JOURNAL OF BACTERIOLOGY, June 1992, p. 3945-3952 Vol. 174, No. 120021-9193/92/123945-08$02.00/0Copyright ©D 1992, American Society for Microbiology

Identification of Novel Loci Affecting Entry of Salmonellaenteritidis into Eukaryotic Cells

BARBARA J. STONE,' CHRISTINE M. GARCIA,' JULIE L. BADGER,' THERESA HASSETT,1RICHARD I. F. SMITH,' AND VIRGINIA L. MILLER' 2*

Department of Microbiology and Molecular Genetics' and Molecular Biology Institute,2University of California at Los Angeles, Los Angeles, California 90024

Received 6 February 1992/Accepted 10 April 1992

There are an estimated 2 million cases of salmonellosis in the United States every year. Unlike the incidenceof many infectious diseases, the incidence of salmonellosis in the United States and other developed countrieshas been rising steadily over the past 30 years, and the disease now accounts for 10 to 15% of all cases of acutegastroenteritis in the United States. The infecting organism is ingested and must traverse the intestinalepithelium to reach its preferred site for multiplication, the reticuloendothelial system. Despite several recentstudies, the genetic basis of the invasion process is poorly understood. An emerging theme from these studiesis that wild-type SalmoneUla organisms probably have several chromosomal loci that are required for the mostefficient level of invasion. In this study, we have identified and characterized 13 TnphoA insertion mutants ofSalmonella enteritidis CDC5 that exhibit altered invasion phenotypes. The mutants were identified by screeninga bank of TnphoA insertions in S. enteritidis CDC5str for their invasion phenotype in three tissue culture celllines (HEp-2, CHO, and MDCK). These 13 mutants were separated into six classes based on their invasivephenotypes in the tissue culture cell lines. Several mutants were defective for entry of some cell lines but notfor others, while two mutants (SM6 and SM7) were defective for entry into all three tissue culture cell lines.This suggests that Salmonella spp. may express more than one invasion pathway. Southern analysis andchromosomal mapping indicated that as many as nine chromosomal loci may contribute to the invasionphenotype. It is becoming clear that the invasive phenotype of Salmonella spp. is multifactorial and morecomplex than that of some other invasive members of the family Enterobacteriaceae.

Salmonella spp. are an important cause of disease world-wide and are responsible for an estimated 40,000 reportedcases of gastroenteritis in the United States annually (5).Salmonellosis costs $50,000,000 per year in the UnitedStates because of lost work hours and medical expenses (5).In addition to gastroenteritis, Salmonella spp. cause entericfever (Salmonella typhi) and septicemia (Salmonella choler-aesuis). In contrast to the incidence of typhoid fever andmany other infectious diseases in developed countries, theincidence of gastroenteritis due to Salmonella spp. has beensteadily increasing for the past 30 years (5, 36). In all threediseases (enteric fever, septicemia, gastroenteritis), the in-fecting organism is food or water borne and must cross theintestinal mucosal barrier to reach its preferred site ofmultiplication, the reticuloendothelial system. Analysis ofthe mechanism Salmonella species use to cross this barrierhas been the topic of several recent reports (7, 10, 11, 13),but the mechanism is still not well understood. What hasbecome clear from these studies is that the genetic basis forepithelial cell entry by Salmonella spp. is complex anddistinct from that of other invasive members of the Entero-bacteriaceae family: Yersinia spp., Shigella spp., and en-teroinvasive Escherichia coli.A single locus from Shigella spp. and individual genes

from Yersinia spp. which are responsible for the invasivephenotype have been identified (18, 25, 29). It appears thatthe genes involved in epithelial cell entry by Shigella spp.are encoded on a 37-kb region of an essential 200-kb viru-lence plasmid (25). Salmonella and Yersinia spp. also havehigh-molecular-weight plasmids required for virulence, but

* Corresponding author.

curing the virulence plasmid from either Yersinia (34) orSalmonella (16) spp. does not affect invasion, suggestingthat the genes required for invasion are located on thechromosomes of organisms in these genera. This has beenclearly demonstrated in the case of Yersinia pseudotubercu-losis and Yersinia enterocolitica, in which chromosomalgenes (invPstb, invent, and ail) have been cloned that individ-ually confer an invasive phenotype on the noninvasive E.coli strain HB101 (18, 29).

Several recent studies have sought to define the geneticbasis of epithelial cell entry by Salmonella spp. Finlay et al.(11) identified mutants of S. choleraesuis that were unable totranscytose polarized MDCK cells; all of these mutants weredefective for cell entry to some degree, entering at levelsbetween 1 and 57% of wild-type levels. Elsinghorst et al. (7)isolated cosmids carrying S. typhi chromosomal DNA thatcould confer an invasive phenotype for Henle 407 cells on E.coli. The level of invasion exhibited by E. coli carrying S.typhi invasion genes was 4 to 7% that of wild-type S. typhi.The homologous region from Salmonella typhimuriumcloned into E. coli did not confer the ability to invade Henle407 cells. Using a slightly different approach, Galan andCurtiss (13) identified a cosmid carrying S. typhimuriumchromosomal DNA that could complement the invasiondefect of a noninvasive S. typhimurium laboratory strain.However, when transferred to E. coli, this cosmid did notconfer an invasive phenotype. The relationship between thisclone and that identified by Elsinghorst et al. (7) is unknown,but comparison of restriction maps and gene locations sug-gests that they are not the same.

Together, these results suggest that wild-type Salmonellaspp. probably have several chromosomal loci that are re-quired to fully express the invasive phenotype. To begin to

3945

on October 29, 2017 by guest

http://jb.asm.org/

Dow

nloaded from

3946 STONE ET AL.

assess the number of chromosomal loci and genes involvedin the Salmonella invasion phenotype, we have generatedTnphoA insertion mutants of Salmonella enteritidis thatexhibit an altered invasion phenotype. The invasion mutantsfell into six classes, and the mutations were not clustered onthe chromosome. The results suggest that S. enteritidis hasseveral loci that influence the invasion phenotype and that S.enteritidis may have several pathways for cell entry.

