INFECTION AND IMMUNITY, Jan. 1994, p. 244-2510019-9567/94/$04.00+0Copyright © 1994, American Society for Microbiology

Cloning and Sequencing of the Genes Encoding Escherichiacoli Cytolethal Distending Toxin

DANIEL A. SCOTTt AND JAMES B. KAPER*

Centerfor Vaccine Development, Division of Geographic Medicine, Department ofMedicine,University ofMaryland School of Medicine, Baltimore, Maryland 21201

Received 23 July 1993/Returned for modification 3 September 1993/Accepted 19 October 1993

Escherichia coli strains expressing cytolethal distending toxin (CDT) cause elongation of CHO cells at 24 h,followed by progressive cellular distention and death for up to 120 h. Similar distention and cytotoxicity areseen in HeLa, HEp-2, and, to a lesser extent, Vero cells. The initial elongation in CHO cells is indistinguishablefrom that caused by E. coli heat-labile toxin (LT). In contrast to those from LT strains, supernatants from thesestrains have no effect on Y-1 adrenal cells. TnphoA was introduced into CDT-positive E. coli E6468/62(086:H34), isolated from a child with diarrhea, and 13 CDT-negative transconjugants were identified. DNAprobes constructed from DNA flanking the TnphoA insertion sites of CDT-negative mutants were used toidentify a CDT-positive clone from an E6468/62 genomic library with a 5.5-kb insert. Exonuclease deletionswere created and assayed in CHO cells. In this manner, a 2.3-kb CDT-active region was defined, and thenucleotide sequence was determined. Sequence analysis identified three open reading frames (ORFs),designated cdtA, cdtB, and cdtC. These contain 711, 819, and 570 bp, respectively, and encode polypeptideswith predicted molecular masses of 25.5, 29.8, and 20.3 kDa, respectively. Each ORF has a putative signalsequence, and there are 4-bp overlaps between cdtA and cdtB and between cdtB and cdtC. The nucleotide andpredicted amino acid sequences have no significant homology with those of any previously reported genes orproteins. By in vitro transcription-translation and an anti-alkaline phosphatase immunoblot, native proteinsand/or fusion proteins corresponding to each ORF were identified.

Tissue culture assays have been used to identify andcharacterize a variety of important enterotoxins and cyto-toxins from diarrheal pathogens (5, 7, 13, 18). In 1987,Johnson and Lior, using strains of Escherichia coli 0128isolated from children with diarrhea, described a new type ofactivity in Chinese hamster ovary (CHO) cells and namedthe responsible factor cytolethal distending toxin (CDT) (10).Supernatants from CDT-positive (CDT') E. coli cause elon-gation of CHO cells at 24 h that is indistinguishable from thatcaused by E. coli heat-labile toxin (LT). With CDT' strains,this elongation is followed by progressive cellular distentionand cytotoxicity for up to 120 h. Similar cellular distentionand cytotoxicity are seen with HeLa cells, Hep-2 cells, and,to a lesser extent, Vero cells. In contrast to LT strains,CDT' strains have no effect on Y-1 adrenal cells. Theauthors noted that, in a CHO cell assay, strains expressingCDT could be confused with enterotoxigenic E. coli strainsexpressing LT. This may explain several previous reports ofE. coli strains that appeared to produce LT, as judged fromelongation of CHO cells, but were not reactive in either Y-1adrenal cells or an immunologic assay for LT (8, 16, 21).CDT activity is destroyed by treatment with trypsin or by

heating (70°C for 15 min) and is retained by a 30,000-molecular-weight filter (10, 11). Rabbit antiserum preparedto a CDT' E. coli strain neutralizes only CDT activity. CDTactivity is not neutralized by antibody to cholera toxin, E.coli LT, E. coli Shiga-like toxin, or Clostridium difficile

* Corresponding author. Mailing address: Center for VaccineDevelopment, University of Maryland School of Medicine, 10 SouthPine St., Baltimore, MD 21201. Phone: (410) 706-5328. Fax: (410)706-6205.

t Present address: Enteric Diseases Program, Naval MedicalResearch Institute, Rockville, MD 20852.

cytotoxin. CDT can be differentiated from Shiga-like toxinby its effect in CHO cells, in which Shiga-like toxin isinactive. Cytotoxic necrotizing factor causes cellular disten-tion similar to that by CDT but with prominent multinucle-ation not seen with CDT (4). CDT activity has been found inCampylobacter jejuni as well as other Campylobacter spp.and Shigella spp. (10, 12).CDT' E. coli have been isolated from patients in Canada

with a variety of diarrheal syndromes (2). In an Indianpediatric study, Bouzari and Varghese reported that 6.4% of249 E. coli isolates belonging to classic enteropathogenic E.coli 0 serogroups expressed CDT activity (3). Anderson etal. isolated CDT' E. coli from the stool of a 3-year-old boypresenting with gastroenteritis and encephalopathy with noother etiology (1). Although these clinical associations areinteresting, there are no data clearly linking CDT activitywith diarrhea or any other clinical syndrome. Our goal in thisstudy was to clone and sequence the gene(s) responsible forthis activity in E. coli as a first step in determining thesignificance of this putative virulence factor in enteric patho-gens.

