kfu/PTS Region and Tissue-Invasive K. pneumoniae • JID 2005:192 (1 July) • 117

M A J O R A R T I C L E

Genomic Heterogeneity in Klebsiella pneumoniaeStrains Is Associated with Primary Pyogenic LiverAbscess and Metastatic Infection

Li-Chen Ma,1 Chi-Tai Fang,2 Cha-Ze Lee,2 Chia-Tung Shun,3 and Jin-Town Wang1,2

1Department of Microbiology, National Taiwan University College of Medicine, and Departments of 2Internal Medicine and 3Forensic Medicine,National Taiwan University Hospital, Taipei, Taiwan

Background. Primary pyogenic liver abscess (PLA) with septic complication by Klebsiella pneumoniae is anemerging infectious disease.

Methods and results. Using DNA microarray hybridization, we identified a 20-kb chromosomal region thatcontained 15 open-reading frames (ORFs), including an iron-uptake system (kfu), a phosphoenolpyruvate sugarphosphotransferase system (PTS), and 6 unknown ORFs. The region was more prevalent among tissue-invasivestrains (35/46) than among noninvasive strains (19/98) ( , x2 test). To test the role played by this regionP ! .0001in pathogenesis, 3 different deletion mutants (NTUH-K2044 [Dkfu], K2044 [DORF7–9], and K2044 [DPTS]) wereconstructed. Only the DkfuABC mutants showed decreased virulence in mice, compared with the wild-type strain.An in vitro assay confirmed the involvement of kfu in iron acquisition. There was a high correlation rate (85%)between the kfu/PTS region and 2 tissue invasion–associated chromosomal regions (allS and magA). Moreover,all 3 regions were present in strains that caused PLA plus endophthalmitis or meningitis.

Conclusion. Our results suggest that chromosomal heterogeneity is present in tissue-invasive K. pneumoniaestrains. A genotype containing all 3 regions is strongly associated with PLA and metastatic infection. These regionsmay serve as convenient markers for the rapid diagnosis of emergent tissue-invasive strains.

Klebsiella pneumoniae is an important hospital-acquired

pathogen that is a frequent cause of urinary tract in-

fection, septicemia, and pneumonia in immunocom-

promised individuals; it is also an important pathogen

with respect to community-acquired infectious dis-

eases, such as community-acquired pneumonia [1]. In

Taiwan, K. pneumoniae–associated primary pyogenic

liver abscess (PLA) has recently become an important

emerging infectious disease [2–8]. This disease is also

a global concern, as is attested by reports from North

America [9], Europe [10, 11], and Asia [12–14].

Fulminant tissue-invasive K. pneumoniae infections

can attack healthy persons who have no history of he-

Received 3 November 2004; accepted 5 February 2005; electronically published25 May 2005.

Financial support: Taiwan National Science Council (grant 93-3112-B-002-002).Reprints or correspondence: Dr. Jin-Town Wang, Dept. of Microbiology, National

Taiwan University College of Medicine, 1, Sec. 1, Jen-Ai Rd., Taipei, Taiwan (wangjt@ntu.edu.tw).

The Journal of Infectious Diseases 2005; 192:117–28� 2005 by the Infectious Diseases Society of America. All rights reserved.0022-1899/2005/19201-0017$15.00

patobiliary disease, and only 50% of patients have a pre-

disposing condition, such as diabetes mellitus [2–8]. PLA

is manifest with other septic metastatic lesions, including

pyogenic meningitis and endophthalmitis, in 10%–12%

of patients [3–5, 7]. Despite aggressive antibacterial strat-

egies, significant morbidity and mortality still exists, es-

pecially in those with diabetes mellitus [2–8].

Using transposon mutagenesis and full genome ex-

pression analysis, we recently identified genomic seg-

ments of 33 kb (magA) and 22 kb (allS), which were

absent in the genome of MGH 78578 and present in

most tissue-invasive strains from patients with PLA,

meningitis, or endophthalmitis [15, 16]. Therefore, it

is probable that these tissue-invasive strains might rep-

resent a specific genotype that harbors specific regions

involved in pathogenesis and prevalence.

DNA microarray technology provides a useful tool

for assessment of the differences and changes in bac-

terial genomes [17, 18]. Therefore, we used this tech-

nology to compare genomic variations between tissue-

invasive strains and noninvasive strains.

118 • JID 2005:192 (1 July) • Ma et al.

Table 1. Prevalences of 3 chromosomal fragments among Klebsiella pneumoniae strains used in the present study.

Strain typeClinical diagnosis

of patient OriginNo.

of isolates

No. of isolates positive for region

20-kbregion

(kfu/PTS)Large

plasmid

22-kbregion(allS)

33-kbregion(magA)

Tissue-invasive strains (n p 46) PLA NTUH (Taiwan) 33 24 30 24 26PLA plus endopthalmitis NTUH (Taiwan) 8 8 7 8 8Menignitis NTUH (Taiwan) 4 2 4 0 2PLA plus meningitis NTUH (Taiwan) 1 1 1 1 1

Noninvasive strains (n p 98) Disease other than PLA NTUH (Taiwan) 33 9 13 2 2Disease other than PLA FEMH (Taiwan) 32 0 30 0 0Disease other than PLA ECKH (Taiwan) 11 3 11 1 1Disease other than PLA ATCC (USA) 22 7 22 0 3

NOTE. ECKH, En Chu Kong Hospital; FEMH, Far Eastern Memorial Hospital; NTUH, National Taiwan University Hospital; PLA, primary pyogenic liver abscess;PTS, phosphoenolpyruvate sugar phosphotransferase system.

PATIENTS, MATERIALS, AND METHODS

K. pneumoniae strains and culture conditions. K. pneumoni-

ae isolates were obtained consecutively from cultures of blood

samples obtained from patients between 1997 and 2003. A total

of 46 tissue-invasive K. pneumoniae strains were obtained from

patients with PLA or meningitis at National Taiwan University

Hospital (NTUH). Of these, 22 were isolated from patients with

diabetes mellitus, and 24 were isolated from patients who had

previously been healthy.

For comparison, 98 noninvasive strains were obtained from

patients with sepsis but without any tissue-invasive disease

(such as PLA, meningitis, and endophthalmitis). Of these

strains, 33 were obtained from patients at NTUH, 32 were

obtained from patients at Far Eastern Memorial Hospital

(FEMH; Banciao, Taiwan), and 11 were obtained from pa-

tients at En Chu Kong Hospital (ECKH; Sansia, Taiwan). Al-

so, 22 strains from the United States, including MGH 78578,

were purchased from the American Type Culture Collection.

