In this study, an improvement in the oligonucleotide-based DNA microarray for the genoserotyping of Escherichia coli is presented. Primer and probes for additional 70 O antigen groups were developed. The microarray was transferred to a new platform, the ArrayStrip format, which allows high through-put tests in 96-well formats and fully automated microarray analysis. Thus, starting from a single colony, it is possible to determine within a few hours and a single experiment, 94 of the over 180 known O antigen groups as well as 47 of the 53 different H antigens. The microarray was initially validated with a set of defined reference strains that had previously been serotyped by conventional agglutination in various reference centers. For further validation of the microarray, 180 clinical E. coli isolates of human origin (from urine samples, blood cultures, bronchial secretions, and wound swabs) and 53 E. coli isolates from cattle, pigs, and poultry were used. A high degree of concordance between the results of classical antibody-based serotyping and DNA-based genoserotyping was demonstrated during validation of the new 70 O antigen groups as well as for the field strains of human and animal origin. Therefore, this oligonucleotide array is a diagnostic tool that is user-friendly and more efficient than classical serotyping by agglutination. Furthermore, the tests can be performed in almost every routine lab and are easily expanded and standardized.
enterohemorrhagic E coli
Shiga toxin-producing E. coli.
Escherichia coli belongs to the family of Enterobacteriaceae. These are gram-negative, facultative anaerobic, often peritrichous bacteria. Because of their oxygen-using metabolism, E. coli bacteria play an important role in the symbiotic relationship of a host organism with its intestinal flora. In addition to commensal intestinal E. coli, an increasing number of obligatory pathogenic variants of E. coli have been described. These include both extra-intestinal pathogenic E. coli strains (uropathogenic variants and those causing neonatal meningitis) and intestinal pathogenic E. coli. The latter are grouped in various pathotypes such as enterotoxigenic E. coli, enteropathogenic E. coli, entero-invasive E. coli, entero-aggregative E. coli and STEC. EHEC is a subgroup of STEC. In humans, infections with EHEC serotypes may result in hemorrhagic or non-hemorrhagic diarrhea, and can be complicated by the HUS . Some STEC strains have a zoonotic potential; ruminants are believed to represent their main reservoir [2-5].
Because specific serogroups are associated with certain clinical syndromes, serotyping has a central place in the epidemiology and surveillance of E. coli infections . Classical antibody-based serotyping uses specific antibodies to discriminate variants of O and the H antigens. However, conventional serotyping is largely restricted to specialized laboratories and not suitable for routine diagnostics because: (i) a complete and standardized set of all known O and H antigens is very costly; (ii) serotyping requires trained and experienced personnel; (iii) agglutination results are not always unambiguous and cross-reactions can occur; (iv) a minority of strains is non-typeable; (v), capsular antigens can mask the O antigens; and (iv) non-motile strains that do not express the flagellar antigen are observed.
Different genetic approaches have been described previously [7-9]. Ballmer et al. developed a diagnostic oligonucleotide-based DNA microarray on the ArrayTube format of Alere Technologies GmbH (Jena, Germany) for detection of 24 of the most epidemiologically relevant O antigens from over 180 known O antigens and for 47 of the 53 different H antigens . This oligonucleotide array is a diagnostic tool that is simpler and better than classical serotyping by agglutination. It improves the diagnosis and makes epidemiological studies of E. coli infection easier. Furthermore, Braun et al. later showed that this approach can also be applied to the closely related genus Salmonella and its genoserotyping scheme , according to the Kauffman and White classification Scheme .
Here, we present an improvement in the oligonucleotide-based DNA microarray for genoserotyping E. coli described by Ballmer et al. . We developed and processed primers and probes for additional 70 O antigen groups by biomathematical methods. The microarray was transferred to a completely new platform, the ArrayStrip format of Alere Technologies GmbH, which allows high through-put tests in 96-well formats and fully automated microarray analysis by imaging. Moreover, the test protocol was optimized. Subsequently, we verified the performance of the more advanced microarray by testing E. coli reference and field strains of human and animal origin.
