) DU1 +Protein disulfide reductase (whiB7) DU2 +Possible glutaredoxin (Rv3198A) DU2 +Protein disulfide reductase (whiB1) DU2 + +Oxidative stress sigma factor (sigH) DU2 + +NADPH quinone reductase (lpdA, Rv3303c) DU2 + +PI3P phosphatase (sapM, Rv3110) DU2 + +

NOTE. The genes within DU1 of BCG Pasteur 1173P2 were determined from BCGList (http://genolist.pasteur.fr/BCGList/). To identify genes within DU2, H37Rv coordinates of the duplicated regionsin each BCG daughter strain were obtained from Brosch et al [14] and included the following: Tokyo 172,3,684,2293,704,932; Danish 1331, 3,567,4593,608,472 and 3,671,5363,709,097; and Pasteur 1173P2,3,590,9023,608,472 and 3,671,5363,690,127. Then TubercuList (http://genolist.pasteur.fr/TubercuList/)was used to identify the genes within the DU2 region of chromosome for each BCG daughter strain.NADPH, nicotinamide adenine dinucleotide phosphate; PI3P, phosphatidylinositol 3-phosphate.

monary tuberculosis (178 cases), met thisstandard. In Indonesia, the effectiveness ofBCG relative to unvaccinated individualsdecreased from +58% to!38% when To-kyo 172 was replaced by Pasteur 1173P2.In Colombia, vaccination effectiveness de-creased from +51% to!18% when Tokyo172 (or British/Glaxo) was replaced byDanish 1331. It is noteworthy that BCGTokyo 172 ranked eighth in the schemeused to prioritize BCG substrains beforethe Chingleput trial [22], well below therank of Pasteur 1173P2 and Danish 1331.BCG Tokyo 172 is also cleared more rap-idly than Pasteur 1173P2 from the organsof mice [28, 29], and its lower virulencerelative to Pasteur 1173P2 and Danish1331 is reflected in fewer adverse effects[30].In developing his argument, Comstock

mentioned 2 other studies that comparedBCG substrains [27]. The first was theChingleput trial, in which neither virulentvaccine was effective. Comstock lamentedthe omission of a substrain that rankedpoorly in animal models while acknowl-edging that it was understandable thatmaking such an odious comparison wasnot politically possible for a World HealthOrganizationsponsored project [27, pS251]. The second study involved new-borns in Hong Kong [5]. Fourteen casesof tuberculosis involving lymph nodes,

meninges, bone, joints, or multiple siteswere observed in 150,000 persons vacci-nated with BCG Pasteur, versus 31 casesin 150,000 recipients of BCG British/Glaxo. Of note, these vaccines ranked sec-ond and 11th, respectively, in animalmodels [22], and thus results in animalsand humans correlated nicely. More re-cently, another study focused on protec-tion against disseminated tuberculosis inearly childhood, comparing intradermalDanish 1331 and percutaneous Tokyo 172[31]. Both vaccines were highly effective,reducing disseminated tuberculosis by87% overall compared with nonvaccin-ated individuals, yet Danish 1331 was 46%more effective than Tokyo 172. These re-sults also correlate well with animal mod-els, for which the vaccines ranked thirdand eighth, respectively [22].In summary, in the 5 human studies in

which BCG daughter strains have beencompared, Tokyo 172 has demonstratedgreater effectiveness against pulmonarytuberculosis, whereas BCG Pasteur andDanish 1331 have demonstrated greaterprotection against disseminated tubercu-losis in early childhood. Furthermore, al-though rankings of BCG substrains in an-imal models [22, 29] correlate poorly withprotection against pulmonary tuberculosisin humans [2527], they correlate wellwith protection against disseminated tu-

berculosis. This dichotomy may reflect arequirement for qualitatively different im-mune responses for preventing pulmonarytuberculosis than disseminated tubercu-losis [12]. Obviously, this type of analysisis limited because of the small number ofstudies that have compared BCG sub-strains in humans.