MGH 78578 was chosen for full genome sequencing at Wash-

ington University (St. Louis, MO; available at: http://www

.genome.wustl.edu/projects/bacterial/). Information pertain-

ing to the 144 strains is listed in table 1. All strains were

identified and cultured according to standard clinical micro-

biology methods [19].

Microarray construction and hybridization. The genomic

library was constructed from a clinical isolate, NTUH-K2044,

obtained from a patient with PLA plus meningitis in a phag-

emid [20]. DNA fragments in phagemids of K. pneumoniae

were amplified by polymerase chain reaction (PCR) with prim-

ers in vectors and were spotted onto a nylon membrane (Roche)

by a computer-controlled XYZ translation system (PM500;

Newport) [21].

Probe preparation and hybridization. Genomic DNA

from 4 PLA strains (NTUH-K2044, A1208, A3021, and

A5011) and 3 noninvasive strains (N3423, N3529, and N5322)

were extracted and were labeled with biotin-16-dUTP (Perkin

Elmer) by a randomly primed polymerization reaction. The

microarray membrane was prehybridized in 2 mL of hybrid-

ization buffer for 4 h at 65�C and was hybridized for 16 h at

68�C. The membrane was washed twice with 2� standard

saline citrate (SSC) containing 0.1% SDS for 5 min at room

temperature and was then washed 3 times with 0.1� SSC

containing 0.1% SDS for 15 min at 65�C each time. Color-

imetry detection and image analysis were then performed as

described elsewhere [22].

Recombinant DNA techniques and plasmids. K. pneu-

moniae deletion mutants were constructed by replacing the

deletion region with a kanamycin (Km) cassette in a double-

crossover integration of chromosomal DNA. All primers used

in the present study are listed in table 2. To generate the

NTUH-K2044 (Dkfu) mutant, a PCR fragment amplified by

primers PVAR KO-1 and PVAR KO-2 was cloned into a

pGEM-T Easy vector (Promega). PVAR KO-1(iPCR) and

PVAR KO-2(iPCR) primers were used for inverse PCR with

plaque-forming-unit polymerase (MBI Fermentas). A blunt-

end Km gene was phosphorylated by use of a polynucleo-

tide kinase (New England Biolabs) and was ligated to the

inverse PCR product to generate the kfu disruption fragment.

We cloned the fragment into a pUT suicide vector [23] con-

taining an EcoRI site and added a second marker (spection-

mycin; Spe) to the ApaLI site. pUT-(Dkfu) was transformed

into wild-type NTUH-K2044 to generate a kfu deletion mu-

tant by conjugation. Deletion clones were selected by Kmr

and Spes. The same procedures were used for open-reading

frames (ORFs) 7–9 and phosphoenolpyruvate sugar phospho-

transferase system (PTS) deletion constructs. All of the deletion

mutants were confirmed by PCR with multiple primer pairs

and sequence determination.

Transcomplementation of Escherichia coli H1443. An

aroB� E. coli H1443 strain is deficient of siderophore for growth

kfu/PTS Region and Tissue-Invasive K. pneumoniae • JID 2005:192 (1 July) • 119

Table 2. Polymerase chain reaction primers used in the present study.

Primer Sequence Description Reference

PVAR KO-1 5′-GCAGCAGATGAATATTCTGG-3′ kfu deletion construct Present studyPVAR KO-1(iPCR) 5′-TCTGGGTGCAGAACCAAATG-3′ kfu deletion construct Present studyPVAR KO-2 5′-TTCCGACGCCAATGCTGATC-3′ kfu deletion construct Present studyPVAR KO-2(iPCR) 5′-ATGGAGTGGTAGTACGTTGG-3′ kfu deletion construct Present studyPVAR KO-3 5′-GGTCCTCGATAATCTCAATC-3′ ORF7–9 deletion construct Present studyPVAR KO-3(iPCR) 5′-TGACCGTCAGTCTCACCATC-3′ ORF7–9 deletion construct Present studyPVAR KO-4 5′-GACGTTGCCGAATGTCCATC-3′ ORF7–9 deletion construct Present studyPVAR KO-4(iPCR) 5′-GCAGCAGGAAGCTCTTGATG-3′ ORF7–9 deletion construct Present studyPVAR g-F 5′-ATTTGCGGCCAATCAGCGTC-3′ PTS deletion construct Present studyPVAR g-R 5′-CGACGCCGATATCAAAGAGT-3′ PTS deletion construct Present studyPVAR h-F 5′-AACCCGACGCTGTTTCATGC-3′ PTS deletion construct Present studyPVAR h-R 5′-GGAACTGGATAAAACGCATC-3′ PTS deletion construct Present studyPVAR d-F 5′-GCGTGCTAAGGTAATTAGGC-3′ kfu operon analysis Present studyPVAR d-R 5′-CGAAAATCTGGTGAAGTCGT-3′ kfu operon analysis Present study6C12E1-Sp63 5′-ACTTCAAACACCAGGATCCG-3′ kfu operon analysis Present study6C12E1-T73 5′-AGGCGTAGCTCTGGACAAAG-3′ kfu operon analysis Present studykfuB-F1179 5′-GAAGTGACGCTGTTTCTGGC-3′ kfu operon analysis Present studykfuC-R649 5′-TTTCGTGTGGCCAGTGACTC-3′ kfu operon analysis Present studyPVAR-F 5′-GCTTTTACCTGTAGCCGTCC-3′ 20-kb region (PVAR) prevalence analysisa Present studyPVAR-R 5′-GCAGGATATGACCATTTTGC-3′ 20-kb region (PVAR) prevalence analysisa Present study7C4-T71 5′-ATCCTGGAAGCTCACCGTTC-3′ 20-kb region (PVAR) prevalence analysisb Present study26D6-T32 5′-ATTTCCACGCGGATACCGTC-3′ 20-kb region (PVAR) prevalence analysisb Present study26A1-T34 5′-ACGTTCGCAGGTTAAAGCTG-3′ Large plasmid prevalence analysis Present study26A1-T73 5′-CCATGTAGCTGCAGATACTG-3′ Large plasmid prevalence analysis Present study28G9-T33 5′-AAGCGTCTGGTTTCCTGGTG-3′ Large plasmid prevalence analysis Present study28G9-T72 5′-TATTTGTCGGCAATTACGGC-3′ Large plasmid prevalence analysis Present study2801769-F 5′-GTACACCAGCGGCCTGTTCTGGGCGCAATC-3′ 22-kb region (all) prevalence analysisa [15]2803831-R 5′-CAGAACAGCATTTACTATGATGTGGTG-3′ 22-kb region (all) prevalence analysisa [15]10E4-2-5F 5′-AGTCGGCCTGGGGTTTAAGG-3′ 22-kb region (all) prevalence analysisb [15]10E4-2-475R 5′-CAGTCAACGTGGCGATTCGC-3′ 22-kb region (all) prevalence analysisb [15]TF-F 5′-GCGATACTGGTAACTGGGTT-3′ 33-kb region (magA) prevalence analysisc [16]M5YM-R 5′-GTCACTTTCTTGGCTTCCTG-3′ 33-kb region (magA) prevalence analysisc [16]M3YM-F 5′-AACCCATCGCATTGGACAGA-3′ 33-kb region (magA) prevalence analysisd [16]M3M-R 5′-ACCTTTTACTCGCGCACCAA-3′ 33-kb region (magA) prevalence analysisd [16]1040-F 5′-CGTTACGAACTTGAACGAGC-3′ 33-kb region (magA) prevalence analysisb [16]936-R 5′-GCAATGGCCATTTGCGTTAG-3′ 33-kb region (magA) prevalence analysisb [16]