MATERIALS AND METHODS
Bacterial strains, growth conditions and genomic DNA extraction
The microarray was verified with a set of reference strains that had previously been serotyped by conventional agglutination in various reference centers (Table 1). For validation of the microarray, 180 clinical E. coli isolates of human origin were used. These included 56 isolates from urine samples, 50 from blood cultures, 39 from bronchial secretions and 35 from wound swabs; all were obtained in the course of routine diagnostic procedures at the Institute for Medical Microbiology and Hygiene, TU Dresden, Dresden, Saxony, Germany. Furthermore, 51 E. coli isolates of animal origin (42 isolates from cattle, 5 from pigs and 4 from poultry) were included; these originated from clinical diagnostics or epidemiological studies.
|Serotype||Strain id and origin||Genoserotype|
|O13:H11||157044†||O13 + O129 + O135:H11|
|O13||131078‡‡||O13 + O129 + O135:H12|
|O17||168481§§||O17 + O44 + O73 + O77 + O106:H18|
|O28:H37||188340††||O28 + O42:H37|
|O28ac:H-||157048†||O28 + O42:H7|
|O42:H37||123113¶||O28 + O42:H37|
|O44||168497§§||O17 + O44 + O73 + O77 + O106:H18|
|O73||168501§§||O17 + O44 + O73 + O77 + O106:H31|
|O77||166246§||O17 + O44 + O73 + O77 + O106:H18|
|O77||129654¶¶||O17 + O44 + O73 + O77 + O106:H18|
|O77||172819¶¶||O17 + O44 + O73 + O77 + O106:H18|
|O106||132486¶¶||O17 + O44 + O73 + O77 + O106:H18|
|O106||168507§§||O17 + O44 + O73 + O77 + O106:H33|
|O107||131129‡‡||O107 + O117:H27|
|O107||168508§§||O107 + O117:H27|
|O117:H16||123101¶||O107 + O117:H16|
|O118||94331§§||O118 + O151:H10|
|O124:H30||157052†||O124 + O164:H30|
|O129||131140‡‡||O13 + O129 + O135:H11|
|O129||168511§§||O13 + O129 + O135:H11|
|O151||168517§§||O118 + O151:H10|
|O164:H-||157058†||O124 + O164:H7|
The strains were cultivated on tryptone yeast agar (VWR International GmbH, Darmstadt, Germany). A full 1 µL loop (diameter 1 mm) of clonal colony material of each strain was picked from solid medium, resuspended in 200 µL lysis reagents (lysis enhancer A2 dissolved in 200 µL lysis buffer A1, Alere Technologies GmbH) and incubated for 30–60 min at 37°C and 550 rpm in a thermomixing device (Eppendorf GmbH, Hamburg, Germany). RNA-free unfragmented genomic DNA was extracted with a Qiagen DNeasy Blood & Tissue kit (Qiagen GmbH, Heiden, Germany) according to the manufacturer's instructions. The DNA concentration was determined spectrophotometrically at 260 nm and finally analyzed for fragmentation by electrophoresis in a 1% non-denaturing agarose gel.
Multiplex linear DNA amplification and labeling for hybridization to prepared ArrayStrips
For multiplex linear DNA amplification, a set of 252 primers was used. One hundred and three of these primers have been described elsewhere . A list of the additional new 149 primers is shown in Table S1. These primers are non-overlapping, but located as close as possible, on the complementary strand, upstream of the position of the oligonucleotide probe. For labeling and biotinylation of the genomic DNA, a site-specific labeling approach was used . A primer elongation reaction was performed using a primer mixture, the HybPLUSKit (Alere) and 0.5–1.5 µg unfragmented RNA-free genomic DNA of the E. coli isolates according to the manufacturer's instructions. The reaction was started with denaturation (5 min, 96°C), followed by 50 cycles of 60 s at 96°C, 20 s at 50°C, and 40 s at 72°C. The sample was then cooled down to 4°C.
