Vol. 57, No. 9INFECTION AND IMMUNITY, Sept. 1989, p. 2674-26820019-9567/89/092674-09$02.00/0Copyright C 1989, American Society for Microbiology

Use of the Promoter Fusion Transposon Tn5 lac To IdentifyMutations in Bordetella pertussis vir-Regulated Genes

ALISON ANN WEISS,* ANGELA R. MELTON, KIM E. WALKER, CECILE ANDRAOS-SELIM, ANDJOHN J. MEIDL

Department of Microbiology and Immunology, Box 678, Virginia Commonwealth University, Richmond, Virginia 23298

Received 31 January 1989/Accepted 5 May 1989

Mutants of Bordetella pertussis deficient in virulence-associated factors were identified by using thetransposon TnS lac. TnS lac is a derivative of TnS which generates promoter fusions for ,B-galactosidase. Tn5lac insertions in the vir-regulated genes of B. pertussis were identified by selecting for kanamycin-resistantmutants that expressed I-galactosidase when the vir-regulated genes were expressed but not when thevir-regulated genes were turned off. Fourteen different mutations in vir-regulated genes were identified. Twomutants were deficient in the production of the filamentous hemagglutinin, two mutants were deficient in theproduction of adenylate cyclase toxin and hemolysin, and one mutant was deficient in the production ofdermonecrotic toxin. One insertion mapped adjacent to the pertussis toxin gene, but the mutant producedpertussis toxin. The phenotypes of the remaining eight mutants were not determined, but the mutants did notappear to be deficient in the production of the 69,000-dalton outer membrane protein (agglutinogen 3) or thecapsule. Screening for mutations in either of the fimbrial genes proved to be problematic since the parentalstrain was found to switch from a fimbriated to a nonfimbriated state at a high frequency, which was suggestiveof the metastable expression of pili in other bacteria. We used Southern blot analysis with a 30-mer specific forthe fimbrial sequences. No bands with the predicted increase in size due to the 12 kilobases from TnS lac wereobserved, which suggests that none of these genes were mutated. Southern blot analysis also revealed that sevenof the eight unidentified mutations mapped to different restriction fragments, which suggests that they could bedeficient in as many as seven different genes.

The goal of our research is to understand the molecularbasis of bacterial disease. Bordetella pertussis, the gram-negative bacterium that causes whooping cough, provides anexcellent model system for several reasons. First, whoopingcough remains an unconquered disease. The current whole-cell vaccine has adverse side effects which have led to adecline in its acceptance in such countries as England andSweden and a subsequent rise in disease (8). However, eventhis imperfect vaccine is beyond the financial means of manydeveloping nations (26).

Second, a unique aspect of the biology of this microbemakes it an ideal model system to study pathogenesis. All ofthe genes required for virulence (perhaps 20) are coordi-nately expressed (35, 38) and are regulated by the growthconditions. High concentrations of MgSO4 can repress theexpression of the virulence genes (11, 15). We have shownthat one gene (called vir for virulence regulating) acts as apositive transcriptional inducer for the expression of thevirulence genes (35, 38). The importance of vir has been seenin animal studies. Inoculation with only 2,000 virulent bac-teria kills a mouse, whereas greater than 107 bacteria with amutation in vir do not even produce disease (40).The virulence genes encode several toxins (pertussis

toxin, adenylate cyclase toxin, dermonecrotic toxin [DNT],and hemolysin), putative attachment factors (filamentoushemagglutinin [FHA] and fimbriae), other factors identifiedonly serologically or by the appearance of proteins onpolyacrylamide gels, and possibly other as yet undiscoveredtraits. Identification of the genes under the control of virpresents an opportunity to study what is required to causedisease from a unique bias, the bacterial point of view.

In this study, mutants were identified by using the trans-

* Corresponding author.

poson Tn5 lac, which is a derivative of Tn5 that retains thetransposition functions and kanamycin resistance, and areporter gene, which is the gene for 3-galactosidase (P-gal)without a functional promoter (14). Expression of P-galoccurs only when Tn5 lac is inserted into a functional gene.TnS lac insertions in the vir-regulated genes of B. pertussiswere identified by selecting for kanamycin-resistant mutantsthat expressed 3-gal when the vir-regulated genes were onbut not when the vir-regulated genes were off. Characteriza-tion of these mutants reveals new information on the molec-ular requirements for pathogenesis and, we hope, providesnew insights for an effective and safe vaccine formulation.

MATERIALS AND METHODSBacterial strains and growth conditions. Escherichia coli

strains were grown on L-agar plates (21); B. pertussis strainswere grown on Bordet-Gengou agar (BGA; BBL Microbiol-ogy Systems, Cockeysville, Md.), as described previously(39), or on X agar. X agar was composed of Stainer-Scholte(SS) broth (32) with 50 mM Tris (pH 7.6) supplemented with1.5% agar-0.15% bovine serum albumin. The bovine serumalbumin helped overcome difficulties in growing B. pertussison semisolid medium without blood.

Isolation of TnS lac insertions in the B. pertussis chromo-some. A suicide plasmid (pUW1019) to deliver Tn5 lac to B.pertussis was made from pRK2013 kan::Tn7 as describedpreviously for pUW964 (39). P1::Tn5 lac was used as thesource of Tn5 lac (14). This plasmid could replicate in E. colibut not in B. pertussis. This plasmid was conjugation profi-cient, and all kanamycin-resistant B. pertussis strains after amating were the result of transposition events into thechromosome. Phenotypic expression of kanamycin resis-tance in B. pertussis was found to have a long lag period, andovergrowth by E. coli reduced the viability of the B. pertus-

2674

on January 10, 2021 by guesthttp://iai.asm

.org/D

ownloaded from

VIRULENCE MUTANTS OF B. PERTUSSIS 2675

sis cells. The use of minimal medium which did not supplythe amino acid requirements of the E. coli donor strainHB101 was found to improve the frequency of mating. Thenalidixic acid-resistant B. pertussis recipient BP338 wasgrown for 24 h on BGA, and the E. coli donor strainHB101(pUW1019) was grown on L agar. Both donor andrecipient strains were harvested in broth prewarmed to 37°Cat an optical density at 600 nm approximately equal to 1.0 (E.coli in L broth and BP338 in SS broth). The strains weremixed at a ratio of 10 times more B. pertussis than E. coli,and plate matings were performed by plating 1.0 ml of themixture on X agar supplemented with 20 mM MgSO4. Theplates were incubated for 5 to 7 h. The cells were harvestedon X-agar plates with 25 ,ug of kanamycin per ml and 30 ,ugof nalidixic acid per ml and incubated at 37°C until thecolonies appeared, which took up to 1 week. Matings withstrain J5(pRP4) (37) were performed as described above forkanamycin selection, and tetracycline was used at 15 p.g/ml.

