JOURNAL OF BACTERIOLOGY, Apr. 1987, p. 1509-15150021-9193/87/041509-07$02.00/0

Vol. 169, No. 4

The Secreted Hemolysins of Proteus mirabilis, Proteus vulgaris, andMorganella morganii Are Genetically Related to Each Other and to

the Alpha-Hemolysin of Escherichia coliVASSILIS KORONAKIS,' MICHAEL CROSS,' BERNARD SENIOR,2 EVA KORONAKIS,' AND COLIN HUGHES'*

Department of Pathology, University of Cambridge, Cambridge CB2 JQP,1 and Department of Microbiology,University ofDundee, Dundee DDi 9Sy,2 United Kingdom

Received 17 September 1986/Accepted 9 January 1987

Secreted hemolysins were extremely common among clinical isolates of Proteus mirabilis, Proteus vulgaris,and Morganella morganii, and hemolytic activity was either cell associated or cell free. Southern hybridizationof total DNA from hemolytic isolates to cloned regions of the Escherichia coli alpha-hemolysin (hly) determinantshowed clear but incomplete homology between genes encoding production of hemolysins in the four species.One of the two E. coli secretion genes, hlyD, hybridized only with DNA from P. vulgaris and M. morganii,which produced cell-free hemolysis, but not with that from P. mirabilis, which showed only cell-associatedactivity. Molecular cloning of the genetic determinants of cell-free hemolytic activity from P. vulgaris and M.morganii chromosomal DNA allowed their functional analysis via inactivation with the transposons TnlOOO andTnS. Both hemolysin determinants were about 7.5 kilobase pairs and comprised contiguous regions directingregulation, synthesis, and specific secretion out of the cell. Transposon mutations which eliminated secretion ofthe Proteus and Morganella hemolysins could be complemented specifically by the E. coli hemolysin secretiongenes hlyB or hlyD. Alignment of the physically and functionally defined hly determinants from P. vulgaris andM. morganii with that of the E. coli alpha-hemolysin confirmed a close genetic relationship but also indicatedextensive evolutionary divergence.

Production of cytotoxic hemolysins is common amongboth gram-negative and -positive pathogenic bacteria, but itsmolecular basis varies greatly (13). In Escherichia coli,secretion of the active hemolysin protein demands transloca-tion across both cytoplasmic and outer membranes, and it isnow evident that the process involved does not, in contrastfor example to the transport of enterotoxins (25, 27), makeuse of a conventional N-terminal signal sequence on thesecreted protein (5) nor does it proceed via a cell lysisanalogous to that which effects release of colicin E (16).

This novel secretion is absolutely dependent on twoproteins, HlyB and HlyD, which are encoded by the four-gene hemolysin (hly) determinant itself (2, 12, 21, 22, 29),and it seems possible that specific export is achieved by thetwo secretory proteins spanning both membranes to promoterecognition and direct translocations of the 107-kilodaltonhemolysin molecule (9, 22). It nevertheless remains unclearhow this might occur, and although the C terminus of thehemolysin protein (HlyA) is now known to be essential (9),little is known of the sequences intimately involved inrecognition and interaction during secretion or indeed incoordinating synthesis and secretion. The secretion proteinHlyB has at least one potential ATP-binding domain analo-gous to those present in a wide range of kinases (3, 7) andbacterial transport proteins (15), provoking the view that thisfeature is central to the active secretion of hemolysin.

Both the organization (4, 18, 20, 21, 24, 28, 30) andnucleotide sequence (6, 14) of the four contiguous hemolysingenes (hlyC, hlyA, hlyB, and hlyD) are rigidly conserved onboth plasmids and chromosomes throughout E. coli strains.The genes nevertheless have a G+C content of about 39%rather than 50%, the percentage typical of the E. coli

* Corresponding author.

genome, and this suggests that the hly genes did not originatein this species. We report here on their relationship to thegenetic determinants of hemolysin production in Proteusmirabilis and Proteus vulgaris, both of which have 39%G+C genomic DNA, and also in Morganella (formerlyProteus) morganii, which has a genomic G+C content of50%, analogous to that of E. coli.

MATERIALS AND METHODSBacteria. Clinical isolates of P. mirabilis, P. vulgaris, M.

morganii, and E. coli were obtained from the United King-dom (Ninewells Hospital, Dundee, and Charing Cross Hos-pital, London), Hungary (University Hospital, Budapest),and the Federal Republic of Germany (University Hospital,Wurzburg). Representative isolates cited in the text arelisted in Table 1. Typing of the strains was performed asdescribed previously (26).Hemolysin characterization. Hemolytic activity was as-

sayed either on brain heart infusion agar containing 2%washed horse erythrocytes (and 100 ,g of p-nitrophenyl-glycerine per ml [Sigma] when swarming was evident) or,after growth in brain heart infusion broth, in a 2% erythro-cyte suspension in 0.85% NaCl plus 20 mM CaCl2 incubatedat 40°C for 15 min. Activity was measured by the release ofhemoglobin (A543) and defined as cell free, cell associated, orintracellular after comparative assay of intact bacterial cul-ture, supernatant fluid from a ceritrifuged culture, andcleared cell extract of washed (10 mM Tris chloride, pH 7.4)bacterial cells (which were sonicated twice [20 s each time]on ice in an ultrasonic disintegrator [MSE Scientific Instru-ments] and centrifuged for 5 min at 1,500 x g).Cosmid cloning, subcloning, restriction mapping, and

transformation. Total cellular DNA was isolated as de-scribed previously (23), and partially digested Sau3A frag-

