Distribution of

Background: Extraintestinal Escherichia coli (E. coli) causing urinary tract infections (UTIs), and often referred to as uropathogenic E. coli (UPEC), are a major contributor to the morbidity of UTIs and associated healthcare costs. UPEC possess several virulence factors (VFs) for infecting and injuring the host. We studied the papG allele distribution, and its association with other VF genes and phylogenetic groups, amongst 836 UPEC and fecal isolates from reproductive age women. Results: The papGII gene was highly prevalent amongst pyelonephritis isolates (68%), whilst the majority, albeit smaller proportion, of cystitis isolates (31%) harboured the papGIII gene. Among the pyelonephritis and cystitis isolates, papG positive isolates on average had higher VF gene scores, and were more likely to belong to phylogenetic group B2, than their negative counterparts. This was mostly due to the contribution of papGII isolates, which on average contained more VF genes than their papGIII counterparts, irrespective of the uro-clinical syndrome. However, the papGII isolates from the pyelonephritis cohort had higher VF gene scores than the cystitis ones, suggesting presence of possible papGII clones with differing inferred virulence potential. Furthermore, papGII isolates were more likely to possess an intact pap gene operon than their papGIII counterparts. Also of note was the high proportion of isolates with the papGI allele which was not associated with other pap operon genes; and this finding has not been described before. Conclusions: The association of the papGII gene with several VF genes compared to the papGIII gene, appears to explain the abundance of these genes in pyelonephritis and cystitis isolates, respectively.

groups as per the revised molecular typing technique. Furthermore, few studies have analysed the relationship between papG genotypes and phylogenetic group status, to shed some light into the epidemiological relationships between the two. Therefore, we studied urinary E. coli isolates from reproductive age women with UTI, and fecal isolates from healthy controls, from the same geographical area (Central West region, NSW, Australia) and time period, in order to define; (i) the papG class and phylogenetic group distribution among the isolates; (ii) the relationship between VF gene distribution and papG class type; (iii) the relationship between papG genotypes and phylogeny; and finally, (iv) the inferred virulence potential of cystitis and pyelonephritis isolates by papG allele genotype status.

Study design and setting
This prospective study included 11 regional hospitals and 23 outpatient centers, in the Central West region of New South Wales (NSW), Australia. It was conducted with the help of volunteer physicians in the participating centers. A guiding protocol for urine collection and a strict clinical diagnostic criteria for defining cystitis or pyelonephritis, was given to each participating physician. Based on the medical history and patient physical examination, physicians recorded the following information in each patient recruited into the study; de-identified patient information: age, clinical UTI syndrome, previous UTI history, and any known underlying host conditions.

E. coli strains and study subjects
Consecutively collected E. coli isolates (n = 601) from mid-stream urine (MSU) specimens of non-pregnant, reproductive-age (i.e., age 15-45 years) women with cystitis (n = 307) or, pyelonephritis (n = 294), as defined below, or from fecal specimens of healthy women without urinary symptoms (n = 235), were studied. Fecal controls were matched for sex and age within the overall age group. The isolates were collected over a 2-year period (June 2009-July 2011), with only one isolate per subject included in the study. To allow a focus on bacterial characteristics, with minimal confounding by host characteristics, patients with known diabetes mellitus, diarrhoea, antibiotic therapy in the last month, menstruation, or urinary tract abnormalities, were excluded. Urinary tract abnormality was defined based on the attending physician's assessment at the time of the index patient encounter. Urological evaluation to exclude inapparent structural or functional genitourinary abnormalities was not done. A diagnosis of cystitis or pyelonephritis required specific manifestations, as recorded by the treating medical practitioner based on a standardized protocol, and a MSU culture yielding ≥ 10 8 cfu/L of E. coli. Cystitisdefining manifestations included dysuria, frequent urination, and/or suprapubic tenderness, without fever or loin pain. Pyelonephritis-defining manifestations included urinary symptoms plus, fever of ≥ 38 °C and flank pain, with or without nausea/vomiting.
Fecal control isolates were collected from consenting volunteers, who lacked UTI-associated manifestations as per self-report. These controls were matched with UTI subjects by age and, as closely as possible, by place of residence within the region. Each volunteer provided a written informed consent and a rectal swab. Rectal swabs were processed within 15 h of collection, and one arbitrarily chosen E. coli colony per specimen was used [27]. E. coli was identified by conventional biochemical tests, and the isolates were stored in 5% glycerol in trypticase soy broth at − 70 °C until further use.