MATERIALS AND METHODS

Bacterial strains and media. CDT' E. coli E6468/62 (sero-type 086:H34) was used for all studies reported here. It is aclinical isolate from a child of <2 years of age with acutediarrhea (kindly provided by Timothy Barrett, Centers forDisease Control, Atlanta, Ga.). It has no detectable plas-mids, does not hybridize with probes for E. coli LT orheat-stable toxin, and does not adhere to HEp-2 cells in anyof the recognized patterns (17). E. coli DH5ao (BRL Inc.,Gaithersburg, Md.) was used as the recipient for all trans-formations. Bacterial strains were grown in Luria broth or

244

Vol. 62, No. 1

E. COLI CYTOLETHAL DISTENDING TOXIN 245

on Luria plates. Ampicillin (200 jig/ml), kanamycin (50,ug/ml), and streptomycin (100 ,ug/ml) were added whenappropriate.CDT assays. CHO cells were maintained at 37°C with 5%

CO2 in Ham's F12 medium plus glutamine supplementedwith 10% fetal bovine serum, penicillin (100 U/ml), andstreptomycin (100 ,ug/ml). Cells were trypsinized, resus-pended at 2 x 104 cells per ml in Ham's F12 medium with 1%fetal bovine serum, penicillin, and streptomycin (gentamicinwas included at 50 ,ug/ml to assay TnphoA mutants whichwere Kmr Smr), and added to 96-well plates at 150 ,il perwell. Bacterial cultures were grown with shaking overnightin Trypticase soy broth with 0.6% yeast extract. Aliquots (15,ul per well) of each test supernatant were added in triplicateto freshly plated CHO cells. Initial studies showed identicalactivity with culture supernatants and sterile culture fil-trates. Assay mixes were incubated for 5 days at 37°C in 5%CO2, fixed with 80% ethanol for 30 min, and stained withGiemsa stain.TnphoA mutagenesis. To obtain chromosomal insertions,

SM1OXpir containing TnphoA in the suicide vector pRT733was conjugated with a streptomycin-resistant derivative ofE6468/62 (E6468/62Smr), and kanamycin- and streptomycin-resistant (KMr Smr) colonies were selected. Transconjugantsthat had alkaline phosphatase (PhoA) activity on mediumcontaining 5-bromo-4-chloro-3-indolylphosphate (40 ,ug/ml)were chosen for further study. It is expected that thesecolonies contain in-frame TnphoA insertions in genes encod-ing proteins that are directed out of the cytoplasm by signalsequences (15). To minimize the number of siblings, no morethan three colonies were selected from each mating. Allselected colonies were assayed for CDT activity in CHOcells.The TnphoA insertion from each CDT-negative (CDT-)

mutant was cloned by digestion of chromosomal DNA withBamHI or SalI (both restriction sites are located down-stream of the Kmr gene in TnphoA), ligation into thecorresponding digest of pBluescriptII SK- (pBSII; Strata-gene Inc., La Jolla, Calif.), and selection for Kmr and PhoAactivity on medium containing 5-bromo-4-chloro-3-in-dolylphosphate. These cloned insertions were used to se-quence the junction of each TnphoA insertion with E6468/62Smr chromosomal DNA and to identify PhoA fusionprotein products in an in vitro transcription-translation sys-tem, described below.

Southern hybridization. Chromosomal DNA from eachCDT- mutant was digested separately with KpnI, MluI, andEcoRV, which have no restriction sites within TnphoA.Digestion products were separated by electrophoresis in0.5% agarose and transferred to nitrocellulose filters. A2.2-kb internal BglII fragment of TnphoA containing thekanamycin resistance gene was isolated by electroelutionand used as a DNA probe under stringent hybridizationconditions by the method of Southern (20). All restrictionfragment probes were labeled with [a-32P]dATP by randomprimer extension (Random-prime DNA Labeling Kit; Boehr-inger Mannheim, Mannheim, Germany). Synthetic oligonu-cleotide probes were end labeled with [_y-32P]dATP by usingpolynucleotide kinase.Genomic library construction. Chromosomal DNA from

E6468/62 was partially digested with Sau3A and separatedby electrophoresis in 0.7% agarose. The 4.5- to 15-kb frac-tion was electroeluted, ligated into pUC19 which had beendigested with BamHI and dephosphorylated, and trans-formed into E. coli DH5a.Two gene probes were constructed from cloned

E6468/62Smr: :TnphoA insertion strains. With a BamHIclone of TnphoA insertion strain DS16 (see Fig. 3), DNAsequence upstream of the junction of TnphoA was deter-mined with a primer homologous to phoA. From this se-quence, an oligonucleotide probe (5'-CACGTCTGCAAGGCACTAC-3'; see Fig. 4) was synthesized and end labeled.An EcoRI clone of TnphoA insertion strain DS17 (see Fig. 3)in pACYC184 was digested with HpaI, yielding three frag-ments (3, 7, and 10 kb). The EcoRI (located upstream of theKmr gene of TnphoA) and HpaI (189 bp from the 3' end ofTnphoA) restriction sites were chosen to avoid inclusion ofphoA and other TnphoA sequences in the probe. Both theoligonucleotide probe and the 3-kb HpaI fragment hybrid-ized with E6468/62, a second CDT' E. coli strain, and twoCDT' Shigella strains. Neither probe hybridized with E.coli DH5ot, prototypic strains of enterotoxigenic or entero-pathogenic E. coli, or Vibno cholerae 01. Both probes wereused to screen colony blots from the E6468/62 genomiclibrary. Probe-positive clones were tested for CDT activityin CHO cells.

Exonuclease deletions. An HindIII fragment from a CDT'gene bank clone was ligated into HindIII-digested pBSII.Exonuclease deletions were made with the Erase-a-Basesystem (Promega Inc., Madison, Wis.). The first set ofdeletions used the SstI and NotI restriction sites in the pBSIImulticloning site. A series of deletions were assayed forCDT activity, and the smallest active clone was used tomake a second set of deletions from the KjpnI and SalIrestriction sites of the pBSII multicloning site. These dele-tions were also assayed for CDT activity and defined thelimits of a CDT-active region for DNA sequencing.DNA sequencing. Exonuclease deletions through the CDT-

active region were sequenced by using M13 forward andreverse primers. The chromosomal DNA junction of eachTnphoA insertion was sequenced with a primer to the phoAgene and the BamHI or Sall clones described in the TnphoAmutagenesis section as templates. Sequencing reactionswere done with [ct-35S]dATP and Sequenase reagents (USB,Cleveland, Ohio). Several regions were also sequenced withthe Taq Dye Primer Cycle Sequencing Kit (ABI, FosterCity, Calif.) with an automated gene sequencer (ABI 3073).Plasmids for sequencing were isolated by alkaline denatur-ation, followed by either centrifugation to equilibrium inCsCl-ethidium bromide density gradients or purification inQiagen columns (Qiagen Inc., Chatsworth, Calif.). DNA andamino acid sequences were analyzed with the GeneticsComputer Group software (6).