GENOMIC DELETIONS IN BCGSUBSTRAINS

Over the past few decades, the genetic evo-lution of BCG has been partly recon-structed. As BCG evolved from M. bovis,it lost the region of difference 1 (RD1).RD1 is absent from the chromosome ofall BCG substrains and encodes a secretionsystem involved in virulence [3234]. Ad-ditional genomic deletions are found insome BCG substrains, and despite the fail-ure of virulent BCG vaccines in Chingle-put, the hypothesis that the decline in theeffectiveness of the BCG vaccine involvedoverattenuation has resurfaced [3537]. Inthe reformulated hypothesis, deletions af-ter RD1 are believed to cause overatten-uation, yet no deletion has been clearlylinked to attenuation [37].

INCREASED EXPRESSIONOF ANTIOXIDANTS IN BCGSUBSTRAINS

The evolution of BCG vaccine also in-cluded the duplication of regions of itsgenome and greater expression of anti-oxidants (Figure 1 and Table 1) [13, 14].Two duplication units, DU1 and DU2, ex-ist. Although DU1 may be unique to BCGPasteur, DU2 is widely distributed, as-sumes different forms, and is triplicatedin some BCG daughter strains. Notableamong the duplicated and highly ex-pressed genes is the sigma factor SigH,which augments the expression of multi-ple antioxidants during oxidative stress,including thioredoxin, thioredoxin reduc-tase, and iron-cofactored superoxide dis-mutase (SodA) [38]. SigH also inducesenzymes that synthesize precursors ofmycothiol [39].The biological reason underlying the

180 CID 2010:51 (15 July) Kernodle

Figure 2. Influence of oxidant-dependent immune signaling on antigen-presentation pathways and clinical outcome. After becoming infected withMycobacterium tuberculosis, the host faces 2 challenges. The first is to halt the lymphohematogenous spread of bacilli and prevent the developmentof miliary tuberculosis; this is accomplished by macrophages and CD4+ T cells that produce interferon g (IFN-g). The second is to prevent foci ofgranulomatous inflammation from expanding and damaging normal tissue; this probably requires cytotoxic T lymphocytes (CTLs) that kill bacteria withininfected macrophages. The induction of CTLs appears to involve Mycobacterium-infected phagocytes that first undergo apoptosis. Then mycobacterialantigens within apoptotic cell fragments are taken up and presented by dendritic cells. Apoptosis-associated cross-priming pathways leading to CTLs(black box) may be the reason that 90% of M. tuberculosisinfected humans never develop pulmonary tuberculosis. In contrast, the small-animalmodels commonly used to evaluate tuberculosis vaccines are not natural hosts for M. tuberculosis. In most of these models, the host can restrictdissemination unless it is starved or has certain genetic defects; however, disease in the lungs progresses and lifelong containment similar to latenttuberculosis in humans does not occur, possibly because of inadequate CD8+ T cell responses. M, macrophages; MHC, major histocompatibilitycomplex; PMNs, polymorphonuclear neutrophils; TB, tuberculosis.

evolution of the BCG vaccine to producemore antioxidants is uncertain; however,antioxidants may enhance mycobacterialgrowth in vitro by detoxifying by-productsof aerobic metabolism [12]. This hypoth-esis is supported by the observation thatgenetic modifications of BCG that reduceantioxidants also reduce its rate of growthin vitro [12]. Furthermore, investigatorswho worked with BCG before it was pre-served as seed lots report that its rate ofgrowth accelerated over time [21, 40].In summary, as BCG was cultivated for

decades, it not only underwent genomicdeletions but also evolved to producemore antioxidants. Oxidants produced byimmune cells augment the activation andapoptosis of phagocytes, and thus micro-bial antioxidants are well positioned to in-terfere with early host responses and thedevelopment of adaptive immunity (Fig-ure 2). These considerations and the factthat antioxidants increased more in Pas-teur 1173P2 and Danish 1331 than in To-

kyo 172 suggest that the decline in theeffectiveness of BCG vaccine against pul-monary tuberculosis might involve in-creased interference with oxidant-depen-dent immune signaling pathways.