NOTE. ORF, open-reading frame; PTS, phosphoenolpyruvate sugar phosphotransferase system.a Outside primers of the region.b Inside primer of the region.c Left junction primer of the 33-kb (magA) region.d Right junction primer of the 33-kb (magA) region.

when cultured in medium containing iron chelator 2,2′-dipyr-

idyl (Sigma) [24, 25]. Transformed E. coli H1443 with plasmid

pBR322::kfuABC, plasmid pSZ1 [25], or pBR322 were grown

overnight at 37�C in Luria-Bertani (LB) broth. Each culture

was diluted in fresh LB broth supplemented with 0.1, 0.2, or

0.5 mmol/L 2,2′-dipyridyl. The growth rate was monitored

spectrophotometrically at 620 nm.

RNA isolation and reverse-transcription (RT) PCR. Total

RNA was isolated from K. pneumoniae cultured at the expo-

nential phase of growth, as described elsewhere [15]. For each

RT, 5 mg of RNA was used with 2 pmol of RT-PCR primer and

200 U of M-MLV reverse transcriptase (Invitrogen). Reaction

mixtures without reverse transcriptase were included as nega-

tive controls. PCR was performed with 10% of each RT reac-

tion volume under 30 cycles of amplification.

Murine experiments. Initially, we infected BALB/cByl

mice intraperitoneally (ip). However, later experiments

showed that intragastric (ig) inoculation had a higher sen-

sitivity to differential virulence [15, 16]. Mice were admin-

istered ig [26] either wild-type NTUH-K2044, MGH 78578,

NTUH-K2044 (Dkfu), K2044 (DORF7–9), or K2044 (DPTS)

(103–106 cfu; 4 mice for each dose). For ig inoculation, we

carefully slipped a 1-mm polyethylene flexible tube (Becton

Dickinson) past the pharynx into the stomach (∼5 cm of

intubation) of each mouse and delivered 0.2 mL (103–106 cfu)

of bacterial suspension. Anesthesia was not used for this pro-

120 • JID 2005:192 (1 July) • Ma et al.

Figure 1. DNA hybridization and colorimetric detection of 4 tissue-invasive and 3 noninvasive Klebsiella pneumoniae strains on a microarray. Circlesshow decreased hybridization signals in noninvasive strains.

Table 3. Sequences of 12 clones with decreased hybridization signals in noninvasive strains.

Clone no. Predicted protein (GenBank accession no.)

Length of comparableamino acid sequence

(% identity)Comparison

with MGH 78578

1 Ferric enterobactin (enterochelin) transport (NP 286317.1) 333 (86) MGH contig 123Hypothetical protein …Hypothetical membrane protein P43 (NP752609.1) 406 (85)Ferrienterobactin-binding periplasmic protein precursor (NP 752610.1) 307 (84)

2 Class II aldolase/adducin domain protein (NP745015.1) 196 (58) MGH contig 112Putative 2-hydroxyacid dehydrogenase (NP879163.1) 269 (47)Probable short-chain dehydrogenase (NP250035.1) 263 (60)Conserved hypothetical protein (NP929872.1) 191 (44)Probable extracellular solute-binding protein (NP406806.1) 250 (63)

3∼10a Iron-uptake system … No similarityPTS …Six uncharacterized hypothetical proteins …

11b CobW (AAR07790) 406 (100) No similarityConserved hypothetical protein (AAR07791) 68 (100)Ferrous ion transporter protein (AAR07792) 549 (100)

12b Putative partitioning (parA) (AAR07854) 296 (100) No similarityPutative partitioning (parB) (AAR07855) 343 (100)Conserved hypothetical protein (AAR07856) 147 (99)

NOTE. PTS, phosphoenolpyruvate sugar phosphotransferase system.a Sequence could be connected to ∼9 kb; see table 4 for further analysis.b Proved, by BLAST analysis, to be localized in the large plasmid of Klebsiella pneumoniae CG43; the position of clone 11 was between 128577 and 131336,

and the position of clone 12 was between 188902 and 191960.

cedure. Mice were monitored for 4 weeks; upon death, the

liver and brain were removed, and histopathological exami-

nation was conducted. Surviving mice were killed at the end

of the 4 weeks. The LD50 was calculated as described elsewhere

[27]. Survival was analyzed by Kaplan-Meier analysis with a

log-rank test; was considered to be statisticallyP ! .05

significant.

Slot-blot hybridization. Ten micrograms of genomic DNA

from K. pneumoniae strains were vacuum blotted onto nylon

membranes. Hybridization was performed for 16 h at 68�C

kfu/PTS Region and Tissue-Invasive K. pneumoniae • JID 2005:192 (1 July) • 121

Figure 2. Genetic organization of the 20-kb kfu/phosphoenolpyruvate sugar phosphotransferase system (PTS) chromosomal region. A 5292-bp fragmentin MGH 78578 was replaced by a 19,640-bp fragment in NTUH-K2044. The locations and orientation of the open-reading frames (ORFs) described inthe present study are indicated by arrows.