In vitro transcription-translation. In vitro transcription-translation studies were performed with purified plasmids orcloned linear DNA fragments. The cloned DNA fragmentswere separated from vector sequences by agarose gel puri-fication of BssHII restriction digests (BssHII sites flank themulticloning site of pBSII). Proteins were labeled with[35S]methionine by using the Prokaryotic DNA-DirectedTranslation Kit (Amersham, Arlington Heights, Ill.). Theincorporation of radioactivity was measured by removing analiquot from each reaction mix, precipitating the proteinswith trichloroacetic acid, and liquid scintillation counting.Transcribed proteins were solubilized in sample buffer (2%sodium dodecyl sulfate [SDS], 2% 3-mercaptoethanol) andseparated by electrophoresis in separate 8, 12.5, and 15%polyacrylamide gels with SDS. Sample volumes were ad-justed so that each lane contained an equal amount ofincorporated radioactivity. Gels were dehydrated by soakingin 90% ethanol-10% glycerol for 3 to 4 h and air-dried priorto autoradiography (23).

VOL. 62, 1994

246 SCOT1T AND KAPER

, _X _

ff 9 0000 0000 t0:,0<-;'-,00 :0E,?ut 0:0.-00 t:>0 00 X, :0_

W; _.^. Di .&. ,. ff X,-D _ -

, ,, _X, -f; ' %:> a 00 .'VEX, 0

.. -0 .t ?000 * t - 0 -__: : : . . A - ; :ta _' Wo 0 lS AS , lt f :5ot) _

:D:f j^z A '; 1, a. iCii. i,:.: .'D: D': C 'z': _

a, ,. - X, __ _

00 ;, jf' 'S- ;'',M',; ffff 0,0 Ck.'o'X.'b;.. .S,i' _ t. af,:

,, it. E:S, >^f

/ F:054,S ; ,0 ,0d, dt X, t);

r s; ^; ->d t,7 t:4, ;$'

20 ' Cs'' , 1',.>.-;' " f;f ' A. ........ 0 fW ' .:0. ,. +' :.e

.E i:

d0"" ' iHi 3),'0 Ato0 ,l @ 0 >0 x i S,-> ,00--23t- ' ' t 0 0 l';.

0 0SP-#0 4fR0jf rkSastHf fff; t000 0

FIG. 1. CDT assays. Photomicrographs of CHO cells stained with Giemsa after 120 h of incubation with supernatant from (A) wild-typeCDT+ E. coli E6468/62, (B) E. coli DHSa(PUC19), (C) E. coli DHSa(PDS1), undiluted, and (D) E. coli DHSa(PDS1), diluted 1:256.

Anti-alkaline phosphatase Western blot (immunoblot).E6468/62Smr::TnphoA mutants were grown overnight andpelleted, and the cells were lysed in sample buffer. Proteinswere separated by electrophoresis in an 8% polyacrylamide-SDS gel and electroblotted to Immobilon-P (Millipore, Bed-ford, Mass.). The blot was incubated sequentially at roomtemperature in the following solutions: phosphate-bufferedsaline (PBS)-3% (wt/vol) milk (5 min), PBS-3% milk withrabbit anti-alkaline phosphatase immunoglobulin G (IgG)diluted 1:500 (5'-3' Inc., Boulder, Colo.) (1 h), and PBS withalkaline phosphatase-conjugated goat anti-rabbit IgG diluted1:1,000 (Sigma Immunochemicals, St. Louis, Mo.) (1 h). Theblot was developed with Western Blue (Promega Inc.) alka-line phosphatase substrate.

Nucleotide sequence accession number. The DNA sequencereported here has been submitted to GenBank (accessionnumber U03293).

RESULTS

CDT assay. As shown in Fig. 1A, supernatants fromE6468/62 cause marked distention of CHO cells identical tothat described previously for CDT (9). Activity was detectedwhen the supernatant was diluted c 1:8 before addition to theassay system. When exposed to supernatants from nontox-igenic strains, the cells grew into a confluent monolayer (Fig.1B).TnphoA mutagenesis. A total of 346 PhoA+ transcon-

jugants were selected from 147 individual filter matings.These mutants were screened in the CHO cell assay, and 13CDT- mutants, designated strains DS10 to DS22, wereidentified. Southern analysis of chromosomal DNA revealedthat each mutant contained a single TnphoA insertion. Foreach of the 13 mutants, the BglII TnphoA fragment probehybridized to a 23-kb KpnI fragment, a 23-kb MluI fragment,

INFECT1. IMMUN.

E. COLI CYTOLETHAL DISTENDING TOXIN 247

A

23.1

9.4 .