MATERUILS AND METHODS

Bacterial strains, phage, plasmids, and tissue culture cells.E. coli HB101 (2), E. coli SM10 Apir (31), and S. enteritidisCDC5 (a clinical isolate from a case of gastroenteritis) and itsderivatives were stored at -70°C in Luria broth (LB) (27)medium containing 25% glycerol or on LB agar plates at 4°C.Antibiotics were used at the following concentrations: strep-tomycin (Str), 100 Rg/ml; ampicillin (Ap), 100 ,ug/ml; kana-mycin (Kn), 40 pLg/ml; gentamicin, 100 ,ug/ml. Strains con-taining plasmids were routinely grown in LB supplementedwith the appropriate antibiotics. A spontaneous streptomy-cin-resistant mutant of S. enteritidis CDC5 was obtained byspreading the bacterial cells collected from a 10-ml overnightculture grown in LB on an LB agar plate containing strep-tomycin. Nine Strr colonies, all of which had the sameinvasion phenotype and growth rate as the parent strain,were obtained. One of these mutants was selected for furtherstudies and was designated CDC5str. Salmonella phage P22HTint was propagated and used in transductions as previ-ously described (6). Plasmid pRT733 (41) is a derivative ofthe suicide vector pJM703.1 (31) that carries TnphoA andconfers Apr and Knr. Human laryngeal epithelium (HEp-2)cells, Chinese hamster ovary (CHO) cells, and Madin-Darbycanine kidney (MDCK) cells were maintained and preparedfor the invasion assay as previously described (9).

Invasion and attachment assays. Either quantitative orqualitative (for screening) invasion assays were performed.S. enteritidis strains were grown at 28°C with aeration for 12to 18 h in LB or in LB containing the appropriate antibiotic.The bacteria (ca. 2.0 x 107) were added to the indicatedmonolayer. The microtiter plates were then incubated at37°C in an atmosphere containing 5% CO2. After 3 h, thetissue culture medium was removed and the cells werewashed three times with phosphate-buffered saline (PBS) toremove nonadherent bacteria. Fresh tissue culture mediumcontaining 100 ,ug of gentamicin per ml was then added, andthe plates were reincubated as described above. After 90min, the medium was removed and the cells were washedtwice with PBS to remove the gentamicin. The tissue culturecells were then lysed to release intracellular bacteria byadding 0.2 ml of 1% Triton X-100 to each well. After 5 min,0.8 ml of LB was added; the final concentration of TritonX-100 was 0.2%. For the qualitative assay, 100 ,ul of thissuspension was plated on the appropriate bacteriologicalmedium. The wild-type strain CDC5 under these conditionsgave nearly confluent growth, whereas invasion-defectivemutants gave individual colonies. For the quantitative assay,the suspension was diluted and plated on the appropriatebacteriological medium to determine viable counts. Quanti-tative invasion assay results are calculated as follows: per-cent invasion = 100 x (number of bacteria resistant togentamicin/total number of bacteria added).Attachment of bacteria was assessed by determining the

total number of cell-associated bacteria. S. enteritidis strainswere cultured at 28°C with aeration for 12 to 18 h in LB.Approximately 2 x 107 bacteria were added to the indicated

monolayer. The microtiter plates were incubated at 37°C inan atmosphere containing 5% CO2. After 3 h, the tissueculture medium was removed and the monolayers werewashed six times with PBS. The tissue culture cells werethen lysed, and viable counts were determined as describedabove.

Isolation of invasion mutants of S. enteritidis CDC5str. Thetransposon TnphoA was used to mutagenize the S. enteriti-dis chromosome (23). TnphoA is a derivative of TnS con-taining a portion of the E. coli phoA gene, which encodesalkaline phosphatase (AP), inserted into the left repeat. AP isinactive unless it is secreted. Fusions expressing AP caneasily be identified by screening colonies on agar platescontaining the chromogenic indicator 5-bromo-4-chloro-3-indolyl phosphate (XP). To generate TnphoA insertion mu-tations in S. ententidis CDC5str, E. coli SM10 Xpir (pRT733)(41) was mated with CDC5str on LB agar plates at 37°C for6 to 8 h. The mating mixture was then streaked onto LB agarplates containing kanamycin (to select for the transposon),streptomycin (to select against the donor E. coli), and XP (toscreen for insertions that result in AP activity). The plateswere incubated aerobically at 37°C. The Knr Strr bluecolonies (AP') were purified and tested individually ininvasion assays with HEp-2 cells; later mutant searchesincluded both HEp-2 and CHO cells.Chromosomal DNA isolation and Southern blot analysis.

Chromosomal DNA was purified from Salmonella spp. bythe protocol described by Mekalanos (26) except that thePronase digestion step was incubated at 42°C rather than at37°C. Restriction endonuclease digestions using EcoRV,HpaI, or SacI were performed according to the recommen-dations of the manufacturer (New England Biolabs). Di-gested chromosomal DNA was then separated by electro-phoresis through a 0.7% agarose gel and transferred tonitrocellulose as described elsewhere (39). A 2.8-kb BglIIfragment from TnphoA containing the kanamycin resistancegene was purified from agarose gels by using GeneClean(BiolOl, LaJolla, Calif.) and used as a probe after beinglabeled with [32P]dCTP by the hexamer priming method (8).Plasmid pYA2219 (14) carrying invABCDE and plasmidpSB52 carrying invFGH were obtained from J. Galan.Sau3AI-digested pYA2219 and pSB52 were labeled with[32P]dCTP by the hexamer priming method (8). Hybridiza-tions were performed under medium stringency, and thefilters were washed as previously described (30).Western blot (immunoblot) analysis. Cell extracts of S.