Ethics approval
The Sydney West Area Health Service (SWAHS) Human Ethics Research Committee and Charles Sturt University Ethics committee approved the study protocol. Guidelines for experimentation at the authors' institutions were followed in the conduct of this clinical research. Since clinical information for patients with UTI was collected anonymously, patient consent was not obtained. However, written informed consent was obtained from each individual in the case of faecal specimens from healthy controls.

Phylotyping and VF genotyping
E. coli phylogenetic groups were identified by the revised Clermont et al. method as previously described [23]. Twenty-two UPEC-associated VF genes (list in Table 1, and Additional file 1: Table S1), encompassing several VF functional categories, were detected using a previously described multiplex PCR-based reverse line blot (mPCR/ RLB) hybridization assay [28]. The aggregate VF gene score was the sum of VF genes detected in an isolate, with multiple pap operon genes counted as one. Such VF gene scores have been shown to predict experimental in vivo virulence [29,30].

Quality control
Each molecular testing was done in duplicate, and with independently prepared DNA lysates of the test isolates and with appropriate positive and negative controls.

Statistical analysis
Comparisons of proportions was done by Chi-square or Fisher's exact tests whilst VF scores were compared using the Mann-Whitney U-test. For comparisons of proportions for different characteristics in the same population, McNemar's test was used. P values < 0.05 were considered significant.

General distribution of papG alleles in E. coli isolates from reproductive-age women with cystitis or pyelonephritis and healthy fecal controls
The majority of pyelonephritis (75%) and cystitis (62%) isolates harboured at least one papG allele gene, as opposed to the lower proportion in fecal controls (35%) (P < 0.001 for each) ( Table 1). The papGII gene was highly prevalent in pyelonephritis isolates (68%), whilst papGIII was the most abundant allele gene, albeit at a lower proportion, in cystitis isolates (30%), and the differences were highly statistically significant (P < 0.001). In contrast to both pyelonephritis and cystitis isolates, a significantly lower proportion of fecal isolates contained either only papGII (13%) or papGIII (19%) genes (P < 0.001), with most of the remainder devoid of any papG gene. The proportion of isolates harbouring a combination of papG allele genes was relatively small, ranging from 3% each in pyelonephritis and fecal isolates, to 13% in cystitis isolates. The gene papGI was relatively rare among the isolates, being detected as the only papG allele in 1% of the cystitis isolates, and in 3% each in combination with either papGII or papGIII, all in cystitis isolates. Likewise, the papGIV gene was rare, being detected in 1% of cystitis isolates only. The most common papG allele type combination was papGII and papGIII, being present in 2%, 7% and 3% of pyelonephritis, cystitis and fecal isolates, respectively ( Table 1). As previously observed, an overwhelming majority of pyelonephritis (80%) and cystitis (70%) isolates belonged to phylogenetic group B2, and a much lower proportion to group D at about 10% each for both clinical syndromes. In contrast to the UTI isolates, a much lower proportion of fecal isolates belonged to group B2 (32%), and a higher proportion to groups A (22%) and B1 (20%). However, the proportion of fecal isolates belonging to group D was almost equal to that in both pyelonephritis and cystitis isolates at 13%. The remaining phylogenetic groups were distributed variously different amongst the three groups, ranging from 1 to 8%.
Distribution of papG alleles in relation to other VF genes among E. coli pyelonephritis isolates from reproductive-age women On average, pyelonephritis isolates harboring papG genes also possessed a higher number of other VF genes, compared to their papG negative counterparts (Table 2). Specifically, all the VF genes tested, save for three (afa, fyuA and kpsIII), were detected in significantly higher proportions in papG positive isolates than negative ones. Consequently, VF gene scores were significantly higher in the papG positive isolates than in the negative ones ( Table 2). In terms of specific papG alleles, papGII + isolates, which were the most prevalent in pyelonephritis isolates, on average harboured more VF genes than their papGIII counterparts, albeit small number of papGIII isolates analysed. Specifically, 12 of the 18 VF genes tested were much more prevalent in papGII + isolates than papGIII + ones. Genes that were much more prevalent in papGIII isolates than papGII included afa (25% vs 5%), traT (83% vs 70%), and usp (75% vs 66%). Consequently, VF gene scores were significantly higher in papGII than papGIII isolates. All the papGII positive pyelonephritis isolates harboured the fimH and papC genes, whilst no gafD and bmaE genes were detected in both papGII and papGIII isolates. Furthermore, no kpsIII gene was detected in papGIII isolates.
In terms of the relationship between papG alleles and phylogeny, a greater proportion of papG positive isolates (86%) was restricted to group B2 compared to 64% for papG negative isolates (P = 0.006). Almost all papGII + isolates (98%) were of phylogenetic group B2 compared to only 58% for papGIII + isolates (P < 0.001). Based on the group D status, a greater proportion (25%) of papGIII + isolates belonged to this group compared to only 2% for papGII isolates, albeit very small number of papGIII + isolates (n = 12) (P = 0.004). None of the papGII + isolates belonged to phylogenetic groups A, B1, Clade 1, and only 1 papGIII isolate each, belonged to these 3 phylogenetic groups. The recently described phylogenetic groups (E, F and clade 1) were evenly distributed amongst papG negative vs. papG positive isolates, and papGII + vs. papGIII + isolates, but with very small numbers of isolates in each subgroup.