6.5

4.3

1 2 3 4 5 6 7 8 9 10 11 12

B

23.1 -

9.4 -

6.5 -

4.3 -

1 2 3 4 5 6 7 8 9 10 11 12FIG. 2. Southern blots hybridized with the [ox-32P]dATP-labeled

2.5-kb BglII fragment of TnphoA. (A) Lanes 1 to 4 and 6 to 11,EcoRV-digested chromosomal DNA from E6468/62Smr cdt::TnphoA mutants DS10, -11, -16, -20, -21, -12, -18, -15, and -19,respectively; lane 5, X digested with HindIll; lane 12, EcoRV-digested E6468/62 chromosomal DNA. (B) Lanes 1 to 4, EcoRV-digested chromosomal DNA from E6468/62 and E6468/62Smrcdt::TnphoA mutants DS14, -13, and -17, respectively; lanes 5 to 12,K.nI-digested chromosomal DNA from E6468/62Smr cdt::TnphoAmutants DS10, -16, -20, -21, -22, -12, -18, and -15, respectively.Sizes are shown in kilobases.

and either a 9-kb or 10-kb EcoRV fragment (partial datashown in Fig. 2).BamHI (pDSxx.3) or Sall clones of each of the 13 mutants

exhibited PhoA activity. The chromosomal DNA junction ofeach TnphoA insertion was sequenced, and sequence anal-ysis showed that all 13 lay within the CDT-active regiondescribed below (Fig. 3).Genomic library and exonuclease deletions. Five colonies

from the E6468/62 genomic library hybridized with both the3-kb HpaI fragment probe and the oligonucleotide probe.Four clones had no activity, but supernatant from one clone,pDS1, was found to completely inhibit CHO cell growth(Fig. 1B). When the supernatant from pDS1 was diluted, itshowed typical CDT activity at dilutions from 1:128 to 1:512(Fig. 1C). Restriction digestion showed that pDS1 containeda 5.5-kb insert.A 4.5-kb HindIII fragment from pDS1 was cloned into

pBSII, and the resulting plasmid, pDS2, expressed CDTactivity at the same titer as pDS1. A series of exonucleasedeletions from the SstI and NotI sites of pBSII were madefrom pDS2 (designated pDS6.x) and assayed for CDT activ-ity. The smallest CDT-active clone (pDS7) was used for asecond series of exonuclease deletions (designated pDS7.x)beginning at the KpnI and SalI sites of pBSII. The smallestCDT-active clones (pDS7.96, pDS7.84, and pDS7.14 in Fig.4) defined the region necessary for expression of CDTactivity, which was subjected to DNA sequencing.DNA sequencing. The exonuclease deletions were used to

sequence both DNA strands through the CDT-active region,which contained 2,305 bp (Fig. 4). Analysis of the DNAsequence identified three potential open reading frames(ORFs), designated cdtA, cdtB, and cdtC (Fig. 3). TheseORFs contain 711, 819, and 570 bp, respectively, andpotentially encode polypeptides with predicted molecularmasses of 25.5, 29.8, and 20.3 kDa, respectively. Each ORFhas a putative signal sequence, ribosome-binding site, andpromoter region (Fig. 4). The first ORF (cdtA) has a putativeGTG translational start identified by visual inspection. Thefirst in-frame ATG in this ORF probably does not representthe actual start site because six PhoA+ TnphoA insertions,DS10 to DS15, lie upstream of this ATG (Fig. 3). This is alsosupported by data presented below regarding the size of

DS13 DS16,17 DS19

DS10,11,12 DS14,15 DS18 DS20,21 DS22

GTG ATG EcoRV BamHI Ec

r'

cdtB

coRV

cdtC

pDS6.52

Eco RI

1.

nDDS7.43pDS7.34

pDS7.220pDS7.84

500 1000I

1500I

2000I .

FIG. 3. Diagram of the 2,305-bp CDT-active region sequenced in this study. The three ORFs cdtA, cdtB, and cdtC are noted by the widearrows, and the region sequenced upstream of cdtA is shown as an open box. DNA flanking the sequenced region is represented by a thinline. The scale below indicates base pairs in the CDT-active region. The position of the putative GTG translational start for cdtA and the firstin-frame ATG in cdtA are shown. pDS7.x designates the exonuclease deletions used in the in vitro transcription-translation studies. pDS6.52is an exonuclease deletion that contains 1.5 kb of E6468/62 DNA upstream of cdtA but has lost the terminal 215 bp of cdtC and is inactivein CHO cells. The location of each TnphoA insertion (DS10 to DS22) is indicated. The presence of more than one number at a site indicatesmultiple independent insertions at exactly the same base pair.

BamHI

/ /

20 kb _//

1.5 kb

PI. . .

VOL. 62, 1994

'ICdtb

248 SCOTT AND KAPER

LIZ pDS7.96UI VIDL FLUUUAAIAAIAAIuAu6AIIAIAAIIAAIIU UI I I IIUUI W t.AI

V D K IC L I A F L C T L I I-. .-.--------. - --_-- ._ - - ---- .-. . . --_ ---- _--------- .-__-.-_-.IAAI.IIIp7 A84 L 35 -IL0I UI u I I I ApIRAAAUAAUAU0ISUU7U4AUALRAUUAAASIAALISIAAAAAAIAAI

pDS7. 84 -35 - 1 0 pDS7. 14 '- RBS TS

T G C S N G I G D S P S P P G K N V E L V G I P G Q G I A V T S N G A T P T L G241 TTACTGGTTGCTCGMTGGGATCGGTGATTCACCTTCACCTCCGGGAAMTGTAGMTTGGTTGGMTCCCTGGACMGGTATTGCAGTGACTTCAMCGGTGCMCTCCMCACTTG

pDS7.91

A N N T E F P E V S I M S T G G A L L T I W A R P V R N W L W G Y T P F D S V N361 GAGCCACACACTGAGTTTCCTGAGTTTCAATATGAGCACTGGTGGGGCGCTGCTTACTATTTGGGCCAGACCTGTTCGTACTGGCTTTGGGGGTATACTCCTTTTGATTCAGTA

F G E N R N W K V V D G K D A G T V K F V N V A Q G T C M E A F K N G V I H N T481 ATTTTGGTGAGMTCGGMCTGGMGGTTGTGGATGGGMMGATGCCGGCACAGTGAMTTTGTTAATGTTGCCCAGGGGACTTGCATGGAGGCCTTTAMAACGGGGTGATACATAATA

