Main

The clinical spectrum of STEC, especially of prototypic E. coli O157 strains, includes mild diarrhea, hemorrhagic colitis, and the classical (enteropathic) HUS (E+ HUS) (1–3). E+ HUS occurs predominantly in young children and in the elderly possibly because of a lack of immunity in these age groups (3, 4). It is characterized by a prodrome of gastroenteritis, frequently with bloody diarrhea, followed by acute microangiopathic hemolytic anemia, thrombocytopenia, and acute renal failure. Occasionally, other organs are also affected (4, 5). E+ HUS accounts for approximately 50% of acute renal failure in young children (4, 6). Outbreaks and sporadic infections by E. coli O157 have been reported from all inhabited continents (6). The largest recorded outbreak of E. coli O157 to date occurred in Sakai City (Japan) in 1996. It led to 12,680 cases of diarrhea and 121 of HUS and dramatically illustrated the pathogenic potential and public health importance of this organism (7). It also demonstrated the continuing need for simple and rapid diagnostic techniques.

The detection of IgM and IgA, and IgG antibodies to STEC LPS O157, but also to non-O157 LPS in serum samples, has emerged as a useful and reliable diagnostic method (8–13), especially when the bacterial isolation fails (13, 14). Two preliminary reports proposed the measurement of saliva antibodies for the diagnosis of infections by E. coli O157 (15, 16). The objective of the current study was therefore to investigate the occurrence and dynamics of the immune response to E. coli O157 LPS in children with HUS as reflected by saliva antibodies and to further explore the utility of a saliva-based, rapid and noninvasive diagnostic technique.

METHODS

Patients and controls.

Between November 1996 and September 1999 22 consecutive patients 0.9 to 14.3 years of age (median 3.5 years; mean ± SD: 4.3 ± 3.2 years) with enteropathic HUS from the Children's Hospital, University of Hamburg were investigated for saliva and serum antibodies against E. coli O157 LPS. HUS was defined as the triad of microangiopathic hemolytic anemia, thrombocytopenia, and acute renal failure according to the criteria of Fong et al.(17). Four additional patients with HUS were excluded from the study because of missing consent or incomplete samples. Twelve patients were female. All 22 patients presented with diarrhea, which was bloody in 20 cases. Nine patients (41%) were dialyzed, eight underwent peritoneal dialysis for 5 to 43 d (median 9 d; mean ± SD: 13 ± 11 d), and one received hemodialysis for 9 d. Eighteen patients received at least one packed red cell transfusion. None of the patients received plasma preparations. For each patient, two age-matched healthy controls (n = 44) were identified during the same time period from various kindergartens and from the endocrinology clinic of the Poliklinik für Kinder- und Jugendmedizin. Blood was drawn from these children for evaluation of their immune status before hepatits B immunization or for delayed or accelerated growth. Their saliva and serum samples were used to establish normal cut-off levels. None of the controls had recent diarrhea or was related to HUS cases. The project has been approved and was conducted in accordance to the ethical guidelines for the Children's Hospital Hamburg-Eppendorf. Parental informed consent was obtained for all children studied.

Serum and saliva samples.

A total of 119 paired saliva and serum samples were simultaneously collected from 22 HUS patients during acute and convalescent stages of the disease. The first sample pair was obtained 5 to 21 d after the onset of diarrhea (median 8 d; mean ± SD: 8.5 ± 3.7 d), which corresponded to 4 d before to 10 d after the clinical manifestation of HUS (median 1 d; mean ± SD: 1.5 ± 2.7 d after HUS). Follow-up sample pairs were collected 7 to 761 d (median 24 d; mean ± SD: 77 ± 121 d) after the onset of diarrhea. Fourteen additional saliva samples were collected during the acute HUS without corresponding serum samples.

For the collection of saliva, a small absorbent pad (Rauscher, Pattensen, Germany) was placed between the patient's cheek and gum. From each patient, at least two pads were used to obtain a minimum of 500 μL saliva. All samples were processed within 30 min after receipt. The pads were centrifuged at 13,000 g in a microliter tube. The supernatant was stored at −20°C until analyzed.