Distribution of papG alleles among E. coli cystitis isolates from reproductive-age women
Similar to pyelonephritis isolates, on average, cystitis papG positive isolates harboured more VF genes than their papG negative counterparts, with only one gene (afa) being statistically much more abundant in the papG negative isolates at 20% vs. 5% (P < 0.001). Likewise, papGII + isolates harboured more VF genes compared to papGIII + ones, with only 2 genes (afa, usp) being much more prevalent in the papGIII isolates, albeit statistically insignificant (P > 0.05), and hence VF gene scores were significantly lower in papGIII isolates compared to papGII ones (Table 3). Due to the high number of isolates harbouring both the papGII and papGIII genes concurrently among cystitis isolates (n = 21), we analysed them further for VF gene carriage distribution. These isolates harboured more VF genes than any other isolates, with an overwhelming majority of the isolates harbouring most of the tested genes. Specifically, at least 77% of these isolates containing both papGII and papGIII genes 19 (26) 113 (51) 104 (52)  harboured 14 of the 18 VF genes analysed. As a result, the VF gene scores of these isolates were the highest among all the isolates studied, pyelonephritis and fecal isolates included. Furthermore, these isolates with a combination papGII and III allele genotype carried the following genes, fimH, focG, papAH, papC, papEF and traT. The gene bmaE, which was not detected in any pyelonephritis isolates, was detected in 5% of papGIII isolates only and none in papGII and papGII + papGIII positive isolates. Similar to pyelonephritis isolates, the majority (75%) of cystitis isolates belonged to phylogenetic group B2, with a greater proportion of papG positive isolates (78%) belonging to this phylogenetic group, as opposed to 70% for papG negative ones. Specifically, all the Table 3 Distribution of phylogenetic groups and virulence factor (VF) genes in relation to papG alleles among E. coli isolates from reproductive-age women with cystitis a The 22 virulence factors analyzed were; papA, P fimbriae structural subunit; papC, P fimbriae assembly; papEF, fimbriae tip pilins; papG, P fimbriae adhesin (and alleles I, II and III); sfaS, S fimbriae; focG, F1C fimbriae; afa/draBC, Afimbrial adhesin (Dr-binding adhesin); fimH, type 1 fimbriae; hlyA, hemolysin; cnf1, cytotoxic necrotizing factor type1; fyuA, ferric yersiniabactin receptor; iutA, aerobactin receptor; iroN, catecholate siderophore receptor; kpsMTII group 2 capsule (with K1 and K2 variants); kpsMTIII, group 3 capsule; traT, serum-resistance associated; ompT, outer membrane protein T (protease); bmaE, M fimbriae; gafD, (G) fimbriae b papG neg, papG gene negative (for all alleles) c papG pos, papG gene positive (for any allele)  papGII isolates, and 91% of papGIII isolates belonged to this group. For group D status, a greater proportion of papGIII isolates (13%) belonged to this group compared to only 4% for papGII isolates, albeit statistically insignificant. The rest of the phylogenetic groups were generally evenly distributed amongst the different subgroups.