enteritidis CDC5str and the invasion mutants were preparedas follows. Cells from 1.5 ml of S. enteritidis strains grown ateither 28 or 37°C with aeration for 12 to 18 h in LB werecollected by centrifugation. The cells were washed twicewith 10 mM Tris (pH 8.0) and suspended in 100 ,ul of 10 mMTris (pH 8.0). The cell suspension was then boiled for 5 minand allowed to cool to ambient temperature before 2 U ofDNase (Promega Biotec) was added. The samples weredigested with DNase at 37°C for 2 h and then stored at-20°C. The protein concentrations of the samples weredetermined by using the Bradford method (BioRad). Foreach sample, 45 ,ug of protein was mixed with sodiumdodecyl sulfate (SDS) sample buffer containing dithiothreitoland boiled for 5 min, and then the proteins were separated byelectrophoresis through a 12.5% SDS-polyacrylamide gel(22). The proteins were then transferred to nitrocellulose andincubated with rabbit polyclonal antibody to AP (gift of R.Taylor) or with a monoclonal antibody to lipoprotein (gift ofJ. Bliska) essentially as described previously (4). Afterincubation with the primary antibody, immunoblots were

J. BACTERIOL.

on October 29, 2017 by guest

http://jb.asm.org/

Dow

nloaded from

ENTRY OF S. ENTERITIDIS INTO EUKARYOTIC CELLS 3947

incubated with a second antibody (Bethesda Research Lab-oratories) conjugated to AP. Antibody binding was visual-ized by incubating the filter with XP and p-nitroblue tetra-zolium chloride as described previously (3). In some

experiments, the cells were separated into membrane andsoluble fractions, before separation on an acrylamide gel, bya modification of the Schnaitman method (37) as describedby Pepe and Miller (32); when alterations in membraneprotein profile were screened for, the proteins were visual-ized by staining with Coomassie blue R-250.LPS analysis. Lipopolysaccharide (LPS) was purified by

the method of Darveau and Portnoy as described by Mar-tinez (24) from S. enteritidis CDC5str and invasion-defectivemutants grown aerobically for 12 to 18 h in LB at 37°C.Mapping invasion-defective loci. To determine the chromo-

somal location of the TnphoA insertions in invasion-defec-tive mutants, the chromosome-TnphoA junction was clonedfrom each mutant and used as a probe against a set ofmapping phages carrying defined portions of the S. typhimu-rium chromosome. To clone the chromosome-TnphoA junc-tion, chromosomal DNA was purified from each mutant anddigested with either BamHI or Sall, as these restrictionendonucleases each have a single site within TnphoA just 3'of the Knr gene. The chromosomal DNA fragments were

then ligated either to pUC19 (Ampr) or pACYC184 (Camr)digested with the same enzyme. Recombinant plasmidscarrying the chromosome-TnphoA junction were identifiedby transforming E. coli DH5cx and selecting for the resis-tance marker of the vector and for kanamycin resistance.Because probes derived from TnphoA alone do not hybridizeto Salmonella chromosomal DNA, these chromosome-TnphoA clones were used directly to probe the set ofmapping phages.Youderian et al. (43) constructed a set of mapped,

locked-in Mud-P22 prophages along the S. typhiinuriuinchromosome. Each of these prophages is capable of pack-aging several minutes of chromosomal DNA adjacent to thesite of prophage insertion after induction of the prophageswith mitomycin C. DNA was purified from the phage parti-cles, and dot blot filters were prepared with 0.5 ,ug of DNAfrom each phage. The filters were hybridized under mediumstringency (30) with probes derived from chromosome-TnphoA junctions of each of the invasion-defective mutants.

RESULTS

Isolation of invasion-defective mutants of S. enteritidis. Weand others have been unable to directly clone and express innoninvasive E. coli genes encoding the invasion phenotypefrom Salmonella strains responsible for gastroenteritis (7,13). The reasons for this are unclear, but the data suggestthat unlike the situation in Y. pseudotuberculosis (17) and Y.enterocolitica (29), at least two chromosomal loci acting inconcert are required to convert a noninvasive E. coli to an

invasive organism. Therefore, the genetic basis for epithelialcell invasion by S. enteritidis was studied by identifyingmutations in S. entenitidis that affect this phenotype. Thenumber of insertion mutants to be screened was narrowed bymaking the assumption that at least some components of theinvasion apparatus would be secreted proteins. To this end,random TnphoA insertion mutants of S. entenitidis were

screened for changes in the invasion phenotype in the tissueculture invasion (TCI) assay. The transposon insertions intothe chromosome of S. enteritidis CDC5str were generated byconjugation of a suicide plasmid carrying TnphoA. Insertionmutants expressing AP activity were examined in the TCI

TABLE 1. Invasion and attachment phenotypes of invasion-defective mutants"

%c, Wild-type activity

Mutant Class HEp-2 CHO MDCK

TCI TCA TCI TCA TCI TCA

SMi I 1.6 59 24.6 70 21.0 102SM2 I 1.5 70 18.3 55 17.3 97SM8 I 2.6 94 24.2 104 26.8 102

SM3 II 8.1 69 2.4 52 2.5 81SM4 II 7.3 87 2.5 59 2.3 88SM5 II 10.5 94 3.1 62 2.6 98

SM6 III 0.2 35 0.05 19 0.07 20

SM7 IV 1.7 50 0.7 40 0.2 40

SM9 V 0.5 89 25.1 112 0.1 100

SM1o VI 40.6 42 44.7 48 15.6 27SMi1 VI 46.1 57 39.8 44 60.6 109SM12 VI 26.6 16 20.5 11 4.2 19SM13 VI 14.9 46 7.0 108 10.5 470

' The amounts of TCI and total cell-associated (TCA) bacteria weredetermined as described in Materials and Methods. The results are expressedas a percentage of wild-type level and represent the average of at least threeseparate assays, each performed in duplicate.

assay with HEp-2 and CHO cells. In this way, from approx-imately 300 AP' insertion mutants screened, 13 invasion-defective mutants were identified.

