Virulence Factors in Urinary Escherichia coli Strains: Phylogenetic Background and Quinolone and Fluoroquinolone Resistance

Abstract

Quinolone- and fluoroquinolone-resistant Escherichia coli strains harbor fewer virulence factors than susceptible strains. The reasons underlying this correlation are incompletely understood. We investigated the phylogenetic background, the presence of the papC, hlyA, and cnf1 (pathogenicity island II_J96-associated), fimA, iss, and iutA genes, and the presence of type 1 fimbriae, P fimbriae, and hemolysin in 243 urinary E. coli isolates resistant only to quinolones (8%), resistant to both quinolones and fluoroquinolones (51%), or susceptible to both drugs (41%). Group B2 accounted for 56% of the isolates, showing a significantly higher prevalence among fluoroquinolone-susceptible strains than among resistant strains (65% versus 50% [P = 0.03]). hly and cnf1 were significantly more associated with susceptibility (P < 0.001) and with group B2 (P < 0.001 for group B2 versus groups A and D). However, within group B2, fluoroquinolone-resistant strains showed lower prevalences of papC, hlyA, and cnf1 than their susceptible counterparts (P < 0.001). In contrast, the incidence of iutA appeared higher for refractory isolates, including group B2, than for susceptible isolates (P < 0.001). Only in group B2 did fluoroquinolone-resistant strains reveal a lesser ability to agglutinate Saccharomyces cerevisiae (7%) than quinolone-resistant (87%) and susceptible (80%) isolates, despite uniform possession of fimA genes. No similar contrast emerged for expression of hemolysin and P fimbriae. Mutations conferring quinolone and fluoroquinolone resistance may thus require a particular genetic background, not strictly correlated with phylogenetic groups. More interestingly, the mutational event itself can affect the expression of type 1 fimbriae, at least in the prevalent and complex B2 strains.

Urinary tract infections (UTIs) in humans are the most frequent bacterial disease, affecting both inpatients and outpatients. Especially the uncomplicated cases are mainly due to extraintestinal pathogenic Escherichia coli (ExPEC) (39). In recent years, management of UTIs has become increasingly problematic due to the emergence of resistance to first-line antibiotics among the causative bacteria, particularly among uropathogenic E. coli (UPEC) strains. This phenomenon involves quinolones (Q) and fluoroquinolones (FQ) (25), drugs of paramount importance in the treatment of several other infectious diseases. Indeed, the renal excretion of these molecules and the availability of oral and parenteral formulations have allowed them to compete with aminoglycosides and beta-lactams in the therapy of complicated UTIs, especially in the hospital setting. Their appropriate spectrum and good tolerability have also led to increased empirical adoption in uncomplicated infections, although their usage for these conditions in outpatients is still under debate.

In E. coli, as in other gram-negative bacteria, DNA gyrase, which codifies type II DNA topoisomerases, is the main target of Q and FQ, and mutation in gyrA is the most common way to acquire resistance. In this connection, several investigations have shown that Q- and FQ-resistant UPEC strains display overall reduced virulence and invade compromised patients. By contrast, susceptible E. coli strains are virulent and affect uncompromised hosts.

At present it is not clear if this relationship is due to the significant association between susceptible strains and phylogenetic group B2, the most virulent group. In other words, it remains undecided whether drug-refractory E. coli strains are intrinsically less virulent bacteria or if they become less virulent following acquisition of the gyrA mutation (18).

There is clear evidence that the relationship among virulence properties of E. coli, phylogenetic background, and antibiotic resistance is a complex phenomenon, resulting from their different interplays. Therefore, we thought that these possibilities were not mutually exclusive.

It is also possible that the geographic source of isolates represents an important additional element to be taken into consideration (9). The analysis of a collection of UPEC strains from a particular region may therefore be useful in order to correlate the patterns of antibacterial resistance with local trends in the human usage of antibiotics and/or consumption of animal products.

To clarify whether the lack of virulence factors (VFs) is directly associated with resistance or, instead, depends on a phylogenetic distribution, we analyzed VFs, resistance phenotype, and biotypes in a collection of 243 E. coli strains. The isolates were from urinary specimens obtained from inpatients and outpatients at our institution during 2005 and 2006. We considered VFs such as fimA and papC genes, whose products (type 1 fimbriae and P pili) are instrumental in the pathogenesis of UTIs. The presence of hlyA and cnf1 was also investigated because of the physical linkage of these genes with papC in pathogenicity island II_J96 (PAI II_J96)-like domains, which are prevalent in UTI isolates (41), whereas iutA and iss were studied because they are known not to be linked to PAIs.