E. coli O157 LPS ELISA for saliva and serum samples.

LPS from E. coli O157:H− strain 493-1, from a child with HUS, was extracted by hot phenol water as described by Westphal and Jann (18). Microtiter plates (96-well format Immunoplates, Nunc, Roskilde, Denmark) were coated with 0.5 μg O157 LPS per well diluted in 10 mM PBS. LPS ELISA plates were interchangeably used for saliva and serum antibody studies. Serum samples were diluted 1:500 in PBS containing 0.1% Tween-20 (Merck, Darmstadt, Germany) and 2% fetal bovine serum (Linaris, Bettingen am Main, Germany) and were tested for IgM and IgA antibodies against E. coli O157 LPS as published previously (10).

Saliva samples were diluted 1:10 in PBS containing 0.1% Tween-20 and 0.5% fetal bovine serum (Linaris, Bettingen am Main, Germany) added to triplicate wells and incubated at 37°C for 2 h. Wells were extensively washed with PBS-Tween-20 1%. Optimized dilutions of rabbit anti-human IgM (1:2000) or IgA (1:9000) alkaline-phosphatase conjugated antibodies (Sigma Chemical Co. Chemicals, St. Louis, MO, USA), in PBS-Tween 0.1% with 0.5% fetal bovine serum, were added to the wells for 60 min at 37°C. p-Nitrophenyl Phosphate (Sigma Chemical Co. Chemicals, St. Louis, MO, USA) was used as substrate. The reaction was stopped with 50 μL 2N NaOH after 30 min at 37°C. The A405 was recorded against a blank well containing substrate solution and 2N NaOH. All saliva samples from patients and controls were tested at least twice for IgM and IgA antibodies. To establish the sensitivity and specificity of the ELISA, homologous anti-O157 LPS and heterologous non-O157 LPS sera from rabbits were used, respectively. The breakpoints were defined as the mean absorbency plus 3 SD of the control specimens as described previously (10, 11). Previously defined serum and saliva samples were run with each assay as positive and negative controls. No difference was observed between the mean optical densities of saliva IgM or IgA anti-E. coli O157 LPS antibodies of control children younger (n = 30) and older than 5 years of age (n = 14);t tests for unpaired samples were performed without a significant result for both (IgM and IgA) ELISA (the normality assumption was tested by Kolmogoroff-Smirnoff test).

E. coli O157 immunoblot for saliva and serum samples.

Selected serum samples, especially those with a weak antibody response by ELISA, were also tested for IgM antibodies by immunoblotting as described (9) using the same O157 LPS preparation. E. coli O157 LPS was separated by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to nitrocellulose membranes (Protran BA85 Cellulose Membrane, Schleicher and Schuell, Dassel, Germany) by electroblotting. Membranes were blocked and cut into strips that were incubated with diluted serum (1:100) or saliva (1:10) for 2 h at room temperature. Bound Igs were detected using 1:5000 dilutions of alkaline-phosphatase-conjugated rabbit anti-human IgM or IgA (Sigma Chemical Co. Chemicals, St. Louis, MO, USA). Previously defined serum samples and saliva samples were run with each assay as positive and negative controls.

STEC isolation and stx genotyping.

Stool samples were collected 5 to 22 d (median 9 d) after the onset of the diarrhea and examined for STEC, including E. coli O157, and traditional enteropathogenic E. coli. STEC O157 were isolated using sorbitol-MacConkey agar and immunomagnetic separation (19, 20). Sorbitol-negative colonies were confirmed biochemically as E. coli and tested for the presence of O157 antigen using a latex slide agglutination test (Oxoid, Wesel, Germany) (19). To detect non-O157 STEC strains, colony sweeps from primary stool cultures on sorbitol-MacConkey agar were screened directly for the presence of stx1 and stx2 gene sequences. The detection of the stx1 and stx2 genes from E. coli O157 and non-O157 strains by PCR was done by using primers KS7/KS8 (21) and GK3/GK4 (22), respectively, according to published methods (23). To identify STEC strains from PCR positive stool samples and to determine the stx genotype, 100 to 200 colonies from each plate were tested by colony blot hybridization with digoxigenin-labeled DNA probes (23). Differentiation of stx2 and stx2c genes was performed by the restriction endonuclease analysis of the PCR products as described by Rüssmann et al.(24). O:H serotyping was performed as previously described (25).

Free fecal Shiga toxin.

Stool filtrates from all 22 HUS patients were investigated for the presence of free fecal cytotoxicity using vero cell monolayers (2).