Determination of "-gal activity. A modification of theprotocol of Miller (23) by using the chromogenic substrateo-nitrophenyl-i-D-galactopyranoside (ONPG) was used forthe determination of ,-gal activity. The bacteria were grownfor 24 h on BGA plates (conditions supporting vir geneexpression) or BGA plates supplemented with 20 or 50 mMMgSO4 (conditions suppressing vir gene expression) andharvested in 3.0 ml ofZ buffer, and the optical density at 600nm was determined with a spectrophotometer (Spec 20). Thecells were lysed by adding sodium lauryl sulfate (finalconcentration, 0.0003%) and a drop of chloroform. To startthe reaction, ONPG was added to a final concentration of 0.8mg/ml. The reaction was allowed to continue at roomtemperature for 10 min and then was stopped by the additionof 1.5 ml of 1 M sodium carbonate. Formation of the yellowproduct was measured at an optical density of 420 nm, andthe background caused by bacterial cell debris was deter-mined at an optical density of 550 nm. Units of activity werecalculated as described by Miller (23).DNA manipulations. Chromosomal DNA was isolated as

described previously (39). Enzyme reactions were per-formed by the specifications of the manufacturers. Southerntransfers were performed by using positively charged nylonmembranes (1). The oligonucleotide probe was synthesizedon a DNA synthesizer (model 380A; Applied BiosystemsInc.). It was purified by thin-layer chromatography with

U,

ac soo

._a

a-c

.,

0

c bea

U)

a

silica gel 60 F254 (E. Merck AG, Darmstadt, Federal Repub-lic of Germany) with n-propanol (55%), NH40H (35%), andwater (10%). The 30-mer was visualized by shortwave UVlight, cut from the plate, eluted with water, extracted threetimes with ether, lyophilized, and suspended in water. Theoligonucleotide was end labeled with 32P by the kinaseexchange reaction (1).

Assay for virulence factors. DNT, hemagglutination byFHA, adenylate cyclase, and pertussis toxin were measuredas described previously (39).

Binding of calcofluor (or cellufluor; Polysciences, War-rington, Pa.) was assayed by gently layering 5 ml of a 0.005%solution onto a fresh plate of cells (24 h). The solution wasallowed to sit for 1 min, the liquid was poured off, and theplate was rinsed two times with 10 ml of water. Thefluorescence was observed by using longwave UV light.

Antigen phenotypes. Monoclonal antibodies, which werekindly provided by Michael Brennan (Bureau of Biologics,Food and Drug Administration, Bethesda, Md.), were usedto determine the serotypes of the strains. We attempted toavoid the ambiguity in the agglutinogen nomenclature byspecifying the products by their biochemical designations.Serotype 3 of Preston et al. (30) was equivalent to thefimbriae recognized by serotype 6 in the scheme of Elderinget al. (7) and Li et al. (17), and we used the 3/6 designation,since serotype 3 of Eldering et al. (7) recognizes a nonfim-brial 69-kilodalton (kDa) protein (5).

Fimbria 2 (22 kDa) was detected by using monoclonalantibody BPD5 (17), and fimbria 3/6 (21.5 kDa) was detectedby using monoclonal antibody BPC10 (17). Agglutinationtests were performed by mixing cells that were grown onBGA harvested in saline to an optical density at 600 nm ofabout 1.0 with a 1/1,500 dilution of serum (fimbria 2) or a1/600 dilution of serum (fimbria 3/6), and 100 ,ul of antibodywas mixed with 100 ,ul of bacteria and 100 ,ul of saline inflat-bottom 24-well microtiter plates. The suspensions wereshaken vigorously at 37°C for 5 min and read with adissecting microscope with an indirect light source. Controlwells containing only bacteria and saline were also tested torule out autoagglutination. Agglutination did not appear tobe a very sensitive assay for fimbriae. In mixing experi-ments, a population of 10% fimbria 2-positive cells mixedwith fimbria-negative cells no longer gave a positive re-sponse by agglutination. In our hands, immunoblotting also

a

a

a

40rn

a

es

co

a

a

.2

U)W)el

10 50 100 500 1000units a-galactosidase in virulent phase

10 so 100 500 1000units a-galactosidase in virulent phase

FIG. 1. (A) 3-Gal activity was determined for Tn5 lac mutants grown on X agar (virulent phase) or on the same medium supplemented with20 mM MgSO4 (modulated avirulent phase). (B) An avirulent-phase derivative was selected on erythromycin plates for each strain tested. Thepaired strains were grown on BGA plates, and ,-gal activity was determined for each strain.

- A4 A

AAA AAA

- A A

A

I/

VOL. 57, 1989

iv

on January 10, 2021 by guesthttp://iai.asm

.org/D

ownloaded from

2676 WEISS ET AL.

gave unsatisfactorily results. The monoclonal antibodies didnot react with boiled, denatured subunits after sodiumdodecyl sulfate-polyacrylamide gel electrophoresis (5, 17),and colony blots with whole bacteria never gave completelynegative results even for the vir mutant controls.

Lipooligosaccharide A, corresponding to agglutinogen 1,was detected by using colony blots as described previously(35) with monoclonal antibody G1OF8C3 (18).

Agglutinogen 3, also identified by monoclonal antibodyBB05 as a 69-kDa outer membrane protein, was detected inWestern blots as described previously (39) by using eithermonoclonal antibody BB05 (24) or monoclonal antibodyBPE3 (5).

Antibiotic sensitivities. Virulent-phase organisms are moresensitive to erythromycin than avirulent phase strains are(38), and the vir mutant derivatives used in this study wereselected on BGA with 0.2 pLg of erythromycin per ml. Theloss of three or more vir-regulated determinants was alwaysverified.