In order to further explore virulence itself, the phenotypic expression of type 1 fimbriae, P fimbriae, and hemolysin was also assessed in this collection.

MATERIALS AND METHODS

Patients and bacterial isolates.

E. coli isolates were collected from July 2005 to November 2006 from urinary specimens that were either sent to the Clinical Microbiology Laboratory of San Martino Hospital by 16 internal medicine, 4 surgery, and 11 day care wards or collected from outpatients directly referred to our center. A single strain from each patient was analyzed.

Samples were derived either from clean-catch, midstream urine or from urinary catheters. Specimens were cultured (10 μl) on MacConkey agar. Lactose-fermenting, indole-positive colonies were evaluated by the BBL Crystal method (Becton Dickinson) to identify E. coli. Bacteria included in the study were from cultures yielding >10⁵ CFU/ml.

Susceptibilities to pipemidic acid, norfloxacin, and ciprofloxacin were tested by disk diffusion techniques, using CLSI (formerly NCCLS) criteria (34).

Phylotyping, virulence genotyping, and phenotyping.

E. coli isolates were grouped into phylogenetic biotypes (A, B1, B2, and D) according to the presence of the chuA and yjaA genes and TspE4.C2 (anonymous DNA fragment) by using multiplex PCR (6). Six ExPEC-associated VFs, fimA, papC, cnf1, hlyA, iutA, and iss (encoding, respectively, type 1 fimbriae, P fimbriae, cytotoxic necrotizing factor-1, hemolysin, the aerobactin iron transport system, and an increased-serum-survival factor), were also sought by using single PCR as described elsewhere (19, 37).

The presence of type 1 fimbriae was determined on the basis of mannose-sensitive agglutination of Saccharomyces cerevisiae, while P fimbriae were evidenced by P blood group-dependent hemagglutination, after in vitro passage to enhance their expression (1, 26). The production of hemolysin was evaluated visually by the appearance of a clear halo on 5% sheep blood agar plates after overnight culture at 37°C.

Statistical analysis.

Statistical analysis was performed by using the Fisher exact and chi-square tests. The threshold for statistical significance was a P value of <0.05.

RESULTS

Characteristics of patients.

A total of 243 UPEC isolates were analyzed. Of these, 191 derived from midstream urine and 52 were from urinary catheters; 150 patients were in the internal medicine ward (32% with urinary catheters) and 10 were in the surgery ward, while 83 specimens came from outpatients (43 in day care wards and 40 in our outpatient department). There were 55 males (23%) and 188 females (77%) (data not shown).

Q and FQ resistance and bacterial characteristics.

Of 243 E. coli isolates, 19 (8%) were resistant only to the Q (pipemidic acid) while 123 (51%) were also resistant to FQ (norfloxacin and ciprofloxacin), while 101 (41%) were susceptible to both classes of drugs.

The incidences of global resistance to Q and FQ according to the different sources of specimens were 52% (no catheter) and 83% (urinary catheter); incidences according to the type of patient were 64% for inpatients (66% of those in the internal medicine ward and 42% of those recovering from surgery) and 46% for outpatients (43% of day care patients and 49% of the patients referred directly to our laboratory) (data not shown).

When the incidence of Q and FQ resistance was correlated with bacterial characteristics, i.e., phylogenetic groups and VFs, the following results were obtained (Table 1). A significant prevalence of phylogenetic group B2 (56%) was apparent with respect to the other biotypes (A, B1, and D) (Table 1). Despite the fact that B2 represented the most frequent group among the susceptible as well as the resistant bacteria, the prevalence of B2 strains was significantly higher among susceptible strains than among FQ-resistant strains (65% versus 50%, respectively; P = 0.03). Among susceptible strains, in fact, the frequencies of groups A, B1, and D were, respectively, 16%, 5%, and 14%, while among FQ-resistant strains these frequencies were 22%, 9%, and 19% (Table 1).

TABLE 1.