Distribution of papG genotypes among E. coli pyelonephritis and cystitis isolates from reproductive-age women
An overwhelming majority of pyelonephritis isolates harbouring only the papGII gene (79%), were positive for most of the other pap genes tested, compared to only 17% for isolates carrying the papGIII gene only (P < 0.001) ( Table 4). The remainder of the isolates that harboured only the papGII gene also harboured at least 2 of the 3 other pap operon genes tested. All the pyelonephritis isolates containing only the papGII gene, had at least one other pap operon gene detected. For isolates harbouring only the papGIII gene, the prevalence of other pap genes was generally similar, mostly at 17% each. All the isolates with a combination genotype of papGII and papGIII genes possessed all the other pap genes tested. Most (82%) of the papG negative pyelonephritis isolates also lacked the other pap genes tested, with the remainder harbouring only one of the other pap operon genes. For cystitis isolates, papGI isolates, which were solely confined to this group, belonged only to one pap genotype, papEF − papAH + papC − , albeit only 4 isolates ( Table 5). The majority of cystitis isolates possessing the papGII gene only (46%), contained all the pap operon genes tested, and hence assumed to have an intact or complete pap operon. The majority of the remaining isolates in this group possessed at least 2 of the other 3 pap genes tested. Cystitis isolates harbouring only the papGIII allele gene, mostly (38%) also contained all the operon genes tested, or harboured at least 2 other pap genes (32%). For cystitis isolates harbouring both papGII and papGIII genes concurrently, most (82%) also contained all the pap genes tested. Likewise, most (65%) of the cystitis isolates devoid of any papG allele genes were also null for other pap operon genes.