Hybridization of the DNA-based serotyping E. coli ArrayStrips
The ArrayStrips, which were spotted with the probes for serotyping (see Table S1), were produced by Alere Technologies GmbH. Details on the layout of the ArrayStrips are provided in the Results Section. For hybridization, a HybPLUSKit (Alere) was used with the following adapted protocol. Each ArrayStrip was initially washed with 200 µL double-distilled water and then with 150 µL of buffer C1 using a thermomixing device (BioShake IQ; Qinstruments GmbH, Jena, Germany; each 5 min, 50°C, 550 rpm). The hybridization sample consisted of 10 µL labeled probe and 90 µL buffer C1. It was transferred into the ArrayStrip and incubated (60 min, 50°C, 550 rpm). The sample was then removed from the tube and the array washed twice (10 min, 45°C, 550 rpm), with buffer C2. Next, 100 µL conjugate solution (1 µL C3 HRP conjugate plus 99 µL C4 conjugation buffer) was added for 10 min at 30°C and 550 rpm followed by a washing step with 200 µL buffer C5 for 5 min at 30°C and 550 rpm. The ArrayStrip was then stained with buffer D1 (100 µL, 10 min, no agitation). After removal of the liquids, it was photographed using an ArrayMate instrument (Alere) and automatically analyzed. Mean signal intensity (mean) and local background were measured for each probe position and values calculated by the following formula: value = 1 − mean/local background. Breakpoints for the interpretation of signals were defined based on a series of experiments with known, characterized reference strains (DSM10728, DSM1058, DSM10720, DSM10784, DSM11753, DSM9025, DSM9026, DSM5212, DSM9027, DSM9031, DSM8701, DSM8702, DSM9028, DSM9030, DSM17076, DSM9034, DSM9033-Leibniz Institut DSMZ-Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH, Braunschweig, Germany). Resulting values below 0.1 were considered negative and above 0.3 as positive. Values between 0.1 and 0.3 were regarded as inconclusive.
Target gene selection and construction of array
The oligonucleotide array-based detection of 24 of the over 180 known O antigens (O antigens 4, 6–9, 15, 26, 52, 53, 55, 79, 86, 91, 101, 103, 104, 111, 113, 114, 121, 128, 145, 157 and 172) by use of the target sequences of wzx (an O antigen flipase) or wzy (an O antigen polymerase) was developed by Ballmer et al. . In the present study, the same strategies were applied to search in public sequence databases (GenBank, EMBL) for additional wzy and wzx sequences and other previously described O-specific sequences, such as wzm, wzt, rfbU, rfbE, isla29, sil-inv, sil-1, sil-2, wbdA, wbdH, wbdM, wbdU, gtrB and gtrX.
Ballmer et al. evaluated probes for 47 of the 53 known H antigens by using the fliC locus-specific sequences . All these probes were re-used for the expanded microarray described herein.
New 23 to 31 bp oligonucleotide probes were designed as described previously . The new probe sequences were designated using a code consisting of the gene name and the serotype or strain number (Table S1; Fig. S1). Seven wzx or wzy gene probes covered multiple O groups as follows: wzx_O13 + O129 + O135/wzy_O13 + O129 +O135 (O13, O129, O135), wzx_O17 + O44 + O73 +O77 + O106/wzy_O17 + O44 + O73 +O77 + O106 (O17, O44, O73, O77, O106), wzx_O28 + O42/wzy_O28 + O42 (O28 and O42), wzx_O107 +O117/wzy_O107 + O117 (O107 and O117), wzx_O118 + O151/wzy_O118 + O151 (O118 and O151), wzx_O124 + O164/wzy_O124 + O164 (O124 and O164) and wzx_O152 + O173/wzy_O152 + O173 (O152 and O173). Additionally, an individual-specific spot for O152 is present. The array comprises a total of 259 different probes, each probe being spotted in duplicate. A staining control (Biotin-Marker), family-specific controls for Enterobacteriaceae (hp_rrs_611, hp_rrs_612, hp_dnaE_613 and prob_gapA_611) and species-specific controls for E. coli (hp_dnaE_612 and gad_10) complete the array. DNA-free spotting buffer was used as a negative control (Fig. S1).