Statistical methods.

t test, χ2 test, and Spearman's rank correlation (rho), were used as indicated. Descriptive measures were the correlation coefficient and the Spearman's rho; the latter was used to test the correlation for significance. Statistical testing was performed using SPSS version 10.0 (SPSS, Inc., Chicago, IL, USA) on a standard PC.

RESULTS

Microbiological studies.

STEC were identified in stool samples of 17 out of 22 (77%) patients. Thirteen of the isolates belonged to serogroup O157 and four to E. coli serogroups O26, O145 (n = 2), and O165, respectively (Table 1). Free fecal Stx (FStx) was demonstrated in stool filtrates of all 17 patients with STEC O157 and non-O157 associated HUS. None of the stool filtrates of the five patients lacking STEC isolates showed free cytotoxic activity. Ten of 13 E. coli O157 isolates harbored the stx2 gene, two the stx1 genes, and in one case the genotype could not be determined. The non-O157 E. coli isolates carried the stx1 gene (E. coli O26 and one E. coli O145 isolate), the stx2 gene (E. coli O145), or the stx2 and stx2c genes (E. coli O165).

Table 1 E. coli O157 LPS IgM and IgA antibodies in serum and saliva*
Full size table

E. coli O157 LPS antibody detection in saliva and serum samples.

Immunoblot results compared favorably with the ELISA results for detecting anti-O157 LPS IgM and IgA antibodies in serum and saliva samples. An example is shown in Figure 1 with an immunoblot for serum and saliva samples from one patient with E. coli O157- associated enteropathic HUS (Fig. 1, lanes 5 and 6) (26, 27).

Figure 1
figure 1

Immunoblot with E. coli O157 LPS as antigen. Blots (lanes):lane 1, low-molecular-weight protein standard (Rainbow prestained; BioRad, Ca, USA) demonstrating the molecular weight in kilo Dalton;lane 2, Rabbit anti-E. coli LPS immune serum, used as positive control;lane 3, positive human control serum;lane 4, negative human control serum. Serum (lane 5) and paired saliva (lane 6) from a child with acute HUS;lane 7, saliva from an age-matched healthy control. Blots were developed with HRP-conjugated anti-rabbit IgG (lane 2) or anti-human IgM (lanes 3–7). The orderly spaced bands, indicated by the bracket and the asterisk, represent LPS molecules having different numbers of repeating units in the O-side chains (26). Note that the molecular weight markers of proteins do not reflect the size of LPS in SDS-PAGE of LPS (27).

Full size image

Breakpoints for the E. coli O157 LPS IgM and IgA ELISA were established using saliva and serum samples from age-matched controls as defined in the methods section. For saliva, the A405 values for the ELISA ranged from 0.000 to 0.095 for IgM, and from 0.000 to 0.026 for IgA. The level of significance (mean plus 3 SD of the controls) was 0.042 for IgM and 0.035 for IgA, respectively. For serum, the A405 values for the ELISA ranged from 0.000 to 0.375 for IgM and from 0.000 to 0.060 for IgA. The level of significance (mean plus 3 SD of the controls) was 0.400 for IgM and 0.030 for IgA, respectively.

Acute phase saliva samples (first and second samples, collected all in the first week after the onset of HUS) of all 13 patients with E. coli O157 isolates revealed optical densities in the IgM ELISA exceeding the cut-off level. Twelve of 13 acute saliva samples also had elevated IgA antibodies to E. coli O157 LPS (Table 1). In addition, four of the five culture-negative cases were positive for saliva IgM and IgA antibodies against E. coli O157. These results closely mirrored the detection of E. coli O157 LPS-specific IgM and IgA in the corresponding serum samples (Table 1). A good correlation was observed between stool culture results and the detection of anti-O157 LPS antibodies (IgM and/or IgA) both in serum and saliva (p = 0.01; χ2 test). Furthermore, a good correlation was found between the A405 value of all 119 paired saliva and serum samples from all 22 HUS patients with a correlation coefficient of 0.569 for IgM and r = 0.650 for IgA. Using the Spearman's rank correlation (rho) it was calculated to be 0.751 for IgM and 0.592 for IgA (p < 0.01 for both; Spearman's rho).