EFFECT OF REDUCING BCGANTIOXIDANTS ONIMMUNOGENICITY

In M. tuberculosis, reducing the activity orsecretion of SodA enhances the activationand apoptosis of mononuclear cells andstrengthens antigen-specific CD8+ T cellresponses [911]. Furthermore, SigH pro-motes lung immunopathology by an un-known mechanism [41]. To determinewhether antioxidants that increased asBCG evolved suppress immune responses,Sadagopal et al [12] eliminated SigH andreduced SodA activity and secretion in anextensively passaged BCG substrain. BCG-specific CD8+ T cell responses suppressedby the parent vaccine were unmasked dur-

ing vaccination with the modified BCGvaccine. Memory immunity was also en-hanced, and the modified BCG grewslower in vitro and survived less well inmice.

RECOGNITION OF A PARADOXAND ITS IMPLICATIONS

The evolution of BCG represents a para-dox. Unlike other live vaccines thatevolved to become more attenuated dur-ing in vitro cultivation, BCG instead be-came more virulent. The literature clearlydocuments that the extensively cultivatedBCG substrains Pasteur 1173P2 and Dan-ish 1331 exhibit greater virulence in ani-mal models than the less extensively cul-tivated Tokyo 172 substrain [12, 22, 28,29]. Furthermore, Danish 1331 is morevirulent thanDanish 1077 (British/Glaxo).The arguments made in this article suggestthat the evolution of the BCG vaccine to-ward greater virulence might also explain

BCG-Induced Immune Suppression CID 2010:51 (15 July) 181

the decline in its effectiveness against pul-monary tuberculosis.Decisions made in accordance with the

overattenuation model of BCG evolutionhave repeatedly failed to improve protec-tion against pulmonary tuberculosis, mostnotably in the Chingleput trial [23] butalso in subsequent studies in which Tokyo172 was replaced with amore virulent vac-cine [2527]. These failures now makesense in the context of a new model ofBCG evolution in which overattenuationnever occurred. Instead, BCG evolved tomake more antioxidants, possibly to fulfilla physiologic need involving the myco-bacterial cell wall [12]. By chance, this co-incided with amycobacterial immune eva-sion strategy and made BCG better atsuppressing host immunity, particularlyCD8+ T cell responses. Although gene du-

plication and presumably other moresubtle mutations underlie most of theincreased expression of antioxidants incurrent BCG vaccines, it is noteworthythat a genomic deletion also contributed.The increased expression of Rv1774, anoxidoreductase, involves the loss of itsrepressor within RD14 [42].The mechanistic and clinical implica-

tions of BCG evolving to become morevirulent by producing more antioxidantsare enormous. First, rather than not sur-viving long enough to induce immunity,the extensively cultivated BCG daughterstrains actively suppress CD8+ T cell re-sponses and immune memory. Becauseinterferon g production by CD4+ T cellsand macrophage responses are less af-fected than CD8+ T cell responses by mi-crobial antioxidants [12], when subse-quently infected with M. tuberculosis, aBCG-vaccinated host still compartmen-talizes infection within granulomata andthus is protected against disseminated tu-berculosis. However, weak CD8+ T cell re-sponses limit the hosts ability to kill in-fected macrophages and resolve granulo-matous foci of infection. An increasingbody of evidence suggests that CD8+ Tcells may help protect against pulmonarytuberculosis [43, 44]. In effect, as BCG

evolved it became better at suppressingimmune responses needed for protectionagainst pulmonary tuberculosis whilelargely retaining responses that protectagainst disseminated tuberculosis in earlychildhood.Second, because of divergent evolution