C D D N S L S Q E F Q L L P S T N G N V L I R S S A L Q T C I R A D Y L S R T I601 CCTGTGATGATMCTCGTTATCTCAGGAGTTTCAGTTACTGCCTTCTACTMTGGTMTGTGCTTATMGMGTAGTGCCTTGCAGACGTGTATMGAGCAGACTATTTMGCAGMCTA

OLigonucleotide probe

L S P F A F T I T L E K C P G A K E E T Q E M L W A I S P P V R A A K P N L I K721 TATTGTCACCGTTTGCTTTTACMTCACCCTTGAGMMTGCCCTGGTGCMAAGMGAMCGCMGAAATGCTATGGGCMTMGTCCACCTGTCAGAGCGGCMAACCAAATCTGATTA

P E L R P F R P L P I P P H D K P D G M E G V *841 AGCCAGAGTTAGACCATTCAGACCATTGCCAATTCCACCTCATGACAACCTGATGGATGGAGGGAGTATGAAAATTATTATTCCTGTTATGATTTTGCCGGGTATTTCTTTTGC

-35 -10 RBS N K K L L F L L N I L P G I S F A

961 AGATTTMGCGATTTTAMGTTGCMCCTGGMTTTGCAGGGTTCMMTGCACCGACAGMAATMATGGMCACACATGTCCGACMCTTGTTACGGGMGTGGTGCTGTTGATATCCTD L S D F K V A T W N L Q G S N A P T E N K W N T H V R Q L V T G S G A V D I L

1081M V Q E A G A V P A S A T L T E R E F S T P G I P M N E Y I W N T G T N S R P Q

1 201 GGAGTTGTTTATATATTTCTCACGTGTTGATGCATTCGCTAACAGAGTAAATCTTGCGATTGTTTCAAMCAGAAGAGCTGATGAGGTGATTGTATTACCTCCTCCAACTGTTGTATCACGE L F I Y F S R V D A F A N R V N L A I V S N R R A D E V I V L P P P T V V S R

1321 ACCGATCATCGGCATTAGMTTGGTMTGATGTTTTCTTCTCMCCCATGCATTGGCGAATCGGGGCGTGGATTCAGGAGCMTTGTMMTAGTGTTTTTGAGTTCTTCMCAGACAAACP I I G I R I G N D V F F S T H A L A N R G V D S G A I V N S V F E F F N R Q T

1441 GGATCCTATMGACAGGCCGCTAACTGGATGATTGCAGGAGATTTTMCCGTTCACCGGCTACACTATTTTCMCTCTTGMCCAGGGATTCGCMTCATGTAMTATTATTGCTCCACCD P I R Q A A N W M I A G D F N R S P A T L F S T L E P G I R N H V N I I A P P

-35 -101561 AGATCCMCGCMGCCAGTGGTGGTGTTCTTGATTATGCAGTAGTTGGMMTTCAGTGAGCTTTGTTCTTCCTCTGTTGAGGGCCTCGTTGTTATTCGGATTATTMGAGGGCMATTGC

D P T Q A S G G V L D Y A V V G N S V S F V L P L L R A S L L F G L L R G Q I A

CTCTGATCATTTTCCGGTTGGCTTTATTCCTGMGAIS D H F P V G F I P G R

RBS N K T V I V F F V L L L T G C A S E P A N Q R NGGAGCAAGAAGATGAAAACAGTTATTGTGTTTTTTGTTTTACTGCTGACAGGTTGTGCTTCTGAACCTGCAAATCAGCGTAATG A R R *

L L T Q F V G N N A P V D P E P S P V L V N I R N V L T G G I I R N P V G S D F1801 CTTCTTACTCAGTTTGTCGGCAACMTGCCCCTGTAGACCCTGMCCCAGTCCAGTATTGGTTMTATCAGMMCGTTCTTACAGGGGGGATMTCCGAAATCCTGTTGGCAGTGACTTT 1920

N V N N W V I S E V K T N D L D L I S A P G G H V Q0I K N P D G N E C F A I L N1921 4TGTAAAT0TTGGGTTATATCTGMGTMMGACTMTGATTTGGATTTGATATCGGCACCGGGAGGGCATGTTCAGATTMMATCCTGATGGCMTGMTGCTTTGCTATTCTMAC2040

G Q L A V A K Q C S E S D R N A L F T F I T S D T G A V Q I K S I G S G Q C L G2041 GGGCAATTGGCAGTGGCT2GCAGTGCTCTGAAAGTGACCGTAACGCATTGTTTACATTTATACCAGTGATACTGGGGCTGTGCMATCMGTCMTAGGMGCGGTCMTGCCTAGGG2160

N G E S I T D F R L K K C V D D L G R P F D T V P P G L L W M L N P P L S P A I2161 8TGGAGAGAGCATTACAGATTTCAGGTTAAATGTGTTGATGATCTTGGGCGTCCTTTTGATACGGTGCCGCCGGGGTTACTCTGGATGCTGATCCACCATTATCTCCGGCATA2280

M S P L T S *2281 ATGTCTCCATTAACGAGCTGATCTG 2305

FIG. 4. Nucleotide sequence of the CDT-active region and predicted amino acid sequence of cdtA, cdtB, and cdtC. The position of theamino acid sequence above or below the nucleotide sequence changes as the reading frame shifts. Underlined areas indicate sequencesidentified as putative translation starts (TS), ribosome-binding sites (RBS), promoters (-10 and -35 boxes), and the oligonucleotide probeused to screen the genomic library. The amino acids constituting putative signal sequences are highlighted with boldface type. The start sitesof exonuclease deletions pDS7.96, pDS7.84, pDS7.14, and pDS7.91 are marked with angle brackets. The first three deletions express CDTat a titer of 1:512, 1:512, and 1:2, respectively, and pDS7.91 is inactive.

cdtA polypeptide and fusion protein products. There is a

4-bp overlap between cdtA and cdtB and between cdtB andcdtC (Fig. 4). The nucleotide and predicted protein se-

quences of all three ORFs were compared with those in theGenBank data base (version 76), but no significant homologywith any previously reported genes or proteins was seen.