Distribution of phylogenetic groups and VFs among 243 Escherichia coli isolates from UTIs according to antimicrobial resistance phenotypes

Phylogenetic group or VF gene	No. (%) of E. coli isolatesa				Pb
	Total (n = 243)	Susceptible (n = 101)	Q resistant (n = 19)	FQ resistant (n = 123)	Susceptible vs Q resistant	Susceptible vs FQ resistant	Q resistant vs FQ resistant
Phylogenetic groups
A	46 (19)	16 (16)	3 (16)	27 (22)
B1	19 (8)	5 (5)	3 (16)	11 (9)
B2	136 (56)	66 (65)	8 (42)	62 (50)		0.030
D	42 (17)	14 (14)	5 (26)	23 (19)
VFs
fimA	222 (91)	96 (95)	18 (95)	108 (88)
papC	67 (28)	49 (48)	5 (26)	13 (11)		<0.001
hlyA	50 (21)	44 (44)	2 (10)	4 (3)	0.009	<0.001
cnf1	47 (19)	42 (42)	2 (10)	4 (3)	0.010	<0.001
iss	75 (31)	31 (31)	12 (63)	32 (26)	(0.009)		[0.002]
iutA	143 (59)	32 (32)	14 (74)	97 (79)	(0.001)	(<0.001)

^aAll susceptible isolates were susceptible to both quinolone and fluoroquinolones. Q resistant, resistant only to quinolone (pipemidic acid). FQ resistant, resistant both to quinolone and to fluoroquinolones (norfloxacin and ciprofloxacin).

^bOnly P values of <0.05(by the Fisher exact test) are shown. P values without parentheses indicate a positive association of the biotype or VF with susceptible isolates; those in parentheses indicate a negative association of the VF with susceptible isolates; those in brackets indicate a negative association of the VF with fluoroquinolone resistance.

Susceptible isolates exhibited significant differences from both Q- and FQ-resistant isolates with regard to the distribution of papC, hlyA, and cnf1. The most significant differences emerged between susceptible and FQ-resistant isolates (P < 0.001): 48%, 44%, and 42% of the susceptible strains possessed papC, hlyA, and cnf1, respectively, whereas 11% of the FQ-resistant strains harbored papC and 3% possessed hlyA and cnf1. On the other hand, susceptible E. coli strains exhibited a significantly lower incidence of iutA (32%) than Q (74%)- and FQ (79%)-refractory strains (P = 0.001 for both comparisons). Q-resistant isolates exhibited a significant association with iss in comparison with susceptible isolates (P = 0.009) and particularly in comparison with FQ-resistant strains (P = 0.002). The presence of fimA did not differ according to the antimicrobial susceptibility phenotype (Table 1).

The Q- and FQ-resistant isolates from urinary catheters did not have lower incidences of VFs than resistant E. coli isolates obtained from midstream urine (data not shown).

Phylogenetic distribution of VFs.

For all VFs considered, a very significant phylogenetic distribution emerged (Table 2). Despite the fact that fimA was detected at least in 83% of the isolates (groups A and D), some significant differences among the biotypes were observed with regard to the presence of fimA, which exhibited an association with group B2 (P = 0.008 for B2 versus A; P = 0.014 for B2 versus D). However, we found the highest differences among the biotypes with regard to the presence of hly and cnf1. In fact, hly and cnf1 were associated with phylogenetic group B2; about one-third of B2 isolates possessed both traits (34% and 33%, respectively). hly and cnf1 were nearly absent in the remaining groups, especially groups A and D (P < 0.001 for B2 versus A and for B2 versus D). In contrast, the presence of papC in group D was similar to that in group B2 (37% of group B2 versus 26% of group D isolates; P, not significant), while it was much lower in groups A and B1 (P < 0.001 for B2 versus A) (Table 2). A significant association was observed between the presence of iss and biotype B1 (P < 0.05 for B1 versus B2 and for B1 versus D). A similar association was observed between the presence of iutA and biotype D (P < 0.05 for B1 versus D) (Table 2).

TABLE 2.