the immune-suppressive capacity of eachBCG substrain differs yet should correlateroughly with the number of passages be-fore the seed lot was preserved. This claimis supported by the greater immunoge-nicity, including CD8+ T cell responses, ofBCG Tokyo 172 than Danish 1331 [45].Third, the virulent and immune sup-

pressive vaccines aremore effective againstdisseminated tuberculosis in early child-hood, possibly because they survive longerin vivo. Vaccine persistence also correlateswith protection in mice [19, 29], yet ifantibiotic treatment is used to clear BCGDanish 1331 from mice, immune re-sponses and protection against dissemi-nation decrease to low levels withinmonths [46]. In effect, vaccines that in-duce the greatest protection against pul-monary tuberculosis may be less effec-tive against disseminated disease in ear-ly childhood because they induce immu-nity that clears the vaccine strain from thehost [12].Fourth, involvement of oxidant-depen-

dent immune responses in protectionagainst pulmonary tuberculosis (Figures 2and 3) makes sense in the context of riskfactors for pulmonary tuberculosis in hu-mans. For example, persons with chronicgranulomatous disease, a genetic diseasein which nicotinamide adenine dinucle-otide phosphate oxidase fails to assembleto produce superoxide, frequently developpulmonary tuberculosis and occasional-ly develop disseminated tuberculosis orBCG-osis [48, 49]. In mice, p47phox de-ficiency causes a survival defect in CD8+

T cells that is partially corrected by addingoxidants [50]. Less severe defects in oxi-dant-generating capacity, such as a weakoxidative burst [51] and/or low polymor-phonuclear neutrophil counts [52], mayunderlie the high rate of pulmonary tu-

berculosis in some racial/ethnic groups.The role played by apoptosis-associatedcross-priming in the induction of CD8+ Tcell responses seems plausible in the con-text of the proapoptotic effects of oxidantsand evidence indicating that oxidation ofphosphatidylserine in membranes of ap-optotic cells enhances the uptake of cellfragments by other phagocytes [53]. A rolefor antigen cross-presentation in humansis suggested by the observation that per-sons with latent tuberculosis (ie, posi-tive for purified protein derivative) ex-hibit predominant macrophage apoptosis,whereas persons cured of active tubercu-losis exhibit greater necrosis [54].Fifth, a model in which the balance be-

tween host-generated oxidants and my-cobacterial antioxidants affects CD8+ Tcell responses also provides insight into thepathogenesis of tuberculosis. By allowingCD4+ T cell responses to develop and lim-

it dissemination while suppressing CD8+

T cell responses, M. tuberculosis ensuresdisease transmission. Instead of dyingquickly, most infected hosts with CD8+

responses inadequate to maintain latenttuberculosis will develop granulomatouslung cavities and infectious aerosols. Athird of the worlds population is infectedwith M. tuberculosis, demonstrating thesuccess of this pathogenesis strategy.Sixth, the increased immune suppres-

sion model of BCG evolution predicts thatthe virulent BCG daughter strains are theleast like the early BCG vaccines that ex-hibited 80% protection against pulmonarytuberculosis [1, 2]. In the new model, thevirulent vaccines are not only expected tobe ineffective against pulmonary tuber-culosis, but their use increases adverse re-actions unnecessarily. This contrasts withthe overattenuationmodel, which predictsthat increased efficacy will be achievedonly at the cost of increased adverse re-actions [36, p 134]. These predictions aretimely in the context of reports that vac-cination with BCG Danish 1331 causesdisseminated BCG disease in almost 1%of human immunodeficiency virusin-fected infants [55].