TABLE 1. 13-Gal activity in TnS-lac insertion mutants

,8-Gal activity on the RatiobStrain following growth mediaa

Permissive Nonpermissive Modulation Mutation

vir-regulated mutantsBPM177BPM178

BPM245BPM246

BPM409BPM410

BPM433BPM434

BPM1579BPM1580

889 7958 125

400 10283 62

2,380 4837 62

1,04078

102116

231 446 36

11 15

3.9 4.8

50 64

10 13

5.2 3.8

RESULTS

Generation of TnS lac fusions in B. pertussis. Wild-type B.pertussis did not produce any 1-gal activity, as determinedby using either ONPG or 5-bromo-4-chloro-3-indole 13-D-galactopyranoside (X-gal). This allowed us to use Tn5 lac togenerate transcriptional fusions to B. pertussis promotersand to monitor promoter activity by assaying for 1-gal. Over1,700 different kanamycin-resistant Tn5 lac insertion mu-tants in B. pertussis were tested for 13-gal expression underpermissive conditions for vir gene expression and nonper-missive conditions for expression of the vir genes by thequantitative ONPG assay. The results for a representativeseries of independent, consecutive, kanamycin-resistant in-sertion mutants are plotted in Fig. 1; this does not includeresults for five mutants which had no detectable 13-galactivity. Each point in Fig. 1 represents 1-gal activity instandard units for a single Tn5 lac insertion mutant in thevirulent phase plotted versus expression in the avirulentphase. Two different methods were used to monitor expres-sion of 13-gal in the avirulent phase: modulation or mutationof the vir gene. 13-gal activity was examined when the samestrain was grown either on medium that supported theexpression of the vir-regulated genes or on medium that wassupplemented with 20 mM MgSO4 (Fig. 1A). This reversiblysuppressed the expression of the vir-regulated genes, acondition called modulation (15). 1-Gal expression of asingle mutant was plotted against the 1-gal expression of itserythromycin-resistant avirulent-phase derivative, whichwas grown on the same plate (Fig. 1B). Previous studieshave shown that erythromycin selects for mutations in thevir regulatory gene, and these mutants do not express thevir-regulated genes even when they are grown under condi-tions permissive for vir gene expression (38). In our screen-ings, we numbered each mutant with a consecutive oddnumber and reserved the following even number for itsavirulent derivative.The 1-gal activities for the different strains differed over

several orders of magnitude (Fig. 1A and B), suggesting thatexpression is determined by the site of insertion and that alarge number of different insertion sites are involved. Onlyabout 3% of the insertion mutants did not produce 1-galactivity. This was unexpected. Random transposition pre-dicts a maximum of only 50% of the cells should be 1-galpositive, since insertion in only one orientation would createa productive 13-gal transcriptional fusion. Another unex-

BPM1809BPM1810

BPM1821BPM1822

BPM2041BPM2042

BPM2055BPM2056

BPM2119BPM2120

BPM2123BPM2124

BPM2859BPM2860

BPM3171BPM3172

BPM3183BPM3184

Pleiotropic mutantsBPM575BPM779BPM2375BPM2377 Hly+BPM2378BPM3447

188 3052 NDC

14362

414112

17534

264100

23464

36693

2,285162

1,747130

92ND

198136

18ND

6764

6968

7279

175146

221217

6.3 3.6

1.6 2.3

2.1 3.7

9.7 5.1

3.9 2.6

3.4 3.7

5.1 3.9

13 14

7.9 13

420 582 0.72389 549 0.71755 630 1.2430 312 1.4153 175483 864 0.56

NDNDND2.8

ND

a 1-Gal activity was determined for each TnS lac mutant (virulent-phasestrains are indicated by odd numbers) and its paired avirulent-phase derivative(indicated by the next larger even number) obtained by selection on erythro-mycin plates. Both strains were grown for 24 h on BGA (permissive for virgene expression) or BGA plus MgSO4 (nonpermissive for vir gene expres-sion). Arbitrary units of 13-gal activity were determined by the method ofMiller (23).

b The ratio is the value for the virulent phase divided by either the value forthe MgSO4-grown, virulent-phase strain (modulation) or the value for theavirulent-phase strain (mutation).

' ND, Not determined.

pected observation was that most strains expressed more13-gal activity in the avirulent phase than they did in thevirulent phase, either under the modulated conditions or, forthe avirulent mutants, under the permissive conditions, as

INFECT. IMMUN.

on January 10, 2021 by guesthttp://iai.asm

.org/D

ownloaded from

VIRULENCE MUTANTS OF B. PERTUSSIS 2677

TABLE 2. Putative TnS lac mutations in vir-regulated genes'

Strain (genotype) Adc Hly FHA DNT 6p9roktDa Cal Ptx Fim2 Fim3/6 LosA

BPM177 (unknown) + + + + + + + - - +BPM245 (unknown) + + + + + + + - - +BPM409 (fhaBI::TnS lac) + + - + + + + - - +BPM433 (adc-l::TnS lac) - - + + + + + - - +BPM1579 (unknown) + + + + + + + - - +BPM1809 (dnt-l::TnS lac) + + + - + + + - - +BPM1821(fhaAJ::Tn5 lac) + + - + + + + - - +BPM2041 (unknown) + + + + + + + - - +BPM2055 (unknown) + + + + + + + - - +BPM2119 (unknown) + + + + + + + - - +BPM2123 (unknown) + + + + + + + + - +BPM2375 (pleiotropic) + + + + + + + + - +BPM2859 (unknown) + + + + + + + - - +BPM3171 (unknown) + + + + + + + - - +BPM3183 (adc-2::TnS lac) - - + + + + + + + +BPM3447 (pleiotropic) + + + + ± + + - + +BP338 (parental) + + + + + + + + - +BP347(vir-1::TnS) - - - - - - - - - +

a Abbreviations: Adc, adenylate cyclase activity; Hly, hemolysin activity; FHA, filamentous hemagglutinin determined by hemagglutination and immunoblot-ting; DNT, dermonecrotic toxin activity assayed in infant mice; 69-kDa protein, vir-specific outer membrane protein corresponding to agglutinogen 3; Cal,calcofluor binding; Ptx, pertussis toxin activity; Fim2, fimbria 2 determined by agglutination; Fim3/6, fimbria 3/6 determined by agglutination; LosA,lipooligosaccharide A corresponding to agglutinogen 1.

shown by most values appearing above the diagonal line inFig. lA and B. In retrospect, this is consistent with theobservations that avirulent-phase strains have a faster dou-bling time and produce larger colonies than do virulent-phase strains (29).A single mutant, strain BPM177, gave ,-gal activity con-

sistent with an insertion in a vir-regulated gene, that is, it fellbelow the diagonal line (Fig. 1A and B) under both condi-tions that suppressed vir gene expression, and to date wehave detected 14 insertions in vir-regulated genes, for afrequency of about 1 per 125 strains tested.Each of the 14 putative mutations was tested by modula-

tion or mutation to ensure that it was in a vir-regulated gene(Table 1), and to be considered to have a mutation in avir-regulated gene, the strain had to fulfill both requirements.Several mutants, including BPM2375 and BPM3447, did notfulfill both requirements but were further characterizedbecause they appeared to be pleiotropic for several vir-regulated genes (see below).