1509

on May 22, 2021 by guest

http://jb.asm.org/

Dow

nloaded from

1510 KORONAKIS ET AL.

TABLE 1. Hemolytic isolates of the Proteeae tribe and E. coli

Hemolytic activity"

Isolate Type" Infection' Intracellular Cell associated Cell free Calciumdependent

P. mirabilis U6450 p3/S1, 13 Chronic UTI + + - -P. vulgaris 76362 pS/S4 Burn graft + + + +M. morganii 227 38X UTI + + + +E. coli 519 04 UTI + + + +

a Proteeae isolates were typed by proticin production/sensitivity.bUTI, Urinary tract infection.cActivity in erythrocyte suspension.

ments of 20 to 35 kilobase pairs (kb) were isolated from NaClgradients (13 ml; 1.25 to 5 M) run for 3.5 h at 32,000 rpm and18°C in an SW40 rotor. Ligation into BamHI-cut cosmidpHC79 was followed by in vitro packaging (Boehringer) andtransfection of E. coli HB101. Recombinant cosmid DNAfrom hemolytic colonies was subjected to preliminary re-striction nuclease analyses and was then subcloned as

HindIII or PstI partial fragments into plasmid vectorpBR325. Restriction endonuclease analyses, deletion map-

ping, gel electrophoresis, and transformation of E. coliHB101 were performed as described previously (23).Transposon mutagenesis. Insertional inactivation of hly

recombinant plasmids was achieved with TnlOOO, isolatedafter conjugation of cointegrates formed with F::TnlOOO(11), and also TnS, delivered by bacteriophage lambda (1,17). Mapping of insertions followed comparative restrictionnuclease analyses.

Southern DNA-DNA hybridization analyses. Total cellularDNA (15 ,ug) was digested to completion overnight with 10 Uof restriction endonuclease per pLg, and fragments were

separated in 0.8% agarose before transfer to nitrocellulosefilters (Schleicher & Schuell). Prehybridization and hybrid-ization were performed at 42°C in 6x SSC (1x SSC is 0.15 MNaCl plus 0.015 M sodium citrate)-0.5% sodium dodecylsulfate-50% formamide-100 p.g of denatured salmon spermDNA per ml, Prehybridization was carried out for 1 h, andhybridization was carried out overnight. After hybridization,filters were washed three times for 1 h in low-stringencyconditions (i.e., at 50°C in 0.lx SSC-0.1% SDS) and dried.Autoradiography took up to 7 days at -70°C with intensify-ing screens. DNA probes were isolated from low-meltingagarose (Bio-Rad) and labeled by nick translation with[ot-32P]dCTP (3,000 Cilmmol; Amersham).Complementation tests. Recombinant DNA derived from

pBR325 and containing either TnS or TnlO00 was introducedby transformation into E. coli HB101 carrying incompatiblerecombinant plasmids derived from the vector pACYC184and carrying, singly or together, hlyB and hlyD from E. coli(21). Selection of the incoming plasmids was effected by thepresence of antibiotics. That both recombinant plasmidswere maintained independently in the recA host was

confimed by plasmid DNA isolation from the hemolyticdouble transformants, retransformation of E. coli HB101,and selection for the transposon-bearing plasmid. No hemo-lytic transformants were observed.

RESULTS

Hemolysin production among Proteeae clinical isolates.Ninety-five strains of the Proteeae tribe were examined, andthese were isolated from a wide range of clinical specimensin four hospitals. There was a high incidence of hemolysin

production among the three species; 94% of P. mirabilis,84% of P. viulgaris, and 56% of M. morganii strains wereclearly hemolytic when grown in brain heart infusion broth.The isolates taken for further examination appeared to betypical and are listed in Table 1.

Hemolysin production during bacterial batch growth. Fourindependent isolates each of P. mirabilis, P. vulgaris, andM. morganii and two of E. coli (including the representativeisolates listed in Table 1) were grown in brain heart infusionbroth, and at intervals throughout batch growth they wereassayed for extracellular (both cell-associated and cell-free)and intracellular hemolytic activities. This was achieved byincubating washed erythrocytes with samples of the wholeculture, supernatant from a centrifuged culture, or sonicatedwashed cells.