Invasion phenotype of invasion-defective mutants. Thirteenputative invasion mutants identified by the qualitative TCIassay screen were subsequently examined quantitatively fortheir TCI phenotypes in HEp-2, CHO, and MDCK cells(Table 1). As a result of these analyses, the 13 mutants fellinto six phenotypic classes. All the mutants except SM3 andSM5 were derived from independent matings. Class I mu-tants (SM1, SM2, and SM8) had a severe defect for invasionof HEp-2 cells but only a slight to moderate defect forinvasion of MDCK or CHO cells. Class II mutants (SM3,SM4, and SM5) had the phenotype opposite that of class Imutants; they were moderately to severely affected in theirinvasion of CHO cells and MDCK cells but had only amoderate defect for HEp-2 cells. The class III mutant (SM6)was of particular interest because it was noninvasive in allcell types tested. The class IV mutant (SM7), like SM6, wasalso severely affected for entry into all cell types but wasabout 10-fold more invasive than SM6. The class V mutant(SM9) had a severe defect for entry into HEp-2 and MDCKcells but only a slight to moderate defect for entry into CHOcells. Class VI mutants (SM10 to SM13) had a slight butreproducible defect in invasion relative to wild type regard-less of which cell line was used for the assay.Attachment phenotype of invasion-defective mutants. To

determine whether the observed defect in invasion was dueto an inability to adhere to the monolayers, the invasion-defective mutants were assayed for attachment to tissueculture cells by measuring total cell-associated bacteria(Table 1); this value included both attached and intracellularbacteria. Interestingly, the LPS mutant, SM13 (see below),is considerably more adherent than wild type for MDCKcells yet has only 10% of the wild-type invasion capability.This confirms previous observations (29) that adherence is

VOL. -174, 1992

on October 29, 2017 by guest

http://jb.asm.org/

Dow

nloaded from

3948 STONE ET AL.

II) 0 T- N4

N0 CO) 'o LO CD I- e(07*-U) 0e o ca X 2 E 2 E X 2 a X EQ en en en) en en CO) en en u) en u) enI; Swf

4 am = ~ sv ss* s. _.

CO,

U)

97 KO-

66 K -

45 K

31 KO

22 K o

FIG. 1. LPS analysis. LPS was isolated from S. enteritidis CDC5and the 13 invasion mutants as described elsewhere (24). Thesamples were separated by electrophoresis through a 15% SDS-polyacrylamide gel and visualized by silver staining.

not sufficicnt for entry to occur. In general, the number ofcell-associated bacteria was slightly less for the mutants thanfor the wild-type strain and was probably due in part to fewerintracellular bacteria. In some cases, notably with SM9,attachment was essentially at wild-type levels for all threetissue culture cell lines. The mutants in classes I to V,therefore, appear to have more-pronounced defects in themechanism(s) of cellular invasion than in cellular attach-ment. In most cases, the class VI mutants had defects inattachment similar in magnitude to their defects in invasion.Together, these data indicate that with the exception of theclass VI mutants, it is unlikely that the alteration in totalcell-associated bacteria could alone account for the changein the invasion phenotype.

Characterization of invasion-defective mutants. To deter-mine whether the invasion-defective phenotype of the mu-tants was a result of secondary effects, the following prop-erties were investigated: growth rate, motility, and cellsurface composition. All 13 invasion-defective mutants had agrowth rate in LB medium similar to that of the wild-typeparental strain CDC5str (data not shown). In addition, theinvasion-defective mutants were as motile as the parentstrain in soft-agar plates (0.4% agar) (data not shown). LPSwas purified from CDC5str and the mutants, separated on a15% SDS-polyacrylamide gel, and visualized by stainingwith silver (Fig. 1). Twelve of the 13 mutants had an LPSprofile similar to that of CDC5str both in the amountsynthesized and in the extent of 0-side chain polymeriza-tion. Only mutant strain SM13 failed to synthesize 0 sidechains, and only a minor invasion defect was observed forthis mutant. It therefore appears that LPS is not a majorcontributor to the invasion phenotype of S. enteritidis.Comparison of the outer membrane protein profiles of

invasion mutants with that of wild type showed the absenceof certain proteins in some of the mutants (Fig. 2). All theclass II mutants (SM3, SM4, and SM5) were missing aprotein of -18 kDa. The class III mutant, SM6, was missinga 52-kDa protein in addition to this 18-kDa protein. Loss oflipoprotein as identified by Western analysis with an anti-body to lipoprotein (data not shown) was evident only inclass VI mutant SM1O. For these mutants, it is likely that theTnphoA insertion is in either the gene or the operon encod-

1 4 K -

FIG. 2. Outer membrane proteins of S. enteritidis CDC5 and theinvasion mutants. Overnight cultures of S. enteritidis CDC5 and the13 invasion mutants were separated into Triton X-100-soluble andinsoluble fractions by the cell fractionation method of Schnaitman(37) as described by Pepe and Miller (32). The Triton X-100-insoluble fraction (containing primarily outer membrane proteins)was separated on a 12.5% SDS-polyacrylamide gel and visualizedby staining with Coomassie blue. Shown on the left are the positionsof the protein size standards in kilodaltons. Proteins that appear tobe missing in the mutants are indicated by a dot.

ing the missing protein(s) or is in a gene that affects synthesisor localization of the missing protein(s).

Identification of the AP fusion proteins. Western analysisusing antibody directed against AP was performed to con-firm that AP fusion proteins were being synthesized and todistinguish mutants within a class if possible (Fig. 3). This

o0 N_ _APO ^

AP C) U) U) U U U U) U) U)

110K O-

84 K 10

47CK S3

33K

24K *_

FIG. 3. Western blot analysis of S. enteritidis CDC5 and theinvasion mutants. Whole-cell extracts were prepared from S. enter-itidis CDC5 and the 13 invasion mutants. For each sample, 45 ,ug ofprotein was separated by electrophoresis through a 12.5% SDS-polyacrylamide gel. The proteins were transferred to nitrocelluloseand incubated with rabbit polyclonal antibody to AP. The filter wasthen incubated with goat anti-rabbit antibody conjugated to AP.Antibody binding was visualized by incubation with XP and p-ni-troblue tetrazolium. Shown on the left are protein size standards inkilodaltons.