Table 4. Characteristics of the open-reading frames (ORFs) and deduced amino acid sequences present in the 20-kb DNA fragment.

ORF no.,name

Productsize, aa ORF product exhibits homology to: Source

Identity,%

GenBankaccession no.

01, ORF1 808 Glycogen phosphorylase Thermosynechococcus elongatus BP-1 50 NP681571.102, ORF2 115 Anti–anti-j factor Rhodobacter sphaeroides 40 ZP00006519.103, ORF3 511 Putative protease Shigella flexneri 2a strain 301 50 NP707066.104, kfuA 338 Iron(III)-binding periplasmic protein Yersinia pestis 78 NP406453.105, kfuB 524 Iron(III)-transport system permease Y. pestis 68 NP406455.106, kfuC 342 Iron(III)-transport ATP-binding protein Y. pestis 58 NP103033.107, ORF7 306 … … … …08, yijO 277 Putative ARAC-type regulatory protein Escherichia coli O157:H7 57 NP290591.109, mmcH 297 Biosynthesis of mitomycin antibiotics Streptomyces lavendulae 29 AF127374.110, frwD 113 PTS fructose-like IIB component 2 Salmonella typhimurium LT2 59 NP462996.111, pflC 328 Pyruvate formate-lyase activating enzyme S. typhimurium LT2 63 NP462995.112, pflD 769 Formate acetyltransferase 2 S. flexneri 2a strain 301 85 NP838933.113, frwB 106 PTS fructose-like IIB component 1 S. typhimurium LT2 92 NP462993.114, frwC 389 PTS fructose-like IIC component S. flexneri 2a strain 301 85 NP709749.115, ptsA 833 Phosphoenolpyruvate-protein phosphotransferase E. coli CFT073 78 833 79 NP756760.1

NOTE. PTS, phosphoenolpyruvate sugar phosphotransferase system.

with each biotin-labeled probe generated by PCR. The gene

encoding 23S rRNA was used as a positive control. Detection

was performed by use of the Southern-Light Chemiluminescent

detection system (Tropix), in accordance with the manufac-

turer’s instructions.

RESULTS

DNA microarray hybridization. A total of 3146 PCR clones

were randomly selected for the microarray. The coverage rate

was ∼88%, according to the formula N p ln (1 � P)/ ln (1 � f )

122 • JID 2005:192 (1 July) • Ma et al.

Figure 3. Polymerase chain reaction primer alignments used in the present study to analyze the prevalences of the kfu/phosphoenolpyruvate sugarphosphotransferase system (PTS) chromosomal region.

Figure 4. A, Schematic diagram of the deletion constructs. Deletion regions (shown by dashed lines) were replaced by a kanamycin (Km) cassette.B and C, Survival of BALB/cByl mice inoculated with various deletion mutants at a dose of 104 cfu (B) or 105 cfu (C). �, wild-type strain; �, NTUH-K2044 (Dkfu); �, NTUH-K2044 (DORF7–9); �, NTUH-K2044 (DPTS); ORF, open-reading frame.

[28]. To test the redundancy of the library, 798 of the 3146

clones were randomly selected for sequencing. These clones

contained 678 distinct sequences, representing a redundancy

rate of 15%.

Comparison of 4 tissue-invasive strains (A1208, A3021,

A5011, and NTUH-K2044) and 3 noninvasive strains (N3423,

N3529, and N5322) revealed 12 clones with significantly de-

creased hybridization signals (defined as �3 SDs of the mean

ratio) in the noninvasive strains (figure 1 and table 3). These

12 clones were sequenced and then compared with the 10�

shotgun sequences of K. pneumoniae MGH 78578 (available at:

http://www.genome.wustl.edu/projects/bacterial/kpneumoniae/).

Of them, there were 10 clones (clones 3–12) that bore no sim-

ilarity in sequence to MGH 78578. Of these 10 clones, 8 over-

lapped and extended to encompass an ∼9-kb fragment, whereas

the other 2 matched to a 200-kb plasmid (pLVPK) of K. pneu-

moniae CG43 [29] (hereafter, “the large plasmid”).

Sequencing of the flanking regions of the 8 clones in NTUH-

K2044. The flanking regions of the ∼9-kb fragment were se-

quenced from the genomic library until both ends matched in

MGH 78578. A 19,640-bp fragment (GenBank accession num-

ber AB115591) that was obtained replaced a 5292-bp fragment

in the genome of MGH 78578 (figure 2). BLAST searches re-

vealed that this fragment contained 15 ORFs (figure 2 and table

4). The common flanking regions of NTUH-K2044 and MGH

78578 respectively contained oppA and an ORF encoding a pu-

tative diogenase b subunit. The overall GC content of this region

was 56.9%, similar to the 57.7% GC content of the remainder

of the genome.

Annotation of this fragment revealed a modular structure

kfu/PTS Region and Tissue-Invasive K. pneumoniae • JID 2005:192 (1 July) • 123

Figure 5. Histological examination of the livers and brains from infected mice. The wild-type strain and deletion mutants were administered to micevia intragastric inoculation at a dose of 104 cfu. Half of the mice inoculated with 104 cfu of the wild-type strain, NTUH-K2044 (DORF7–9), or K2044(DPTS) were dead within 2 weeks; however, all of the mice inoculated with 104 cfu of K2044 (Dkfu) or MGH 78578 remained healthy until they werekilled at the end of 4 weeks. The livers from the mice inoculated with the wild-type strain (a), K2044 (DORF7–9) (b), K2044 (DPTS) (c), or K2044(Dkfu) (d) were examined. The first 3 had severe liver abscesses, whereas the livers from the mice inoculated with K2044 (Dkfu) had normal histologicalstructure. Brain abscesses could also be found in the mice inoculated with the wild-type strain (e), whereas the brains from the mice inoculated withK2044 (Dkfu) had normal histological structure (f). The stain used was hematoxylin-eosin. Original magnification, �100.

with 4 regions (figure 2 and table 4). Region 1 (293–4704)

carried ORF1–3. These ORFs showed homology to proteins

involved in glycogen phosphorylase (ORF1), an anti–anti-j fac-

tor (ORF2), and a putative protease (ORF3). Region 2 (5104–

8743) harbored ORF4–6. These ORFs exhibited high homology

to the bacterial ferric iron–uptake system, which is a bacterial

ABC iron transport system [30]. They include Sfu of Serratia

marcescens [25, 31], Hit of Haemophilus influenzae [32], Yfu

of Yersinia pestis [33], Afu of Actinobacillus pleuropneumoniae

[34], and Fbp of Neisseria gonorrhoeae [35]. Therefore, these

ORFs were putatively designated kfuA, kfuB, and kfuC, respec-

tively (“kfu” stands for Klebsiella ferric iron uptake). Region 3

(8990–11,786) contained ORF7–9. ORF7 displayed no signif-

icant homology to any sequences in the database. However,

ORF8 revealed similarity to the gene yijO, which encodes a

putative ARAC-type regulatory protein in E. coli. ORF9 showed

homology to a protein involved in the biosynthesis of mito-

mycin [36]. Region 4 (11,786–19,553) extended from ORF10

to ORF15. These 6 ORFs showed high homology to the PTS.