Distribution of VFs among 243 Escherichia coli isolates from UTIs according to phylogenetic groups

VF	No. (%) of E. coli isolates					Pa
	Total (n = 243)	Group A (n = 46)	Group B1 (n = 19)	Group B2 (n = 136)	Group D (n = 42)	A vs B2	A vs D	B1 vsB2	B1 vs D	B2 vs D
fimA	222 (91)	38 (83)	19 (100)	130 (95)	35 (83)	0.008				0.014
papC	67 (28)	4 (9)	1 (5)	51 (37)	11 (26)	<0.001	(0.045)	0.004
hlyA	50 (21)	2 (4)	1 (5)	46 (34)	1 (2)	<0.001		0.014		<0.001
cnf1	47 (19)	0 (0)	1 (5)	45 (33)	2 (5)	<0.001		0.014		<0.001
iss	75 (31)	19 (41)	10 (53)	36 (26)	10 (24)			[0.030]	[0.039]
iutA	143 (59)	26 (56)	7 (37)	82 (60)	28 (67)				(0.049)

^aOnly P values of <0.05 (by the Fisher exact test) are shown. P values without parentheses indicate a positive association of the VF with biotype B2; parentheses indicate a positive association of the VF with biotype D; brackets indicate a positive association of the VF with biotype B1.

Expression of VFs in relation to phylogenetic background.

In order to assess the capability of E. coli isolates for phenotypic expression according to biotype, we evaluated the presence of type 1 fimbriae, P fimbriae, and hemolysin through mannose-sensitive agglutination, mannose-independent hemagglutination, and detection of hemolysis (Table 3). No isolates without fimA, papC, and hlyA genes showed the expected phenotype (data not shown). The number of isolates with phenotypic expression of type 1 fimbriae, P fimbriae, or hemolysin was calculated as a percentage of the number of isolates carrying the corresponding gene fimA, papC, or hlyA, respectively. The results showed that fimA was present in most isolates and type 1 fimbriae were expressed by the different groups with no significant difference. There was no difference in hemolysin expression among the four biotypes, though hlyA was associated only with biotype B2 (Table 2). The expression of P fimbriae in biotypes B2 and D (29% and 18%, respectively) did not permit a statistical correlation because of the small numbers in group D.

TABLE 3.

Frequency of VF expression in relation to phylogenetic groups among Escherichia coli isolates from UTIs carrying fimA, papC, and hlyA

Phenotype or gene	No. of E. coli isolatesa
	Total	Group A	Group B1	Group B2	Group D
Phenotypic expression
Type 1 fimbriae	115 (52)	18 (47)	12 (63)	63 (48)	22 (63)
P fimbriae	18 (27)	1 (25)	0 (0)	15 (29)	2 (18)
Hemolysin	47 (94)	2 (100)	1 (100)	43 (93)	1 (100)
Gene presence
fimA	222	38	19	130	35
papC	67	4	1	51	11
hlyA	50	2	1	46	1

^aNumbers in parentheses are percentages of strains carrying the single VF gene fimA, papC, or hlyA that express type 1 fimbriae, P fimbriae, or hemolysin, respectively. The presence of type 1 fimbriae, P fimbriae, and hemolysin was evaluated according to the occurrence of mannose-sensitive agglutination of Saccharomyces cerevisiae, hemagglutination, and hemolysis. No significant P value (<0.05) emerged from the comparisons of the four E. coli groups in relation to the expression of any VF.

Virulence genotype in relation to antimicrobial resistance.

The results from Tables 1 and 2 highlight a preliminary connection among susceptibility, VFs linked to PAI II_J96, and phylogenetic group B2, which, however, does not explain why 50% of the resistant isolates belong to B2 but only 3% have hlyA and cnf1. Therefore, to assess whether the absence of these virulence genes is also specifically related to drug resistance, we evaluated the frequency of each VF considered (fimA, papC, hlyA, cnf1, iss, and iutA) in susceptible or resistant (to Q and/or FQ, globally considered) E. coli isolates within each phylogenetic group (A, B1, B2, and D) (Table 4). The results indicate that within each biotype, as in the total strains (Table 1), the prevalence of the fimA gene was identical in susceptible and resistant E. coli isolates. papC, associated with group D as well as with B2 (Table 2), showed the same prevalence in susceptible (29%) and resistant (25%) group D isolates. hlyA and cnf1 were nearly absent in biotypes A, B1, and D (Table 2), independently of the phenomenon of resistance. Thus, besides the papC distribution in group D, it was possible to compare the frequencies of papC, hlyA, and cnf1 according to the resistance phenotype only within group B2. In susceptible E. coli B2 strains, the incidences of papC, hlyA, and cnf1 were, respectively, 68%, 62%, and 61%, whereas in resistant B2 isolates, the incidence was 9% for papC and 7% for hlyA and cnf1 (P < 0.001).

TABLE 4.