_X u X O ° N acm

0C.) co

J. BACTERIOL.

..W.. M", - I'll Z.-

A& ." 1. Ai .. .. 46 iii I" i"

on October 29, 2017 by guest

http://jb.asm.org/

Dow

nloaded from

ENTRY OF S. ENTERITIDIS INTO EUKARYOTIC CELLS 3949

aumX

X

n am en e0 0 en U) (a

23.1 -

9.4

6.6 -

4.4 -

23 -

2.0 -

FIG. 4. Southern hybridization analysis of S. enteritidis CDC5and the invasion mutants. Chromosomal DNA was purified from S.enteritidis CDC5 and the invasion mutants as described in Materialsand Methods and then digested with restriction endonucleaseEcoRV. DNA fragments were separated by electrophoresis in a

0.7% agarose gel, transferred to nitrocellulose, and probed with a

fragment derived from TnphoA (see Materials and Methods fordetails). Shown on the left are the positions of the fragments derivedfrom bacteriophage DNA digested with HindIII in kilobases.

polyclonal serum had weak cross-reactivity to parentalstrain CDC5str. Fusion proteins ranging in size from >110 to49 kDa could be identified in all mutants except SM13. Theamount of AP produced by SM13 may be below the level ofdetection with this antibody; consistent with this is theobservation that this mutant is very pale blue on platescontaining XP. In addition to a larger product, most mutantshad a protein similar in size to AP that was recognized by theantibody; it is quite common to see AP size breakdownproducts from AP fusion proteins (23). Expression of theinvasion genes from Yersinia spp. is regulated by tempera-ture (20, 33, 42), so the effect of growth temperature on

expression of the fusion proteins was tested. The same

fusion products were produced at similar levels regardless ofwhether the cells were grown at 37 or 28°C (data not shown).Class I mutants, SM1, SM2, and SM8, had the same invasionphenotype, and the locations of the TnphoA insertion inthese three strains could not be distinguished by Southernanalysis with the TnphoA probe (see below), yet the hybridprotein produced by SM1 was a different size from thoseobserved from SM2 and SM8. These data together suggestthat these mutants may have insertions in neighboring genes

or at different places within the same gene, with the insertionin SM1 being promoter proximal.

Southern hybridization analysis and mapping of invasion-defective mutants. To determine the number of TnphoAinsertions in each mutant and to obtain an estimate of thenumber of chromosomal loci contributing to the invasionphenotype, Southern hybridization analysis of the mutantswas performed by using a probe containing sequences inter-nal to TnphoA (Fig. 4). Parental strain CDC5str had no

sequences recognized by this probe, indicating that thehybridization observed in the mutants was due to TnphoAsequences only. This probe was used to carry out Southernanalysis of chromosomal DNA isolated from the mutantsdigested with three different restriction enzymes (Fig. 4; datanot shown). Two of the restriction endonucleases, EcoRVand SacI, do not cut within TnphoA, while HpaI cuts 186 bpupstream of the 3' end of TnphoA. Mutants SM3 and SM6appeared to have two insertions of TnphoA. The insertionsin SM6 were subsequently separated by phage P22 transduc-tion; the invasion phenotype was associated with only one of

the two TnphoA insertions. The class I mutants (SMI, SM2,and SM8) appeared to have TnphoA inserted on the samefragment regardless of which restriction enzyme was used,suggesting that the insertions in these mutants are in thesame locus. Likewise, the three class II mutants (SM3, SM4,and SM5) also had TnphoA inserted in the same fragment,but this fragment was distinct from that observed for theclass I mutants. Southern hybridization analysis of all theother TnphoA insertion mutants revealed unique combina-tions of fragments when results from the three restrictionendonucleases were compared.To more clearly define the number of invasion loci iden-

tified by the TnphoA invasion-defective mutants, the ap-proximate chromosomal location of TnphoA was determinedfor each mutant. The chromosome-TnphoA junction wascloned from SM1, SM2, SM5 to SM9, and SM11 to SM13.These clones were then used as probes against a set ofmapping phages constructed and described by Youderian etal. (43) (Table 2). The class I mutants mapped to the 57- to60-min region of the chromosome. A previously character-ized invasion locus, inv (13), also maps to this region.Therefore, Southern hybridization analysis was performedusing probes derived from the inv locus (data not shown).Only class I mutants (SM1, SM2, and SM8) appeared to haveTnphoA insertions in this area. The nucleotide sequences ofthe chromosome-TnphoA junctions of SM1, SM2, and SM8were determined (data not shown). Comparison of thesesequences with the inv locus indicated that these threemutants have TnphoA insertions into invH (1). The class IImutants, represented by SM5, mapped to the 25- to 28.5-minregion of the chromosome. Each of the remaining six mu-tants mapped to different regions of the chromosome. Al-though we were unable to map the insertion in SM10, webelieve, on the basis of its invasion phenotype, its loss oflipoprotein, and Southern hybridization analysis, that theTnphoA insertion in this strain is distinct from the othermutations. Thus, it appears that S. enteritidis has at leastnine chromosomal loci that influence the invasion pheno-type. Some of these loci may actually have several genescontributing to invasiveness, as has been demonstrated forthe invA locus (13).

