Comparison of Virulence Gene Profiles of Escherichia coli Strains Isolated from Healthy and Diarrheic Swine

Abstract

A combination of uni- and multiplex PCR assays targeting 58 virulence genes (VGs) associated with Escherichia coli strains causing intestinal and extraintestinal disease in humans and other mammals was used to analyze the VG repertoire of 23 commensal E. coli isolates from healthy pigs and 52 clinical isolates associated with porcine neonatal diarrhea (ND) and postweaning diarrhea (PWD). The relationship between the presence and absence of VGs was interrogated using three statistical methods. According to the generalized linear model, 17 of 58 VGs were found to be significant (P < 0.05) in distinguishing between commensal and clinical isolates. Nine of the 17 genes represented by iha, hlyA, aidA, east1, aah, fimH, iroN_{E. coli}, traT, and saa have not been previously identified as important VGs in clinical porcine isolates in Australia. The remaining eight VGs code for fimbriae (F4, F5, F18, and F41) and toxins (STa, STb, LT, and Stx2), normally associated with porcine enterotoxigenic E. coli. Agglomerative hierarchical algorithm analysis grouped E. coli strains into subclusters based primarily on their serogroup. Multivariate analyses of clonal relationships based on the 17 VGs were collapsed into two-dimensional space by principal coordinate analysis. PWD clones were distributed in two quadrants, separated from ND and commensal clones, which tended to cluster within one quadrant. Clonal subclusters within quadrants were highly correlated with serogroups. These methods of analysis provide different perspectives in our attempts to understand how commensal and clinical porcine enterotoxigenic E. coli strains have evolved and are engaged in the dynamic process of losing or acquiring VGs within the pig population.

The acquisition of virulence genes is believed to provide an evolutionary pathway to pathogenicity. As a genetically diverse group, most strains of Escherichia coli are harmless commensals of mammals (31, 54), but others are capable of causing either intestinal or extraintestinal disease (47). Manifestation of clinical symptomology and pathology appears to be closely associated with the possession of certain virulence gene combinations in E. coli (27, 35). For instance, diarrheagenic E. coli strains are classified on the basis of their virulence properties as enterotoxigenic (ETEC), enteropathogenic (EPEC), enterohemorrhagic (EHEC), enteroinvasive (EIEC), and enteroaggregative (EaggEC) (43). In general, these pathotypes have in common various virulence gene combinations for attachment and elaboration of hemolysins and enterotoxins, but there are considerable polymorphism and sequence variation in the molecular identities of genes that code for these virulence factors (8, 42).

Indeed, PCR analysis has revealed that even commensal E. coli isolates possess some of these virulence genes (10, 19). However, mere possession of a single or a few virulence genes does not endow a strain with pathogenic status unless that strain has acquired the appropriate virulence gene combination to cause disease in a specific host species (25). At present, it is debatable whether isolates that have one or a few virulence genes represent pathogenic clones that have lost virulence genes or are commensals in the process of acquiring them. Using a more phylogenetic approach, in 2000 Clermont (15) described a three-gene combination to differentiate between strains in the ECOR collection that are pathogens (phylogenetic groups B2 and D) and those members that are mainly commensals (phylogenetic groups A and B1). Surprisingly, these three genes alone were capable of providing a phylogenetic classification that closely mirrored similar groupings based on a more complex genetic analysis by multilocus enzyme electrophoresis. Furthermore, these relationships were established with a population of assembled clones, primarily of commensal origin, and in the absence of a panel of functionally accredited virulence genes.

Diarrhea in pigs can be caused by a number of pathogens, with transmissible gastroenteritis virus, rotavirus (RV), ETEC, Clostridium perfringens, and Isospora suis being the most common (61). Of these, ETEC strains are recognized as the most common cause of neonatal diarrhea (ND) in 0- to 4-day-old piglets with accompanying high morbidity and mortality rates (61). Strains of E. coli isolated from piglets with ND are mucoid (A-type capsule), often nonhemolytic, and usually confined to serogroups O8, O9, O20, O64, and O101 (23, 58). These strains have been classified as atypical “class 2” ETEC as they possess fimbrial adhesins belonging to F4, F5, F6, or F41 and are generally LT⁻ (heat-labile toxin) and ST⁺ (heat-stable toxin) (23, 29, 57, 58). Older preweaning pigs, as well as postweaning (PW) animals up to 12 weeks of age, are affected by hemolytic ETEC (22). These strains are frequently represented by classic serogroups, including O8, O138, O139, O145, O141, O149, and O157, and are considered to be typical “class 1” ETEC (22, 58). They express F4 fimbrial adhesin in association with heat-labile enterotoxin LT alone or in combination with heat-stable enterotoxin STa and/or STb (40, 56, 65). More recently, Benz and Schmidt (5-7) have reported the detection of a class of diffusely adhering E. coli (DAEC) strains from piglets with diarrhea and its adhesin virulence factor, AIDA. Specific combinations of virulence genes encoding virulence factors such as adhesins and enterotoxins are the characteristic signature of pathogenic ETEC isolates.

Unlike commensals, extraintestinal pathogenic E. coli (ExPEC) strains do not establish long-term symbiotic relationships with their host (38, 65). Like strains that cause intestinal infections in humans, ExPEC strains possess virulence genes that have a range of functions, including attachment/invasion, toxin production, iron scavenging, and immune evasion (34, 37). ExPEC strains are divided into three major pathotypes (4): (i) uropathogenic (UPEC) strains that cause urinary infections, (ii) strains that cause neonatal meningitis (MENEC), and (iii) strains that cause septicemia (4). ExPEC strains possess virulence gene combinations that are distinctive from those found in their counterparts that cause intestinal disease. For instance, UPEC strains are more likely to possess P pili, S pili, afimbrial adhesin, and toxins such as hemolysin and cytotoxic necrotizing factor 1 (34, 37).

The main objectives in this study were to apply a wider array of virulence genes known to occur in intestinal and extraintestinal E. coli pathotypes associated with both human and animal disease and to optimize uni/multiplex combinatorial PCR assays for their detection in porcine isolates. The assays were then used to determine the presence of these genes in E. coli strains isolated from clinical cases of ND and postweaning diarrhea (PWD) to explore the possibility that clinical isolates can be identified by their virulence gene combinations. E. coli isolates from healthy pigs (commensals) were also included in the virulence gene analysis to assess whether there exists an evolutionary and phylogenetic relationship between pathogens and commensals modeled principally on their virulence gene repertoire. A mathematical model involving principal coordinate (PCO) analysis was used to visualize these relationships.