182 CID 2010:51 (15 July) Kernodle

Figure 3. Determination of the outcome of infection by the balance between host-generated oxidants and microbial antioxidants. In this model,90% of humans generate enough oxidants during early infection to activate cross-priming pathways of antigen presentation (A). The cytotoxic Tlymphocyte responses heal granulomatous foci of infection and provide ongoing immune surveillance to prevent the development of pulmonarytuberculosis (Figure 2). However, in 10% of persons the host-generated oxidants are insufficient to overcome suppression by antioxidants, resultingin weak CD8+ cell responses and eventually in the development of active tuberculosis. The goal of vaccination is to prevent active tuberculosis fromdeveloping in the 10% of persons so predisposed (B ). The original bacille Calmette-Guerin (BCG) vaccine reduced pulmonary tuberculosis by 80%,which may reflect greater cross-priming compared with natural infection, perhaps from the loss of genes within region of difference 1 that suppressthe production of oxidants by antigen-presenting cells [47]. Alternatively, by inoculating thousands of vaccine bacilli during vaccination, more poly-morphonuclear cells with highly potent oxidant-generating capacity are activated than after inhalation of a few tubercle bacilli. Because the increasein antioxidants in BCG daughter strains is roughly proportional to the number of times they were passaged, vaccination effectiveness remained about+50% with Tokyo 172 yet decreased to as low as !38% with Pasteur 1173P2 [2527]. The increased immune suppression model predicts that BCGcan be modified to exhibit diminished activity of antioxidants, thereby inducing strong immunity that prevents pulmonary tuberculosis. In effect, byreducing BCG antioxidants the redox balance during early infection shifts, and cross-priming can now occur in some persons with low oxidant-generatingcapacity. This should enable them to develop immune responses more like those that develop in the 90% majority during natural infection. Then ifinfection occurs the vaccination-induced immune responses are recalled and help to prevent active tuberculosis. The development of memory immunityis probably the reason that protection was observed for 6 decades in recipients of early BCG vaccines [2]. TB, tuberculosis.

BCG-Induced Immune Suppression CID 2010:51 (15 July) 183

Seventh, if increased antioxidant pro-

duction is the primary reason for the de-

cline in the effectiveness of the BCG vac-

cine against pulmonary tuberculosis, then

the simplest way to restore its effective-

ness is to reduce BCG antioxidants (Fig-

ure 3). If the new model is correct, such

modifications should enhance protection

against pulmonary tuberculosis in hu-

mans. Table 1 and Figure 1 show some of

the antioxidant genes that underwent du-

plication and increased expression as BCG

evolved and thus represent high-priority

targets. Reconstructing a BCG vaccine that

exhibits protection comparable or supe-

rior to the early BCG vaccines should be

easier to accomplish with Tokyo 172 than

with virulent BCG substrains. It is note-

worthy that prioritization of vaccines for

human trials will need to be different from

the algorithm used for the Chingleput

study [22, 23]. Because the immune re-

sponses of mice and other small animals

commonly used in vaccination-challenge

experiments are inadequate to induce la-

tent tuberculosis, more relevant models

are needed. In the context of controlling

tuberculosis globally, it is more important

to target the pulmonary, contagious form

of tuberculosis than disseminated tuber-

culosis. Thus, vaccines that induce strong

immunity should be favored even if they

are less effective against dissemination

than virulent vaccines. Secondary immune

responses in a memory-immune model

[12] may be particularly useful in identi-

fying vaccines likely to exhibit greater pro-

tection against pulmonary tuberculosis.

Finally, this new hypothesis does not

mean that other hypotheses regarding the

suboptimal effectiveness of the BCG vac-

cine against pulmonary tuberculosis do

not also contribute. However, when it

comes to understanding the effect of phe-

notypic differences amongBCG substrains

on protection against pulmonary tuber-

culosis, we failed to grasp a paradox and

had it backward all along.

Acknowledgments

Financial support. David E. Rogers Profes-sorship in Medicine. The financial support did notinfluence the opinions expressed in this essay.

Potential conflicts of interest. D.S.K. is listedas an inventor on issued patents and patent ap-plications for a technology for enhancing the im-munogenicity of bacterial vaccines by reducing theactivity of antiapoptotic microbial enzymes. Thetechnology has been assigned to Vanderbilt andthe US government as represented by the US De-partment of Veterans Affairs.