PTS catalyzes translocation with concomitant phosphorylation

of sugars and hexitols and regulates metabolism in response to

the availability of carbohydrates [37]. Some PTS proteins have

been tentatively linked with bacterial virulence [38–40]. This

20-kb region was designated “the kfu/PTS region.”

Prevalences of the kfu/PTS region and the large plasmid

among K. pneumoniae strains. Because sequences of clones

1 and 2 were present in the noninvasive strain MGH 78578

(table 3), we studied the prevalences of the kfu/PTS region and

the large plasmid among the clinical isolates to find regions

that are specific to tissue-invasive strains. Genomic DNA ex-

tracted from 46 tissue-invasive and 98 noninvasive strains was

used for PCR analysis. We used the inside and outside primers

to detect the presence of the kfu/PTS region (figure 3), on the

following basis: If a clinical strain contained the region being

tested, then the primer pairs for the flanking regions (outside

primers) should fail to amplify the fragments; however, the

inside primers should amplify products with a predicted length.

Conversely, if the clinical strain did not contain the region,

124 • JID 2005:192 (1 July) • Ma et al.

Figure 6. A, Promoter sequence of the Klebsiella pneumoniae kfu operon. The potential ribosome binding site (RBS) is underscored. The box indicatesthe conserved Fur-binding sequences (Fur box) [41]. The transcription initiation start site (predicted by the Berkeley Drosophila Genome Project witha score of 1.0 [available at: http://www.fruitfly.org/seq_tools/promoter.html/]) and translation initiation condon are indicated by boldface. B, Geneticorganization of the kfu locus. The thick black line represents chromosomal DNA, and the open arrows represent kfu open-reading frames. Ethidiumbromine–stained agarose gels show transcriptional analysis of the kfu locus. Each panel contains products with the same primer pairs for polymerasechain reaction (PCR) with K. pneumoniae chromosomal DNA as the template (lane 1), for reverse transcription PCR with K. pneumoniae RNA as thetemplate (lane 2), and for the same PCRs containing no reverse transcriptase, which generate no product (lane 3). Bars represent the correspondingtarget mRNA segments for each primer pairs used; arrows indicate the expected size of the RT-PCR products. C, Growth experiments with the wild-type strain, the NTUH-K2044 (Dkfu) deletion mutant, MGH 78578, and a noninvasive strain, N5322, in Luria-Bertani (LB) broth containing 0.2 mmol/L 2,2′-dipyridyl (DIP). D, Functional assay of the kfuABC operon of K. pneumoniae in LB broth containing 0.2 mmol/L DIP. Growth was monitoredspectrophotometrically at 620 nm. The cultures were inoculated with fresh precultures to an OD620 nm of 0.01. The optical-density values are the meansof 3 experiments; error bars indicate SDs. H1443 pBR322 is the Escherichia coli aroB mutant strain transformed with pBR322 vector only; H1443 pSZ1is the E. coli aroB mutant strain transformed with pBR322 containing an intact sfuABC as a positive control; and H1443 pBR322::kfu is the E. coliaroB mutant strain transformed with pBR322 containing an intact kfuABC.

then PCR with the outside primer pairs would be positive,

whereas PCR with the inside primers would be negative. We

also used the inside sequences of clones 11 and 12 to detect

the existence of the large plasmid.

The prevalence of the kfu/PTS region was significantly higher

in tissue-invasive strains than in noninvasive strains (35/46 vs.

19/98; , x2 test) (table 1). However, no significantP p .0001

correlation was observed between the large plasmid and clin-

ically invasive disease (42/46 vs. 76/98; , x2 test).P p .077

Analysis of deletion mutants. Three deletion mutants

(NTUH-K2044 [Dkfu], K2044 [DORF7–9], and K2044 [DPTS])

were constructed by use of a suicide vector (figure 4A). The

kfu/PTS Region and Tissue-Invasive K. pneumoniae • JID 2005:192 (1 July) • 125

Figure 7. Polymerase chain reaction (PCR) analysis and slot-blot hybridization for 3 chromosomal regions. A, PCR primer alignments used in theanalysis of the prevalences of 2 chromosomal regions among tissue-invasive and noninvasive Klebsiella pneumoniae strains. B, Slot-blot analysesperformed to confirm the PCR results. Hybridization was performed by use of probes located inside the 3 regions. Lanes 1–6 represent 6 tissue-invasive strains; lanes 7–10 represent 4 noninvasive strains. The 23S rRNA gene (upper panel), NTUH-K2044 (lane 11), and the PCR product of eachprobe (lane 12) served as positive controls. PTS, phosphoenolpyruvate sugar phosphotransferase system.

growth of the mutants was compared with that of the wild-

type strain in nutrient-rich, undefined LB medium. However,

no significant differences were found.

The effect on virulence of each mutation was investigated in

a mouse model (figure 4B and 4C). All of the mice inoculated

with 103 cfu survived. Half of the mice died when inoculated

with 104 cfu of the wild-type strain, K2044 (DORF7–9), or K2044

(DPTS); there were no significant differences in survival among

these 3 groups (K2044 [DORF7–9] vs. wild-type strain, P p

; K2044 [DPTS] vs. wild-type strain, ; log-rank test).76 P p .67

(figure 4C). In contrast, all of the mice inoculated with K2044

(Dkfu) at the doses of 104–106 cfu survived and appeared to be

healthy after 4 weeks. The survival of the mice inoculated with

K2044 (Dkfu) differed significantly from that of the mice inoc-

ulated the wild-type strain ( , log-rank test) (figure 4C).P p .0067

When inoculated with the wild-type strain, K2044 (DORF7–

9), or K2044 (DPTS) at the highest dose (106 cfu), most of the

mice had died by 7 days after infection, with no obvious signs

before death and with no pathological changes detected on

histological examination; they were considered to have died of

septic shock. At lower inoculation doses (104–105 cfu), most

mice died between day 12 and 21 after infection, with signs of

lethargy, labored breathing, or trembling 1–2 days before death.