MATERIALS AND METHODS

Bacterial strains.

Table 1 lists 75 E. coli strains used in this study. These were obtained from scouring neonatal and weaner piglets as well as healthy animals. The ND and PWD strains were obtained from diarrhea samples cultured on blood agar plates following previously described protocols (20, 22). The commensal isolates represent a subset of E. coli previously isolated from the luminal contents of the duodenum, ileum, colon, and fecal samples of healthy pigs (19). Additional strains used as a reference source of VGs for PCR analysis are summarized in Table S1 in the supplemental material. The identity of all E. coli isolates was confirmed by a positive indole test, with no growth on Simmons citrate agar and growth on minimal lactose agar plates (19).

TABLE 1.

List of 75 E. coli strains from clinical cases of porcine neonatal and postweaning diarrhea and commensal isolates from healthy pigs

Identification code	Serogroup	Origin of isolate	Hemolytic phenotypef	Phylogenetic statusg		Identification code	Serogroup	Origin of isolate	Hemolytic phenotypef	Phylogenetic statusg
Neonatal						38	O141K85ac	Fecalb	β-h	A
1	O8:F−	Fecala	Nh	A		39	O141K85ac	Fecalb	β-h	A
2	O8:F41	Fecalb	Nh	A		40	O141K85ac	Fecalb	β-h	A
3	O8:K99	Fecalb	Nh	A		41	O141K85ac	Fecalb	β-h	B2
4	O8:K88	Fecalb	Nh	D		42	O141K85ac	Fecalb	β-h	A
5	O8:K88	Fecalb	Nh	A		43	O149:K88	Fecald	β-h	A
6	O8:K88	Fecalb	Nh	B2		44	O149K88	Fecalb	β-h	A
7	O8:K99	Fecalb	Nh	B1		45	O149K88	Fecalb	β-h	A
8	O9:F41	Fecalb	Nh	A		46	O149K88	Fecalb	β-h	A
9	O9:F41	Fecalb	Nh	A		47	O149K88	Fecalb	β-h	A
10	O9:F41	Fecalb	Nh	A		48	O149K88	Fecalb	β-h	A
11	O9:987P	Fecalb	Nh	A		49	O149K88	Fecalb	β-h	A
12	O20:K99	Fecalb	Nh	A		50	O149K88	Fecalb	β-h	A
13	O20:K99	Fecalb	Nh	A		51	O149:K88	Fecalb	β-h	A
14	O64:K99	Fecala	Nh	B1		52	O149:K91:Hnt	Fecalc	β-h	A
15	O99:Ksnt:H38	Fecalc	Nh	B1
16	O101:F41	Fecalb	Nh	A		Commensal
17	O101:K88	Fecala	Nh	B1		53	O8:H−	Ileume	β-h	A
18	O101:Ksnt:H11	Fecalc	Nh	A		54	O18a,b:H−	Ileume	nh	B1
19	O109:Knst:H11	Fecalc	Nh	B1		55	O40:H25	Fecale	nh	A
20	O157:K88	Fecalb	β-h	A		56	O40:H25	Fecale	nh	D
						57	O75:H−	Fecalc	nh	A
Weaner						58	O75:H−	Fecalc	nh	A
21	O8G7	Fecalb	β-h	A		59	O77:H−	Duodenume	nh	B1
22	O8G7	Fecalb	β-h	D		60	O82:H8	Colone	nh	A
23	O8G7	Fecalb	β-h	D		61	O106:H−	Duodenume	nh	B1
24	O8G7:K88	Fecala	β-h	A		62	O121:H21	Fecalc	nh	A
25	O45	Fecalc	β-h	B1		63	O126:H−	Fecalc	nh	A
26	O138	Fecald	β-h	D		64	O130:H−	Fecalc	nh	B1
27	O138:K81:Hnt	Fecalc	β-h	D		65	O130:H11	Facalc	nh	B1
28	O139	Fecalb	β-h	D		66	OR:H−	Fecale	nh	A
29	O141:K85ab	Fecald	β-h	A		67	OR:H38	Fecale	nh	A
30	O141:K85ab	Fecalb	β-h	A		68	Ont/R:H−	Ileume	nh	A
31	O141ac	Fecald	β-h	A		69	Ont:Hnt	Ileume	nh	B1
32	O141:K85ac	Fecalb	β-h	A		70	Ont:H−	Duodenume	nh	A
33	O141:K85ab:H4	Fecalc	β-h	A		71	Ont:H−	Duodenume	nh	A
34	O141:K85ac:H−	Fecalc	β-h	A		72	Ont:H−	Fecale	nh	A
35	O141:K85ac	Fecalc	β-h	B1		73	Ont:H11	Fecalc	nh	B1
36	O141:K85ac	Fecalc	β-h	B1		74	Ont:H27	Ileume	nh	A
37	O141K85ac	Fecalb	β-h	A		75	Ont:H45	Duodenume	nh	B1

^aSchool of Veterinary Science, The University of Queensland.

^bDepartment of Primary Industries, Victoria.

^cRegional Veterinary Laboratory, Elizabeth Macarthur Agriculture Institute.

^dChris Richards and Associates, Bendigo, Victoria.

^eDixit et al. (19).

^fnh, nonhemolytic; β-h, beta-hemolytic.

^gClermont et al. (15).

Bacterial serotyping.

Both O serotyping and H serotyping of E. coli were performed using previously reported methods (9, 12).

Hemolysis.

Hemolysis was determined by streaking E. coli isolates onto blood agar containing 10% sheep blood and incubating them at 37°C for 24 h. A clear zone around colonies where the blood cells had been utilized was characteristic of beta-hemolytic E. coli.

Maintenance of bacteria.

E. coli strains were stored at −80°C in Luria-Bertani (LB) broth containing 20% glycerol. Bacteria were recovered from frozen stocks and plated on LB agar and were never subcultured more than twice before DNA extraction.