Distribution of VFs within each phylogenetic group according to antimicrobial resistance phenotype among 243 Escherichia coli isolates from UTIs

VF	No. (%)a of group A isolates (n = 46)		Pb	No. (%) of group B1 isolates (n = 19)		No. (%) of group B2 isolates (n = 136)		P	No. (%) of group D isolates (n = 42)
	S (n = 16)	R (n = 30)		S (n = 5)	R (n = 14)	S (n = 66)	R (n = 70)		S (n = 14)	R (n = 28)
fimA	14 (87)	24 (80)		5 (100)	14 (100)	65 (98)	65 (93)		12 (86)	23 (82)
papC	0 (0)	4 (13)		0 (0)	1 (7)	45 (68)	6 (9)	<0.001	4 (29)	7 (25)
hlyA	2 (12.5)	0 (0)		0 (0)	1 (7)	41 (62)	5 (7)	<0.001	1 (7)	0 (0)
cnf1	0 (0)	0 (0)		0 (0)	1 (7)	40 (61)	5 (7)	<0.001	2 (14)	0 (0)
iss	5 (31)	14 (47)		1 (20)	9 (64)	22 (33)	14 (20)		3 (21)	7 (25)
iutA	4 (25)	22 (73)	(0.004)	0 (0)	7 (50)	22 (33)	60 (86)	(<0.001)	6 (43)	22 (79)

^aPercentage of susceptible or resistant isolates with the indicated VF. S, susceptible to both Q and FQ; R, resistant. In each phylogenetic group (A, B1, B2, or D), resistant E. coli isolates include Q-resistant and FQ-resistant strains.

^bP values (by the Fisher exact test) are for comparisons of isolates susceptible and resistant to the drugs within in a single biotype. Only P values of <0.05 are shown. Parentheses indicate a negative association of the VF with susceptible isolates. No comparison involving groups B1 and D yielded a P value of <0.05.

Our results also indicate that in each phylogenetic group, the iss trait showed no significant differences between susceptible and resistant isolates, whereas the iutA trait, in groups A and B2, showed a significant shift toward the phenotype of resistance (in biotype A, P = 0.004; in biotype B2, P < 0.001) (Table 4).

Virulence phenotype in relation to antimicrobial resistance.

To evaluate whether the capability for phenotypic expression in E. coli isolates, which has been shown not to be related to biotypes (Table 3), is instead related to resistance, we analyzed the presence of type 1 fimbriae, P fimbriae, and hemolysin in susceptible and resistant (to Q and/or FQ, globally considered) strains within groups A, B1, B2, and D. The number of isolates with phenotypic expression of type 1 fimbriae, P fimbriae, or hemolysin was calculated as a percentage of the number of isolates carrying the corresponding gene fimA, papC, or hlyA, respectively (Table 5). Only in group B2 did the expression of type 1 fimbriae greatly decrease, from 80% in susceptible strains to 17% in resistant strains (P < 0.001).

TABLE 5.

Association of VF expression with antimicrobial resistance phenotypes within each phylogenetic group among 243 Escherichia coli isolates from UTIs carrying fimA, papC, or hlyA

Phenotype or gene	No. of E. coli isolatesa
	Group A		Group B1		Group B2		Group D
	S	R	S	R	S	R	S	R
Phenotypic expression
Type 1 fimbriae	7 (50)	11 (46)	5 (100)	7 (50)	52 (80)	11 (17)b	8 (67)	14 (61)
P fimbriae	0	1 (25)	0	0	14 (31)	1 (17)	1 (25)	1 (14)
Hemolysin	2 (100)	0	0	1 (100)	39 (95)	4 (80)	1 (100)	0
Gene presence
fimA	14	24	5	14	65	65	12	23
papC	0	4	0	1	45	6	4	7
hlyA	2	0	0	1	41	5	1	0

^aNumbers in parentheses are percentages of strains carrying the single VF gene fimA, papC, or hlyA that express type 1 fimbriae, P fimbriae, or hemolysin, respectively. The presence of type 1 fimbriae, P fimbriae, and hemolysin was evaluated according to the occurrence of mannose-sensitive agglutination of Saccharomyces cerevisiae, hemagglutination, and hemolysis. S, susceptible to both Q and FQ; R, resistant. In each phylogenetic group (A, B1, B2, or D), resistant isolates include Q-resistant and FQ-resistant strains.