Large liver and/or brain abscesses were observed in this group,

and septic shock with possible organ failure was considered to

be the cause of death. However, there was no histological change

in the livers and brains obtained from the mice inoculated with

K2044 (Dkfu) at doses of 104–106 cfu (figure 5). Because the

manifestations of K. pneumoniae infection observed in BALB/

cByl mice were very similar to those observed in patients, we

did not try other murine strains.

Functional analysis of the kfu system in K. pneumoniae.

K. pneumoniae kfuABC was found to be preceded by a putative

19-bp Fur box consensus sequence (figure 6A). The 3 genes of

kfu were transcribed in the same direction and had short in-

tergenic sequences (maximum 21 bp between kfuA and kfuB),

suggesting that they were in a single transcriptional unit. RT-

PCR analysis confirmed that the 3 genes were transcribed as a

single operon (figure 6B).

Comparison of the growth rates of the wild-type strain, K2044

(Dkfu), MGH 78578, and a noninvasive strain, N3423, in an

iron-chelated medium revealed slightly lower growth for K2044

(Dkfu) and MGH 78578 (figure 6C). The E. coli aroB mutant

carrying kfuABC or sfuABC grew well under conditions of iron

126 • JID 2005:192 (1 July) • Ma et al.

Table 5. Distribution of the 3 chromosomal regions amongtissue-invasive Klebsiella pneumoniae strains, according to theclinical disease of the patients.

Disease, no. of isolateswith distribution pattern

Distribution pattern

20-kbregion

(kfu/PTS)

22-kbregion(allS)

33-kbregion(magA)

PLA (n p 33)22 + + +3 � + +1 + � +1 + � �

6 � � �

Meningitis (n p 4)2 + � +2 � � �

PLA plus endophthalmitis (n p 8) + + +PLA plus meningitis (n p 1) + + +

NOTE. Results were confirmed by both polymerase chain reaction andslot-blot hybridization (table 1). PLA, primary liver abscess; PTS, phospho-enolpyruvate sugar phosphotransferase system; +, region present; �, regionabsent.

limitation. However, the growth of H1443 with vector only

(pBR322) was significantly inhibited (figure 6D).

Prevalence and correlation of the kfu/PTS, magA, and allS

regions in K. pneumoniae clinical isolates. We compared the

presence of the kfu/PTS, magA, and allS regions by PCR. Figure

7A shows the location of the primers used in this analysis. The

criteria used to detect the prevalences of the allS and kfu/PTS

regions were the same, because they both replaced a small frag-

ment in the noninvasive strains. The PCR results were in com-

plete agreement with those of slot-blot hybridization (figure

7B). When the data were combined with those on the kfu/PTS

region, the presence of these 3 regions was strongly associated

with the clinically invasive strains more than with the clinically

noninvasive strains (for the kfu/PTS region, 35/46 vs. 19/98,

[ , x2 test]; for the allS region, 33/46 vs. 3/98 [P! .0001,P ! .0001

x2 test]; and for the magA region, 37/46 vs. 6/98 [P! .0001, x2

test]). All 3 chromosomal regions correlated well with one an-

other, with coincidence in 39 (31 positive and 8 negative for all

3 regions) of 46 isolates (85%) (table 5).

The distribution of the 3 chromosomal regions was further

analyzed according to the clinical disease of the patients (table

5). Strains from the patients with PLA plus endophthalmitis

or meningitis were positive for all 3 regions. However, when

strains from the 4 patients with meningitis but without PLA

were examined, 2 of the strains were found to be negative for

all 3 regions, and 2 of the strains were found to be positive for

the kfu/PTS and magA regions but not for the allS region.

DISCUSSION

In the present study, microarray comparison of the genomic

DNA of 4 tissue-invasive K. pneumoniae strains and 3 non-

invasive strains identified 12 clones with significantly different

hybridization signals. Two clones that matched the sequence

in MGH 78578 were probably the result of experimental var-

iations, because both were present in the 2 groups of strains.

Another 2 clones were located in a large plasmid. However, the

prevalences of the large plasmid were invariant in the 2 groups

of strains. The remaining 8 clones connected together and fur-

ther extended to encompass a 20-kb fragment (table 3). Only

the deletion mutant NTUH-K2044 (Dkfu) showed decreased vir-

ulence in vivo. Functional studies confirmed the operon’s in-

volvement in iron uptake. Because our array was estimated to

cover only 88% of the entire K. pneumoniae genome, we could

miss some regions specific to tissue-invasive strains; hence, the

allS and magA regions were not identified by this method.

The 20-kb kfu/PTS region was markedly more prevalent

among tissue-invasive strains. The kfu/PTS region encoded an

iron-uptake system involved in virulence and fit the size range

(10–200 kb) of a pathogenicity island. However, the GC content

of this region was similar to the remainder of the whole genome,

and no mobile elements or insertion sequences were noted [42].

Thus, the fundamental nature of the heterogeneity of this spe-

cific region in K. pneumoniae awaits further study [43].

Acquisition of nutrients such as iron to sustain growth in

the host environment is essential for bacterial pathogens to

establish an infection. Iron uptake is also critical to pathogenesis

as a vital cofactor for many components of microbial antiox-

idative stress defense, including such components as superoxide

dismutase, catalase, and peroxidase [44]. We have shown here

that the kfu operon was present in most of the genomes of the

tissue-invasive K. pneumoniae strains we examined and was

absent from most of the genomes of the noninvasive strains.

It is possible that the presence of a functional kfu operon might

principally or secondarily modulate virulence in vivo and so

provide a strong competitive advantage to those strains that

harbor it.

There was a significant difference in the prevalence of the

kfu/PTS region between noninvasive control strains from NTUH

and FEMH. Strains from FEMH were all isolated from patients

with nosocomial infection, whereas strains from NTUH were

all isolated from patients with community-acquired infection.

This could be the reason for the genetic difference. In addition,

there would be an epidemiological difference between strains

from a medical center (NTUH) and a those from a community

hospital (FEMH).