DNA extraction.

All isolates were prepared by inoculating a single colony into 1 ml of LB broth and incubated at 37°C with shaking (∼100 rpm) overnight. DNA was extracted using a Promega DNA purification kit.

Virulence genes.

A group of 58 VGs (Table 2), reported in the literature to be associated with different E. coli pathotypes, were selected as the panel to be used in our analysis (Table 3) (3, 15, 17, 20, 28, 34, 44, 50, 52, 70, 71).

TABLE 2.

List of 58 virulence genes reported in the literature to be associated with different E. coli pathotypes associated with human and animal disease

Virulence gene(s)/activity	E. coli pathotype	Description/function
Adhesins
afa/draBC	ExPEC	Central region of Dr antigen-specific fimbrial and afimbrial adhesin operons (e.g., AFA, Dr, and F1845)
aidA	DAEC	Adhesin involved in diffuse adherence, consisting of AIDA-I (orfB) and AIDAc (orfBc)
aah	DAEC	Autotransporter adhesin heptosyltransferase encoding AAH protein which modifies AIDA-I adhesin
bfpA	EPEC	Type IV bundle-forming pili
bmaE	ExPEC	M-agglutinin subunit
eaeA	EPEC, EHEC	Intimin
faeG	ETEC	F4 fimbrial adhesin
fanC	ETEC	F5 fimbrial adhesin
fasA	ETEC	F6 fimbrial adhesin
fedA	ETEC	F18 fimbrial adhesin
F41	ETEC	Fimbrial adhesin
fimH	ExPEC	d-Mannose-specific adhesin, type 1 fimbriae
focG	ExPEC	Pilus tip molecule, F1C fimbriae (sialic acid specific)
iha	EHEC	Novel nonhemagglutinin adhesin (from O157:H7 and CFT073)
nfaE	ExPEC	Nonfimbrial adhesin I assembly and transport
paa	EPEC	Porcine A/E-associated gene
papA	ExPEC	Major structural subunit of pilus associated with pyelonephritis (P fimbriae), defines F antigen
papC	ExPEC	Pilus assembly, central region of pap operon
papEF	ExPEC	Minor tip pilins, connect PapG to shaft (PapA)
papG	ExPEC	Gal(1-4)Gal-specific pilus tip adhesin molecule
papG allele I	ExPEC	(Rare) J96-associated papG variant
papG allele II	ExPEC	Pyelonephritis-associated papG variant
papG allele III	ExPEC	Cystitis-associated (prs or pap-2) papG variant
papG allele l′	ExPEC	papG variant identified in canine urine
saa	EHEC	STEC autoagglutinating adhesin
sfa/focDE	ExPEC	Central region of sfa (S fimbriae) and foc (F1C fimbriae) operons
sfaS	ExPEC	Pilus tip adhesin, S fimbriae (sialic acid specific)
Toxins
cdt	EPEC, ExPEC, NTEC	Cytolethal distending toxin
cdtB	ExPEC	Cytolethal distending toxin
cnf1	ExPEC, NTEC	Cytotoxic necrotizing factor 1
univcnf	ExPEC	Universal primer for cytotoxic necrotizing factor 1
cvaC	ExPEC	Colicin V, conjugative plasmids (traT, iss, and antimicrobial resistance)
east1	EaggEC	EaggEC heat-stable enterotoxin
exhA	EPEC, EHEC	Enterohemolysin
hlyA	ExPEC	α-Hemolysin
LT	ETEC	Heat-labile toxin
STa	ETEC	Heat-stable enterotoxin a
STb	ETEC	Heat-stable enterotoxin b
stx1	EHEC	Shiga toxin I
stx2	EHEC	Shiga toxin II
Capsule synthesis
kpsMTII	ExPEC	Group II capsular polysaccharide synthesis (e.g., K1, K5, and K12)
kpsMTIII	ExPEC	Group III capsular polysaccharide synthesis (e.g., K3, K10, and K54)
kpsMT K1	ExPEC	Specific for K1 (group II) kpsMT
kpsMT “K5”	ExPEC	Specific for non-K1 and non-K2 group II kpsMT
rfc	ExPEC	O4 lipopolysaccharide synthesis
Siderophores
fyuA	ExPEC	Yersinia siderophore receptor (ferric yersiniabactin uptake)
ireA	ExPEC	Iron-regulated element, a siderophore receptor
iroNE. coli	ExPEC	Novel catecholate siderophore
iutA	ExPEC	Ferric aerobactin receptor (iron uptake/transport)
Invasins
ibeA	ExPEC	Invasion of brain endothelium
ipaH	EIEC	Invasion plasmid antigen
Additional virulence genes
chuA	EHEC	Gene required for heme transport in EHEC O157:H7
iss	ExPEC	Serum survival gene
ompT	ExPEC	Outer membrane protein A and T (protease)
PAI	ExPEC	Pathogenicity-associated island, provides mechanism for coordinate horizontal transfer of VF genes between lineages
TSPE4C2		Anonymous DNA fragment
traT	ExPEC	Surface exclusion, serum survival
yjaA		Identified in E. coli K12, function currently unknown

TABLE 3.

Summary of 12 multiplex (I to XII) and 6 uniplex (XIII to XVIII) primer sets for the amplification of the 58 virulence genes in this study^a