The region of DNA required for CDT activity was more

closely defined by the activity of several nested deletions.CDT activity is expressed at high titer by pDS7.96 (Fig. 4),which contains the entire 2,305-bp region, and remains highfor pDS7.84 (bp 156; Fig. 3 and 4). Activity falls to a titer of1:2 for pDS7.14 (bp 198; Fig. 4), and pDS7.91 (bp 258; Fig.

4) is inactive. Deletion of 215 bp from the 3' end of cdtC(pDS6.52, Fig. 3) also results in the loss of CDT activity.

In vitro transcription-translation. To confirm the presenceof three ORFs and identify protein products for each, we

performed in vitro transcription-translation studies with thefour nested deletions diagrammed in Fig. 3. Compared withthe vector alone, two new proteins were seen at 28 to 30 kDa(Fig. 5) with pDS7.84, which includes all three ORFs (Fig.5). The upper protein (ca. 30 kDa) is not expressed bypDS7.220, from which the 5' region of cdtA was deleted.Neither of these two proteins is expressed by pDS7.34, fromwhich cdtA and the 5' region of cdtB were deleted. The same

121

120

240

360

480

600

720

840

960

1080

1200

1320

1440

1560

1680

1681 1800

TATTnAATAnTTTTnnnnnnAATATAAAnAATTATATTTnAnTATnrTnTTTnTn&rTrTnnnAATA&TAArnAnrATTATAATAAArTTnTTTnTTTTTrTnnTTTrnrATTTrrTrATItIIWUA I AUI I I t UUUWAA I Pt I AAAt3AA I A A I I UAU I A I Ut. I Ul I t

TA&TrTTnTTnnTAATATTTTrnTTnrTTTnTTnTTTrTATTTTTTTATAAAnAAnAnnTnnTnrAnAnn&nnAAATArAr.Tir.r.ATAAAAAArTAATTnrATTTTTnTnrArArTTATAAI t3lA I I I I U I ttAtAt.I IAIAA

GATGGTTCAGGAGGCAGGGGCAGTACCAGCTTCTGCAACGTTGACTGAGCGAGAATTTAGCACTCCTGGTATTCCGATGAATGAGTATATCTGGAATACCGGAACCAATAGTCGCCCACA

INFECT. IMMUN.

E. COLI CYTOLETHAL DISTENDING TOXIN 249

1053

69.8

43.3

28.3

105

69.8

43t.30i$,

43.3

1 2 3 4 5 6 7 8 9 10FIG. 5. Autoradiograph of an 8% polyacrylamide-SDS gel con-

taining [35S]methionine-labeled protein products of in vitro tran-scription-translation. The reaction mixes in lanes 3 to 6 containedthe exonuclease deletions diagrammed in Fig. 3, and those in lanes7 to 10 contained BamHI clones of TnphoA insertions. Lanes: 1, noDNA; 2, pBSII; 3, pDS7.84; 4, pDS7.220; 5, pDS7.34; 6, pDS7.43;7, pDS10.3; 8, pDS16.3; 9, pDS21.3; 10, pDS22.3. Sizes are shownin kilodaltons.

two proteins were expressed by using cloned linear DNAfragments (BssHII fragment) from pDS7.84 and subse-quently not expressed in the same pattern by using linearfragments from pDS7.220 and pDS7.34 (data not shown).These two proteins are approximately the size of the pre-dicted protein products of cdtA (25.5 kDa) and cdtB (29.9kDa) except that cdtB was predicted to encode the larger ofthe two, which is the opposite of what was seen. No proteincorresponding to the 20.3-kDa predicted product of cdtCcould be identified either on the 8% polyacrylamide-SDS gelshown in Fig. 5 or with separate 12.5 and 15% polyacryl-amide-SDS gels (data not shown).To identify alkaline phosphatase fusion proteins, BamHI

clones of four TnphoA insertions (DS10, DS16, DS21, andDS22, Fig. 3) were used in the same in vitro transcription-translation system. As shown in Fig. 3, these include twoinsertions in cdtA flanking the first in-frame ATG and oneeach in cdtB and cdtC. There are BamHI sites -20 kbupstream of cdt and 288 bp upstream of the cdtC start codon.Compared with the vector alone, new fusion proteins wereseen in lanes 7 to 10 in Fig. 5. Each of these was theappropriate size for fusion proteins between the N-terminalportion of each ORF upstream of the TnphoA insertion andPhoA (alkaline phosphatase, 48.3 kDa). These products arepredicted to be 52.6, 68.2, 65.3, and 52.2 kDa for DS10,DS16, DS21, and DS22, respectively. Of particular impor-tance is that, with pDS22.3, a fusion product was identifiedfor cdtC, the only one of the three ORFs for which noproduct was found with the cloned gene alone.

Anti-alkaline phosphatase Western blot. With whole-celllysates of the E6468/62Smr cdt::TnphoA mutants, fusionproteins were identified for one or more insertions in each ofthe three ORFs (Fig. 6). The size of each corresponded tothe sizes of fusion proteins between the N-terminal portionsof CdtA, CdtB, or CdtC and alkaline phosphatase predictedby the location of each TnphoA insertion.