All strains tested synthesized and secreted active hemo-lysin in the logarithmic growth phase, and all showed adramatic loss of hemolytic activity in the late logarithmicphase. No intracellular activity was detected as the cellsentered stationary phase, but extracellular activity some-times persisted, slowly declining. Typical activities areshown in Fig. 1. Although some variation was seen in themaximum activities attained by isolates of a particular spe-cies, the large differences depicted among the four specieswere a clear and regular feature, with the members of theProteeae being more hemolytic than were the two E. colistrains examined (this was seen not only with maximalintracellular levels but also with extracellular activity).Two types of hemolytic activity. There were two types of

secreted hemolytic activity. That of all the M. morganiistrains was seen to equal extents in both whole-culture andcell-free assays. This hemolysis was visible around bacterialcolonies and was reminiscent of E. coli hemolysis. That of allP. mirabilis strains was seen (at high levels) only in whole-culture assays; i.e., activity was secreted but still associatedwith the cells (this was not seen on blood agar plates aroundgrowing colonies). Both these types of activity were pro-duced by isolates of P. vulgaris, the majority (ca. 60%)showing cell-associated activity. Activity of the a-hemolysinof uropathogenic E. coli is dependent on the presence ofCa2 + in the erythrocyte assay, whereas others (e.g., thegram-positive cytolysins) are not (12, 13). The two types ofhemolytic activity produced by members of the Proteeaealso differed in this respect; the cell-free activity of M.morganii and normally that of P. vulgaris were strictlydependent on the presence of Ca2+ (maximal activity wasattained with 10 mM Ca2+), and no hemolysis was seenwhen EDTA was added to the assay. In contrast, thecell-bound activity typical of P. mirabilis and P. vulgaris wasnot influenced by the addition of 20 mM Ca2+ or 10 mMEDTA; i.e., the activity is Ca2+ independent (Table 1).Southern DNA-DNA hybridization analyses. Hybridization

J. BACTERIOL.

on May 22, 2021 by guest

http://jb.asm.org/

Dow

nloaded from

SECRETED HEMOLYSINS 1511

-c

n

LJC.J

-j

C)

TIME (h) TIME (h)

FIG. 1. Hemolytic activity throughout batch growth. (a) Intracellular hemolytic activity of E. coli 519 (0), M. morganii 227 (D), P.mirabilis U6450 (*), and P. vulgaris 76362 ( * ). Growth is indicated (0). (b) Secretion of hemolysin by M. morganii 227 during batch growth.Intracellular activity (0), cell-free extracellular activity (*), and total culture activity (D). Growth is indicated (0).

of nick-translated E. coli alpha-hemolysin geHindIII-cleaved total DNA from the represerspecies cited above was performed at low saltThe restriction nuclease fragments used asmost of the hly determinant (12, 14) and are c2. The hybridization results (Fig. 2) indichomology, but this was only seen with

c

Excoi Ml determinant IEJA....... ..... ...__

I i i

a b c

2

2a b c

2 4 407 ...1 * i

4371 0- _ 6. _ }w

...

13:53_ _

10768- C-

8724

603*0-

FIG. 2. Low-stringency Southern hybridizatiortal cellular DNA was hybridized with hly probesrecombinant plasmid pANN202-312. The three nu

used as probes are shown at the top in relation tothe E. coli hly determinant. Panels: 1, BamHI-BgEcoRI-EcoRI (3,310 bp); 3, HindIII-PstI (840 bp). Iand hlyA (open bars) dictate intracellular hemolythlyB and hlyD (shaded bars) determine secretionDNA from the following: a, P. vulgaris 76362; b,c, P. mirabilis U6460.

ne probes with amounts of total DNA (15 ,ug per track), high levels ofitatives of each specific activity, and 3 to 7 days of autoradiography. Theconcentration. intensity of hybridization was about 1/20 of that seen whenprobes cover total DNA was taken from E. coli strains, e.g., strain 519.

lepicted in Fig. All three Proteeae DNAs hybridized reproducibly in a.ate significant number of HindIII fragments with the 3,650-base-pair (bp)relatively high BamHI-BglII fragment (probe 1). This covers 2,490 bp of the

3,074-bp hlyA hemolysin gene and also 150 bp of the 507-bphlyC gene, which, together with hlyA, determines active

a D intracellular hemolytic activity in E. coli. This may indicatethat these strains carry more than one copy of this region.

3 F-- Strains of P. mirabilis and P. vulgaris showed a similarpattern of hybridization which was quite different from thatofM. morganii. No extra hybridization was obvious with the

3 3,310-bp EcoRI-EcoRI fragment (probe 2) carrying 970 bp ofa b c (3') hlyA, the entire 2,123-bp secretion gene hlyB, and 115 bp

of hlyD, the second secretion gene. Again, several fragmentshybridized. In contrast, hybridization with the 840-bpHindIII-PstI probe fragment (probe 3) carrying only thecentral region of the 1,436-bp hlyD gene was found in onlyone HindIII fragment of DNA from P. vulgaris and M.