Discussion
Our aim was to understand the role played by papG alleles in UTI pathogenesis, specifically their association with particular UTI syndromes, other VF genes and phylogenetic groups. Our findings demonstrate, in a stringently selected reproductive age women population by region, time and UTI syndrome, that papGII allele is strongly associated with pyelonephritis, with 68% of the isolates harboring this gene, whilst papGIII was more common in cystitis isolates, albeit at a lower level of influence, with a majority of 31% of the cystitis isolates harboring this allele only. These findings are in agreement with those of several other studies in different jurisdictions [9,10,31]. However, despite the stringent selection of the isolates for inclusion in the present study according to the uro-clinical category, it is possible that some misclassification of the uro-clinical syndrome might have occurred. The predominance of the papGII gene in the studied pyelonephritis isolates, is in line with the demonstrated abundance of papGII iso-receptors in human kidney tissue, indicating an important role for this allele in ascending UTI infection as previously observed [32]. The papGII gene has been associated epidemiologically with pyelonephritis and urinary-source bacteremia in directed, usually PCR-based studies [33], and was demonstrated experimentally, with varying degrees of rigor, to contribute to kidney infection in murine and monkey models [34][35][36]. However, contrary to these findings, is the reported dominance of papG class II gene in pediatric cystitis isolates, albeit small number of isolates studied [37,38], suggesting that associations of papG alleles with specific clinical syndromes may depend on the specific population studied, including age, gender, and geographical locale [37,38]. It is thus important that comparisons of studies take into consideration VF gene carriage population dynamics.
On average, pyelonephritis and cystitis isolates, irrespective of papG allele gene status, more often belonged to phylogenetic group B2, and to a lesser extent group D, had higher VF gene scores, than fecal isolates which were mostly confined to groups A and B1. Our results are consistent with previous evidence of the domination of E. coli phylogenetic group B2 among extraintestinal pathogenic E. coli (EXPEC), and suggest that phylogenetic group B2 may be the main source, and hence presumably the original source of many VF genes in EXPEC [39]. These EXPEC VF genes are understood to be mainly inherited vertically within evolutionary ancestries, but can also be transferred horizontally between lineages, on gene bocks that contain multiple contiguous VF genes, commonly referred to as "pathogenicity-associated islands" (PAIs), or through plasmids [9,39]. Although on average the UTI isolates had higher VF gene scores than the fecal isolates as previously observed [28,[40][41][42], the pyelonephritis isolates contained more individual VF genes, and consequently had higher aggregate VF gene scores, than the cystitis (or fecal) isolates, suggesting that VF gene repertoire plays a significant role in ascending UTI pathogenesis [43].
VF gene distribution by papG gene status revealed that papG positive pyelonephritis isolates on average contained more VF genes, and hence higher VF scores, than their papG negative counterparts, implying an association of papG with several other VF genes. This may be due to the fact that the pap operon can be located on chromosomal or PAIs which contain other VF genes, and hence are transmitted as a block of VF genes [6,44]. The high VF gene carriage amongst the pyelonephritis isolates was mostly due to the contribution of papGII + isolates, which on average carried more VF genes than their papGIII counterparts as evidenced by higher VF scores, with 16 out of 18 VF genes analysed being much more abundant in the papGII isolates. This suggests an association of papGII with a wide variety of VF genes, compared to papGIII, and hence a possible increased inferred virulence potential of such strains, and increased capacity to cause ascending UTI. Likewise, the overall trend for cystitis isolates in relation to papG gene vs. other VF gene carriages, followed a similar pattern to pyelonephritis isolates. Taken together, these results strongly suggest that papG is an important VF gene in UTI pathogenesis, especially given that it is involved in the attachment of E. coli to the epithelial cells of the urogenital tract [45].
Interestingly, although papGIII gene was the most prevalent allele in cystitis isolates (31%), on average, papGII cystitis isolates had significantly higher VF gene scores (P = 0.03), suggesting increased inferred virulence potential for such strains as also demonstrated in the pyelonephritis cohort. This suggests an increased association of papGII with a wide range of VF genes, possibly due to their phylogenetic group B2 status as most of the papGII isolates derived from this phylogenetic group. However, even among the papGII isolates, pyelonephritis isolates had higher overall VF gene carriage than cystitis ones (data not shown), probably indicating presence of papGII clones within phylogenetic group B2, with differing levels of virulence, based on other factors, including possibly VF gene copy number, and other host and bacterial factors, as previously observed by us [41,42]). Notably, predominance of the papGII allele among avian pathogenic E. coli isolates with high homology to human isolates has been previously reported [46], and also similarity between human and avian E. coli strains representing zoonotic potential has been demonstrated [47], suggesting that horizontal gene transfer of pathogenicity elements from chickens to humans may play a role in UTI pathogenesis [48].