DISCUSSION

CDT activity is characterized by progressive cellulardistention and cytotoxicity in CHO cells and other cell lines.Using a CDT' E. coli strain from a child with diarrhea, wehave cloned the genes necessary for expression of CDT

28.3

1 2 3 4 5 6 7 8 9

FIG. 6. Anti-alkaline phosphatase Western blot with whole-celllysates from E6468/62Smr cdt::TnphoA mutants DS11, DS12, DS13,DS14, DS15, DS16, DS20, and DS22 and original strain E6468/62(lanes 1 to 9, respectively). Sizes are shown in kilodaltons.

activity into an E. coli K-12 laboratory strain. Inactivation ofCDT activity by TnphoA mutagenesis and exonucleasedigestion of the cloned genes defined a 2.3-kb region that isrequired for CDT activity.DNA sequence analysis of this 2.3-kb CDT-active region

revealed three ORFs, designated cdtA, cdtB, and cdtC. Thepresence of three ORFs is supported by in vitro transcrip-tion-translation and Western blot analyses, which identifiednative proteins and PhoA fusion proteins for cdtA and cdtB.For cdtC, a protein product was not visualized with thenative gene, but a PhoA fusion protein was visualized in bothsystems for the TnphoA insertion in this ORF. The discrep-ancy in size between the predicted protein products of cdtAand cdtB and the proteins visualized by in vitro transcrip-tion-translation studies is not readily explained. The pre-dicted isoelectric points of CdtA and CdtB (6.5 and 10.1,respectively) are very different, but this should not affecttheir mobilities in the denaturing gels used. It is possible thatthe translational start of cdtA is further upstream than theGTG that we have identified, but an appropriate site is notobvious. Identification of the actual translational start willrequire primer extension studies.The cdt genes are organized with 4-bp overlaps between

the stop codons and translational starts of cdtA and cdtB andbetween those of cdtB and cdtC. This arrangement is similarto the 4-bp overlap between ctxA4 and ctxB and between toxAand toxB, which code for the A and B subunits of choleratoxin and E. coli LT, respectively (14, 22).Each ORF begins with one or two charged amino acids

followed by a hydrophobic region and potential signal pep-tidase cleavage site, suggesting a signal sequence (Fig. 4).This is consistent with CDT activity being present in thesupernatant and with PhoA+ TnphoA insertions in each ofthe three ORFs.Using consensus sequences, we have assigned potential

promoter regions and ribosome-binding sites to each ORF(Fig. 4). Because we can visualize a polypeptide product foreach ORF in the in vitro transcription-translation systemwhen only the promoter for that ORF is present (exonucle-ase deletions pDS7.84 and pDS7.220 for cdtA and cdtB,respectively, and a BamHI clone of TnphoA insertion DS22for cdtC), it appears that each ORF has a functional pro-moter. However, conclusions about the cdt promoter(s)must await mRNA analysis.

VOL. 62, 1994

250 SCOTT AND KAPER

The translational start site for cdtA was a matter of someuncertainty and prompted us to sequence this region severaltimes with two different sequencing systems and a variety oftemplates and primers. The first in-frame ATG is at bp 396,and although this gives rise to an ORF with two in-frameTnphoA insertions, it does not explain the presence of sixPhoA+ TnphoA insertions upstream of this ATG. Further-more, the predicted protein product for this start site (19.1kDa) is smaller than those identified by either in vitrotranscription-translation or the alkaline phosphatase West-ern blot. A GTG translational start site at bp 201 fits bestwith the present data. A translational start at this site isupstream of, and in-frame with, all eight TnphoA insertionswithin cdtA. The predicted native protein product from thisstart site (25.5 kDa) and PhoA fusion products match wellwith those identified by in vitro transcription-translation andWestern blot. A protein beginning at the GTG has a charac-teristic signal sequence consistent with the extracellularlocation of CDT activity and active TnphoA insertions.There are putative ribosome-binding sites and promoter

sequences upstream of the GTG that are not found upstreamof the first ATG. A promoter region upstream of the GTG isalso consistent with the dramatic drop in CDT activity seenwith pDS7.14 (Fig. 4), which produces fully active toxin butat a greatly reduced titer of 1:2, compared with 1:512 forclones that contain the putative promoter region. Sequenceanalysis with the program TESTCODE shows that codonusage becomes very nonrandom at approximately bp 200(data not shown), which also supports GTG as the transla-tional start. Although ATG is by far the most commontranslational start in E. coli, GTG is the second mostcommon, accounting for 6.8% of the translational starts in alarge series of well-characterized E. coli genes (19).We have not found any significant homology between the

cdt genes and any previously reported genes at either thenucleotide or the amino acid level. A search for proteinmotifs with the program MOTIFS also did not yield anymatches. Thus, the sequence analysis yields no clues as tothe possible mechanism of action of this toxin.

It is clear that cdtC is necessary for activity, because aTnphoA insertion in cdtC abolishes activity and an exonu-clease deletion (pDS6.52 in Fig. 3) which contains all of cdtAand cdtB but does not contain the terminal 215 bp at the 3'end of cdtC is inactive. Because of possible polar effects, wecannot be absolutely sure from either transposon insertionsor exonuclease deletions that either cdtA or cdtB is neces-sary for activity. However, since cdtB is still expressed witha cdtA deletion (pDS7.220 in Fig. 3) but no activity ispresent, it is likely that both cdtA and cdtB are necessary.Although the E. coli isolate used in this study was from a

patient with acute diarrhea, there are no strong epidemio-logic data linking CDT-producing E. coli to diarrhea, andCDT has not yet been shown to be a virulence factor. Thecloning and sequencing of the genes encoding CDT will nowallow us to pursue epidemiological and molecular pathogen-esis studies to establish the significance of CDT in disease.

ACKNOWLEDGMENTSThese investigations were supported by grant AI21657 from the

National Institute of Allergy and Infectious Diseases. While thiswork was performed, Daniel A. Scott, U.S. Navy, was an infectiousdiseases fellow at the University of Maryland School of Medicine ina full-time outservice training position supported by the U.S. NavyHealth Sciences Education and Training Command.We thank Patricia Guerry for assistance with automated DNA

sequencing and Stephen Savarino for review of the manuscript.