References

1. Bjartveit K. Olaf Scheel and Johannes Heim-beck: their contribution to understanding thepathogenesis and prevention of tuberculosis.Int J Tuberc Lung Dis 2003; 7:306311.

2. Aronson JD, Aronson CF, Taylor HC. Atwenty-year appraisal of BCG vaccination inthe control of tuberculosis. Arch Intern Med1958; 101:881893.

3. Aronson NE, Santosham M, Comstock GW,et al. Long-term efficacy of BCG vaccine inAmerican Indians and Alaska Natives: a 60-year follow-up study. JAMA 2004; 291:20862091.

4. Fine PEM, Carneiro IAM, Milstien JB, Cle-ments JD. Issues relating to the use of BCGin immunization programs: a discussion doc-ument. Geneva, Switzerland: World HealthOrganization Department of Vaccines and Bi-ologicals, 1999. Document WHO/V&B/99.23.

5. Ten Dam HG. BCG vaccination. In: Tuber-culosis: a comprehensive international ap-proach. Reichman LB,Hershfield ES, eds.NewYork: Marcel Dekker, 1993:251274.

6. Trunz BB, Fine P, Dye C. Effect of BCG vac-cination on childhood tuberculous meningitisand miliary tuberculosis worldwide: a meta-analysis and assessment of cost-effectiveness.Lancet 2006; 367:11731180.

7. World Health Organization. Global tubercu-losis control: surveillance, planning, financ-ing. Geneva, Switzerland: World Health Or-ganization, 2009. DocumentWHO/HTM/TB/2009.411.

8. Andersen P, Doherty TM. The success andfailure of BCG: implications for a novel tu-berculosis vaccine. Nat Rev Microbiol 2005;3:656662.

9. Edwards KM, Cynamon MH, Voladri RK, etal. Iron-cofactored superoxide dismutase in-hibits host responses to Mycobacterium tuber-culosis. Am J Respir Crit Care Med 2001; 164:22132219.

10. Kurtz S, McKinnon KP, Runge MS, Ting JP,Braunstein M. The SecA2 secretion factor ofMycobacterium tuberculosis promotes growthin macrophages and inhibits the host immuneresponse. Infect Immun 2006; 74:68556864.

11. Hinchey J, Lee S, Jeon BY, et al. Enhancedpriming of adaptive immunity by a proapo-

ptotic mutant of Mycobacterium tuberculosis.J Clin Invest 2007; 117:22792288.

12. Sadagopal S, Braunstein M, Hager CC, et al.Reducing the activity and secretion of micro-bial antioxidants enhances the immunogenic-ity of BCG. PLoS One 2009; 4:e5531.

13. Brosch R, Gordon SV, Buchrieser C, Pym AS,Garnier T, Cole ST. Comparative genomicsuncovers large tandem chromosomal dupli-cations in Mycobacterium bovis BCG Pasteur.Yeast 2000; 17:111123.

14. Brosch R, Gordon SV, Garnier T, et al. Ge-nome plasticity of BCG and impact on vaccineefficacy. Proc Natl Acad Sci U S A 2007; 104:55965601.

15. Irvine KN. B.C.G. vaccination in theory andpractice. London, England: Blackwell Scien-tific Publications, 1949.

16. Dubos RJ, Pierce CH. Differential character-istics in vitro and in vivo of several substrainsof BCG. I: Multiplication and survival in vitro.Am Rev Tuberc 1956; 74:655666.

17. Dubos RJ, Pierce CH. Differential character-istics in vitro and in vivo of several substrainsof BCG. II: Morphologic characteristics invitro and in vivo. Am Rev Tuberc 1956; 74:667682.

18. Dubos RJ, Pierce CH, Schaefer WB. Differ-ential characteristics in vitro and in vivo ofseveral substrains of BCG. III: Multiplicationand survival in vivo. Am Rev Tuberc 1956;74:683698.