Since we had a reasonable number of cystitis isolates that had a combination genotype of papGII and papGIII genes concurrently (n = 21), we analysed these isolates in relation to VF gene carriage. These isolates on average had the highest VF gene scores, with all VF genes tested, save for two (gafD and sfaS), having a prevalence of at least 67% each in these isolates. Hence, as per murine and mice studies [29], the inferred virulence capacity for such isolates was very high, and as such, we expected these strains to be confined to the pyelonephritis group, which was not the case. A further analysis of the patients from which the isolates originated revealed that they were all early UTI cases, suggesting that the bacterial strain might not have had enough time to cause ascending UTI. However, it is rather surprising that we did not find a high number of isolates containing both alleles amongst the pyelonephritis isolates.
We further analysed the association between the different papG gene alleles and other pap operon genes since they are encoded on the same pap operon. Among pyelonephritis isolates, papG negative isolates were more likely to be negative for all other pap genes tested (82%), suggesting complete absence of the pap operon or possible deletion [49], and hence inability to produce the P fimbria, or only positive for papC (8%), suggesting presence of an incomplete or truncated pap operon and hence incapacity to express P fimbria. This finding indicates that other VF adhesin genes or other bacterial or host factors may also be involved in ascending UTI pathogenesis, albeit on a smaller scale.
In the case of cystitis isolates, the majority (65%) of papG negative isolates, albeit lower proportion compared to the pyelonephritis isolates (82%), harbored no other pap genes, and hence were presumed to lack capacity to express P fimbriae. This finding suggests that for the pathogenesis of cystitis, other adhesins may be involved, including afa, sfaS, fimH, bmaE, gafD. Indeed, when these isolates were analysed for other adhesins, they, on average harbored a higher and wider variety of adhesins compared to their pap operon positive counterparts (data not shown). Many microorganisms have the genetic capacity to express different adhesins, providing access to multiple receptors and therefore increasing their pathogenicities [50]. Specifically, about 14% of our papG negative cystitis isolates were positive for all pap genes tested, suggesting possible deletion of the papG gene or presence of a yet to be described papG allele.
The majority of papGII positive pyelonephritis isolates (79%) were also positive for all the other pap operon genes tested, and hence had the ability to express P fimbriae under the right conditions [51]. In contrast, the pyelonephritis papGIII gene isolates demonstrated a limited capacity to express P fimbriae, as only 17% of the isolates presumably carried a complete pap operon, whilst the rest had incomplete operons with basically a uniform spread of other pap operon genes, mostly at 17% for each of the individual pap genes tested. This probably explains why papGIII isolates were more likely to be detected in cystitis isolates, as such isolates would have lacked the capacity to produce P fimbriae which is considered important in ascending UTI pathogenesis. It is not clear why the papGII gene was associated with a complete pap operon in pyelonephritis isolates, which calls for further studies to investigate this finding.
When compared by papG allele status, the majority of cystitis papGII + (46%) and papGIII + (38%) isolates contained all the other pap genes tested, and hence considered to have the capacity to express P fimbria. However, this was much lower than was the case with pyelonephritis isolates, where an overwhelming majority of the papGII strains had a presumed complete pap operon. These findings again seem to highlight that the capacity to express P fimbria is important in ascending UTI pathogenesis as observed by others [52,53]. Furthermore, a large majority of isolates that contained both alleles concurrently (82%), contained a presumed intact pap operon. Although such isolates were, in the present study, confined to the cystitis isolates, we think that they had the capacity to cause ascending UTI over time as most of them were limited to early UTI cases.
Although the above argument that papGII operons from pyelonephritis isolates are more likely to be complete than papGII or papGIII operons from cystitis isolates, is plausible from a pathogenicity perspective, caution must be exercised in the interpretation of the findings. Firstly, the interpretations are based on the amplification of a limited region covering about 1 kb of the 9 kb pap operon, which limits the strength of the argument about the completeness of the pap operon. It is reasonable to argue that insertions or other recombination events are more likely to occur in the rest of the 8 kb region that was not investigated. Furthermore, some minor sequence variants have been described in the primer binding regions of papA and papE genes, and thus one can also hypothesize that such variants could have been present in the negative isolates. However, from an epidemiological viewpoint, we think that such minor variants would not have much effect on the overall picture. Although