C- _morganii, and no homology was detected with DNA from P.mirabilis.

C- Isolation of hly determinants from P. vulgaris and M.morganii. P. vulgaris 76362 and M. morganii 227 producing

_- cell-free hemolysins appeared to carry no extrachromosomal_ DNA (data not shown). Cosmid cloning was therefore used

to isolate the genetic determinants of hemolysins secreted bythese strains. Several hemolytic colonies were obtained aftercloning of partial Sau3A fragments into the cosmid vector

i. HindIII-cut to- pHC79 and transfection of packaged recombinant cosmidfrom the E. coli DNA into E. coli HB101. Transfections derived from theclease fragments same strain showed no significant differences in hemolyticthe four genes of phenotype on blood agar and had very similar restriction. ccli6genes hlyC patterns. Recombinant DNA was isolated from representa-

ic activity; genes tives of the two types (pcos763-7 and pcos227-5, respec-.iLanes are total tively) and subcloned via HindIII and PstI partial digestsM. morganii 227; into the plasmid vector pBR325.

Recombinant pBR325 plasmids from hemolytic E. coli

VOL. 169, 1987

on May 22, 2021 by guest

http://jb.asm.org/

Dow

nloaded from

1512 KORONAKIS ET AL.

C A B DE.coli hly determinant

HP )A E HI I T I IIpANN202-3 12

pMO227-508

pVU763-710

p BgH[I:|:e;~~~~~~~~~. .. -f-

PHcA p P HP N E K ffi P4[ A Kl14I Hc Hc A E IPA..A1vt I I1Ill I I [vL .I-I tt++It H4 ++* ft H t *+

2hy~hyp(4)

1II 1! I t1hyp(2)

I

0 1 2 3 46 1

S r 6 8 9 Kb

FIG. 3. Physical and functional maps of the hly determinants. The maps of the hly determinants of the recombinant plasmids pVU763-710and pMO227-508 are aligned with each other and that of the published E. coli hly determinant on pANN202-312 (12, 14). The secretion genesof E. coli, hlyB and hlyD, are shown (shaded bars at top), and the two E. coli transcription start sites (18, 21, 30) are indicated by the arrowsabove the genes. Functional regions of the Proteeae hly determinants were defined by deletion and transposon inactivation, the latter withpublished maps of Tn5 (17) and TnlO00 (11). Insertion sites of the transposons TnS ( 11 ) and TnlO00 ( 1j ) are indicated by the perpendiculararrows under the lines, and their effect is indicated by open (no hemolytic activity) or closed (intracellular activity only) symbols. The hypinsertions caused hyperhemolytic activity, and insertions or deletions in the hatched regions had no effect on hemolysin production. Theplasmid vector DNA (pANN202-312, pMO227-508, and pVU763-710) is indicated by shaded bars. Restriction endonuclease sites: Av, AvaI;A, AccI; B, BamHI; Bg, BglII; C, CMlI; E, EcoRI; H, Hindlll; Hc, HindII; K, KpnI; N, NdeI; P, PstI; Pv, PvuII; R, EcoRV; S, SphI; X,XbaI. (There were no BclI, SmaI, or SalI sites.) Nuclease sites which appear to be common to more than one of the determinants areencircled; those common to all three determinants are boxed. Sites present in pANN202-312 but not in the cloned chromosomal hlydeterminant of pSF4000 are marked by an asterisk (i.e., all others shown are present in both these E. coli hly determinants [6, 12, 14]).

HB101 transformants were isolated and digested with eitherHindIII or PstI, and one of the smallest representatives ofeach was taken for detailed analysis. These were pVU763-710 and pMO227-508, which were stable and carried insertsof 8.6 and 9.6 kb, respectively.

Physical and functional comparison of the hly determinants.E. coli HB101 strains carrying each of the recombinantDNAs pVU763-710 and pMO227-508 were subjected tomutagenesis with the transposons TnS and Tn1000. Fromeach strain, 40 to 50 insertion mutants were obtained whichwere nonhemolytic on blood agar, and these were checkedfor residual hemolytic activity, particularly after cell sonica-tion. In each case, these mutants were of two kinds: thoselacking all activity, secreted or intracellular, and those whichretained normal (or slightly elevated) intracellular activitywhile having lost all extracellular activity. These findingsdemonstrated the separate functions of synthesis and secre-tion encoded by both determinants.

Detailed restriction nuclease maps of the cloned hly deter-minants pVU763-710 and pMO227-508 were constructed andwere used to map the sites of transposon insertions. Thesedefined the extent of the regions coding for synthesis andsecretion and allowed (Fig. 3) an alignment of the twohemolysin determinants with each other and also the previ-ously published E. coli alpha-hemolysin determinant of therecombinant DNA pANN202-312 (12, 14).