19. Dubos RJ, Pierce CH. Differential character-istics in vitro and in vivo of several substrainsof BCG. IV: Immunizing effectiveness. AmRev Tuberc 1956; 74:699717.

20. Yamamoto S, Yamamoto T. Historical reviewof BCG vaccine in Japan. Jpn J Infect Dis2007; 60:331336.

21. Bunch-Christensen K, Ladefoged A, Guld J.The virulence of some strains of BCG forgolden hamsters: further studies. Bull WorldHealth Organ 1970; 43:6570.

22. Ladefoged A, Bunch-Christensen K, Guld J.Tuberculin sensitivity in guinea-pigs aftervaccination with varying doses of BCG of12 different strains. Bull World Health Or-gan 1976; 53:435443.

23. Tuberculosis Prevention Trial. Trial of BCGvaccines in South India for tuberculosis pre-vention: first report. Bull World Health Organ1979; 57:819827.

24. Wiegeshaus EH, Harding G, McMurray D,Grover AA, Smith DW. A co-operative eval-uation of test systems used to assay tubercu-losis vaccines. Bull World Health Organ 1971;45:543550.

25. Comstock GW. Identification of an effectivevaccine against tuberculosis. Am Rev RespirDis 1988; 138:479480.

26. Comstock GW. Evaluating vaccination ef-fectiveness and vaccine efficacy by means ofcase-control studies. Epidemiol Rev 1994; 16:7789.

27. ComstockGW. Simple, practicalways toassessthe protective efficacy of a new tuberculosis

184 CID 2010:51 (15 July) Kernodle

vaccine. Clin Infect Dis 2000; 30(suppl 3):S250S253.

28. Sher NA, Chaparas SD, Pearson J, ChirigosM. Virulence of six strains of Mycobacteriumbovis (BCG) in mice. Infect Immun 1973; 8:736742.

29. Lagranderie MR, Balazuc AM, Deriaud E, Le-clerc CD, Gheorghiu M. Comparison of im-mune responses of mice immunized with fivedifferent Mycobacterium bovis BCG vaccinestrains. Infect Immun 1996; 64:19.

30. Milstein JB, Gibson JJ. Quality control of BCGvaccine by WHO: a review of factors that mayinfluence vaccine effectiveness and safety. BullWorld Health Organ 1990; 68:93108.

31. Mahomed H, Kibel M, Hawkridge T, et al.The impact of a change in bacille Calmette-Guerin vaccine policy on tuberculosis inci-dence in children in Cape Town, South Africa.Pediatr Infect Dis J 2006; 25:11671172.

32. Mahairas GG, Sabo PJ, Hickey MJ, Singh DC,Stover CK. Molecular analysis of genetic dif-ferences between Mycobacterium bovis BCGand virulent M. bovis. J Bacteriol 1996; 178:12741282.

33. Lewis KN, Liao R, Guinn KM, et al. Deletionof RD1 fromMycobacterium tuberculosismim-ics bacille Calmette-Guerin attenuation. J In-fect Dis 2003; 187:117123.

34. Hsu T, Hingley-Wilson SM, Chen B, et al. Theprimary mechanism of attenuation of bacillusCalmette-Guerin is a loss of secreted lyticfunction required for invasion of lung inter-stitial tissue. Proc Natl Acad Sci U S A2003; 100:1242012425.

35. Behr MA, Wilson MA, Gill WP, et al. Com-parative genomics of BCG vaccines by whole-genome DNA microarray. Science 1999; 284:15201523.

36. Behr MA, Small PM. Has BCG attenuated toimpotence? Nature 1997; 389:133134.

37. Mostowy S, Tsolaki AG, Small PM, Behr MA.The in vitro evolution of BCG vaccines. Vac-cine 2003; 21:42704274.

38. Mehra S, Kaushal D. Functional genomics re-veals extended roles of the Mycobacterium tu-berculosis stress response factor jH. J Bacteriol2009; 191:39653980.