Characterization of the TnS lac insertion mutations invir-regulated genes. We attempted to determine the pheno-types of the mutants (Table 2). Strain BPM3171 was shownto have a TnS lac insertion in or near the pertussis toxin geneby Southern blot analysis with KpnI-restricted chromosomalDNA (Fig. 2). The probe, pUW972, was derived from thepertussis toxin mutant BP357, which contained a TnS inser-tion in the structural gene (22, 39). Plasmid pUW972 con-tained the 4,696-base-pair EcoRI to BamHI region of thepertussis toxin, which has been sequenced independently byNicosia et al. (27) and Locht and Keith (20). In addition, itcontained the entire Tn. transposon. This probe shouldhybridize to two fragments corresponding to the two KpnIfragments in the pertussis toxin gene (8 and 12 kilobases[kb]) and to a third fragment corresponding to the TnS lacinsertion fragment, unless the insertion was in the pertussistoxin gene. There were no KpnI sites in Tn5 lac; therefore,the fragment containing TnS lac must be greater than 12 kb,the size of the transposon. These predictions were met for allof the insertions except BPM3171, in which only two homol-ogous fragments were seen. The 8-kb pertussis toxin band

was missing, and the band containing Tn5 lac was 20 kbwhich was consistent with the prediction that the 12-kbtransposon inserted into the 8-kb fragment.

Culture supernatants prepared from BPM3171 produced apositive reaction for pertussis toxin, the characteristic clus-tering appearance in the CHO cell assay (10), suggesting thatthis mutant is not negative for pertussis toxin. However, thewell-characterized mutants BP357 and BP356 (39) had TnSinsertions within the structural gene for pertussis toxin (20,22, 27) and yet they still elicited a positive CHO cell reaction

A B C D E F G H I J K L

w

S~~~~~~~~~~~

4*a*.~ -

FIG. 2. Chromosomal digests cut with Kpnl were hybridizedwith plasmid pUW972, which contained both pertussis toxin andTn5 sequences. Lanes: A, BPM2055; B, BPM2859; C, BPM2041; D,BPM2119; E, BPM177; F, BPM3447; G, BPM1579; H, BPM2123; I,BPM3171; J, BPM245; K, Hindlll-digested lambda DNA standard;L, cosmid clone pUW13-4 containing the adenylate cyclase gene.The arrows denote the lambda bands that contained the cos site (27,23, and 4 kb, from top to bottom, respectively) which hybridized tothe cos site that was also present on pHC79.

VOL. 57, 1989

on January 10, 2021 by guesthttp://iai.asm

.org/D

ownloaded from

2678 WEISS ET AL.

A B C D E F G H I J K L M N

P R S T U V W X Y Z

FIG. 3. Detection of FHA with a Coomassie brilliant blue-stained gel of whole bacterial protein separated by sodium dodecylsulfate-polyacrylamide gel electrophoresis. Lanes: A, myosin, 200-kDa protein standard; B, blank; C, BP338; D, BP347; E, BP348; F,BP356; G, BP357; H, Tohama; I, BP325; J, BP370 (34); K, 18323(35); L, BPM177; M, BPM245; N, BPM409; 0, BPM3447; P,BPM3183; Q, BPM3171; R, BPM2859; S, BPM2375; T, BPM2123;U, BPM2119; V, BPM2055; W, BPM2041; X, BPM1821; Y,BPM1809; Z, myosin. The arrowheads denote myosin.

(39). Studies are in progress to determine whether thisinsertion is in or only adjacent to the pertussis toxin operon.

Strains BPM433 and BPM3183 were both nonhemolyticand deficient in adenylate cyclase enzyme activity (Table 2).It has been shown that several genes in an operon are

required for both hemolysis and adenylate cyclase (3, 6, 9).Strains BPM409 and BPM1821 were deficient in the pro-

duction of FHA, as shown by their failure to hemagglutinatesheep erythrocytes and their failure to cross-react withantisera to this protein (Table 2). The presence of the220-kDa protein believed to be FHA also was not apparent ineither of these strains (Fig. 3, lanes N and X) or in the virmutant BP347 (Fig. 3, lane D). Two virulent-phase B.pertussis strains, BP325 and BP370 (Fig. 3, lanes I and J,respectively), were found to be spontaneous FHA mutants.BP370 was used by Black et al. (4) in a study of theimmunogenicity of a pertussis toxin mutant.

It is interesting that the activities of the two promoterfusion mutants were so different, with that of BPM409 beingabout 10-fold higher than that of BPM1821 (Table 1). At leastthree transcriptional units are required for FHA production(35), and we mapped these two insertions to different tran-scriptional units.One mutant, BPM1809, appeared to be deficient in the

production of DNT. Virulent-phase bacteria, but notBPM1809 and avirulent-phase bacteria, produced character-istic lesions when they were injected subcutaneously. Adilution series was performed to determine the minimalnumber of bacteria that gave a positive reaction (Table 3).Wild-type strain BP338 elicited a positive reaction when itwas diluted to an optical density at 600 nm of 0.01, whereasstrain BPM1809 was negative at an optical density at 600 nmof 1.5, suggesting that there was at least a 150-fold reductionin DNT activity in mutant BPM1809.

Characterization of the serotypes of the mutants. Recently,the identity of many of the serotype antigens has beendetermined. We characterized the insertion mutants usingmonoclonal antibodies corresponding to different serotypeswhich were produced by Michael Brennan and coinvestiga-tors at the Bureau of Biologics. Avirulent-phase strains werenontypeable with several of the different typing sera, sug-gesting that they do not express those antigens (5, 17, 18).