Full intracellular activity was determined in both Proteeaegenera by a region of ca. 4.2 kb, and this was followedimmediately by ca. 3.3 kb dictating secretion of the activehemolysis. In addition, several TnS and Tn1000 insertions(occurring at a frequency of about 1% of the Hly- mutantphenotype) caused considerable increases in the hemolyticactivity of both determinants. They mapped very closely toeach other at the end of the cloned determinants andadjacent to the region encoding synthesis of active intracel-lular hemolysin. These mutated (hyp) loci aligned approxi-mately with the regulatory sequences of the E. coli hlyC-hlyA-hlyB transcriptional unit and will be described in detailelsewhere.Alignment of the regions dictating synthesis and secretion

allowed a comparison of restriction nuclease sites through-out the three hly coding regions. Of the 26 restrictionnuclease sites mapped within the E. coli determinants, 7were apparent in one of the Proteeae hly determinants, and4 were apparently shared by both. Comparison of the twoProteeae determinants revealed 5 of the 25 restriction nucle-ase sites in the hly determinant of pVU763-710 to be presentin pMO227-508. This restriction nuclease site correspon-dence clearly suggests basic identity between the three hlydeterminants. Nevertheless, the degree of nuclease sitecoincidence is not comparable to that apparent betweenindependently isolated E. coli hly determinants; e.g., the

I I

J. BACTERIOL.

I I

., . 6

Hc HpH Hc OB P Hc H

.....1-4- I I 1. . IIT,l .Iflll I

I h ill 11 I

i A A

on May 22, 2021 by guest

http://jb.asm.org/

Dow

nloaded from

SECRETED HEMOLYSINS 1513

pANN202-312 hly determinant (originally plasmid borne[12]) and the cloned E. coli chromosomal hlv determinant ofrecombinant DNA pSF4000 (6, 30) share 24 of the 26 sitescited (Fig. 3).The contiguous regions encoding the synthesis and secre-

tion of the two hemolysins are comparable in size to eachother and to the hly determinant of E. coli, and their closeorganizational similarity is further indicated by two morepoints. Polar Tn5 insertions in the ca. 3.3-kb secretoryregion had no effect on the level of intracellular activity,indicating that secretory functions are transcribed indepen-dently or distal to those determining synthesis. A similartranscriptional organization is also indicated by the align-ment of the hyp mutations in both Proteeae determinantswith the transcription initiation region of the pANN202-312hlyC-hlyA-hlyB transcriptional unit.

Revision of the Southern hybridization data of Fig. 2 in thelight of the maps supports these contentions. Homology inone P. vulgaris 1.4-kb HindIII fragment when probed withE. coli probe 2 correlated with the presence of such afragment aligned with the central part of E. coli hlyA. Thiswas also seen when the overlapping probe 1 was hybridized,and the stronger hybridization in this instance could beexpected because of the additional 1.4-kb HindIll fragmentspanning the 5' end of hlyA and the 3' half of hiyC. Similarly,the hybridization of a ca. 3.8-kb HindIII fragment with bothprobes (2 and 3) spanning the downstream secretion genes ofE. coli was consistent with the alignment of such a fragmentin the P. vulgaris determinant. However, the observedhybridization of a HindIII fragment of comparable size (ca.3.8 kb) to probe 1 was not predicted from the map of P.vulgaris hly, and such additional hybridization was alsoevident with probe 2 (i.e., hybridization was seen with adouble band of this size). This indicates homology with E.coli hly in a further region not isolated on the recombinantDNA pVU763-710. Reexamination of the hybridizations tothe HindIll-cut total cellular DNA of M. morganii revealedsimilar features. A fragment in the region of the 6.5-kb sizemarker hybridized to all three E. coli probes and corre-sponded to the right-hand HindIII fragment of pVU763-710.Nevertheless, after allowing for the hybridization with theunknown HindlIl fragment (>2.5 kb) which includes theregion to the left of the single HindlIl site in M. morganii

TABLE 2. Complementation of Proteeae secretion mutations

Complementation by plasmidb:Mutant plasmida(Proteeae genes) pLG579 pLG595 pLG575 ACYC184

(hlyB) (hlyD) (hlyB, hlyD) P

pVU763-710/T35 + - + -(hlyB: :TnlO00)

pVU763-710/T4 - + + -(hlvD: :TnIOOO)

pMO227-205/T11 + - + -(hlyB: :Tn5)

pMO227-205/T3 - + + -(hlyD: :TnS)

aPlasmids, derived from pBR325, bearing secretion-defective hlv determi-nants from P. vulgaris or M. morganii. The Proteeae genes inactivated wereputatively designated hlyB or hlyD after alignment of the three hlv determi-nants (Fig. 3).bPlasmids, derived from pACYC184, bearing one or both E. coli hly

secretory genes (shown in parentheses) (21). +, Complementation restoredthe secreted activity to the level determined by the parental nonmutatedrecombinant plasmids; -, no change in the secretion-negative phenotypedetermined by the Proteeae recombinant plasmids; i.e., they retained intra-cellular activity only.

hly, there is still the 1.4-kb HindlIl fragment which showshomology and which is clearly not present in pMO227-508.The homologous fragments in the Proteeae total cellular