How K. pneumoniae enter into the bloodstream and liver

has not yet been documented. However, bacterial cells would

enter the bloodstream through M cells [45] or as a result of

kfu/PTS Region and Tissue-Invasive K. pneumoniae • JID 2005:192 (1 July) • 127

minor mucosal injuries of the gastrointestinal tract. Larger ab-

scesses were found in the mice infected via ig inoculation,

whereas, in the mice infected via ip inoculation, microabscesses

were found. This finding supports the hypothesis that K. pneu-

moniae gains entry into the bloodstream via the gastrointestinal

tract, because all venous returns in the gastrointestinal tract

were collected via a portal vein into liver. Many bacteria were

likely trapped in the liver by Kupffer cells; however, the tissue-

invasive strains were resistant to phagocytosis and serum killing.

This may result in the subsequent formation of liver abscesses.

The present study reinforces our awareness of the vital role

that bacterial virulence factors play in the pathogenesis of PLA

and metastatic infection. According to the prevalences of the

3 specific chromosomal regions in tissue-invasive strains, we

were able to classify the 54 invasive strains into 3 groups: group

1 strains contained all 3 regions, group 2 strains contained 1–

2 regions, and group 3 strains contained none of the regions.

When all 3 regions were absent, the strains were only occa-

sionally capable of causing PLA. The most interesting finding

was that all of the strains from patients with PLA plus metastatic

infection contained all 3 regions. Most of the group 3 strains

were noninvasive (table 1), suggesting that they might be less

virulent than the group 1 strains. That the group 1 strains had

an increased ability to invade tissue may well relate to their

enhanced survival and ability to compete for nutrients. The

kfu/PTS region could enrich the ability of bacteria to secure

iron, even in the relatively iron-deficient conditions of the hu-

man host. The allS region could help bacteria to compete for

nitrogen sources via the allantoin-utilizing ability [15]. More-

over, the magA region could render bacteria resistant to phago-

cytosis by polymorphonuclear leukocytes and serum killing [16].

Therefore, the group 1 strains might be classified as a specific

genotype that is associated with PLA and metastatic infection.

Two strains from 4 patients with meningitis but without PLA

were negative for all 3 regions, and the remaining 2 strains were

negative for the allS region. These results suggest that different

pathogenic mechanisms could be at work in patients with men-

ingitis plus PLA and in those with meningitis only.

In conclusion, we have identified a 20-kb chromosomal kfu/

PTS region in K. pneumoniae that is associated with iron ac-

quisition and virulence. This region is more prevalent in tissue-

invasive strains. Strains containing the kfu/PTS, allS, and magA

regions are strongly associated with PLA and metastatic infec-

tion and may be classified as a specific genotype. However, a

different virulence mechanism seems to be at work in patients

with meningitis but without PLA. Therefore, the kfu/PTS re-

gion, as well as the allS and magA regions, can be exploited as

a genetic marker for rapid molecular diagnosis and for tracing

the source of these emergent tissue-invasive strains.

Acknowledgment

We thank Dr. Volkmar Braun, Universitat Tubingen (Tubingen, Ger-many), for providing Escherichia coli strain H1443 and pSZ1 plasmid.

References

1. Podschun R, Ullmann U. Klebsiella spp. as nosocomial pathogens:epidemiology, taxonomy, typing methods, and pathogenicity factors.Clin Microbiol Rev 1998; 11:589–603.

2. Chang SC, Fang CT, Hsueh PR, Chen YC, Luh KT. Klebsiella pneu-moniae isolates causing liver abscess in Taiwan. Diagn Microbiol InfectDis 2000; 37:279–84.

3. Cheng DL, Liu YC, Yen MY, Liu CY, Wang RS. Septic metastatic lesionsof pyogenic liver abscess: their association with Klebsiella pneumoniaebacteremia in diabetic patients. Arch Intern Med 1991; 151:1557–9.

4. Chiu CT, Lin DY, Liaw YF. Metastatic septic endophthalmitis in pyo-genic liver abscess. J Clin Gastroenterol 1988; 10:524–7.

5. Fung CP, Chang FY, Lee SC, et al. A global emerging disease of Klebsiellapneumoniae liver abscess: is serotype K1 an important factor for com-plicated endophthalmitis? Gut 2002; 50:420–4.

6. Ko WC, Paterson DL, Sagnimeni AJ, et al. Community-acquired Kleb-siella pneumoniae bacteremia: global differences in clinical patterns.Emerg Infect Dis 2002; 8:160–6.

7. Liu YC, Cheng DL, Lin CL. Klebsiella pneumoniae liver abscess associ-ated with septic endophthalmitis. Arch Intern Med 1986; 146:1913–6.

8. Wang JH, Liu YC, Lee SS, et al. Primary liver abscess due to Klebsiellapneumoniae in Taiwan. Clin Infect Dis 1998; 26:1434–8.

9. Saccente M. Klebsiella pneumoniae liver abscess, endophthalmitis, andmeningitis in a man with newly recognized diabetes mellitus. Clin In-fect Dis 1999; 29:1570–1.

10. Cahill M, Chang B, Murray A. Bilateral endogenous bacterial endo-phthalmitis associated with pyogenic hepatic abscess [letter]. Br J Oph-thalmol 2000; 84:1436.

11. Casanova C, Lorente JA, Carrillo F, Perez-Rodriguez E, Nunez N. Kleb-siella pneumoniae liver abscess associated with septic endophthalmitis[letter]. Arch Intern Med 1989; 149:1467.

12. Lee KH, Hui KP, Tan WC, Lim TK. Klebsiella bacteraemia: a reportof 101 cases from National University Hospital, Singapore. J Hosp In-fect 1994; 27:299–305.

13. Ohmori S, Shiraki K, Ito K, et al. Septic endophthalmitis and meningitisassociated with Klebsiella pneumoniae liver abscess. Hepatol Res 2002;22:307–12.

14. Assantachai P, Luengrojanakul P, Phimolsarnti R, Charoenlarp K, Ra-tanarapee S. Liver abscess in polycystic liver disease. J Med Assoc Thai1995; 78:210–6.

15. Chou HC, Lee CZ, Ma LC, Fang CT, Chang SC, Wang JT. Isolation ofa chromosomal region of Klebsiella pneumoniae associated with allantoinmetabolism and liver infection. Infect Immun 2004; 72:3783–92.