^bThe P value (by Fisher's exact test) for comparison of expression of type 1 fimbriae by susceptible and resistant isolates within group B2 is <0.001, indicating a positive association of VF expression with susceptible isolates. No other comparison yielded a P value of <0.05.

In groups A, B1, and D, the low level of expression of type 1 fimbriae seems not to be dependent on the phenomenon of resistance (P values for comparisons of susceptible and resistant isolates are not significant) (Table 5).

From these results, in the biotypes carrying the papC gene (biotypes B2 and D), it emerged that the low global expression of P fimbriae, 27% (Table 3), does not depend on the resistance phenomenon; in fact, the P values for comparisons of susceptible and resistant isolates are not significant.

It was impossible to evaluate the different levels of expression of hlyA between susceptible and refractory E. coli strains in either group B2 or non-B2 groups because of the very low incidence of the gene in groups A, B1, and D and in resistant isolates, including those of biotype B2.

Virulence phenotype and genotype of group B2 in relation to classes of resistance.

Since among the resistant (globally considered) E. coli strains in group B2, 11 isolates (17%) still agglutinate S. cerevisiae (Table 5), we investigated the incidences of type 1 fimbriae, P fimbriae, and hemolysin in biotype B2 in order to determine whether diversity between the two different classes of resistance, resistance to Q and resistance to FQ, might provide a possible explanation (Table 6).

TABLE 6.

Association of VF expression with resistance to quinolone and fluoroquinolones among Escherichia coli isolates from UTIs belonging to phylogenetic group B2 and carrying fimA, papC, and hlyA

Phenotype or gene	No. of E. coli isolatesa
	Q resistant	FQ resistant
Phenotypic expression
Type 1 fimbriae	7 (87)	4 (7)b
P fimbriae	0	1 (33)
Hemolysin	2 (100)	2 (66)
Gene presence
fimA	8	57
papC	3	3
hlyA	2	3

^aNumbers in parentheses are percentages of strains carrying the single VF gene fimA, papC, or hlyA that express type 1 fimbriae, P fimbriae, or hemolysin, respectively. The presence of type 1 fimbriae, P fimbriae, and hemolysin was evaluated according to the occurrence of mannose-sensitive agglutination of Saccharomyces cerevisiae, hemagglutination, and hemolysis.

^bThe P value (by Fisher's exact test) for comparison of expression of type 1 fimbriae by Q-resistant and FQ-resistant isolates is <0.001, indicating a positive association of VF expression with isolates resistant to Q. No other comparison yielded a P value of <0.05.

The additional results showed that of the 11 group B2 resistant strains showing mannose-sensitive agglutination, 7 strains were resistant to pipemidic acid alone and only 4 were also resistant to FQ, thus modifying the percentage of resistant E. coli B2 strains retaining the expression of type 1 fimbriae (7% of FQ-resistant strains). The percentage of S. cerevisiae agglutination among the 7 B2 strains resistant only to Q (87%) was similar to that detected for the susceptible strains (80%) (Table 5).

Despite the small sample size of the resulting Q resistance subgroup, which makes the statistical significance uncertain, an additional trend with an interesting biological significance emerged from the results. An important difference in the expression of type 1 fimbriae appears to exist, in fact, between the two classes of resistance (87% for Q resistance versus 7% for FQ resistance; P < 0.001). In contrast, no differences in the detection of hemagglutination or hemolysis emerged between the classes of resistance.

The results relating to phenotypic expression of type 1 fimbriae led us to investigate more closely the presence of VF genes in Q-resistant versus FQ-resistant E. coli isolates within biotype B2 (Table 7). The Q-resistant and FQ-resistant strains differed with regard to the presence of the iss and papC genes, which were more prevalent among the former (62% of Q-resistant versus 14% of FQ-resistant strains had iss; P = 0.006). Although a statistical limitation due to the small sample size still exists, these data confirmed the percentages of iss presence in Q-resistant versus FQ-resistant strains (P = 0.002) that were obtained without separating the phylogenetic backgrounds (Table 1).

TABLE 7.

Association of VF genes with resistance to quinolone and fluoroquinolones among Escherichia coli isolates from UTIs within phylogenetic group B2

VF	No. (%) of E. coli isolates
	Q resistant (n = 8)	FQ resistant (n = 62)	Pa
fimA	8 (100)	57 (92)
papC	3 (37)	3 (5)	0.017
hlyA	2 (25)	3 (5)
cnf1	2 (25)	3 (5)
iss	5 (62)	9 (14)	0.006
iutA	5 (62)	55 (89)

^aOnly P values (by Fisher's exact test) of <0.05 are shown. P values for comparison of Q-resistant and FQ-resistant isolates indicate positive associations of VFs with isolates resistant to Q.