DNAs which are not present in the recombinant pBR325maps were not seen when the recombinant cosmids pcos763-7 and pcos227-5 were hybridized with the same E. coli hlyprobes (data not shown).Complementation of Proteeae mutants by E. coli recombi-

nant DNA. Further evidence of the functional relatedness ofthe three recombinant hly DNAs from E. coli, P. vulgaris,and M. morganii was obtained from complementation tests(Table 2). Secretion-defective mutant derivatives ofpVU763-710 and pMO227-508 (both pBR325 recombinantDNAs) were transformed into recA E. coli HB101 containingcompatible pACYC184-derived plasmids bearing the ex-pressed functional hlyB and hlyD secretion genes of E. coli(21). The transposon mutants, either TnS or TnlOOO, werechosen on the basis of their map locations (Fig. 3) clearlyaligned with either the hlyB or the hiyD genes of E. coli. Thechosen TnlOOO insertions in pVU763-710, T35 and T4,mapped at 6.15 (ca. 150 bp proximal to the EcoRV nucleasesite common to E. coli hiyB) and at 7.5 kb (corresponding tothe center of E. c oli hlyD), respectively. The Tn5 locations inthe selected Tll and T3 mutants of pMO227-508 were,respectively, at 6.1 kb (central hlyB in E. coli) and 7.2 kb,less than 100 bp proximal to the ultimate Hindll nuclease sitewhich aligns within hlyD of E. coli.

In every case (Table 2), the secretion defects of the mutantProteeae hly determinants were complemented to full activ-ity by plasmids carrying together the functional hlyB andhlyD secretion genes of E. coli (no effect was observed afterintroduction of the parent vector plasmid pACYC184). Fur-thermore, when the functional E. coli genes were presentseparately, they were still able to fully complement theProteeae mutations but only those predicted from the mapalignments. The presence of E. coli hlyB did not complementputative Proteeae hlyD mutations, nor did E. coli hlyDrestore secreted activity in putative Proteeae hlyB mutants.

DISCUSSION

The cytotoxic alpha-hemolysin (HlyA) of E. coli is se-creted across both gram-negative membranes without the aidof an N-terminal hydrophobic signal sequence (5). This iseffected by a specific envelope transport system formed bytwo proteins, HlyB and HlyD, which are encoded by genesof the contiguous four-gene hemolysin determinant (12, 21,28, 30). Examination of the highly conserved hly determinantsequence in E. coli has revealed surprisingly that the fourgenes have a G+C content of 39% rather than the expected50% of the genome (6, 14). This suggested to us a possiblecommon ancestry with the uncharacterized hemolysin deter-minants of Proteus, a genus of the family Enterobacteria-ceae having this low G+C genomic content. To investigatethis possibility and open the way to further elucidation of thesecretion mechanism, we examined the phenotypes andgenotypes of hemolysins from three different species ofProteeae.The synthesis of active intracellular hemolysin during

batch growth follows a similar pattern in P. vulgaris, P.mirabilis, M. morganii, and E. coli, although hemolysinactivity in the 12 isolates of Proteeae was considerablygreater than that seen in the two strains of E. coli tested.Maximal activity was attained in the mid-logarithmic phaseand rapidly decayed as culture growth ended. This sharppulse was more pronounced in the Proteeae strains. Se-

VOL. 169, 1987

on May 22, 2021 by guest

http://jb.asm.org/

Dow

nloaded from

1514 KORONAKIS ET AL.

creted activity in all species followed closely that seen

intracellularly, although activity often persisted longer, pre-sumably resulting from the formation of more stable com-

plexes (12). Extracellular hemolytic activity was cell free inthe case of all M. morganii and roughly 40% of the P.vulgaris strains, but in all P. mirabilis isolates and theremaining P. vulgaris strains, activity was found only inassociation with the intact bacterial cells. The possibilitythat the two different types of hemolysin detected in thevarious isolates were fundamentally the same but differed intheir secretory capacity was supported by hybridization oftotal cellular DNA to regions of the E. coli alpha-hemolysindeterminant. Homology was evident with DNA from strainsthat produced both types of activity, but hybridization was

in every case considerably lower than that seen with totalDNA of wild-type hemolytic E. coli. Although DNA from P.mirabilis, P. vulgaris, and M. morganii all hybridized withE. coli probes spanning hlyA, the hemolysin structural gene,and hlyB, the principal secretion gene, when probing was

performed with hlyD, the second transport gene, no homol-ogy was seen with P. mirabilis DNA. Hybridization oc-

curred with a number of restriction endonuclease fragmentswhich were not observed in the hly recombinant DNA,suggesting that Proteeae isolates may perhaps carry morethan one hly determinant. In E. coli, such multiple copies are

associated with specific insertion elements, which are alsoimplicated in the evolution of virulence gene clusters (19).