Primer set (reference[s]) and virulence genea	Primer name	DNA sequence (5′→3′)	Amplified product (bp)	Primer concn
I (34)
PAI	RPAi-F	GGACATCCTGTTACAGCGCGCA	930	0.6 μM
	RPAi-R	TCGCCACCAATCACAGCCGAAC
papAH	PapA-F	ATGGCAGTGGTGTCTTTTGGTG	720	0.6 μM
	PapA-R	CGTCCCACCATACGTGCTCTTC
fimH	FimH-F	TGCAGAACGGATAAGCCGTGG	508	0.6 μM
	FimH-R	GCAGTCACCTGCCCTCCGGTA
kpsMTIII	kpsII-F	GCGCATTTGCTGATACTGTTG	392	0.6 μM
	kpsII-R	CATCCAGACGATAAGCATGAGCA
papEF	kpsII-F	GCGCATTTGCTGATACTGTTG	336	0.6 μM
	kpsII-R	CATCCAGACGATAAGCATGAGCA
ibeA	ibe10-F	AGGCAGGTGTGCGCCGCGTAC	170	0.6 μM
	ibe10-R	TGGTGCTCCGGCAAACCATGC
II (34)
fyuA	FyuA-F	TGATTAACCCCGCGACGGGAA	880	0.6 μM
	FyuA-R	CGCAGTAGGCACGATGTTGTA
bmaE	bmaE-F	ATGGCGCTAACTTGCCATGCTG	507	0.6 μM
	bmaE-R	AGGGGGACATATAGCCCCCTTC
sfa/focDE	sfa 1-F	CTCCGGAGAACTGGGTGCATCTTAC	410	0.6 μM
	sfa 2-R	CGGAGGAGTAATTACAAACCTGGCA
iutA	AerJ-F	GGCTGGACATCATGGGAACTGG	300	0.6 μM
	AerJ-R	CGTCGGGAACGGGTAGAATCG
papG allele III	Allele III-F	GGCCTGCAATGGATTTACCTGG	258	0.6 μM
	Allele III-R	CCACCAAATGACCATGCCAGAC
kpsMT K1	K1-fc-F	TAGCAAACGTTCTATATTGGTGC	153	0.6 μM
	KpsII-R	CATCCAGACGATAAGCATGAGCA
III (34)
hlyA	hly-F	AACAAGGATAAGCACTGTTCTGGCT	1,177	0.6 μM
	hly-R	ACCATATAAGCGGTCATTCCCGTCA
rfc	rfc-F	ATCCATCAGGAGGGGACTGGA	788	0.6 μM
	rfc-R	AACCATACCAACCAATGCGAG
nfaE	nfaE-F	GCTTACTGATTCTGGGATGGA	559	0.3 μM
	nfaE-R	CGGTGGCCGAGTCATATGCCA
papG allele I	Allele I-F	TCGTCTCAGGTCCGGAATTT	461	0.3 μM
	Allele I-R	TGGCATCCCCCAACATTATCG
kpsMTII	kpsII-F	GCGCATTTGCTGATACTGTTG	272	0.3 μM
	kpsII-R	CATCCAGACGATAAGCATGAGC
papC	PapC-F	GTGGCAGTATGAGTAATGACCGTTA	200	0.3 μM
	PapC-R	ATATCCTTTCTGCAGGGATGCAATA
IV (34)
cvaC	ColV-C-F	CACACACAAACGGGAGCTGTT	680	0.6 μM
	ColV-C-R	CTTCCCGCAGCATAGTTCCAT
cdtB	cdt-a1-F	AAATCACCAAGAATCATCCAGTTA	430	0.6 μM
	cdt-a2-R	AAATCTCCTGCAATCATCCAGTTTA
	cdt-s1-F	GAAAGTAAATGGAATATAAATGTCCG
	cdt-s2-R	GAAAATAAATGGAACACACATGTCCG
focG	FocG-F	CAGCACGGCAGTGGATACGA	360	0.6 μM
	FocG-R	GAATGTCGCCTGCCCATTGCT
traT	TraT-F	GGTGTGGTGCGATGAGCACAG	290	0.6 μM
	TraT-R	CACGGTTCAGCCATCCCTGAG
papG allele II	Allele II-F	GGGATGAGCGGGCCTTTGAT	190	0.6 μM
	Allele II-R	CGGGCCCCCAAGTAACTCG
V (34)
papG allele I	pG-F	CTGTAATTACGGAAGTGATTTCTG	1,190	0.6 μM
	pG1"-R	TCCAGAAATAGCTCATGTAACCCG
papG alleles II and III	pG-F	CTGTAATTACGGAAGTGATTTCTG	1,070	0.6 μM
	pG-R	ACTATCCGGCTCCGGATAAACCAT
afa/draBC	Afa-F	GGCAGAGGGCCGGCAACAGGC	559	0.3 μM
	Afa-R	CCCGTAACGCGCCAGCATCTC
cnf1	cnf1-F	AAGATGGAGTTTCCTATGCAGGAG	498	0.3 μM
	cnf2-R	CATTCAGAGTCCTGCCCTCATTATT
sfaS	SfaS-F	GTGGATACGACGATTACTGTG	240	0.3 μM
	SfaS-R	CCGCCAGCATTCCCTGTATTC
kpsMT K5	K5-F	CAGTATCAGCAATCGTTCTGTA	159	0.6 μM
	kpsII-R	CATCCAGACGATAAGCATGAGCA
VI (34, 70)
univcnf	CONCNF-F	ATCTTATACTGGATGGGATCATCTTGG	1,105	0.6 μM
	CONCNF-R	GCAGAACGACGTTCTTCATAAGTATC
iha	IHA-F	CTGGCGGAGGCTCTGAGATCA	827	0.6 μM
	IHA-R	TCCTTAAGCTCCCGCGGCTGA
iroNE. coli	IRONEC-F	AAGTCAAAGCAGGGGTTGCCCG	665	0.6 μM
	IRONEC-R	GACGCCGACATTAAGACGCAG
ompT	OMPT-F	ATCTAGCCGAAGAAGGAGGC	559	0.6 μM
	OMPT-R	CCCGGGTCATAGTGTTCATC
Allele I′	ALLELE I′-F	CTACTATGTTCATGCTCAGGTC	479	0.6 μM
	ALLELE I′-R	CCTGCATCCTCCACCATTATCGA
iss	ISS-F	CAGCAACCCGAACCACTTGATG	323	0.6 μM
	ISS-R	AGCATTGCCAGAGCGGCAGAA
ireA	IRE-F	GATGACTCAGCCACGGGTAA	254	0.6 μM
	IRE-R	CCAGGACTCACCTCACGAAT
VII (51)
ehxA	ehxA-F	GCATCATCAAGCGTACGTTCC	534	20 pmol
	ehxA-R	AATGAGCCAAGCTGGTTAAGCT
eaeA	eaeA-F	GACCCGGCACAAGCATAAGC	384	20 pmol
	eaeA-R	CCACCTGCAGCAACAAGAGG
stx2	stx2-F	GGCACTGTCTGAAACTGCTCC	255	20 pmol
	stx2-R	TCGCCAGTTATCTGACATTCTG
stx1	stx1-F	ATAAATCGCCTATCGTTGACTAC	180	20 pmol
	stx1-R	AGAACGCCCACTGAGATCATC
VIII (20)
eltA	LTA-1	GGCGACAGATTATACCGTGC	696	3.2 pmol
	LTA-2	CCGAATTCTGTTATATATGTC
fasA	F6-Fw	TCTGCTCTTAAAGCTACTGG	333	3.2 pmol
	F6-Rv	AACTCCACCGTTTGTATCAG
estII	STb-1	ATCGCATTTCTTCTTGCATC	172	3.2 pmol
	STb-2	GGGCGCCAAAGCATGCTCC
IX (20)
faeG	F4-Fw	GGTGATTTCAATGGTTCG	764	3.2 pmol
	F4-Rv	ATTGCTACGTTCAGCGGAGCG
fanC	F5-Fw	TGGGACTACCAATGCTTCTG	450	3.2 pmol
	F5-Rv	TATCCACCATTAGACGGAGC
estI	STa1	TCTTTCCCCTCTTTTAGTCAG	166	3.2 pmol
	STa2	ACAGGCAGGATTACAACAAAG
X (20)
fedA	FedA-1	GTGAAAAGACTAGTTTATTTC	510	3.2 pmol
	FedA-2	CTTGTAAGTAACCGCGTAAGC
F41	F41-Fw	GAGGGACTTTCATCTTTTAG	431	3.2 pmol
	F41-Rv	AGTCCATTCCATTTATAGGC
XI (44)
aah (orfA)	UN19	CTGGGTGACATTATTGCTTGG	370	10 pmol
	UN20	TTTGCTTGTGCGGTAGACTG
aidA AIDA-I (orfB)	UN21	TGCAAACATTAAGGGCTCG	450	10 pmol
	UN22	CCGGAAACATTGACCATACC
aidA AIDAc (orfB)	UN23	CAGTTTATCAATCAGCTCGGG	543	10 pmol
	UN24	CCACCGTTCCGTTATCCTC
XII (15)
chuA	ChuA.1	GACGAACCAACGGTCAGGAT	279	20 pmol
	ChuA.2	TGCCGCCAGTACCAAAGACA
yjaA	YjaA.1	TGAAGTGTCAGGAGACGCTG	211	20 pmol
	YjaA.2	ATGGAGAATGCGTTCCTCAAC
TSPE4.C2	TspE4C2.1	GAGTAATGTCGGGGCATTCA	152	20 pmol
	TspE4C2.2	CGCGCCAACAAAGTATTACG
XIII (71)
east1	east 11a	CCATCAACACAGTATATCCGA	111	50 pmol
	east 11b	GGTCGCGAGTGACGGCTTTGT
XIV (17)
cdt	CDT3A	GAGTTATTCCTTCCCCAGGC	108	50 pmol
	CDT3B	CAAAGGCATCAACAGCAGAA
XV (3)
paa	M155-F1	ATGAGGAAACATAATGGCAGG	350	50 pmol
	M155-R1	TCTGGTCAGGTCGTCAATAC
XVI (50)
saa	SAADF	CGTGATGAACAGGCTATTGC	119	50 pmol
	SAADR	ATGGACATGCCTGTGGCAAC
XVII (52)
ipah	ipaHIII	GTTCCTTGACCGCCTTTCCGATACCGTC	600	50 pmol
	ipaHIV	GCCGGTCAGCCACCCTCTGAGATAC
XVIII (28)
bfpA	EP1	AATGGTGCTTGCGCTTCGTGC	326	50 pmol
	EP2	GCCGCTTTATCCAACCTGGTA