Serotype 1 has been shown to be composed of lipooli-gosaccharide (18). Monoclonal antibody G10F8C3 recog-nizes the A form of the lipooligosaccharide. This antigenicdeterminant was strongly expressed by both virulent- and

TABLE 3. DNT activity

Strain Dose Lesions'(optical density)'BP338 (wild type) 0.10 ++

0.06 +0.03 +0.01 +

BP347 (avirulent) 1.51.00.5

BPM1809 1.51.00.50.20.1

"Mice were injected with a bacterial suspension in SS broth at the indicateddensity at 600 nm.

b The lesions were scored as follows: ++++, greater than 1 cm, veryblue-black, often accompanied by death; + + +, greater than 1 cm, blue-black;++, about 1 cm. blue-black; +, less than 1 cm, red lesions; ±, faint rednessat the site of injection; -, no response.

avirulent-phase bacteria and was expressed by all of themutants (Table 2), as determined by colony blots. Li et al.(18) have described strains that are negative for lipooligosac-charide A, and they do not correlate with the avirulent-phasestrains, suggesting that this monoclonal antibody does notrecognize a vir-specific determinant.

Characterization of the 69-kDa protein. An antigen pro-duced by virulent-phase B. pertussis was recognized bymonoclonal antibody BB05. This monoclonal antibody wasoriginally produced against the 68-kDa protein produced byBordetella bronchiseptica, which has been shown to be aprotective immunogen in experimental B. bronchisepticainfections (24). The corresponding protein in B. pertussis hasbeen shown to have a molecular mass of 69 kDa. Thisprotein has been shown to be a vir-regulated gene thatcorresponds to Eldering serotype 3 (5, 7).

Several well-characterized mutants from other studies (39)and all of the TnS lac mutants from this study were screenedfor the presence of the 69-kDa antigen. All of the strains withTn5 lac insertion mutations in the vir-regulated genes pro-duced the 69-kDa protein in Western blots (immunoblots)when they were tested with either the original monoclonalantibody BB05 or monoclonal antibody BPE3 (data notshown). Mutants deficient in FHA, pertussis toxin, hemoly-sin, and adenylate cyclase toxin all produced similaramounts of a protein with the same molecular size as that ofthe wild-type strain. However, the avirulent-phase strainBP347 did not produce any cross-reacting material, nor did aBordetella avium isolate. Previous studies have associatedthe BB05 antigen with adenylate cyclase (28); however, it isunlikely that it is the structural protein for adenylate cyclase,since such mutants as BPM433, BP348, and BPM3181 thatwere deficient in adenylate cyclase all produced the full-sizeprotein.

Characterization of fimbria production. B. pertussis pro-duces two antigenically distinct fimbriae (pili); type 2 iscorrelated with the presence of fimbriae with a subunitmolecular mass of 22 kDa, and type 3/6 is correlated with thepresence of fimbriae with a subunit molecular mass of 21.5kDa (17). For this study, we typed the fimbriae by aggluti-nation with monoclonal antibodies.The expression of fimbriae was found to be variable. B.

pertussis Tohama has been reported to be serotype 1.2.3 (17,

INFECT. IMMUN.

on January 10, 2021 by guesthttp://iai.asm

.org/D

ownloaded from

VIRULENCE MUTANTS OF B. PERTUSSIS 2679

A B C D E F G H I J

* 411*~~~~~~~~~~~~~~~4,._

FIG. 4. Southern blot with the fimbrial gene probe, Sall-digestedchromosomal DNA was probed with the 30-mer GACGACGGCACCATCGTCATCACCGGCACC. Lanes: A, BPM177; B, BPM245;C, BPM1579; D, BPM2041; E, BPM2055; F, BPM2119; G,BPM2123; H, BPM2375; I, BPM2859; J, BPM3447. The arrowsdenote the 4.4-, 2.3-, and 2.0-kb HindIll-digested lambda DNAstandards and the 1.3-, 0.87-, and 0.60-kb HaeIII 4X174 standards(from top to bottom, respectively).

18), with the 2 referring to fimbria 2 expression. Interest-ingly, the isolate in our laboratory was found to be devoid offimbriae. BP338, a nalidixic acid-resistant mutant of ourTohama isolate and the parental strain for all of our mutants,was found to be positive for fimbria 2. Twenty singlecolonies of BP338 were tested, and three were found to benegative for fimbria 2. After one passage, the fimbria 2-positive isolates remained positive, and the fimbria 2-nega-tive colonies remained negative.The fimbrial serotype of the TnS lac mutants is summa-

rized in Table 2. It is possible to conclude that isolates thatexpress a fimbria do not have mutations in that gene, but itis not possible to conclude that a negative strain has amutation in the fimbrial gene. Interestingly, two mutants,BPM3183 and BPM3447, expressed fimbria 3/6. Neither theTohama nor the BP338 strains expressed this fimbria, sug-gesting that metastable expression of this fimbria also oc-curs.

The high proportion of fimbria-negative mutants suggestedthat our selection process might enrich for the fimbria-negative phenotype. One possibility was that the fimbriaemay interfere with the mating process used to introduce thetransposon. This was tested by mating both fimbria 2-positive and -negative isolates of BP338 with an E. coli donorcontaining plasmid RP4. We observed the same frequency(about 30 x 105/ml) of kanamycin-resistant or tetracycline-resistant transconjugants when either type of recipient wasused, suggesting that the fimbriae did not influence theefficiency of this process.To determine whether any of the Tn5 lac insertion muta-

tions mapped within the structural genes of the fimbriae andto circumvent the problem of variable expression of thefimbrial gene, an oligonucleotide probe, 5'-GACGACGGCACCATCGTCATCACCGGCACC, was synthesized.This DNA sequence encodes the amino-terminal end of thefimbria 2 protein (19). The probe was labeled with 32P andused to probe Sall digests of chromosomal DNA. No changein the hybridization pattern was observed for two fragments(1.7- and 0.9-kb fragments) (Fig. 4), suggesting that thesegenes did not suffer a Tn5 lac insertion. Mooi et al. (25) useda mixture of four different oligonucleotides and discovered athird fragment of 3.9 kb, in addition to the 1.7- and 0.9-kbfragments. The three fragments are believed to correspondto the genes for fimbriae 2 and 3/6, as well as a third, as yetundefined, fimbria. Perhaps one of the remaining threeoligonucleotides is required for efficient hybridization to thisthird fragment.

