Shiga toxin-producing escherichia coli infections in Norway, 1992–2012: characterization of isolates and identification of risk factors for haemolytic uremic syndrome

Background Shiga toxin-producing E. coli (STEC) infection is associated with haemolytic uremic syndrome (HUS). Therefore Norway has implemented strict guidelines for prevention and control of STEC infection. However, only a subgroup of STEC leads to HUS. Thus, identification of determinants differentiating high risk STEC (HUS STEC) from low risk STEC (non-HUS STEC) is needed to enable implementation of graded infectious disease response. Methods A national study of 333 STEC infections in Norway, including one STEC from each patient or outbreak over two decades (1992–2012), was conducted. Serotype, virulence profile, and genotype of each STEC were determined by phenotypic or PCR based methods. The association between microbiological properties and demographic and clinical data was assessed by univariable analyses and multiple logistic regression models. Results From 1992 through 2012, an increased number of STEC cases including more domestically acquired infections were notified in Norway. O157 was the most frequent serogroup (33.6 %), although a decrease of this serogroup was seen over the last decade. All 25 HUS patients yielded STEC with stx2, eae, and ehxA. In a multiple logistic regression model, age ≤5 years (OR = 16.7) and stx2a (OR = 30.1) were independently related to increased risk of HUS. eae and hospitalization could not be modelled since all HUS patients showed these traits. The combination of low age (≤5 years) and the presence of stx2a, and eae gave a positive predictive value (PPV) for HUS of 67.5 % and a negative predictive value (NPV) of 99.0 %. SF O157:[H7] and O145:H?, although associated with HUS in the univariable analyses, were not independent risk factors. stx1 (OR = 0.1) was the sole factor independently associated with a reduced risk of HUS (NPV: 79.7 %); stx2c was not so. Conclusions Our results indicate that virulence gene profile and patients’ age are the major determinants of HUS development. Electronic supplementary material The online version of this article (doi:10.1186/s12879-015-1017-6) contains supplementary material, which is available to authorized users.


Background
Shiga toxin-producing Escherichia coli (STEC), also called verocytotoxin-producing E. coli (VTEC), can lead to mild, self-limiting diarrhoea, haemorrhagic colitis or the life threatening complication haemolytic uremic syndrome (HUS). Children less than five years of age, the elderly, and immunocompromised persons, are most susceptible to STEC infection as well as to severe complications [1]. An association between the Shiga Toxin-encoding gene stx2, particularly the subtypes stx2a, stx2c, and stx2d, and development of HUS has been described [2][3][4][5][6][7][8][9][10]. Several other virulence factors that contribute to the pathogenicity of STEC have been identified, such as eae (E. coli attaching and effacing) encoding intimin and the plasmid-borne ehxA encoding enterohaemolysin [11]. In several parts of the world O157 is the predominating STEC serogroup, and this variant has most frequently been associated with HUS and outbreaks [12][13][14][15][16]. In other countries, however, like continental Europe and Scandinavia, non-O157 serogroups are dominating [3,[17][18][19][20]. The involvement of STEC in serious outbreaks combined with a high disease burden per case [21] makes STEC a significant challenge to public health.
In 1995, STEC infection was made mandatory notifiable to the Norwegian Surveillance System for Communicable Diseases (MSIS) (http://www.msis.no), and in 2006 diarrhoea-associated HUS became notifiable. Norway has implemented strict guidelines for prevention and control of STEC infection, in which 3-5 negative stool cultures are required for high-risk groups [22].
A few previous studies have investigated risk factors for development of HUS, however these studies have mainly focused on clinical and demographic parameters among patients infected with either serogroup O157 or O104 [23][24][25]. Although it is well documented that the presence of stx2 and eae as well as being a child are risk determinants of HUS development, few studies have performed multivariable analyses of both O157 and non-O157 STEC to identify factors independently associated with HUS. Furthermore, knowledge of factors independently associated with reduced risk of HUS is sparse.
The main aim of the present study was to identify microbiologic and patient-related criteria differentiating HUS STEC from non-HUS STEC, in order to obtain information enabling revision of the strict control and prevention measures presently employed in Norway. The second aim was to describe human STEC infections in Norway during two decades , to compare with studies from other countries and contribute to our understanding of this infection in general.

Surveillance of STEC infections in Norway
Epidemiological and clinical information about STEC infection in Norway from 1992 through 2012 was obtained from MSIS at the National Institute of Public Health (NIPH), which has received mandatory notifications from medical microbiological laboratories and physicians in the country since 1995 (http://www.msis.no/). During this period, 513 STEC infections were notified ( Fig. 1) (annual mean, 0.54 cases per 100.000 populations). Cases were recorded as domestic if the patients did not report foreign travel in the incubation period, and as imported if the patients became ill while being abroad or shortly after their return home.