^aThe forward and reverse primer sequences for each virulence gene, together with the size of the exception amplicon products, are also shown.

PCR analysis.

A series of 12 multiplex PCR sets (Table 3, sets I to XII) was adapted from published protocols and optimized for the amplification of 52 VGs. The remaining six VGs were individual PCR amplifications represented by sets XIII to XVIII (Table 3). PCR conditions were as described in Table S2 in the supplemental material. PCRs were conducted using a PC960 air-cooled thermal cycler (Corbett Research) with program cycles listed in Table S3 in the supplemental material. Amplicons were visualized by electrophoresis (80 V, 500 mA for 2.5 h for sets I to VI and 1.5 h for sets VII to XVIII) in 2% agarose gels prepared in 0.5× Tris-borate-EDTA (TBE) buffer (45 mM Tris base, 45 mM boric acid, 10 mM EDTA, pH 8) containing 4 μl of 5 μg/ml ethidium bromide. Amplicons were sized with corresponding 100-bp DNA markers (New England Biolabs) and processed in a Gel Doc system (Bio-Rad).

ECOR assignment.

E. coli isolates were assigned to one of the four main groups identified in the ECOR collection (32, 45) by the method of Clermont (15). Any strains that failed to yield amplicons for the three Clermont genes—chuA, yjaA, and TSPE4.C2—by PCR were further identified using the BBL Crystal enteric/nonfermenter identification system (Becton Dickinson) according to the manufacturer's protocol. When confirmed to be E. coli, these isolates were then classified as ECOR group A strains (26).

GLM.

The generalized linear model (GLM) was used to assess the significance of differences between groups of isolates (ND, PWD, and commensals) for each gene and to rank the relative importance of these genes based on the deviance value contributed to by group differences. Consider a model that can relate a dependent variable, x_jr (value 0 = absent or 1 = present), to the group parameters (β) according to a logistic function

(1)

(2)

where G is an N × t design matrix (values = 0 or 1), with N as the number of observations and t as the number of E. coli groups. A logit transformation was used to linearize the above relationship to the following form:

(3)

The group parameters (β) were then estimated using the maximum likelihood estimation, and the deviance value (chi square) was determined using a likelihood ratio test. Details of this method were described by Cox in 1970 (16). The comparisons between group means were determined by forming factors of group contrasts, and chi-square values associated with these contrasts were calculated.

Cluster analysis.