our findings seem to suggest association of specific papG alleles with specific UTI syndromes, contrasting findings, such as the presence of strains with a combination of both papGII and papGIII genes amongst cystitis strains, and presence of a significant proportion of papGII isolates among cystitis strains, suggest that epidemiologic associations between individual papG alleles and specific clinical syndromes must be interpreted cautiously because such associations may be due primarily to other bacterial properties that are differentially linked with particular papG allele, possibly as part of a PAI rather than to any specific pathogenetic role of the papG allele itself. Like many other E. coli VF genes, the pap operon encoding P fimbria lies on PAIs [9,54], which are large horizontally transferable genetic elements assumed to play an important role in the evolution of pathogenic E. coli [9,54].
In 25% and 38% of pyelonephritis and cystitis isolates, respectively, both the major subunit gene (papA) and the distal adhesin gene (papG) were absent, suggesting presence of truncated P fimbrial operons as previously observed [55] or presence of minor variants. At least 95% of the papGII pyelonephritis isolates were positive for both fimH and other pap operon genes tested, and about 50% of them were positive for pap operon genes, fimH, and sfa/foc, implying capacity of these strains to express a wide variety of adhesins. This is in agreement with studies demonstrating that type 1, P, S, F1C, and Dr fimbriae, attach to different sites within the human kidney [32], and thus strains endowed with a diverse range of fimbrial types are more likely to have overall success during renal colonization [32]. Furthermore, P and type 1 fimbriae appear to act in synergy to promote colonization of kidney [56].
Although the papGI allele is relatively rare amongst clinical urinary isolates, it was surprisingly detected in 7% of the present isolates as follows; as the only papG allele in 1% of the cystitis isolates; in 3% each in combination with either papGII or papGIII, and finally as part of a concurrent combination with both papGII and papGIII genes in 1% of pyelonephritis isolates. To the best of our knowledge, this is one of the few, if any, studies in which the papGI allele was this much abundant, and may be attributed to differences in the distribution of papG alleles by geographical location, population cohort or other yet to be elucidated factors. The genetic make-up, or other aspects of the papGI + strains need to be more specifically determined to see if they belong to the same clonal group. Similarly, the papGIV gene, which has been rarely described in previous studies, was detected in only 1% of our isolates, and was only confined to the cystitis subgroup. More studies are needed to clarify the role of this papG allele in UTI pathogenesis.
The identification of a small proportion of cystitis isolates (3%) with the papGI and papGIII allele class combination highlights that this papG genotype, although rarely encountered among clinical isolates, is not limited to source strain J96 as was previously assumed [57,58]. Previous reports indicate that this genotype is characteristic of a J96-like clonal group of E. coli strains of serotype O4:H5:F13 [55,58], and some of these strains have caused cystitis, pyelonephritis, urosepsis, and bacteraemia of unknown origin, in parts of Europe and the United States of America (USA) [37,38]. Both the papGI and papGIII alleles are associated with a papA molecule of the F13 serotype [55,58], and may explain the present finding in which isolates that were papGI allele positive, also contained the papAH gene.
Strengths of this study include the large number of well-characterized cystitis, pyelonephritis, and fecal isolates, from the same geographical region and time period. This is important since human-associated E. coli strains can vary dramatically by region and over time [59,60]. Other strengths include the extensive array of bacterial traits studied and the analysis of their distribution by phylogenetic group and uro-clinical syndrome.
Study limitations include the use of multiple comparisons, which can increase the chance of type 1 errors [61]. However, we regard our study as being of an exploratory nature rather than definitive, requiring future confirmatory studies, and is also designed to generate further hypotheses for future studies. Furthermore, virulence level was inferred based on molecular attributes, not in vivo assessment, and presence-absence testing for a defined set of VF genes can overlook other potentially important determinants of cystitis and pyelonephritis, including unrecognized VF genes [62], minor sequence variants of described VF genes [63,64], or differences in VF gene expression [65]. Another limitation of the study is the possible overlap in the classification of isolates by uro-clinical syndrome into the 2 groups of cystitis and pyelonephritis. However, this is somehow compensated for by the large number of isolates studied. And finally, due to limited budget, whole genome sequencing was not performed, which could have shed some light into possible association of papG alleles with specific clones, presence of novel alleles, and whether the pap operons were truly disrupted.

Conclusions
Findings from this study suggest that the association of the papGII gene with several other VF genes may explain its predominance in E. coli strains from pyelonephritis cases, as opposed to the predominance of the papGIII gene in cystitis cases.