Initial validation of the diagnostic microarray for the new O antigen groups
The performance for the new O antigen groups of the microarray was tested for each spot with a series of reference strains (Table 1). The system detected the available sequences for O antigens with a high degree of reliability (Table S1). The wzx or wzy gene probes had a specificity and sensitivity of 100%. All 70 new O antigen groups were correctly identified with each corresponding reference strain. Unspecific reactions were rare and limited to a small number of spots exclusively representing O antigen-specific genes other than wzx or wzy. Most cross-reactions were found for the target sequences isla29-O145_11, sil-inv-O145_11, rfbU-O157_11, sil-inv-O157_11, sil1-O157_11 and sil2-O157_11. False-positive signals on spots of these target sequences were found in 24.6% of the reference strains. Correct identification of 47 H antigen groups has already been demonstrated by Ballmer et al. . In the present study, all serotyped H antigen groups of the reference strains were identified correctly.
Further validation of the diagnostic microarray using E. coli field strains
For further validation of the microarray, 180 clinical E. coli isolates of human origin (56 from urine, 50 from blood cultures, 39 from bronchial secretions and 35 from wound swabs) were tested. All isolates were characterized as E. coli by VITEK-II (bioMérieux, Nuertingen, Germany/Marcy l'Etoile, France), a commercially available system for routine identification of clinically relevant bacteria based on metabolic profiles and for susceptibility tests using a miniaturized agar dilution approach. However, classical antibody-based serotyping was not performed on these isolates. All isolates were identified as E. coli; and 132 of the 180 isolates (73.3%) were assigned unambiguously to 26 O antigen groups (O1, O2, O4, O6–O9, O11, O15, O18, O21, O22, O24, O25, O58, O63, O75, O78, O79, O83, O86, O91, O101, O130, O143 and O150) using the wzx or wzy gene probes. Most isolates were grouped as O25 (24 isolates), O8 (15 isolates), O2 (13 isolates), O6 (13 isolates), O9 (11 isolates), and O101 (8 isolates). E. coli isolates of the other O groups were detected much less frequently (Table 2). For 10 isolates positive signals were observed with DNA gene probes for several serogenotypes (wzx_O107 + O117/wzy_O107 + O11, one isolate; wzx_O17 + O44 + O73 + O77 + O106/wzy_O17 + O44 + O73 + O77 + O106, nine isolates). This occurred because the sequences of the wzx/wzy genes of these O groups are so similar that exact separation is not possible. Additional positive reactions, besides the wzx and wzy gene probes, were found in only 6 of the 180 isolates (3.3%). Here, positive signals were obtained for the wzx_O9/wzy_O9 and wz_O101_11 gene probes. It should be noted that this wz gene probe is not an allele of wzx/wzy (Table 2). No signals were detected for 32 isolates (17.8%). Six probes (isla29-O145_11, sil-inv-O145_11, rfbU-O157_11, sil-inv-O157_11, sil1-O157_11 and sil2-O157_11) showed strong cross-reactions with about 65% of all strains.
|Number||Isolates from urine||Isolates from bronchial secretions||Isolates from blood cultures||Isolates from wound swabs|
|O antigen group (wzx/wzy)|
|O017 + O044 + O073 + O077 + O106||9||4||1||2||2|
|O107 + O117||1||0||0||1||0|
|H antigen group (fliC)|
For 176 of the 180 E. coli field strains (97.8%), the H antigen groups were unambiguously identified. The isolates were assigned to 25 H antigen groups (H1, H2, H4–H10, H12, H16, H18–H21, H25, H27, H28, H30, H34, H40, H41, H42, H45 and H51). Most isolates were grouped as H4 (39 isolates), H9 (21 isolates), H1 (20 isolates), H7 (16 isolates), H18 (14 isolates) and H6 (13 isolates). E. coli isolates of the other H antigen groups were detected much less frequently (Table 2). Cross-reactions were not observed. For just four isolates no H antigen groups could be identified because no signals were obtained with the fliC probes. All results for the strains of human origin are shown in Table 2.