An agglomerative hierarchical algorithm was used to establish cluster relationships between E. coli strains essentially as described by Kaufman and Rousseeuw in 1990 (34a). Let x be an n × v data matrix with elements x_ir having values of 0 = absent or 1 = present where n is the number of isolates and v is the number of genes. A simple matching coefficient was calculated as follows: m₁₂ = 1 − d²ij/v and the sum of squared distances between all pairs of isolates to form a dissimilarity matrix, Q, with the element given by q_ij = −1/2 d²ij, where d²ij = Σ(x_ir − x_jr)².

The between-cluster dissimilarity, d(K1,K2), was then determined by the complete linkage method (the furthest neighbor method): that is

(4)

Once the longest distance was established for the first cluster, the same calculations were performed to compute the second longest cluster, and this was repeated until all cluster combinations were set within the framework defined by the first cluster.

PCO analysis.

The technique of PCO analysis was used to convert the data matrix of 58 VGs (or dimensions) into 2 or 3 major coordinates (reduced number of dimensions). The conversion of 58 dimensional data (v₁ … v₅₈) to the principal coordinates (PCO₁ … PCO_k) is represented by multiple linear equations as follows:

(5)

(6)

(7)

The coefficients (a₁,j, etc.) are latent vectors and are derived from an association matrix of “simple matching” coefficients and calculated via the principal component analysis technique. Gower (26a) in 1966 introduced the principal coordinate analysis by implementing the conventional principal component analysis on the squared distance matrix Q, which has latent root λ = [λ₁ λ₂ λ₃ … λ_n]^T and latent vector a, such that Qa = λa.

The methodology described in reference 42 was used to derive the latent root λ and the latent vector a. All computations were carried out using GenStat release 7.1 (64).

RESULTS

Bacterial serogroups and phenotypes.

The 75 porcine isolates were characterized for serogroup, hemolytic phenotype, and phylogenetic status as shown in Table 1. Three main serogroups represented by O8, O9, and O101 predominated in ND isolates, while PWD isolates were represented primarily by O8G7, O141, and O149.

All the commensal isolates belonged to serotypes that did not match either ND or PWD strains. Almost half of these were not typeable by currently available protocols. All ND and commensal isolates, with one exception each, were nonhemolytic on blood agar. In contrast, PWD isolates were all beta-hemolytic.

Phylogenetic status based on the Clermont virulence gene combination.

The phylogenetic groupings of porcine ETEC were established using Clermont's three-gene PCR for chuA, yjaA, and TSPE4.C2 (Table 1). On this basis, 90% and 81% of ND and PWD isolates, respectively, were classified as belonging to phylogenetic groups A (chuA yjaA TSPE4.C2 mutant or chuA yjaA⁺ TSPE4.C2 mutant) and B1 (chuA yjaA TSPE4.C2⁺) together with almost all the commensal strains. The phylogenetic group B2 was represented by only one isolate each in the ND and PWD collection. There was only one group D representative in ND (isolate 4) and commensal (isolate 56) isolates compared to five in the PWD assemblage (isolates 22, 23, and 26 to 28).

Distribution of virulence genes.

The identification of 58 VGs by PCR in 18 uni/multiplex sets is depicted in Fig. 1. Prior to this study, E. coli strains isolated from clinical cases of ND and PWD were analyzed for the presence of F4, F5, F6, F18, F41, STa, STb, and LT (Fig. 2a). Other VGs, including AIDA (aah, aidA, and AIDA^c), eaeA, stx₂, east1, and cdt, are found in both porcine and human isolates. As a consequence of this investigation, an additional 25 VGs can now be used to classify porcine clinical E. coli (Fig. 2b). Of the 25 VGs identified in porcine isolates, 19 of these VGs are primarily associated with ExPEC isolates from humans (34), with 4 associated with EHEC (iha, chuA, saa, and ehxA) and 1 associated with EPEC (saa).

FIG. 1.

Ethidium bromide-stained 2% agarose gel showing amplified bands corresponding to the 58 virulence genes associated with the representative pathotypes of E. coli known to cause intestinal and extraintestinal disease in humans and animals. Note that multiplex V contains an internal control and is identified as papG II/III.

FIG. 2.

The 58 virulence genes analyzed in this study that had previously been reported in porcine ETEC infections and human clinical isolates (a). Of the 58 virulence genes analyzed, 25 additional genes have now been identified in the porcine ETECs in this study (b). The superscript numbers represent the chi-square significance ranking.

Prevalence of significant virulence genes in commensal, ND and PWD isolates.

In this analysis, 17 out of 58 VGs were found to be significantly different (P < 0.05) for isolates from the three different sources of origin (Table 4). The chi-square ranking varied from 48.3 for STb to 7.5 for stx₂. Nine of these 17 genes represented by iha, hlyA, aidA, east1, aah, fimH, iroN_{E. coli}, traT, and saa have not been previously identified as important virulence genes in the identification of porcine diarrheagenic isolates. Not surprisingly, the prevalence of virulence genes encoding STb, STa, F18, LT, Stx2, F4, F5, and F41 was considered to be significant in differentiating between clinical and commensal isolates.

TABLE 4.

Prevalence of statistically significant genes (P < 0.05) by chi-square analysis of virulence genes in E. coli from scouring neonatal and weaner pigs and healthy pigs

Virulence gene	% of E. coli strains with gene			Chi-square value (P)a
	Neonatal	Weaner	Commensal
STb	25	81.3	0	48.33
iha	15	81.3	4.3	44.93
hlyA	10	71.9	4.3	37.55
STa	35	68.8	0	34.43
aidA (orfB)	10	56.3	0	30.12
F18	10	46.9	0	23.04
east1	10	40.6	17.4	18.82
LT	15	37.5	0	15.81
fimH	100	62.5	69.6	14.28
iroNE. coli	25	0	0	14.25
aah (orfA)	20	56.3	13	13.79
F4	5	37.5	4.3	11.25
traT	95	65.6	56.5	10.15
F41	0	0	17.4	9.98
F5	25	6.3	0	9.07
saa	40	9.4	13	7.56
stx2	5	34.4	13	7.47

^aA total of 17 of the 58 virulence genes were able to differentiate between the three groups based on chi-square analysis.

Cluster analysis—an association between virulence gene signatures and serogroups.