Molecular cloning of total cellular DNA from P. vulgaris76362 and M. morganii 227 (both of which produce cell-freehemolytic activity) gave a greater insight into the relation-ship of the hly genes to each other and to those of E. coli. Inboth cases, the region needed for hemolysin synthesis andsecretion was ca. 7.5 kb, and insertional inactivation withtransposons TnS and TnlOOO revealed that, analogous to E.coli hly, each determinant consisted of two contiguousstretches of ca. 4.2 and 3.3 kb encoding, respectively, thesynthesis of active hemolysin and its secretion out of thecell. That these analogous functions are dictated by closelyrelated genes is indicated by the coincidence of severalrestriction nuclease sites in the three hly determinants whenaligned on the basis of transposon mutant phenotypes.

In addition, transposon mutations which caused a loss ofsecretion function in the two Proteeae hly determinantscould be complemented by the presence of functional E. colisecretion genes hlyB and hlyD. Moreover, complementationoccurred only when the E. coli secretion gene present, hlyBor hlyD, was that corresponding to the region mutated in thealigned Proteus or Morganella determinant. This points to a

close functional relatedness between the components of thesecretion systems, not only the secretion proteins but alsothe relevant signal(s) on the hemolysin proteins.

Nevertheless, about two-thirds of the restriction siteswere unique to a single determinant, and only about 15%were common to all three hly determinants. This presents a

great contrast to the high degree (about 95%) of restrictionsite homology seen between hly determinants from differentisolates of E. coli. The data also indicate that the hlyD geneof E. coli is absent from P. mirabilis but present in the othertwo species, and this could explain the production by P.mirabilis strains of secreted hemolytic activity which re-mains associated with the cell and is not released into themedium. This seems to be an attractive possibility becausein E. coli the hlyD gene is believed to be transcribedseparately from the hlyC-hlyA-hlyB transcriptional unit (18,21) and therefore may not be a constant presence in entero-bacterial hly determinants. This might indicate that the

protein HlyD can be substituted by a different protein toform a critical secretory channel with HlyB, or alternatively,HlyB may be able to function independently to a large extentin the secretion process.The data suggest that the Proteeae and E. coli hly deter-

minants are closely related but have diverged significantly.In view of the G+C content of the E. coli hly genes (39%), itseems possible that they may have originated in Proteus spp.and spread to E. coli and M. morganii. We hope that thecomparison of conserved and divergent protein sequenceswill help define regions and interactions of critical impor-tance in the novel secretion of enterobacterial hemolysins.Recent reports (8, 10) of striking identity between E. coliHlyB and the mammalian p-glycoprotein which confersmultiple drug resistance on tumor cells suggest that suchstudies may also have relevance to eucaryotic export.

ACKNOWLEDGMENTS

We thank the Medical Research Council, United Kingdom, forfinancing this work.We thank Nigel Mackman for recombinant DNA.

LITERATURE CITED1. Berg, D., J. Davies, B. Allet, and J. D. Rochaix. 1975. Transpo-

sition of R factor genes to bacteriophage lambda. Proc. Natl.Acad. Sci. USA 72:3628-3631.

2. Cavalieri, S. J., G. A. Bohach, and I. S. Snyder. 1984. Esche-richia coli a-hemolysin: characteristics and probable role inpathogenicity. Microbiol. Rev. 48:326-343.

3. Debouck, C., A. Riccio, D. Schumperli, K. McKenney, J. Jeffers,C. Hughes, M. Rosenberg, M. Heusterpreute, F. Brunel, and J.Davidson. 1985. Structure of the galactokinase gene of Esche-richia coli, the last (?) gene of the gal operon. Nucleic AcidsRes. 13:1841-1853.

4. De la Cruz, F., J. C. Zabala, and J. M. Ortiz. 1983. Hemolysisdeterminant common to Escherichia coli strains of different 0serotypes and origins. Infect. Immun. 41:881-887.

5. Felmlee, T., S. Pellett, E.-Y. Lee, and R. A. Welch. 1985.Escherichia coli hemolysin is released extracellularly withoutcleavage of a signal peptide. J. Bacteriol. 163:88-93.

6. Felmlee, T., S. Pellett, and R. A. Welch. 1985. Nucleotidesequence of an Escherichia coli chromosomal hemolysin. J.Bacteriol. 163:94-105.

7. Fry, D. C., S. A. Kuby, and A. S. Mildvan. 1986. ATP-bindingsite of adenylate kinase: mechanistic implications of its homol-ogy with ras-encoded p21, F1-ATPase, and other nucleotide-binding proteins. Proc. Natl. Acad. Sci. USA 83:907-911.

8. Gerlach, J. H., J. Endicott, P. Juranka, G. Henderson, F.Sarangi, K. Deuchars, and V. Ling. 1986. Homology betweenP-glycoprotein and a bacterial hemolysin transport protein sug-gests a model for multidrug resistance. Nature (London)324:485-489.