DISCUSSION

Both in vivo (40) and in vitro (15, 21, 38) studies indicatethat Salmonella spp. are capable of invading and passingthrough eukaryotic cells (10). Recently, genetic and molec-ular analyses of this process were initiated. Finlay et al. (11)isolated several mutants of S. choleraesuis defective intranscytosis of polarized MDCK cells. All of these mutantswere down 2- to 50-fold for invasion of MDCK cells com-pared with wild type. These mutants also had a defect inadherence that roughly paralleled the invasion defect ofindividual mutants. Almost half of these mutants had alter-ations in their LPS. Why we obtained so few LPS mutantscompared with Finlay et al. (11) is not clear. This couldreflect differences between S. enteritidis and S. choleraesuisor differences in the initial screening protocol (i.e., we usedthe invasion phenotype, whereas Finlay et al. [11] used thetranscytosis phenotype).Using an alternate approach, Elsinghorst et al. (7) and

Galan and Curtiss (13) directly cloned invasion loci byscreening for clones that conferred the invasion phenotypeon E. coli or complemented a S. typhimurium invasionmutant, respectively. The clone obtained by Elsinghorst etal. (7) from S. typhi allowed E. coli HB101 to invade Henle

u) 0 01

o zU)l- U)lcz a a a a

0 sn oo (A en

VOL. 174, 1992

on October 29, 2017 by guest

http://jb.asm.org/

Dow

nloaded from

3950 STONE ET AL.

TABLE 2. Chromosomal locations of invasion mutantSa

MutantClass Strong hybridization Moderate hybridization Map position Locus or additionalMutant Class Strong hybridization Moderate hybridization (m hntp(min) phenotype

SMi I cysHIJt574::MudP proU1888::MudP 57.0-60.0 invHnadB226::MudPnadB226::MudQpurG2I49::MudP

SM2 I cvsHIJ1574::MudP proUI888::MudP 57.0-60.0 invHnadB226::MudPnadB226::MudQpurG2149::MudP

SM8 I cysHIJ574::MudP proU1888::MudP 57.0-60.0 invHnadB226::MudPnadB226::MudQpurG2149::MudP

SM3 II ND ND ND pagCSM4 II ND ND ND pagCSM5 II purBl879::MudQ aroD561::MudQ 25.0-28.5 pagC

putA1019: :MudQSM6 III cysHIJ1574::MudQ cysG1573::MudQ 60.0-72.8SM7 IV purE2155::MudQ NMH 12.0

purE2154: :MudPSM9 V pnuE421::MudQ metE2131::MudQ 84.0-86.7SM1O VI ND ND ND LipoproteinSMi1 VI melAB::MudQ purD1874::MudQ 89.2-93.0SM12 VI nadC218::MudQ nadC220::MudP 3.5SM13 VI pyrF2690::MudQ tre-152::MudQ 33.0-35.9 LPS

zea-3666::MudPhisD9950::MudP

"Chromosomal locations of invasion mutants were determined from hybridization of excision-defective phage DNA (Materials and Methods; 43) to probesmade from the TnphoA junction of each mutant. ND, not determined; NMH, no moderate hybridization observed for this mutant.

407 cells at 4 to 7% of the efficiency of the parental S. typhistrain. This result suggested either that the invasion genes(or their products) were not fully active in E. coli or that S.typhi has additional invasion factors. Four regions of thisclone were required for the invasion phenotype. Interest-ingly, the corresponding region from S. typhimurium clonedinto E. coli did not promote invasion of Henle 407 cells.Galan and Curtiss (13) identified a recombinant clone thatcomplemented the invasion defect of a S. typhimuriumlaboratory strain. However, this clone did not confer theinvasion phenotype on E. coli. Insertions of TnphoA intothree genes carried on this clone, invA, invB, and invD,resulted in failure to fully restore the invasion phenotype ofthe S. typhimurium mutant. invA and invB form an operonwith invC; invD, while carried on the same cosmid, is notpart of the invABC operon. This invasion locus was recentlyshown by Southern hybridization analysis to be present inother Salmonella spp., including S. enteritidis (14), and tocontain additional genes, invEFGH (12). An invA mutationwas recombined onto the chromosome of a virulent S.typhimuriun strain, and the dose resulting in death of 50% ofmice infected orally with the invA mutant was found to be50-fold higher than that of the wild-type strain (13). How-ever, the invA mutant was still able to infect and kill mice.This result also suggested that additional entry mechanismsmay exist in Salmonella spp.The phenotypes of the 13 invasion-defective mutants of S.

enteritidis CDC5str described here provided additional evi-dence that more than one entry pathway exists. By screeningour mutants in several cell lines (HEp-2, CHO, and MDCK),mutants defective for entry of some cell lines but not otherswere identified. For example, the class I mutants SM1, SM2,and SM8 were 50- to 100-fold less invasive than wild type forHEp-2 cells but were nearly as invasive as wild type for

MDCK cells. In contrast, the class V mutant SM9 was morethan 200-fold less invasive than wild type for both HEp-2 andMDCK cells but retained 25% of wild-type activity for CHOcells. The class III and IV mutants SM6 and SM7 areparticularly interesting because they were noninvasive for allthree cell types and thus may be affected in steps common toall entry pathways or may have mutations in regulatorygenes.

Southern hybridization analysis and chromosomal map-ping showed that at least nine chromosomal loci affect theinvasion phenotype of S. enteritidis; their chromosomal maplocations and additional phenotypes are summarized inTable 2. The class I mutants (SM1, SM2, and SM8) mappedto 57.0 to 60.0 min on the S. typhimurium chromosome. Theinv locus described by Galan and Curtiss also maps to thisregion (12, 13). Comparison of the chromosomal sequenceupstream of the TnphoA insertions in SMI, SM2, and SM8with the sequence of the inv locus indicated that all threeclass I mutants have TnphoA insertions into invH (1). Theclass II mutant SM5 mapped to 25 to 28.5 min, a region thatcontainspagC, a gene previously implicated in macrophagesurvival (28). These mutants are missing an outer membraneprotein of 18 kDa which is the size of PagC, suggesting thatclass II mutants have a TnphoA insertion in pagC (35). Theclass VI mutants all have TnphoA insertions at sites distinctfrom each other and from the sites in other classes ofinvasion mutants. In general, these mutants had a relativelyminor invasion defect which was paralleled by a similardefect in attachment. While it is possible that these muta-tions directly affect invasion, an alternative explanation isthat the invasion defect observed in this class of mutants isan indirect effect of the mutation. Consistent with thisinterpretation is the observation that two of these mutants,SM10 and SM13, were missing components of the cell