DISCUSSION

Our findings confirmed the complexity of the association between antibacterial resistance and the properties of virulence in Escherichia coli.

Previous investigations clearly demonstrated that resistance of E. coli strains to several antimicrobial agents, e.g., ampicillin, aminoglycosides, and co-trimoxazole, does not significantly correlate with the presence of fewer VFs than in their susceptible counterparts (16, 23, 31). We therefore focused only on the characteristics of Q- and FQ-resistant UPEC strains and examined whether our results mimic or differ from those obtained in other regional settings.

We evaluated, in UPEC strains, the two different opinions currently under debate: the first assuming that the low frequency of certain VFs precedes resistance (22, 23, 33) and the second assuming that, vice versa, the low frequency of these VFs follows the acquisition of resistance (8, 16, 44). We explored a possible coexistence of both mechanisms.

As previously found in other studies (21), our results show that group B2 is globally the most frequent E. coli biotype in UTIs. The same prevalence does not hold true among Q- and FQ-refractory strains in studies that also consider different extraintestinal sources (17, 22).

A greater prevalence of biotype B2 was detected in susceptible than in FQ-resistant E. coli strains, and the association of certain virulence determinants with group B2 and with susceptible isolates was confirmed as well (22, 23, 33).

We also confirmed previous knowledge about the genes mostly present in all isolates (fimA), those nearly absent both in non-B2 isolates and in resistant isolates (hlyA and cnf1, which show a physical linkage in PAI II_J96) (4, 38, 41), and the intermediate percentages of papC presence (8, 16, 22, 23, 33, 44).

We might therefore agree with the idea that the higher incidence of VFs among susceptible E. coli isolates would depend on their phylogenetic distribution and that VFs would be intrinsic bacterial characteristics. Nevertheless, this relationship does not clarify the fact, previously found (33), that despite the predominance of group B2 (50%), FQ-resistant UPEC strains did not harbor VFs such as hlyA (3%), cnf1 (3%), and papC (11%).

This discrepancy was explained by the finding of lower incidences of papC, hlyA, and cnf1 among refractory group B2 strains than among susceptible B2 strains, which seems to be due to a loss of the corresponding PAI, probably as a result of the mutation that causes resistance, as already assessed through the same approach (16). However, among our clinical urinary isolates, the high presence of susceptible E. coli strains lacking papC (52%), hlyA (56%), and cnf1 (58%) might imply that the “lack” of PAI II_J96 does not require the mutation leading to resistance. Experimental approaches suggested that even the “loss” of PAI, obtained in the presence of ciprofloxacin, does not require the mutation in the codon for Ser-83 of the gyrA gene in E. coli HC14366 and HC109 (40). Furthermore, in the clinical E. coli isolates that were rendered FQ resistant through selection in the presence of ciprofloxacin, no loss of PAIs occurred (20, 31). Moreover, E. coli 536 shows a loss of PAIs that appears without gyra mutation and without any antibiotics in the medium (32), indicating that their instability could be exacerbated by several stimuli (32), of which the presence of quinolone is just one (29).

Thus, we speculated that in UPEC, the lack of PAIs does not arise from the gyrA mutation but, instead, both are consequences of chromosomal characteristics. Although the instability and the loss of PAI₅₃₆ and PAI II_J96 have been demonstrated to be peculiar to single PAIs themselves (15, 41), we thought that these phenomena also required a specific genetic background. In E. coli, in fact, the occurrence of multiple insertional events (4) and the acquisition of VFs (10) have been shown to be restricted by a specific genetic structure.