Characterization of calcofluor binding. Leigh et al. (16)have detected capsular mutants in Agrobacterium tume-faciens by using the laundry whitener calcofluor, a fluores-cent stain for cellulose and polysaccharides. In earlier stud-ies a capsule in virulent-phase B. pertussis but not inavirulent-phase B. pertussis has been described (12), but thenature of this capsule has not been defined. We tested B.pertussis for calcofluor binding and found that increasedfluorescence was observed in virulent-phase strains but notin avirulent-phase strains (Fig. 5). All of the Tn5 lac mutantswere as fluorescent as the parental strain, suggesting that nomutant was deficient in this vir-specific phenotype (Table 2).A B. avium strain was negative for this phenotype.

Characterization of mutants pleiotropic for vir-regulatedgene expression. Several mutants were initially selected

. -

FIG. 5. Calcofluor-binding assay. Fluorescence was observed under longwave UV light. Blots: A, BP338; B, BP347; C, BPM2375; D, B.avium W; E, BPM2055; F, BPM2056; G, BPM2041; H, BPM2042; I, BPM409; J, BPM410.

VOL. 57, 1989

on January 10, 2021 by guesthttp://iai.asm

.org/D

ownloaded from

2680 WEISS ET AL.

because they were deficient in hemolysis and were laterfound to be pleiotropic for the expression of several vir-regulated genes. 13-Gal expression was not modulated byMgSO4 in any mutants of this class (Table 1). The prototypemutant in this class was BPM2375 (Tables 1 and 2). We referprimarily to studies that have been performed with thisstrain; however, all the other mutants in this class hadsimilar properties.

All of these mutants were sensitive to erythromycin;however, all of them appeared to be "locked" in the vir+state. Selection of wild-type BP338 on erythromycin yieldedvir mutants at a rate of about 1 in 106, and most of thevir-regulated mutants had a similar rate. BPM2375 andBPM3447, however, did not yield any erythromycin-resis-tant derivatives when greater than 1010 bacteria were platedonto erythromycin.BPM2375 grew slowly and produced very small colonies.

This mutant was unstable, and large hemolytic revertantswere observed even if selection for kanamycin resistancewas maintained. We presume that these kanamycin-resistanthemolytic derivatives were due to second site mutations.One of the hemolytic revertants of BPM2375 was chosen forfurther characterization and was called BPM2377. Thishemolytic derivative can switch to the avirulent state, asevidenced by a high efficiency of plating on erythromycin;and BPM2378, an avirulent derivative of BPM2377, wasselected. It appeared to be similar to all other avirulentstrains. This series of strains was characterized for 1-galexpression in the presence and absence of MgSO4. BPM2377did not appear to be modulated by MgSO4; however, thetwofold-greater expression by the vir+ strain BPM2377 ver-sus that by the vir mutant strain BPM2378 is suggestive thatthe insertion was in a vir-regulated gene. Similar results wereobtained with BPM3447; however, this mutant was lessstable and was not characterized as extensively. BPM2375appeared to express reduced levels of all of the vir genes forwhich quantitative assays were available (Table 2).

Determining the number of independent mutations gener-ated by TnS lac. The actual number of genes regulated by viris unknown, so it was of interest to determine how many ofthe mutations we isolated actually contained insertions inindependent genes. For six of the mutants, genotypes forfour independent genes or operons were easily assignable.The phenotypes of the other eight mutants remain to bedetermined, but we know they were not deficient in severalwell-characterized vir-regulated genes. It is important toknow whether they represent one or several mutant classes.We concentrated our effort on mapping the insertions forwhich the phenotype could not be determined, based on theassumption that they were in sites independent of those forwhich a phenotype could be assigned.The large size of TnS lac (12 kb) made analysis of the

insertion site difficult. KpnI is one of the few restrictionenzymes that did not cut within TnS lac; however, it did notseem to cut frequently in the B. pertussis chromosome.Thus, the chromosomal fragments containing TnS lac were

very large. Each mutant contained a single copy of TnS lacin its chromosome, as determined by Southern blots probedwith TnS sequences (Fig. 2). As described above, the twosmaller bands corresponded to the intact pertussis toxingene, and the band greater than 12 kb corresponded to theTnS lac fusion fragment. The size of the third fragment wasdifferent for most of the mutants, suggesting a differentinsertion site. In addition, the mutations were mapped withSmaI and Sall. Both cut once within TnS lac, generating twofragments when they were probed with Tn5 lac. Since

TABLE 4. Mapping of restriction fragments containing Tn5 lac inmutants with unknown phenotypes

Molecular size (kb) of the followingStrain restriction fragments:

KpnI Smala SalIb

BPM177 38 14.2 12.9BPM245 23 NDb 12.9BPM1579 34.5 18.3 15BPM2041 40.5 14.7 14BPM2055 60 28.4 13.1BPM2119 41 15.7 13.1BPM2123 ND 15.7 13.1BPM2375 14 ND 13.5BPM2859 41.5 12.4 14.3

" The enzyme cut Tn5 lac once; the molecular size was the sum of the twochromosomal restriction fragments homologous to Tn5 lac.

b ND, Not determined.

insertions can occur at different sites in the same gene, themolecular masses of two SmaI and two Sall fragments wereadded to determine whether the insertion corresponded to aunique fragment. The sizes of the TnS lac KpnI fragments,determined from Fig. 2 and several other experiments, andthe SmaI and Sall fragments are summarized in Table 4. Thedata suggest that all of the mutations, except for those inmutants BPM2119 and BPM2123, mapped to different re-striction fragments and perhaps to different genes.

DISCUSSIONWe used a promoter fusion transposon, TnS lac, to iden-

tify mutants in vir-regulated genes by screening for mutantsthat expressed ,B-gal when the vir-regulated genes wereexpressed but not when they were repressed. We identified14 putative mutations in vir-regulated genes as isolates thatexpressed more 3-gal activity when the strains were in thevirulent phase. However, all mutants had low-level activitywhen they were in the avirulent phase. Numerous studieshave failed to detect any protein products of the vir-regu-lated genes in avirulent-phase strains (5, 11, 12, 15, 17, 29,35, 38, 39), and our studies of transcription of the vir-regulated genes have shown that avirulent-phase mutants donot produce detectable transcripts of the vir-regulated genes(A. R. Melton and A. A. Weiss, manuscript in preparation).A possible explanation is that internal TnS lac sequences canserve as promoters in B. pertussis; however, we did notdemonstrate that an internal TnS lac promoter is responsiblefor this phenomenon. Alternatively, transposition may notbe random in B. pertussis, and insertion preferentially oc-curs in active genes in one orientation. A nonrandom orien-tation preference has been observed for another TnS deriv-ative, Tn5-VB32, in which 77% of the insertions that werecharacterized generated promoter fusions (2).Knapp and Mekalanos (13) have used Tn5 ISSOL::phoA, a

transposon which detects fusions in secreted factors, toidentify several vir-activated as well as vir-repressed genes.We did not identify any vir-repressed genes among the 1,700genes that we screened, possibly because the majority of thefusions appeared to be vir repressed; that is, they had moretranscriptional activity when they were in the Vir state. Inthe studies of Knapp and Mekalanos (13), the secretedproduct from the vir-repressed gene was detected in strainsthat were in both the Vir+ and Vir- state, suggesting that theloss of the vir gene product increases the expression of agene that is already on. This is in contrast to the vir-inducedgenes, which have an absolute requirement for vir.