When only initial saliva samples of patients who manifest antibody to O157 in their respective sera (≥3 SD above the mean of healthy controls) were evaluated, a positive correlation was found for IgA (r = 0.642;p < 0.01), but not for IgM (Spearman's rank correlation;Fig. 2A and B). The correlation of positive serum samples (≥3 SD above the mean of healthy controls) to the paired saliva samples for follow-up samples was r = 0.361 for IgM (n = 50), and r = 0.540 for IgA (n = 30) (Spearman's rho;p < 0.01 for both) (Fig. 2C and D). However, of all 17 HUS patients showing elevated IgM antibodies in their first serum samples, 15 (88%) had also elevated IgM antibodies in the corresponding saliva sample. Two additional patients showed increased anti-O157 IgM in their follow-up saliva samples. Of 14 HUS patients showing elevated IgA antibodies in their first serum sample, 11 (79%) had also elevated IgA antibodies in the corresponding saliva sample, while one also remained negative in follow-up saliva samples, and two showed measurable anti-O157 IgA in follow-up saliva samples.

Figure 2
figure 2

Results represent all serum samples showing elevated E. coli O157 LPS-specific IgM or IgA levels (≥3 SD above the mean of healthy controls) measured by ELISA in correlation to the paired saliva samples. Initial serum samples showing antibodies for IgM (A) and for IgA (B), and follow-up samples having detectable IgM (C), and IgA (D) in serum samples in correlation to the corresponding saliva samples. A positive correlation was found for IgA for initial and follow-up samples and for follow-up samples of IgM (all three p < 0.01; Spearman's rho), but data were not significant for IgM for initial samples. Results are presented as SD-units as described in the Methods section. The cut-off level for positive ELISA readings was set at 3 SD above the mean of control samples for both assays. Breakpoints of the A405 values were 0.042 and 0.035 for saliva IgM and IgA, respectively, and 0.400 and 0.030 for serum IgM and IgA, respectively. ♦, initial samples; ▪, follow-up samples.

Full size image

Two patients showed an IgA seroconversion, i.e. changing from negative to positive ELISA results in serum samples during follow-up. Both patients were already positive for IgA in their first saliva sample.

None of the patients with non-O157 STEC isolates showed elevated anti-O157 LPS saliva or serum antibodies (Table 1) confirming the specificity of both serum and saliva ELISA. The sensitivity of the saliva O157 LPS ELISA, calculated from the STEC O157 culture-confirmed cases (n = 13), was 100% for IgM and 92% for IgA (Table 1) for all samples (initial and follow-up samples). For the first collected saliva sample, the sensitivity was 92% for IgM and IgA, respectively.

The first available saliva sample, collected 5 to 21 d after the onset of diarrhea (median 8 d), which corresponded to 4 d before to 10 d after the clinical manifestation of HUS (median 1 d; mean ± SD: 1.5 ± 2.7 d), showed in all but one instance (n = 16) E. coli O157 LPS-specific IgM and/or IgA. In 13 of 17 patients with O157-associated HUS (13 confirmed by culture, 4 by serology), the first available saliva sample was positive for both IgM and IgA anti-O157 LPS. Conversion of saliva anti-O157 LPS IgM and IgA was observed as late as 21 d after the onset of diarrhea in one case (which corresponded to 1 wk after the onset of HUS). Two of three remaining patients demonstrated either conversion of saliva IgM (6 d after the onset of HUS) or saliva IgA (13 d after the onset of HUS) (Fig. 3), while one patient remained negative for anti-O157 IgA in saliva also during follow-up.

Figure 3
figure 3

Saliva anti-O157 LPS IgM and IgA antibody reactivity of the 17 E. coli O157-associated HUS cases (13 culture-confirmed and 4 confirmed by elevated anti-O157 LPS antibodies in serum samples) relative to the onset of HUS. Results represent all patients tested at the respective intervals.

Full size image

Saliva anti-O157 LPS antibody levels peaked 1 to 3 wk after the onset of diarrhea, median 11 d [IgM], and 10 d [IgA].

Examples of four individual saliva IgM and IgA antibody kinetics are shown in Fig. 4.

Figure 4
figure 4

Individual kinetics of E. coli O157 LPS specific saliva IgA and IgM antibodies in children with enteropathic HUS. The cut-off levels for positive ELISA readings were defined as 3 SD-units above the mean of the controls for both assays. Results are presented as SD-units as described in the Methods section. Breakpoints of the A405 values were 0.042 and 0.035 for saliva IgM and IgA, respectively, and 0.400 and 0.030 for serum IgM and IgA, respectively.