REFERENCES1. Anderson, J. D., A. J. MacNab, W. R. Gransden, S. M. Damm,W. M. Johnson, and H. Lior. 1987. Gastroenteritis and enceph-alopathy associated with a strain of Escherichia coli 055:K59:H4 that produced a cytolethal distending toxin. Pediatr.Infect. Dis. J. 6:1135-1136.

2. Bisicchia, R., R. Ciammarughi, A. Caprioli, V. Falbo, and F. M.Ruggeri. 1985. Toxin production and haemagglutination instrains of Escherichia coli from diarrhoea in Brescia, Italy. J.Hyg. (London) 95:353-361.

3. Bouzari, S., and A. Varghese. 1990. Cytolethal distending toxin(CLDT) production by enteropathogenic Escherichia coli(EPEC). FEMS Microbiol. Lett. 59:193-198.

4. Caprioli, A., V. Falbo, F. M. Ruggeri, L. Baldassarri, R.Bisicchia, G. Ippolito, E. Romoli, and G. Donelli. 1987. Cyto-toxic necrotizing factor production by hemolytic strains ofEscherichia coli causing extraintestinal infections. J. Clin. Mi-crobiol. 25:146-149.

5. Donta, S. T., H. W. Moon, and S. C. Whipp. 1974. Detection ofheat-labile Escherichia coli enterotoxin with the use of adrenalcells in tissue culture. Science 183:334-336.

6. Genetics Computer Group. 1991. Program manual for the GCGpackage. Genetics Computer Group, Inc., Madison, Wis.

7. Guerrant, R. L., L. L. Brunton, T. C. Schnaitman, L. I. Rebhun,and A. G. Gilman. 1974. Cyclic adenosine monophosphate andalteration of Chinese hamster ovary cell morphology: a rapid,sensitive in vitro assay for the enterotoxins of Vibio choleraeand Escherichia coli. Infect. Immun. 10:320-327.

8. Honda, T., M. Arita, Y. Takeda, and T. Miwatani. 1982. Furtherevaluation of the Biken test (modified Elek test) for detection ofenterotoxigenic Escherichia coli producing heat-labile entero-toxin and application of the test to sampling of heat-stableenterotoxin. J. Clin. Microbiol. 16:60-62.

9. Johnson, W. M., and H. Lior. 1987. Production of Shiga toxinand a cytolethal distending toxin (CLDT) by serogroups ofShigella spp. FEMS Microbiol. Lett. 48:235-238.

10. Johnson, W. M., and H. Lior. 1987. Response of Chinesehamster ovary cells to a cytolethal distending toxin (CDT) ofEscherichia coli and possible misinterpretation as heat-labile(LT) enterotoxin. FEMS Microbiol. Lett. 43:19-23.

11. Johnson, W. M., and H. Lior. 1988. A new heat-labile cytolethaldistending toxin (CLDT) produced by Campylobacter spp.Microb. Pathogen. 4:115-126.

12. Johnson, W. M., and H. Lior. 1988. A new heat-labile cytolethaldistending toxin (CLDT) produced by Escherichia coli isolatesfrom clinical material. Microb. Pathogen. 4:103-113.

13. Konowalchuk, J., J. I. Speirs, and S. Stavric. 1977. Veroresponse to a cytotoxin of Escherichia coli. Infect. Immun.18:775-779.

14. Lockman, H., and J. B. Kaper. 1983. Nucleotide sequenceanalysis of the A2 and B subunits of Vibrio cholerae entero-toxin. J. Biol. Chem. 258:13722-13726.

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

16. Morris, G. K., M. H. Merson, D. A. Sack, J. G. Wells, W. T.Martin, W. E. Dewitt, J. C. Feeley, R. B. Sack, and D. M.Bessudo. 1976. Laboratory investigation of diarrhea in travelersto Mexico: evaluation of methods for detecting enterotoxigenicEscherichia coli. J. Clin. Microbiol. 3:486-495.

17. Nataro, J. P., J. B. Kaper, R. Robins Browne, V. Prado, P. Vial,and M. M. Levine. 1987. Patterns of adherence of diarrheagenicEscherichia coli to HEp-2 cells. Pediatr. Infect. Dis. J. 6:829-831.

18. O'Brien, A. D., G. D. LaVeck, M. R. Thompson, and S. B.Formal. 1982. Production of Shigella dysenteriae type 1-likecytotoxin by Escherichia coli. J. Infect. Dis. 146:763-769.

19. Rudd, K. E., and T. D. Schneider. 1992. Compilation of E. coliribosome-binding sites, p. 17.12-17.45. In J. H. Miller (ed.), Ashort course in bacterial genetics: a laboratory manual andhandbook for Escherichia coli and related bacteria. Cold SpringHarbor Laboratory Press, Cold Spring Harbor, N.Y.

20. Southern, E. M. 1975. Detection of specific sequences among

INFECT. IMMUN.

E. COLI CYTOLETHAL DISTENDING TOXIN 251

DNA fragments separated by gel electrophoresis. J. Mol. Biol.98:503-517.

21. Takeda, Y., and J. R. Murphy. 1978. Bacteriophage conversionof heat-labile enterotoxin in Escherichia coli. J. Bacteriol.133:172-177.

22. Yamamoto, T., T. Tamura, T. Yokota, and T. Takano. 1982.

Overlapping genes in the heat-labile enterotoxin operon origi-nating from Escherichia coli human strain. Mol. Gen. Genet.188:356-359.

23. Yan, Y. 1990. A simple and inexpensive method for dryinghigh-percentage polyacrylamide gradient gels. BioTechniques8:381-382.

VOL. 62, 1994