Our results, moreover, showed that Q- and FQ-resistant isolates not harboring papC, hlyA, and cnf1 are positively associated with the iutA gene, which is not located on PAIs. Since a similar relationship also emerged for iss and, through the investigations cited above, for other genes harbored outside the PAIs, such as bmaE, gafD, ireA, and cvaC, the hypothesis of a direct linkage between iutA and a resistance gene, occurring on the same plasmid, was discarded. Furthermore, in Escherichia coli, a plasmid-mediated aerobactin system and plasmid-mediated resistance to ciprofloxacin cannot have a considerable effect at the population level because of their rarity (42, 45). Thus, we believed iutA to be a chromosomal gene. The prevalence of iutA among clinical refractory isolates, even those in group B2, and the prevalence of iss among Q-resistant isolates, as well as the nearly total absence of genes linked to PAI II_J96, led us to believe that mutations in gyrA strictly require specific genetic characteristics. These are the genetic background of groups A, B1, and D and of a particular subgroup, or pathotype (5), of B2 resembling non-B2 isolates. From our results it emerged, in fact, that biotype B2 shows characteristics of flexibility, previously described as a bacterial mosaic-like genetic tool (12). In contrast, both susceptible and resistant isolates in non-B2 groups are more homogeneous, since they typically do not harbor PAI genes, hlyA, and cnf1. Since papC possesses a less consistent colocalization on PAIs (4), this VF shows an intermediate percentage of presence in group D isolates and in refractory isolates. We thus believe that in E. coli the association between Q and FQ resistance and the paucity of PAI-related genes does not depend on a phylogenetic distribution but instead is direct. A specific chromosomal background, only partially corresponding to the phylogenetic background, could therefore precede gyrA mutation.

We also demonstrated, for the resistant clinical E. coli isolates, decreased agglutination of S. cerevisiae, already previously described (2, 44) and explained through gyrA mutation affecting transcriptional events (3, 27, 28). Since we assessed the mannose-sensitive agglutination within each biotype, we found this phenomenon to be exclusive of group B2, and we could thus explain the more significant results with respect to those obtained without phylogenetic distinction in the previous works.

The hypothesis that lower expression of type 1 fimbriae would be an intrinsic characteristic of a group of strains was rejected. The different groups, in fact, showed the same capacity for phenotypic expression, and in particular, resistant E. coli non-B2 isolates, not carrying hlyA and cnf1, showed functional type 1 fimbriae. It thus emerged that low phenotypic expression of type 1 fimbriae depends neither on the genetic nor on the phylogenetic background but strictly on the phylogenetic background of B2 plus resistance. Among resistant E. coli strains, it was almost exclusively non-B2 groups that expressed type 1 fimbriae, except for B2 “quinolone-only”' resistant isolates, which carry a single mutation, conferring a low-level resistance that allows them to retain susceptibility to FQ (11, 43) and fimbrial phenotypic expression. This result would explain the relatively high number (11 strains; 17%) of resistant isolates still expressing type 1 fimbriae and would confirm that in UPEC strains, low expression of certain VFs follows acquisition of a certain type of resistance.

A common finding emerging from several aspects of our study is the complexity of lifestyle (of being and behaving) of E. coli isolates belonging to the B2 phylogenetic group. B2 appears to be the only biotype showing a strict relationship between resistance and low virulence, resulting from two different mechanisms that act simultaneously. This characteristic of group B2 can determine the relationship globally found in UPEC strains, because it concerns the most frequent biotype. We also supposed that the prevalence of group B2 among both uropathogenic and commensal E. coli isolates (21, 46), found recently to be independent of VF carriage (35), would be due to a superior adaptability, perhaps corresponding to the previously named “plasticity” of the species (5, 7, 15). In contrast, in sites other than the bladder and the gut, the presence of B2 strains, corresponding to disease (21) and to experimental lethality (24), depends on a high level of shared virulence factors.

Our results showing the high incidence of ciprofloxacin-resistant urinary E. coli isolates among outpatients and, especially, the same finding in the fecal microflora of children (36) (supposed to be naïve hosts with regard to FQ contact) make the natural mechanisms of resistance not very intelligible. However, this aspect led us to believe that, for B2 isolates, the advantage arising from an assumed instability is not, or not only, resistance itself. In fact, the reduced expression of adhesins, such as type 1 fimbriae, can be not a mere loss to pay out but, instead, a gain for the bacterial fine-tuning in the environment. Given the evolutionary success of E. coli B2 strains, the exclusive lack of type 1 fimbriae among the resistant strains could be regarded as a phase variation-like phenomenon (30), as can the low expression of P fimbriae among susceptible isolates. The lack of fimbrial antigens, like the absence of VF traits linked to PAI II_J96, is irreversible but would be adapted to an appropriate context in which bacteria, without particular damage and through avoidance of host defenses, achieve new niches where they colonize or cause chronic infections (13, 14, 15), spreading possible resistance.