Characterization of STEC isolates
All isolates were obtained from the National STEC Culture Collection at the Reference Laboratory for Enteropathogenic Bacteria at the NIPH, which receives all presumptive STEC isolates from medical microbiological laboratories throughout Norway.
We selected one isolate per patient and per outbreak, except for one patient from whom two isolates were included since they showed different virulence gene profiles and genotypes. Likewise, only one isolate was selected if isolates with identical genotype were received within the same time period from two patients with different surnames, but living within the same municipality or county, had attended the same child-care facility, or had identical travel history. Nine isolates were from asymptomatic carriers. In total, 334 STEC isolates from 333 patients (64.8 %, 333/513) were included in the study (Fig. 1).
Eighty-three out of the 334 STEC isolates, all from patients living in one of the 19 counties in Norway, have been described previously [26].

Sorbitol-fermenting (SF) E. coli O157
All STEC isolates belonging to serogroup O157 were analysed with a multiplex PCR (M-PCR) specific for SF E. coli O157 [27] in order to distinguish SF O157 STEC from the classical non-sorbitol fermenting (NSF) O157 STEC.

stx subtyping
Subtypes of stx1 were identified by PCR as described by Scheutz et al. [8]. Subtyping of stx2 was performed with one of the two following methods. The first method used PCR restriction fragment length polymorphism (RFLP) followed by electrophoresis (modification of Russmann et al. [36] and Jelacic et al. [37]), and sequencing [7], in which all stx2d positive isolates were verified by a stx2d specific PCR [38]. The second method determined stx2 subtypes by PCR as described by Scheutz et al. [8].

Statistical methods
Statistical analyses were performed using the computer program SPSS® release 20.0.0 (IBM SPSS Software, International Business Machines Corp., Armonk New York). Univariable analyses were performed with the procedures for cross tables (dichotomous variables) or comparison of means (continuous variables) as appropriate. Binary logistic regression was implemented for multivariable analyses. The results are reported as odds ratios with 95 % confidence intervals and two-tailed p values. A p value of ≤ 0.05 was considered statistically significant. Positive and negative predictive values were calculated as described by Altman & Bland [42].

Ethical considerations
At the NIPH, all STEC strains are routinely collected for disease surveillance and outbreak detection. The current study is based on descriptive analysis of bacterial isolates from the strain collection and the microbiological characteristics so obtained could only be combined with the sex, age, clinical outcome, hospitalization, travel history, and seasonality for the patients from which the strains were isolated. Ethical approval was therefore not required. Also, the Norwegian Communicable Disease Control Act and its companying regulations (https://lovdata.no/dokument/NL/ lov/1994-08-05-55?q=Smittevernloven) obliges the NIPH to perform national surveillance of communicable diseases, including STEC infection. For these reasons, consent was not obtained from the patients to analyze the bacterial samples for this research project. developed HUS (11.1 %), and isolates from 36 of them were submitted to the National Reference Laboratory at NIPH. The major outbreaks reported in Norway during 1992 through 2012 are presented in Additional file 1.

Demographics and clinical presentation
The mean age of the 333 patients selected for the study was 24.6 years and 39.0 % (130/333) were ≤ 5 years old. Diarrhoea was the most frequent clinical manifestation, and 35.7 % (119/333) reported bloody diarrhoea (Table 1). Twenty-five of the patients (7.5 %) developed HUS (Additional file 2). Furthermore, 49.8 % had reportedly acquired their infection in Norway (   Table 2). None of the stx1 positive isolates harboured more than one stx1 subtype. stx1a was the most frequently detected subtype. Of the patients with stx2 positive STEC, 100 (48.3 %) contained stx2c and 85 (41.1 %) carried stx2a . Nearly threefourths of the patients had an eae positive STEC and the majority of the cases harboured STEC with ehxA ( Table 2).
Patients with O157 and non-O157 STEC did not differ with regard to presence of stx1 (Table 2). However, when No information on hospitalization was available for 37 patients. These patients carried STEC with the following stx genes; stx1a (n = 17), stx1c (n = 1), stx1d (n = 2), stx2a (n = 6), stx2b (n = 9), stx2c (n = 10), and stx2g (n = 2) (n = number of patients) (some patients carried STEC with more than one stx subtype) c From one HUS patient two different non-O157 STEC isolates were included. One with stx1a and another with stx1a + stx2a and this patient was included both in the stx1 and stx1a groups as well as in the stx2 and stx2a groups d Including 21 patients with STEC harbouring more than one stx2 subtype; stx2a + stx2c (n = 19), stx2a + stx2d (n = 1), and stx2c + stx2d (n = 1) e One HUS patient carried a STEC with both stx2a + stx2c and this patient was included within the stx2a group as well as in the stx2c group f ND: not determined stx1 was stratified according to subtypes, stx1c was more frequently detected among patients with non-O157 STEC, and neither stx1c nor stx1d was found in any of the patients with O157 isolates ( Table 2, Additional  file 4).
Nearly all patients with O157 isolates carried stx2, while less than half of their non-O157 counterparts had this gene. Among the stx2 subtypes, stx2c and stx2a + stx2c were both more common in O157 (p < 0.005 for each), whereas stx2b and stx2g were only detected in the non-O157 group ( Table 2, Additional file 4).
Both eae and ehxA were more frequently detected in O157 compared to non-O157 (Table 2).
Among non-O157 isolates, O103 was more likely to carry stx1 than the other serogroups combined, while in O145 or O121 stx1 was infrequent. All O91, O113, and O146 isolates were eae negative, and only two of the O117 isolates carried this gene (Additional file 3).
Patients with stx1 positive STEC showed a reduced risk of hospitalization, whereas the contrary was seen for patients with stx2, eae, and ehxA (Table 2).