INFECT. IMMUN.

on January 10, 2021 by guesthttp://iai.asm

.org/D

ownloaded from

VIRULENCE MUTANTS OF B. PERTUSSIS 2681

Several of the mutants identified in this study were defi-cient in previously well-characterized virulence factors suchas adenylate cyclase toxin and hemolysin, FHA, or DNT,verifying that the Tn5 lac insertions were truly identifyingvir-regulated genes. None of the insertions we characterizedwere in the fimbrial genes. This was not too surprising, sincewe were screening for transcriptional fusions and expressionof the fimbriae appeared to switch off and on. These obser-vations suggest that metastable expression of fimbriae can

occur in B. pertussis, as has been observed for type 1 pili inE. coli (31) or pilus expression in Neisseria gonorrhoeae(36). The molecular basis of the phase variation in B.pertussis has not been established yet, and it may or may notoccur by similar mechanisms.

In this study we defined a new vir-associated trait: in-creased binding of calcofluor. In A. tumefaciens calcofluorbinding is correlated with the presence of a carbohydratecapsule (16). We have not characterized further the bio-chemical nature of the increased calcofluor binding byvirulent-phase B. pertussis, but we hope to do so in futurestudies.Seven of the eight as yet undefined mutants appeared to

map to independent chromosomal locations. An estimate ofthe number of vir-regulated genes could be made. Severalmutants were associated with four well-defined Bordetellaproducts. These included pertussis toxin, the adenylatecyclase and hemolysin operon, FHA, and a new mutationdescribed in this study, that for DNT. At least four factorshave not been defined by mutagenesis. These include the69-kDa outer membrane protein, the two fimbriae, and thecapsule, as defined by calcofluor-binding experiments. If weassume that seven of the undefined insertion mutants are

different, there are at least 15 vir-regulated genes. This isprobably an underestimate of the total number of vir-regu-lated genes. Since we did not obtain mutants with some ofthe known characteristics, we assume that we also did notobtain mutants with some of the unknown characteristics.

B. pertussis seems to have a few secrets left to reveal.Characterization of these mutants for their biochemicalnature and for their virulence in animal models will help shedlight on what is and what is not required by this organism inthe disease process.

ACKNOWLEDGMENTS

This work was supported by Public Health Service grant AI-23695from the National Institute of Allergy and Infectious Diseases andby a fellowship from the Pew Foundation to A.A.W.We acknowledge the contribution of Pam Hurst to this project.

Monoclonal antibodies were provided by Michael Brennan. Wethank L. Kroos and D. Kaiser for providing us with TnS lac andJohn Monaco for critically reading this manuscript and assisting uswith the photography.

LITERATURE CITED

1. Ausubel, F. M., R. Brent, R. E. Kingston, D. D. Moore, J. G.Seidman, J. A. Smith, and K. Struhl. 1987. Current protocols inmolecular biology. John Wiley & Sons, Inc., New York.

2. Bellofatto, V., L. Shapiro, and D. A. Hodgson. 1984. Generationof a Tn5 promoter probe and its use in the study of geneexpression in Caulobacter crescentus. Proc. Natl. Acad. Sci.USA 81:1035-1039.

3. Berg, D. E., A. Weiss, and L. Crossland. 1980. Polarity of TnSinsertion mutations in Escherichia coli. J. Bacteriol. 142:439-446.

4. Black, W. J., J. J. Munoz, M. G. Peacock, P. A. Schad, J. L.Cowell, J. J. Burchall, M. Lim, A. Kent, L. Steinman, and S.

Falkow. 1988. ADP-ribosyltransferase activity of pertussis toxinand immunomodulation of Bordetella pertussis. Science 240:656-659.

5. Brennan, M. J., Z. M. Li, J. L. Cowell, M. E. Bisher, A. C.Steven, P. Novotny, and C. R. Manclark. 1988. Identification ofa 69-kilodalton nonfimbrial protein as an agglutinogen of Borde-tella pertussis. Infect. Immun. 56:3189-3195.

6. Brownlie, R. M., J. G. Coote, and R. Parton. 1986. Complemen-tation of mutations in Escherichia coli and Bordetella pertussisby B. pertussis DNA cloned in a broad-host-range cosmidvector. J. Gen. Microbiol. 132:3221-3229.

7. Eldering, G., C. Hornbeck, and J. Baker. 1957. Serologicalstudy of Bordetella pertussis and related species. J. Bacteriol.74:133-136.

8. Fine, P. E. M., and J. A. Clarkson. 1987. Reflections on theefficacy of pertussis vaccines. Rev. Infect. Dis. 9:866-883.

9. Glaser, P., H. Sakamoto, J. Belialou, A. Ullmann, and A.Danchin. 1988. Secretion of cyclolysin, the calmodulin-sensitiveprotein of adenylate cyclase-haemolysin bifunctional protein ofBordetella pertussis. EMBO J. 7:3997-4004.

10. Hewlett, E. L., K. T. Sauer, G. A. Myers, J. L. Cowell, and R. L.Guerrant. 1983. Induction of a novel morphological response inChinese hamster ovary cells by pertussis toxin. Infect. Immun.40:1198-1203.

11. Idigbe, E. O., R. Parton, and A. C. Wardlaw. 1981. Rapidity ofantigenic modulation of Bordetella pertussis in modified horni-brook medium. J. Med. Microbiol. 14:409-418.

12. Kasuga, T., Y. Nakase, K. Ukishima, and K. Takatsu. 1953.Studies on Haemophilus pertussis. Part I. Antigen structure ofH. pertussis and its phases. Kitasato Arch. Exp. Med. XXVI:121-134.