39. Manganelli R, Voskuil MI, Schoolnik GK,Dubnau E, Gomez M, Smith I. Role of theextracytoplasmic-function j factor jH in My-cobacterium tuberculosis global gene expres-sion. Mol Microbiol 2002; 45:365374.

40. Aronson JD, Schneider P. The problem ofstandardization of BCG vaccine. Am J PublicHealth 1950; 40:533544.

41. Kaushal D, Schroeder BG, Tyagi S, et al. Re-duced immunopathology and mortality de-spite tissue persistence in a Mycobacteriumtuberculosis mutant lacking alternative j fac-tor, SigH. Proc Natl Acad Sci U S A 2002;99:83308335.

42. Alexander DC, Behr MA. Rv1773 is a tran-scriptional repressor deleted from BCG-Pas-teur. Tuberculosis (Edinb) 2007; 87:421425.

43. Woodworth JS, Behar SM. Mycobacterium tu-berculosis-specific CD8+ T cells and their rolein immunity. Crit Rev Immunol 2006; 26:317352.

44. Carranza C, Juarez E, Torres M, Ellner JJ, SadaE, Schwander SK. Mycobacterium tuberculosisgrowth control by lungmacrophages andCD8cells from patient contacts. Am J Respir CritCare Med 2006; 173:238245.

45. Davids V, Hanekom WA, Mansoor N, et al.The effect of bacille Calmette-Guerin vaccinestrain and route of administration on inducedimmune responses in vaccinated infants. J In-fect Dis 2006; 193:531536.

46. Olsen AW, Brandt L, Agger EM, van PinxterenLAH, Andersen P. The influence of remaininglive BCG organisms on the maintenance ofimmunity to tuberculosis. Scand J Immunol2004; 60:273277.

47. Sinha A, Singh A, Satchidanandam V, Nata-rajan K. Impaired generation of reactive ox-ygen species during differentiationof dendriticcells (DCs) by Mycobacterium tuberculosis se-

cretory antigen (MTSA) and subsequent ac-tivation of MTSA-DCs by mycobacteria re-sults in increased intracellular survival. J Im-munol 2006; 177:468478.

48. Lee PP, Chan KW, Jiang L, et al. Susceptibilityto mycobacterial infections in children withX-linked chronic granulomatous disease: a re-view of 17 patients living in a region endemicfor tuberculosis. Pediatr Infect Dis J 2008; 27:224230.

49. Bustamante J, Aksu G, Vogt G, et al. BCG-osis and tuberculosis in a child with chronicgranulomatous disease. J Allergy Clin Immu-nol 2007; 120:3238.

50. Donaldson M, Antignani A, Milner J, et al.p47phox-deficient immune microenviron-ment signals dysregulate naive T-cell apop-tosis. Cell Death Differ 2009; 16:125138.

51. Siddiqi M, Garcia ZC, Stein DS, Denny TN,Spolarics Z. Relationship between oxidativeburst activity and CD11b expression in neu-trophils and monocytes from healthy individ-uals: effects of race and gender. Cytometry2001; 46:243246.

52. Martineau AR, Newton SM, Wilkinson KA,et al. Neutrophil-mediated innate immune re-sistance to mycobacteria. J Clin Invest 2007;117:19881994.

53. Arroyo M, Modriansky M, Serinkan FB, et al.NADPH oxidase-dependent oxidation andexternalization of phosphatidylserine dur-ing apoptosis in Me2SO-differentiated HL-60cells: role in phagocytic clearance. J Biol Chem2002; 277:4996549975.

54. Gil DP, Leon LG, Correa LI, et al. Differentialinduction of apoptosis and necrosis in mono-cytes from patients with tuberculosis andhealthy control subjects. J InfectDis 2004; 189:21202128.

55. Hesseling AC, Johnson LF, Cotton MF, et al.Disseminated bacille Calmette-Guerin diseasein HIV-infected South African infants. BullWorld Health Organ 2009; 87:505511.