J. BACTERIOL.

on October 29, 2017 by guest

http://jb.asm.org/

Dow

nloaded from

ENTRY OF S. ENTERITIDIS INTO EUKARYOTIC CELLS 3951

surface and outer membrane (lipoprotein and LPS, respec-tively) that are likely to perturb the interaction and presen-tation of other outer membrane and cell surface components.The remaining three mutants (SM6, SM7, and SM9) were

greatly reduced in their capacities to enter eukaryotic cellsand thus probably have mutations in genes whose productsdirectly influence this phenotype. The TnphoA insertions inthese mutants map to different locations on the chromosomeand do not fall into the previously identified inv locus ofSalmonella spp. We have designated these three new inva-sion loci sinA (Salmonella invasion), sinB, and sinC for themutations present in SM6, SM7, and SM9, respectively. Notall of the loci identified here necessarily encode productsthat directly interact with the eukaryotic cell, as has beendemonstrated for invasin of Y. pseudotuberculosis (18, 19).Rather, the products of these genes may play a role insecretion, localization, or assembly of an "invasion factor"or may possibly play a role in regulating expression of aninvasion factor. The mutant SM6 is a good candidate for aregulatory mutant because several proteins are missing fromits membrane protein profile.

It is becoming clear that the invasion phenotype of Sal-monella spp. is multifactorial, involving several unlinkedchromosomal genes. This is in striking contrast to theinvasion phenotype of another invasive member of theEnterobacteriaceae, Yersinia spp. (17, 29). The early evi-dence appears to suggest that the invasive capacities ofSalmonella and Yersinia spp. evolved independently. Al-though Yersinia spp. also appear to have several entrymechanisms, single genes (either invpstb, invent, or ail) canconfer an invasive phenotype on E. coli. Homologs of invand ail of Yersinia spp. have not been identified in Salmo-nella spp., and no single gene or locus conferring an invasivephenotype on E. coli has been cloned from a Salmonellastrain that causes gastroenteritis. The reasons for this are notclear, but they could include problems with expression of thecloned invasion gene in E. coli or a need for the products ofunlinked genes to work together. Cloning and further anal-ysis of the invasion genes and their products should shedlight on the molecular basis of entry into eukaryotic cells bySalmonella spp.

ACKNOWLEDGMENTS

We thank C. Collins, J. Pepe, S. Kinder, S. Valone, and M.Wachtel for critically reading the manuscript. We also thank J.Galan for providing probes from the inv locus and checking forhomology to the inv locus and S. Libby and F. Heffron for sendingus the set of mapping phage.

B. J. Stone is a recipient of Public Health Service NationalResearch Service Award GM-07104. V. L. Miller is a Pew Scholar inthe Biomedical Sciences.

REFERENCES

1. Altmeyer, R. M., J. K. McNern, J. C. Bossio, I. Rosenchine,B. B. Finlay, and J. E. Galan. Submitted for publication.

2. Bachmann, B. J. 1972. Pedigrees of some mutant strains ofEscherichia coli K-12. Bacteriol. Rev. 36:525-557.

3. Blake, M. S., K. H. Johnson, G. J. Russell-Jones, and E. C.Gotschlich. 1984. A rapid, sensitive method for detection ofalkaline phosphatase-conjugated anti-antibody on Westernblots. Anal. Biochem. 136:175-179.

4. Burnette, W. N. 1981. "Western blotting": electrophoretictransfer of proteins from sodium dodecyl sulfate-polyacryl-amide gels to unmodified nitrocellulose and radiographic detec-tion with antibody and radioiodinated protein A. Anal. Bio-chem. 112:195-203.

5. Cohen, M. L., and R. V. Tauxe. 1986. Drug-resistant Salmonella

in the United States: an epidemiologic perspective. Science234:964-969.

6. Davis, R. W., D. Botstein, and J. R. Roth. 1980. Advancedbacterial genetics, p. 78, 87. Cold Spring Harbor Laboratory,Cold Spring Harbor, N.Y.

7. Elsinghorst, E. A., L. S. Baron, and D. J. Kopecko. 1989.Penetration of human intestinal cells by Salmonella: molecularcloning and expression of Salmonella typhi invasion determi-nants in Escherichia coli. Proc. Natl. Acad. Sci. USA 86:5173-5177.

8. Feinberg, A. P., and B. Vogelstein. 1983. A technique forradiolabeling DNA restriction endonuclease fragments to highspecific activity. Anal. Biochem. 132:6-13.

9. Finlay, B. B., and S. Falkow. 1987. A comparison of microbialinvasion strategies of Salmonella, Shigella, and Yersinia spe-cies. UCLA Symp. Mol. Cell. Biol. 64:227-243.

10. Finlay, B. B., B. Gumbiner, and S. Falkow. 1988. Penetration ofSalmonella through a polarized MDCK epithelial cell mono-layer. J. Cell Biol. 107:221-230.

11. Finlay, B. B., M. N. Starnbach, C. L. Francis, B. A. Stocker, S.Chatfield, G. Dougan, and S. Falkow. 1988. Identification andcharacterization of TnphoA mutants of Salmonella that areunable to pass through a polarized MDCK epithelial cell mono-layer. Mol. Microbiol. 2:757-763.

12. Galan, J. E. Personal communication.13. Galan, J. E., and R. Curtiss III. 1989. Cloning and molecular

characterization of genes whose products allow Salmonellatyphimunium to penetrate tissue culture cells. Proc. Natl. Acad.Sci. USA 86:6383-6387.

14. Galan, J. E., and R. Curtiss III. 1991. Distribution of the invA,-B, -C, and -D genes of Salmonella typhimurium among otherSalmonella serovars: invA mutants of Salmonella typhi aredeficient for entry into mammalian cells. Infect. Immun. 59:2901-2908.

15. Giannella, R. A., 0. Washington, P. Gemski, and S. B. Formal.1973. Invasion of HeLa cells by Salmonella typhimurium: amodel for study of invasiveness in Salmonella. J. Infect. Dis.128:69-75.