An agglomerative hierarchical algorithm was employed to form a gradual nesting of individual isolates into larger clusters. On this basis, all 75 isolates were distributed into one of two main branches (Fig. 3), A and B, which effectively partitioned neonatal diarrhea/commensal isolates from postweaning diarrhea strains, respectively. However, three ND isolates (no. 6 [O8:K88], 20 [O157:K88], and 15 [O99:Ksnt:H38]) and three commensal isolates (no. 53 [O8:H−], 56 [O40:H25], 66 [OR:H−]) were located in the B branch, while two PWD isolates (no. 24 [O8G7] and 25 [O45]) were in the A branch. Within the B branch, PWD isolates in B1, B2 to B4, and B5 were highly associated with their respective serogroup clusters: O8:G7, O141:K85, and O149:K88.

FIG. 3.

An agglomerative hierarchical algorithim was used to derive a cluster analysis dendrogram to establish the relationship between individual E. coli strains isolated from clinical cases of porcine neonatal and postweaning diarrhea and commensal isolates from healthy pigs based on the possession or nonpossession of the 17 significant virulence genes as determined by chi-square analysis.

Ownership of the 17 significant VGs by each strain is shown in Table S4 in the supplemental material. Two genes, fimH and traT, were for the most part present in both A and B branches. The traT gene was missing in almost all B3 members, while fimH was not detected in B5 strains. Unlike branch B strains, branch A strains were missing VGs corresponding to the STb, iha, STa, aidA, F18, east1, and LT genes that were signature genes for PWD strains. Virulence genes such as saa and iroN_{E. coli} were detected in some of the ND strains but were absent from PWD isolates (all members of branch B). A number of other ExPEC VGs that were not statistically significant were also detected occasionally in ND isolates. These genes include the bmaE, cdtB, cvaC, iutA, papA, papG allele I, papC, and K1 genes, which were never identified in PWD or commensal isolates.

Genetic relationships between commensal and clinical E. coli clones from scouring neonatal piglets and weaners.

PCO analysis was applied using the 17 defined chi-square-significant genes (Fig. 4) and the full panel of 58 VGs (Fig. 5). In both cases, 16 out of 20 ND isolates were clustered in quadrant 2 (Q2). Similarly, 22 out of 23 commensal isolates were also located in quadrant 2. Eighteen of the 32 PWD isolates were located in Q3, and 11 were located in Q4. Analyzing the ND, PWD, and commensal isolates with PCO using the 17 significant VGs resulted in a tighter cluster between ND and commensal isolates which resulted in isolate 1 moving from borderline Q2 to Q3 in the 58-VG analysis to Q1 in the 17-VG analysis. The analysis of the 17 VGs of the PWD isolates resulted in isolate 31 moving from Q1 to Q2. PWD isolates clustered in Q3 and Q4 (Fig. 4b) were also grouped coordinately with their serogroup.

FIG. 4.

Principal coordinate analysis using the 17 significant virulence genes (chi-square analysis, P < 0.05) to establish a spatial relationship between the 75 E. coli clones and their clinical origin. The coordinate values of scouring neonates (a [n = 1 to 20]), weaners (b [n = 21 to 52]), and commensals (c [n = 53 to 75]) are shown as a composite in panel d.

FIG. 5.

Principal coordinate analysis using all 58 virulence genes to establish a spatial relationship between the 75 E. coli clones and their clinical origin. The coordinate values of scouring neonates (a [n = 1 to 20]), weaners (b [n = 21 to 52]), and commensals (c [n = 53 to 75]) are shown as a composite in panel d.

DISCUSSION

Intestinal or extraintestinal strains of clinical and commensal E. coli isolates from different animal species, with or without clinical disease, have been found to be extremely diverse in their genetic makeup. Such diversity has been demonstrated by different analytical methods, including multilocus enzyme electrophoresis, multilocus sequencing typing (36, 62, 67, and 69), and the presence of virulence genes by PCR. Over time, genetic diversity has been exploited by selection and adaptation so that pathogenic strains have tended to become host specific, with strains identified in scouring pigs being phenotypically and serotypically different from those that cause diarrhea in humans (30). Within each host species, genetic differences can still be found in different pathogenic isolates. For example, E. coli strains associated with intestinal disease are genotypically different from those that cause extraintestinal disease. In pigs, a similar difference can be found between isolates responsible for ND and PWD. A still unanswered question is how do virulence genes continue to be acquired or lost in each “adapted” individual's evolutionary pathway? Commensal E. coli isolates on the other hand have attracted very little attention because they are not overtly involved in causing disease. These silent autochthonous bacteria, while nonpathogenic, could potentially also harbor virulence genes but are incapable of causing disease because they lack the appropriate virulence gene combinations.

To obtain a better understanding about the role of virulence genes in the pathogenicity of porcine isolates, a total of 58 virulence genes were assembled and their presence or absence in commensal, ND, and PWD isolates was analyzed. As shown in Fig. 6, 15 genes out of 58 were shared in common between clinical and commensal isolates; 11 genes were common between ND and PWD isolates. Nine ND, two PWD, and three commensal isolates also carried specific virulence genes that were not shared. Exclusion or inclusion of virulence genes as depicted in the Venn diagram can be biased in favor of minority genes because even a single gene that has occurred only once in a single isolate can be partitioned and included in the diagram. Notwithstanding this limitation, it is clear that a number of commensal isolates have acquired one or more virulence genes. For example, isolates 64 and 65, which belong to serogroup O130, may not possess pathogenic functionality because they have a nonpathogenic serogroup or lack the appropriate combination of virulence genes. However, it is conceivable that over time, further virulence gene acquisitions, particularly entire pathogenicity islands, can reequip such a commensal with the potential to develop into a pathogen.

FIG. 6.

Distribution of additional virulence genes identified in the ND, PWD, and commensal E. coli isolates.