Discrimination between HUS and non-HUS STEC
All 25 HUS patients, except three, were ≤ 5 years. Compared to non-HUS cases, patients with HUS were more often hospitalized and showed bloody stools (Table 1).
All HUS patients carried STEC with stx2, eae, and ehxA, however only stx2 and eae were significantly associated with HUS, while ehxA was not ( Table 2 and 3). stx2a was present in 24/25 HUS STEC (including one with stx2a + stx2c), whereas the last case carried stx2c only. Both stx1 and stx2c were negatively associated with HUS (Table 2 and 3). The combination of age ≤ 5 years and STEC possessing stx2a and eae showed strong One HUS patients had two STEC isolates; one with stx1a + stx2a and another with stx1a only association to HUS development with PPV of 64.7 % and NPV of 99.0 %. Additionally, this combination gave the highest sensitivity (88.0 %) and specificity (96.1 %) of all determinants investigated (data not shown).
Three multivariable models were fitted, with and without including potentially protective factors. In the first model, two factors were found to be independently related to increased risk of developing HUS: stx2a (OR = 92.7, CI = 10.7-803.5) and age ≤ 5 years (OR = 12.2, CI = 3.2-46.7). In this model, the following factors were not independently associated with HUS: SF O157 and O145:H?, although they were significant in the univariable analysis (Table 3). No first-order interactions were detected in the model. All HUS patients with bloody diarrhoea carried STEC with stx2a, and bloody diarrhoea was therefore not included in the model containing stx2a. Furthermore, eae, stx2, and hospitalization could not be modelled since all HUS patients showed this trait (for one HUS patient no information on hospitalization was available).