16. Gulig, P. A., and R. Curtiss III. 1987. Plasmid-associatedvirulence of Salmonella typhimurium. Infect. Immun. 55:2891-2901.

17. Isberg, R. R., and S. Falkow. 1985. A single genetic locusencoded by Yersinia pseudotuberculosis permits invasion ofcultured animal cells by Escherichia coli K-12. Nature (London)317:262-264.

18. Isberg, R. R., and J. M. Leong. 1988. Cultured mammalian cellsattach to the invasin proteins of Yersinia pseudotuberculosis.Proc. Natl. Acad. Sci. USA 85:6682-6686.

19. Isberg, R. R., and J. M. Leong. 1990. Multiple 1 chain integrinsare receptors for invasin, a protein that promotes bacterialpenetration into mammalian cells. Cell 60:861-871.

20. Isberg, R. R., A. Swain, and S. Falkow. 1988. Analysis ofexpression and thermoregulation of the Yersinia pseudotuber-culosis inv gene with hybrid proteins. Infect. Immun. 56:2133-2138.

21. Kihlstrom, E. 1977. Infection of HeLa cells with Salmonellatyphimurium 395 MS and MR10 bacteria. Infect. Immun. 17:290-295.

22. Laemmli, U. K. 1970. Cleavage of structural proteins during theassembly of the head of bacteriophage T4. Nature (London)227:680-685.

23. Manoil, C., and J. Beckwith. 1985. TnphoA: a transposon probefor protein export signals. Proc. Natl. Acad. Sci. USA 82:8129-8133.

24. Martinez, R. J. 1989. Thermoregulation-dependent expressionof Yersinia enterocolitica protein I imparts serum resistance toEscherichia coli K-12. J. Bacteriol. 171:3732-3739.

25. Maurelli, A. T., B. Baudry, H. d'Hauteville, T. L. Hale, and P. J.Sansonetti. 1985. Cloning of plasmid DNA sequences involvedin invasion of HeLa cells by Shigella flexneri. Infect. Immun.49:164-171.

26. Mekalanos, J. J. 1983. Duplication and amplification of toxingenes in Vibrio cholerae. Cell 35:253-263.

VOL. 174, 1992

on October 29, 2017 by guest

http://jb.asm.org/

Dow

nloaded from

3952 STONE ET AL.

27. Miller, J. H. 1972. Experiments in molecular genetics. ColdSpring Harbor Laboratory, Cold Spring Harbor, N.Y.

28. Miller, S. I., A. M. Kukral, and J. J. Mekalanos. 1989. Atwo-component regulatory system (phoPphoQ) controls Salmo-nella typhimunum virulence. Proc. Natl. Acad. Sci. USA 86:5054-5058.

29. Miller, V. L., and S. Falkow. 1988. Evidence for two genetic lociin Yersinia enterocolitica that can promote invasion of epithelialcells. Infect. Immun. 56:1242-1248.

30. Miller, V. L., J. J. Farmer III, W. E. Hill, and S. Falkow. 1989.The ail locus is found uniquely in Yersinia enterocolitica sero-types commonly associated with disease. Infect. Immun. 57:121-131.

31. Miller, V. L., and J. J. Mekalanos. 1988. A novel suicide vectorand its use in construction of insertion mutations: osmoregula-tion of outer membrane proteins and virulence determinants inVibrio cholerae requires toxR. J. Bacteriol. 170:2575-2583.

32. Pepe, J. C., and V. L. Miller. 1990. The Yersinia enterocoliticainv gene product is an outer membrane protein that sharesepitopes with Yersinia pseudotuberculosis invasin. J. Bacteriol.172:3780-3789.

33. Pierson, D. E., and S. Falkow. 1990. Nonpathogenic isolates ofYersinia enterocolitica do not contain functional inv-homolo-gous sequences. Infect. Immun. 58:1059-1064.

34. Portnoy, D. A., S. L. Moseley, and S. Falkow. 1981. Character-ization of plasmids and plasmid-associated determinants ofYersinia enterocolitica pathogenesis. Infect. Immun. 31:775-782.

35. Pulkkinen, W. S., and S. I. Miller. 1991. A Salmonella typhi-

murium virulence protein is similar to a Yersinia enterocoliticainvasion protein and a bacteriophage lambda outer membraneprotein. J. Bacteriol. 173:86-93.

36. Rubin, R. H., and L. Weinstein. 1977. Salmonellosis: microbio-logic, pathologic, and clinical features. Intercontinental MedicalBook Corp., New York.

37. Schnaitman, C. A. 1971. Solubilization of the cytoplasmicmembrane of Eschenchia coli by Triton X-100. J. Bacteriol.108:545-552.

38. Small, P. L. C., R. R. Isberg, and S. Falkow. 1987. Comparisonof the ability of enteroinvasive Escherichia coli, Salmonellatyphimurium, Yersinia pseudotuberculosis, and Yersinia entero-colitica to enter and replicate within HEp-2 cells. Infect. Im-mun. 55:1674-1679.

39. Southern, E. M. 1975. Detection of specific sequences amongDNA fragments separated by gel electrophoresis. J. Mol. Biol.98:503-517.

40. Takeuchi, A. 1967. Electron microscope studies of experimentalSalmonella infection. I. Penetration into the intestinal epithe-lium by Salmonella typhimunum. Am. J. Pathol. 50:109-136.

41. Taylor, R. K., C. Manoil, and J. J. Mekalanos. 1989. Broad-host-range vectors for delivery of TnphoA: use in geneticanalysis of secreted virulence determinants of Vibrio cholerae.J. Bacteriol. 171:1870-1878.

42. Wachtel, M. R., and V. L. Miller. Unpublished results.43. Youderian, P., P. Sugiono, K. L. Brewer, N. P. Higgins, and T.

Elliot. 1988. Packaging specific segments of the Salmonellachromosome with locked-in Mud-P22 prophages. Genetics 118:581-592.

J. BACTERIOL.

on October 29, 2017 by guest

http://jb.asm.org/

Dow

nloaded from