9. Gray, L., N. Mackman, J.-M. Nicaud, and I. B. Holland. 1986.The carboxy-terminal region of haemolysin 2001 is required forsecretion of the toxin from Escherichia coli. Mol. Gen. Genet.205:127-133.

10. Gros, P., J. Croop, and D. Housman. 1986. Mammalianmultidrug resistance gene: complete cDNA sequence indicatesstrong homology to bacterial transport proteins. Cell 47:371-380.

11. Guyer, M. S. 1983. Uses of the transposon gamma delta in theanalysis of cloned genes. Methods Enzymol. 101:362-369.

12. Hacker, J., and C. Hughes. 1985. Genetics of Escherichia colihemolysin. Curr. Top. Microbiol. Immunol. 118:139-162.

13. Hacker, J., and C. Hughes. 1985. Genetic analysis of bacterialhemolysin production. Bull. Inst. Pasteur 83:149-165.

14. Hess, J., W. Wels, M. Vogel, and W. Goebel. 1986. Nucleotidesequence of a plasmid-encoded hemolysin determinant and itscomparison with a corresponding chromosomal hemolysin se-

quence. FEMS Microbiol Lett. 34:1-11.

J. BACTERIOL.

on May 22, 2021 by guest

http://jb.asm.org/

Dow

nloaded from

SECRETED HEMOLYSINS 1515

15. Higgins, C. F., 1. Hiles, G. P. Salmond, D. Gill, J. Downie, I.Evans, I. B. Holland, L. Gray, S. Buckel, A. Bell, and M.Hermodson. 1986. A family of ATP-binding subunits coupled tomany distinct biological processes in bacteria. Nature (London)323:448-450.

16. Jakes, K. S., and P. Model. 1979. Mechanism of export of colicinEl and colicin E3. J. Bacteriol. 138:770-778.

17. Jorgensen, R. A., S. J. Rothstein, and W. S. Reznikoff. 1979. Arestriction enzyme cleavage map of TnS and location of a regionencoding neomycin resistance. Mol. Gen. Genet. 177:65-72.

18. Juarez, A., C. Hughes, M. Vogel, and W. Goebel. 1984. Expres-sion and regulation of the plasmid-encoded hemolysin determi-nant of Escherichia coli. Mol. Gen. Genet. 197:196-203.

19. Knapp, S., J. Hacker, I. Then, D. Muller, and W. Goebel. 1984.Multiple copies of hemolysin genes and associated sequences inthe chromosomes of uropathogenic Escherichia coli strains. J.Bacteriol. 159:1027-1033.

20. Mackman, N., and I. B. Holland. 1984. Functional characteriza-tion of a cloned haemolysin determinant from E. coli of humanorigin, encoding information for the secretion of a 107K poly-peptide. Mol. Gen. Genet. 196:123-134.

21. Mackman, N., J.-M. Nicaud, L. Gray, and I. B. Holland. 1985.Genetical and functional organization of the Escherichia colihaemolysin determinant 2001. Mol. Gen. Genet. 201:282-288.

22. Mackman, N., J.-M. Nicaud, L. Gray, and I. B. Holland. 1986.Secretion of haemolysin by Escherichia coli. Curr. Top. Micro-

biol. Immunol. 125:159-181.23. Maniatis, T., E. F. Fritsch, and J. Sambrook. 1982. Molecular

cloning: a laboratory manual. Cold Spring Harbor Laboratory,Cold Spring Harbor, N.Y.

24. Muller, D., C. Hughes, and W. Goebel. 1983. Relationshipbetween plasmid and chromosomal hemolysin determinants ofEscherichia coli. J. Bacteriol. 153:846-851.

25. Palva, E. T., T. R. HIirst, S. J. S. Hardy, J. Holmgren, and L.Randall. 1981. Synthesis of a precursor to the B subunit ofheat-labile enterotoxin in Escherichia coli. J. Bacteriol. 146:325-330.

26. Senior, B. W. 1977. Typing of Proteus mirabilis strains byproticine production and sensitivity. J. Med. Microbiol.10:7-17.

27. So, M., and B. J. McCarthy. 1980. Nucleotide sequence of thebacterial transposon Tn1681 encoding a heat-stable (ST) toxinand its identification in enterotoxinogenic Escherichia colistrains. Proc. Natl. Acad. Sci. USA 77:4011-4015.

28. Stark, J. M., and C. W. Shuster. 1983. The structure of clonedhemolysin DNA from plasmid pHlyl85. Plasmid 10:45-54.

29. Wagner, W., M. Vogel, and W. Goebel. 1983. Transport ofhemolysin across the outer membrane of Escherichia coli re-quires two functions. J. Bacteriol. 154:200-210.

30. Welch, R. A., R. Hull, and S. Falkow. 1983. Molecular cloningand physical characterization of a chromosomal hemolysin fromEscherichia coli. Infect. Immun. 42:178-186.

VOL. 169, 1987

on May 22, 2021 by guest

http://jb.asm.org/

Dow

nloaded from