Discussion
Low age (≤5 years), and the presence of STEC with stx2a and eae were identified as risk factors for HUS development. An association between HUS and these parameters has been seen in several studies [2-10, 16-18, 43], however, only a few have explored this by multivariable analysis [2,7,44]. The high NPV (99.0 %) obtained for this combination of determinants indicates that the likelihood of developing HUS is very low when all these factors are negative. However, not all patients with these three risk factors developed HUS (PPV of 64.7 %), emphasizing that other strain characteristics or host specific factors, like the patient's immunocompetence, also are important to consider when assessing a patients' risk for developing HUS. Bloody diarrhoea has previously been identified as a risk factor for HUS [2], and a similar association was achieved in our study, although not proven as an independent risk factor. Interestingly, when only including protective factors in a multivariable model, stx2c was independently associated with reduced risk of HUS, but not when both protective and risk factors were included in the same model. The role of stx2c in HUS pathogenesis has been debated and it has been speculated that stx2c merely assists stx2a during development of this severe complication [7]. However, our results indicated that stx2c neither was sufficient nor necessary for HUS development. In one of the two HUS patients with stx2c positive STEC, stx2c and stx2a were both present, whereas in the second case stx2c was the sole stx gene detected. It is possible, though, that this isolate had lost the stx2a encoding bacteriophage, a phenomenon previously described in isolates from HUS patients [45,46]. stx1 was independently related to reduced risk of HUS in two multivariable models. This has to our knowledge never been shown before, although such an association has been suggested [3,16,26,[47][48][49][50]. None of the HUS patients carried STEC with stx1 as the sole stx gene present. The single HUS patient with stx1 yielded two O111:[H8] isolates, one with stx1a + stx2a and the other with stx1a only, indicating that stx2a was the stx gene responsible for HUS development. Recently, it was demonstrated that STEC O111:H8 strains frequently lose their stx2 encoding bacteriophage during in vitro growth, suggesting that this loss may occur in vivo as well [3,51]. Moreover, stx1 showed a low PPV for HUS, a finding which further emphasises that stx1 was not a key factor for HUS development.
In contrast to some authors [3,15,16,52], but in concordance with others [2,7], we did not find any significant difference between STEC O157 and non-O157 regarding HUS. Interestingly, of the O157 STEC isolated from HUS patients, SF O157 was the dominating variant, despite the fact that NSF O157 was the most frequent STEC detected in Norway. A high frequency of SF O157 in HUS cases has also been reported from other European countries [9,53] and it has been suggested that patients with STEC SF O157 more often develop HUS compared to patients with NSF O157 [54,55]. Furthermore, SF O157 and O145 (particularly O145:H?) were the only serogroups associated with HUS in our univariable analyses, although they were not significant in the multivariable models. All STEC O145:H? and SF O157 cases were domestically acquired, indicating a reservoir of these bacteria in Norway. Both serogroups have previously been responsible for HUS outbreaks in our country [56].
Our results confirm that the severity of STEC illness depends strongly on the virulence gene profile of the infecting STEC as well as the patients' age, unlike serogroup affiliation [2,57,58]. Nevertheless, exceptions exist and therefore clinical findings and the epidemiological situation of each STEC case have to be considered before proper control and prevention measures can be implemented.
In Norway infections with non-O157 STEC were more common than infections with O157 isolates, in accordance with findings from several other countries [3,[17][18][19][59][60][61]. Expectedly, the proportion of STEC O157 declined compared to non-O157 from approximately 2007, when improved methods for detecting stx/Stx were implemented in the majority of clinical microbiological laboratories in Norway [3,16,50,61,62]. In contrast to reports from other countries, more than half of the STEC O157 infections in Norway were imported and no seasonal differences between O157 and non-O157 infections were seen [16,63,64]. Since ruminants are the main reservoir of STEC O157 [65], the low prevalence of STEC O157 among ruminants in Norway might explain these findings [66][67][68][69]. Non-O157 infections were more frequently seen in children (≤5 years) and were more often domestically acquired than O157 infections. Contact with ruminants has previously been identified as the strongest risk factor for non-O157 infection in young children [70] and the following data indicate that this might be the case also in Norway: A national survey of Norwegian sheep flocks [69] showed that as many as 17.3 % (85/491) of the flocks carried non-O157 E. coli considered to be human pathogens (unpublished data). Also, non-O157 STEC outbreaks associated with sheep contact or eating mutton have been reported in Norway [71,72].
There are some limitations to our study. Firstly, we did not examine a consecutive series of STEC isolates from the National STEC Culture Collection, but selected one STEC per patient and per outbreak. Therefore a correct incidence of STEC isolates was not achieved. However, the main aim of our study was to define factors discriminating HUS-STEC from non-HUS STEC. Inclusion of all STEC isolates would have given a biased contribution of the different parameters due to overrepresentation of isolates involved in outbreaks. Secondly, it is likely that non-O157 STEC were underestimated before 2007 since the sensitivity of diagnostic methods were suboptimal at that time. Although the laboratory methods have improved, non-O157 STEC isolation is still a diagnostic challenge due to lack of a selective growth media with sufficient sensitivity. Thirdly, the number of HUS cases included in the study was too low to identify other than the strongest risk factors. Finally, the available clinical information did not permit detailed analysis of patient-related factors such as underlying illnesses, antibiotic treatment, and co-infections, all of which have been considered as putative risk factors for HUS.

Conclusions
Our results showed that the characteristics of the Norwegian STEC isolates were in concordance with data from other countries. However, some country specific characteristics were unravelled. Multivariable regression analyses identified low age (≤5 years) and the presence of stx2a as independent risk factors for HUS development. Additionally, all HUS STEC carried eae. On the other hand, stx1 was independently associated with reduced risk of HUS. Hence, the virulence profile and the patients' age -but not particular serogroups -were the essential determinants discriminating HUS STEC from non-HUS STEC. The results achieved from the current study will contribute, together with previous published knowledge, to revision of the strict national guidelines for prevention and control of STEC infections currently applied in Norway. Nevertheless, it should be emphasized that in addition to the risk factors identified, the clinical presentation of each patient and the epidemiological context also should be taken into account before advice of control and prevention can be given.