13. Knapp, S., and J. J. Mekalanos. 1988. Two trans-acting regula-tory genes (vir and mod) control antigenic modulation in Bor-detella pertussis. J. Bacteriol. 170:5059-5066.

14. Kroos, L., and D. Kaiser. 1984. Construction of TnS lac, atransposon that fuses lacZ expression to exogenous promoters,and its introduction into Myxococcus xanthus. Proc. Natl.Acad. Sci. USA 81:5816-5820.

15. Lacey, B. W. 1960. Antigenic modulation of Bordetella pertus-sis. J. Hyg. 58:57-93.

16. Leigh, J. A., E. R. Singer, and G. C. Walker. 1985. Exopolysac-charide-deficient mutants of Rhizobium meliloti that form inef-fective nodules. Proc. Natl. Acad. Sci. USA 82:6231-6235.

17. Li, Z. M., M. J. Brennan, J. L. David, P. H. Carter, J. L.Cowell, and C. R. Manclark. 1988. Comparison of type 2 andtype 6 fimbriae of Bordetella pertussis by using agglutinatingmonoclonal antibodies. Infect. Immun. 56:3184-3188.

18. Li, Z. M., J. L. Cowell, M. J. Brennan, D. L. Burns, and C. R.Manclark. 1988. Agglutinating monoclonal antibodies that spe-cifically recognize lipooligosaccharide A of Bordetella pertussis.Infect. lmmun. 56:699-702.

19. Lively, I., C. J. Duggleby, and A. Robinson. 1987. Cloning andnucleotide sequence analysis of the serotype 2 fimbrial subunitgene of Bordetella pertussis. Mol. Microbiol. 1:203-209.

20. Locht, C., and J. M. Keith. 1986. Pertussis toxin gene: nucleo-tide sequence and genetic organization. Science 232:1258-1264.

21. Maniatis, T., E. F. Fritsch, and J. Sambrook. 1982. Molecularcloning: a laboratory manual. Cold Spring Harbor Laboratory,Cold Spring Harbor, N.Y.

22. Marchitto, K. S., J. J. Munoz, and J. M. Keith. 1987. Detectionof subunits of pertussis toxin in Tn5-induced Bordetella mutantsdeficient in toxin biological activity. Infect. Immun. 55:1309-1313.

23. Miller, J. H. 1972. Experiments in molecular genetics, p.352-355. Cold Spring Harbor Laboratory, Cold Spring Harbor,N.Y.

24. Montaraz, J. A., P. Novotny, and J. Ivanyi. 1985. Identificationof a 68-kilodalton protective protein antigen from Bordetellabronchiseptica. Infect. Immun. 47:744-751.

25. Mooi, F. R., H. G. J. van der Heide, A. R. ter Avest, K. G.Welinder, I. Lively, B. A. M. van der ZeiJst, and W. Gaastra.1987. Characterization of fimbrial subunits from Bordetellaspecies. Microb. Pathog. 2:473-484.

VOL. 57, 1989

on January 10, 2021 by guesthttp://iai.asm

.org/D

ownloaded from

2682 WEISS ET AL.

26. Muller, A. S., J. Leeuwenburg, and D. S. Pratt. 1986. Pertussis:epidemiology and control. Bull. W.H.O. 64:321-331.

27. Nicosia, A., M. Perugini, C. Franzini, M. C. Casagli, M. G.Borri, G. Antoni, M. Almoni, P. Neri, G. Ratti, and R. Rappuoli.1986. Cloning and sequencing of the pertussis toxin genes:operon structure and gene duplication. Proc. Natl. Acad. Sci.USA 83:4361-4635.

28. Novotny, P., A. P. Chubb, K. Cownley, and J. A. Montaraz.1985. Adenylate cyclase activity of a 68,000-molecular-weightprotein isolated from the outer membrane of Bordetella bron-chiseptica. Infect. Immun. 50:199-206.

29. Peppler, M. S. 1982. Isolation and characterization of isogenicpairs of domed hemolytic and flat nonhemolytic colony types ofBordetella pertussis. Infect. Immun. 35:840-851.

30. Preston, N. W., N. Surapatana, and E. J. Carter. 1982. Areappraisal of serotype factors 4, 5 and 6 ofBordetella pertussis.J. Hyg. 88:39-46.

31. Spears, P. A., D. Schauer, and P. E. Orndorff. 1986. Metastableregulation of type 1 piliation in Escherichia coli and isolationand characterization of a phenotypically stable mutant. J. Bac-teriol. 168:179-185.

32. Stainer, D. W., and M. J. Scholte. 1971. A simple chemicallydefined medium for the production of phase I Bordetella pertus-sis. J. Gen. Microbiol. 63:211-220.

33. Standfast, A. F. B. 1958. Some factors influencing the virulence

for mice of Bordetella pertussis by the intracerebral route.Immunology 2:123-134.

34. Stibitz, S., W. Black, and S. Falkow. 1986. The construction ofa cloning vector designed for gene replacement in Bordetellapertussis. Gene 50:133-140.

35. Stibitz, S., A. A. Weiss, and S. Falkow. 1988. Genetic analysis ofa region of the Bordetella pertussis chromosome encodingfilamentous hemagglutinin and the pleiotropic regulatory locusvir. J. Bacterial. 170:2904-2913.

36. Swanson, J., S. Bergstrom, K. Robbins, 0. Barrera, D. Corwin,and J. M. Koomey. 1986. Gene conversion involving the pilinstructural gene correlates with pilus+ " pilus- changes inNeisseria gonorrhoeae. Cell 47:267-276.

37. Weiss, A. A., and S. Falkow. 1982. Plasmid transfer to Borde-tella pertussis: conjugation and transformation. J. Bacteriol.152:549-552.

38. Weiss, A. A., and S. Falkow. 1984. Genetic analysis of phasechange in Bordetella pertussis. Infect. Immun. 43:263-269.

39. Weiss, A. A., E. L. Hewlett, G. A. Myers, and S. Falkow. 1983.TnS-induced mutations affecting virulence factors of Bordetellapertussis. Infect. Immun. 42:33-41.

40. Weiss, A. A., E. L. Hewlett, G. A. Myers, and S. Falkow. 1984.Pertussis toxin and extracytoplasmic adenylate cyclase as viru-lence factors in Bordetella pertussis. J. Infect. Dis. 150:219-222.

INFECT. IMMUN.

on January 10, 2021 by guesthttp://iai.asm

.org/D

ownloaded from