Additionally, 51 E. coli isolates of animal origin were tested. Complete classical antibody-based O:H serotyping results were known for 37 of the 42 cattle isolates. For the E. coli strains isolated from pigs and poultry, only the O antigen groups had been determined by antibody-based serotyping. A very good concordance was found between the results of classical antibody-based serotyping and the DNA-based genoserotyping by oligonucleotide microarray. Different results were only obtained for the O antigen groups of three E. coli strains. One O136:H49 isolate showed a positive signal with the DNA probes wzx_O6_11/wzy_O6_11 (Table 3) whereas fliC for H49 was correctly identified. A second discrepancy, a clearly positive signal with the K-12-specific DNA gene probes (wzx_O150-K12/wzy_O150-K12), was detected for an E. coli isolate that had originally been reported as O69:H− (Table 3). Furthermore, no signal for any wzx/wzy target sequence was detected for an O98:H-, although DNA probes for O98 (wzx_O98/wzy_O98) are present on the microarray (Table 3). The microarray revealed the O groups of five strains that had not been identified by classical antibody-based serotyping (rough forms or untypeable). Because the target sequences for these O antigen groups are not yet known and therefore not covered by the microarray, correct positive signals for wzx/wzy DNA probes were lacking on the microarray for six O165:H25, two O156:H25, one O136:H- and one O84:H- strain. Three of the five strains without classical serotyping data were characterized as O40:H2, whereas only the H groups (H5 and H8) were identified in the remaining two isolates. Full accordance of the O antigen identifications by classical antibody-based serotyping and by microarray was found for all strains isolated from pigs and poultry. There was also full accordance for H antigen group identifications obtained with the two different methods. Importantly, for 17 strains that had been reported to be non-motile, clear results for a fliC gene, which apparently was not expressed, were found (Table 3). Unspecific reactions were rare, being limited to a small number of spots in seven strains (13.7%) exclusively representing O antigen-specific genes other than wzx or wzy. All results are listed in Table 3. The datasets of mean values, background values and calculated signals of detected spots are shown in Tables S2, S3, and S4.
|Number||Host||O antigen group (wzx, wzy)||H antigen group (fliC)||Classical antibody-based serotyping|
|02||Cattle||28 + 42||37||O28:H37|
|15||Cattle||107 + 117||27||O107:H27|
|34||Cattle||O118 + O151||16||Ont:H16|
We here present an improved version of the oligonucleotide-based DNA microarray for E. coli genoserotyping described by Ballmer et al. . We have expanded the microarray to cover 94 O antigen groups (O antigens 1– 4, 6–9, 11, 13, 15, 17, 18, 21, 22, 24–29, 32, 35, 40, 42, 44, 45, 52, 53, 55, 56, 58, 63, 66, 70, 71, 73, 75, 77–79, 81, 83, 85–87, 91, 98, 101, 103–107, 111–115, 112ac, 117–119, 121, 123, 124, 126–130, 135, 138, 139, 141, 143, 145–152, 157, 159, 164, 167, 168, 172–174 and 177). Ballmer et al. evaluated probes of 47 of the 53 known H antigens (H antigens 1–12, 14–16, 18–21, 23–34, 37–43, 45, 46, 48, 49, 51–54, and 56) . We used all these probes in the improved microarray. In addition, we transferred the microarray to the ArrayStrip format of Alere Technologies GmbH, which allows high through-put tests in 96-well formats and fully automated microarray analysis during imaging. Initial validation against agglutination procedures as the “gold standard,” showed a sensitivity and specificity of 100% using the wzx or wzy gene probes for the new O antigen groups in addition to the panel of probes. All new 70 O antigen groups were correctly identified by each corresponding reference strain. Also, we correctly identified all H antigen groups of serotyped reference strains. Furthermore, validation of the microarray with field strains of human and animal origin