According to the GLM, nine virulence genes represented by iha, hlyA, aah, aidA, east1, fimH, iroN_{E. coli}, traT, and saa, in addition to eight normally associated with pathogenic ETEC (coding for STb, STa, F18, LT, Stx2, F4, F5, and F41), were found to be significant in distinguishing between commensal, ND, and PWD clones. Iha has been described as a nonhemagglutinin adhesin found in O157:H7 and CFT073 (63). AIDA (adhesin-involved-in-diffuse-adherence) consists of the autotransporter adhesin heptosyltransferase (aah, formally orfA) gene which encodes the 44.8-kDa protein AAH, which modifies AIDA-I (orfB) by 19 heptose residues. AIDA-I remains covalently associated with the bacterial surfaces and is responsible for diffuse adherence (DA) patterns when E. coli cells attach to HeLa and HEp-2 cell culture lines (5-7). saa codes for an STEC autoagglutinating adhesin (50). fimH codes for d-mannose-specific adhesin or type 1 fimbriae (59, 60). hlyA α-hemolysin is a member of the RTX family of cytotoxins which is phenotypically observed on sheep blood agar (washed and unwashed) (11). east1 codes for an enteroaggregative E. coli heat-stable enterotoxin (53). iroN_{E. coli} codes for a novel catecholate siderophore. traT is one of the F factor genes involved in encoding the outer membrane protein; the TraT lipoprotein is responsible for surface exclusion activity (1, 2, 13).

One of the main limitations inherent with Venn diagrams is the inability to evaluate relationships between strains based on virulence gene combinations. An agglomerative hierarchical algorithm was used to establish cluster relationships between isolates based on mutual possession or nonpossession of all possible virulence gene combinations out of the 17 significant virulence genes examined in this study. As shown in Fig. 3, isolates clustered on the basis of virulence genes were grouped according to their serogroups and this in turn enabled PWD isolates to be distinguished from ND and commensal strains. A similar cluster analysis utilizing all 58 virulence genes yielded a clustering pattern that was essentially similar to that obtained with the panel of 17 significant virulence genes. The pattern of ownership of these 17 virulence genes provided an opportunity to identify subclusters that were closely associated with the serogroup of member strains (see Table S4 in the supplemental material). Thus, the B1, B2 to B4, and B5 strains subclustered in accordance with serogroups O8, O141, and O149, respectively. It is interesting to note in terms of gene acquisitions that ND and commensal isolates carried primarily ExPEC virulence genes, while the sources of PWD virulence genes were from a range of pathotypes including ETEC, ExPEC, EHEC/EPEC, EaggEC, and DAEC.

Principal coordinate analysis provided another method of examining relationships between individual isolates by reducing 58 virulence genes or dimensions to three principal coordinates, or three dimensions, that can be plotted along the x, y, and z axes. While the three-dimensional plots can provide a better spatial perspective, the two-axis plots shown in Fig. 4 and 5 provide valuable insight into the relationship between individual clones. The spatial relationship supports the previous analytical methods that ND, PWD, and commensal isolates can be segregated on the basis of their virulence gene signatures. Unlike the dendrogram analysis, PCO plots have positioned the commensal isolates within the ND quadrant showing that there is greater similarity between these because they have fewer virulence genes in common compared to PWD isolates. Two-dimensional PCO plots depict Q2 primarily containing ND and commensal isolates, while Q3 and Q4 are dominated by PWD isolates. This clustering of ND and commensal isolates together and the distinct separation of the PWD isolates were apparent in the PCO analysis using only 17 significant genes. A similar computation using the full panel of 58 virulence genes did not significantly alter the clustering pattern.

Until this study, only eight virulence genes (coding for F4, F5, F6, F18, F41, STa, STb, and LT) were recognized and used routinely to characterize porcine ETEC (20, 39, 48, 49). More recently, additional virulence genes have been identified in porcine ETEC, including aidA and aah (44), cdt (17), and east1 (14). This study has revealed a further 25 virulence genes from other pathotypes that have not previously been identified in porcine isolates. Of these, 19 are closely associated with ExPECs, and four members of the virulence genes in this group represented by fimH, hlyA, iroN_{E. coli}, and traT play a significant role in distinguishing between porcine commensal, ND, and PWD isolates (Table 4). The deployment of GLM and PCO methods of analysis was mandated by an accumulation of relatively large numbers of virulence genes acquired in the course of this study. In the past, when only eight virulence genes were used to characterize porcine ETEC, differences or similarities were recorded as percentages of occurrence of each gene in an assembly of isolates (18, 24, 46, 48, and 68). Such data essentially failed to demonstrate relationships between isolates and also did not relate virulence genes to serogroups and pathotypes. It is important to acknowledge that ownership of a virulence gene is not equivalent to its expression. Genes that get turned on to facilitate infection are usually dependent upon environmental queues emanating from community microorganisms (competition stress) or the host (immune stress) (33, 55).

A conclusion that can be drawn from this study is that virulence genes cannot fully define each pathotype. Instead, the virulence genes associated with each pathotype contribute to its functionality in causing characteristic symptomatology that typifies the disease syndrome. For instance, ND strains probably have a survival advantage over PWD strains because they do not require fimbriae to colonize the intestinal wall of the neonate with its immature and underdeveloped immune system. Under a different set of selection pressures, clones that possess disadvantageous virulence genes could be eliminated and replaced with other clones. Such changes can be associated with the increasing use of E. coli immunization of sows to generate protective antibodies in the colostrums (69). This would change the intestinal environment and profoundly alter the capacity of both commensals and pathogens to colonize the intestinal environment. The presence of virulence genes from human isolates in porcine E. coli strains should not be interpreted as a process of active virulence gene acquisition. The data suggests that both porcine commensal and clinical strains have over the years acquired and maintained these genes as part of a survival mechanism to engender greater diversity and hence increase their survival capability in the host animal.

Although the analytical techniques adopted in this study have focused on a panel of virulence genes representing different pathotypes of intestinal and extraintestinal E. coli, they can be applied to other gene panels, including those of genes that encode antibiotic resistance. As additional gene panels become available, the data matrix can be expanded and processed utilizing GLM, clustering, and PCO as the sequence format for biometric interrogation. The tedium of data acquisition using multiple gene analysis by PCR can be offset to some extent by multiplexing. Nevertheless, compared to DNA microarrays, PCR is relatively cost-effective (labor and capital) and allows a higher throughput for multiple clonal analyses. Furthermore, unlike DNA hybridization, the level of sensitivity and specificity of multiplex PCR is affected to a lesser extent by DNA concentrations and plasmid copy numbers. Unfortunately, virulence gene detection by PCR or microarray will still lack the level of definition provided by multilocus sequence polymorphism and cannot be used to detect subtle point mutations.