16. Fang CT, Chuang YP, Shun CT, Chang SC, Wang JT. A novel virulencegene in Klebsiella pneumoniae strains causing primary liver abscess andseptic metastatic complications. J Exp Med 2004; 199:697–705.

17. Fitzgerald JR, Musser JM. Evolutionary genomics of pathogenic bac-teria. Trends Microbiol 2001; 9:547–53.

18. Kim CC, Joyce EA, Chan K, Falkow S. Improved analytical methodsfor microarray-based genome-composition analysis. Genome Biol 2002;3:RESEARCH0065.

19. Abbott SL. Klebisella, Enterobacter, Citrobacter, Serratia, Plesiomonas,and other Enterobacteriaceae. In: Murray PR, Baron EJ, Jorgensen JH,Pfaller M, Yolken RH, eds. Manual of clinical microbiology. 8th ed.Washington, DC: American Society for Microbiology, 2003:684–700.

20. Fang CT, Chen HC, Chuang YP, Chang SC, Wang JT. Cloning of acation efflux pump gene associated with chlorhexidine resistance inKlebsiella pneumoniae. Antimicrob Agents Chemother 2002; 46:2024–8.

21. Chen JJ, Wu R, Yang PC, et al. Profiling expression patterns and iso-

128 • JID 2005:192 (1 July) • Ma et al.

lating differentially expressed genes by cDNA microarray system withcolorimetry detection. Genomics 1998; 51:313–24.

22. Ang S, Lee CZ, Peck K, et al. Acid-induced gene expression in Heli-cobacter pylori: study in genomic scale by microarray. Infect Immun2001; 69:1679–86.

23. Herrero M, de Lorenzo V, Timmis KN. Transposon vectors containingnon-antibiotic resistance selection markers for cloning and stable chro-mosomal insertion of foreign genes in gram-negative bacteria. J Bac-teriol 1990; 172:6557–67.

24. Zimmermann L, Angerer A, Braun V. Mechanistically novel iron (III)transport system in Serratia marcescens. J Bacteriol 1989; 171:238–43.

25. Angerer A, Klupp B, Braun V. Iron transport systems of Serratia mar-cescens. J Bacteriol 1992; 174:1378–87.

26. Okamoto M, Nakane A, Minagawa T. Host resistance to an intragastricinfection with Listeria monocytogenes in mice depends on cellular im-munity and intestinal bacterial flora. Infect Immun 1994; 62:3080–5.

27. Reed LJ, Muench H. A simple method of estimating fifty percent end-points. Am J Hyg 1938; 27:493–7.

28. Winnacker E-L. Genomic libraries. In: Weller MG, ed. From genes toclones: introduction to gene technology. Weinheim, Germany: VCHVerlagsgesellschaft mbH, 1987:383–5.

29. Chen YT, Chang HY, Lai YC, Pan CC, Tsai SF, Peng HL. Sequencingand analysis of the large virulence plasmid pLVPK of Klebsiella pneu-moniae CG43. Gene 2004; 337:189–98.

30. Tomii K, Kanehisa M. A comparative analysis of ABC transporters incomplete microbial genomes. Genome Res 1998; 8:1048–59.

31. Angerer A, Gaisser S, Braun V. Nucleotide sequences of the sfuA, sfuB,and sfuC genes of Serratia marcescens suggest a periplasmic-binding-protein-dependent iron transport mechanism. J Bacteriol 1990; 172:572–8.

32. Adhikari P, Kirby SD, Nowalk AJ, Veraldi KL, Schryvers AB, MietznerTA. Biochemical characterization of a Haemophilus influenzae peri-plasmic iron transport operon. J Biol Chem 1995; 270:25142–9.

33. Gong S, Bearden SW, Geoffroy VA, Fetherston JD, Perry RD. Char-acterization of the Yersinia pestis Yfu ABC inorganic iron transportsystem. Infect Immun 2001; 69:2829–37.

34. Chin N, Frey J, Chang CF, Chang YF. Identification of a locus involvedin the utilization of iron by Actinobacillus pleuropneumoniae. FEMSMicrobiol Lett 1996; 143:1–6.

35. Adhikari P, Berish SA, Nowalk AJ, Veraldi KL, Morse SA, MietznerTA. The fbpABC locus of Neisseria gonorrhoeae functions in the per-iplasm-to-cytosol transport of iron. J Bacteriol 1996; 178:2145–9.

36. Mao Y, Varoglu M, Sherman DH. Molecular characterization and anal-ysis of the biosynthetic gene cluster for the antitumor antibiotic mi-tomycin C from Streptomyces lavendulae NRRL 2564. Chem Biol 1999;6:251–63.

37. Postma PW, Lengeler JW, Jacobson GR. Phosphoenolpyruvate:car-bohydrate phosphotransferase systems of bacteria. Microbiol Rev 1993;57:543–94.

38. Kok M, Bron G, Erni B, Mukhija S. Effect of enzyme I of the bacterialphosphoenolpyruvate:sugar phosphotransferase system (PTS) on vir-ulence in a murine model. Microbiology 2003; 149:2645–52.

39. Tan MW, Rahme LG, Sternberg JA, Tompkins RG, Ausubel FM. Pseu-domonas aeruginosa killing of Caenorhabditis elegans used to identify P.aeruginosa virulence factors. Proc Natl Acad Sci USA 1999; 96:2408–13.

40. Turner AK, Lovell MA, Hulme SD, Zhang-Barber L, Barrow PA. Iden-tification of Salmonella typhimurium genes required for colonizationof the chicken alimentary tract and for virulence in newly hatchedchicks. Infect Immun 1998; 66:2099–106.

41. Escolar L, Perez-Martin J, de Lorenzo V. Opening the iron box: tran-scriptional metalloregulation by the Fur protein. J Bacteriol 1999; 181:6223–9.

42. Hacker J, Kaper JB. Pathogenicity islands and the evolution of mi-crobes. Annu Rev Microbiol 2000; 54:641–79.

43. Lai YC, Yang SL, Peng HL, Chang HY. Identification of genes presentspecifically in a virulent strain of Klebsiella pneumoniae. Infect Immun2000; 68:7149–51.

44. Griffiths E. The iron-uptake systems of pathogenic bacteria. In: BullenJJ, Griffiths E, eds. Iron and infection. New York: John Wiley & Sons,1987:69–137.

45. Lu L, Walker WA. Pathologic and physiologic interactions of bacteriawith the gastrointestinal epithelium. Am J Clin Nutr 2001; 73:S1124–30.