yielded excellent results. We unambiguously identified more than 70% of the O antigen groups and nearly 98% of the H antigen groups of the E. coli human isolates. For the remaining strains, we could not include DNA probes in the microarray because the target sequences are unknown. We found no cross-reactions either using the wzx/wzy DNA probes or for H antigen group determinations by fliC gene probes. Moreover, we found a high degree of concordance between the results of classical antibody-based serotyping and DNA-based genoserotyping by oligonucleotide microarray for the E. coli strains of animal origin. We obtained different results regarding the O antigen groups for only three strains. It is possible that the classical antibody-based serotyping was incorrect for these strains. It is also possible that the cultures were polyclonal, that is contaminated with a second strain, and that one clone was tested serologically whereas another one was subjected to genotyping.
Genes selected for detection of the O antigen, that is, wzx and wzy, are the most promising target sequences for an oligonucleotide DNA microarray-based typing approach, the results of which correlated strongly with those of the classical agglutination method. The agreement of the results of the genotyping assay for wzx, wzy, and fliC with those of serotyping is so high that the microarray technology may replace serotyping as the “gold standard” once relevant sequence data for all serotypes are available. The alternative genes proved to be less specific. We detected most cross-reactions for the target sequences isla29-O145_11, sil-inv-O145_11, rfbU-O157_11, sil-inv-O157_11, sil1-O157_11 and sil2-O157_11. We found false-positive signals for these target sequences in a quarter of the reference strains. Among clinical isolates of human origin, we detected up to 65% cross-hybridization signals with unrelated serotypes. This indicates that the genes of the regions selected for probe design are less suitable than the highly discriminatory wzx and wzy genes.
In summary, genoserotyping is very important from a scientific and practical view. We and other authors have previously shown this for E. coli  and other bacterial species [5, 11, 14-16]. In this study, we detected an excellent match between the results of classical antibody-based serotyping and DNA-based genoserotyping both in validation of the new 70 O antigen groups as well as for the field strains of human and animal origin. The advantages of geroserotyping are obvious, namely: (i) unambiguous results for more than half of all known O antigen groups and nearly all H antigen groups within a few hours, starting from a single colony; (ii) parallel detection of O and H antigens of a single E. coli isolate in a single experiment; and (iii) simultaneous typing of a considerable number of E. coli isolates by use of the ArrayStrip format. Thus, the O and H oligonucleotide microarray is an easy and fast technique and provides a helpful and powerful tool for characterization of pathogenic and non-pathogenic E. coli.
We thank F. Gunzer (Institute for Medical Microbiology and Hygiene), E. Bingen (Service de Microbiology, Paris, France), H. Hächler (Institute for Food Safety and Food Hygiene, Zürich, Switzerland), S. Blum (Kimron Veterinary Institute, Bet Dagan, Israel), S. Chappell (Animal Health and Veterinary Laboratories Agency, Weybridge, UK), L. Beutin (Federal Institute for Risk Assessment, Berlin, Germany), and T. Lindbäck (Norwegian School of Veterinary Science, Oslo, Norway) for kindly providing reference strains.
S. Monecke, I. Engelmann, P. Slickers, S. Braun, and R. Ehricht are employees of Alere Technologies, the company that manufactures the microarrays also used in this study. This had no influence on study design, data collection, and analysis, and did not alter the authors' adherence to all Microbiology and Immunology's policies on sharing data and materials. The other authors do not declare any conflict of interest.