Full size image

One out of 44 (2.3%) control children had elevated saliva IgM; another revealed slightly elevated serum IgA antibodies. None of the control samples showed increased saliva IgA or serum IgM antibodies. The specificity of the saliva-based O157 LPS ELISA, based on the results obtained with the samples from the control children and from the patients with non-O157 isolates, was 98% for salivary IgM and 100% for salivary IgA.

DISCUSSION

Infection by E. coli O157 induces a vigorous immune response that is reflected in an acute rise of serum levels of O-group-specific IgA, IgM and, although less consistently, IgG antibodies (8, 10, 11, 28). Here we show that children with E. coli O157-associated HUS also demonstrate easily detectable salivary IgM and IgA class antibodies. Specifically, saliva samples of all 13 patients with E. coli O157 isolate and 3 HUS patients with serological evidence of E. coli O157 infection, collected in the first week after the onset of HUS, showed elevated IgM by ELISA (Fig. 3). Fourteen of these samples were also positive for E. coli O157 LPS-specific IgA (Fig. 3). We further show that salivary antibodies can be conveniently measured and used for serodiagnostic purposes. Specific antibodies have been demonstrated previously in saliva from patients with viral, bacterial, and parasitic infections (29–33). Preliminary data, presented first by our laboratory (15) and by Chart and Jenkins (16), suggested the potential merit of this approach in patients with E. coli O157 infections. The aim of the present study was to systematically investigate whether saliva would be suitable for the diagnosis of infections by E. coli O157 in children with HUS using an indirect immunoenzyme assay. Here we report that E. coli O157 LPS IgM and/or IgA can be reliably detected in “acute” saliva from patients with HUS secondary to E. coli O157. In fact, we observed a close correlation between serum and saliva anti-E. coli O157 antibodies (p < 0.01; Spearman's rank correlation). Furthermore, paired serum and saliva samples from patients with non-O157 stool isolates yielded negative O157 LPS ELISA results, demonstrating once more the specificity of the assay. It is of note that Chart and Jenkins (16) reported a positivity rate in saliva samples of 62% (13/21) patients with anti-O157 LPS antibodies in serum samples. In our patient population 88% (15/17), patients showing elevated IgM antibodies in their first serum samples had also detectable IgM in the corresponding saliva samples. The apparent discrepancy to our findings may be due to varied reasons, e.g. the timing of the sample collection, the underlying disease (HUS versus hemorrhagic colitis), or the transportation conditions of saliva samples. All three items were addressed in our current study. All except one saliva sample was collected within 1 wk after the onset of HUS, they were processed within a short period of time, and the study was restricted to HUS patients. For diagnostic purposes, we recommend measurement of the IgM antibody response in saliva samples in the first week after the onset of HUS, which usually corresponds to the second or third week after the onset of diarrhea, based on the investigations reported in this paper. Additional samples can be investigated a week later if the first sample is equivocal. Occasionally, sequential saliva sampling may increase the yield of positive samples. This was demonstrated in our study in two patients, with negative IgM antibodies initially, while subsequent samples showed an anti-O157 antibody response, collected 6 and 7 d, respectively, after the onset of HUS.

The reason for the correlation for the O157 LPS-specific IgM ELISA for the first positive serum samples to the corresponding saliva samples being not significant was the wide spread of A405 values in saliva as it is shown in Figure 2A, expressed in SD-units. The results of the present study are encouraging in that they suggest that the use of a simple, noninvasive method, namely saliva sampling, may replace more traditional serological techniques in the setting of E+ HUS. This approach offers a valuable alternative in small—and often severely anemic—children. Furthermore, it may be used as a convenient screening tool for epidemiologic studies, e.g. in outbreaks, and the evaluation of vaccine trials. Future studies are necessary to explore the applicability of saliva-based serology in patients with hemorrhagic colitis and contacts of HUS patients. As with the detection of serum antibodies to E. coli O157, the utility of this method depends on the predominance of E. coli O157 in many regions. The induction of saliva antibodies by emerging non-O157 STEC serotypes, belonging to E. coli O-groups 26 and 111, among others (11, 34–38), appears likely, but has yet to be shown.