I.K. Toth, Scottish Crop Research Institute, Invergowrie, Dundee DD2 5DA, UK (e-mail: email@example.com).
A number of phenotypic and molecular fingerprinting techniques, including physiological profiling (Biolog), restriction fragment length polymorphism (RFLP), enterobacterial repetitive intergenic consensus (ERIC) and a phage typing system, were evaluated for their ability to differentiate between 60 strains of Erwinia carotovora ssp. atroseptica (Eca) from eight west European countries. These techniques were compared with other fingerprinting techniques, random amplified polymorphic DNA (RAPD) and Ouchterlony double diffusion (ODD), previously used to type this pathogen. Where possible, data were represented as dendrograms and groups/subgroups of strains identified. Simpson's index of diversity (Simpson's D) was used to compare groupings obtained with the different techniques which, with the exception of Biolog, gave values of 0·46 (RFLP), 0·39 (ERIC), 0·83 (phage typng), 0·82 (RAPD) and 0·26 (ODD). Of the techniques tested, phage typing showed the highest level of diversity within Eca, and this technique will now form the basis of studies into the epidemiology of blackleg disease.
The soft rot erwinias, Erwinia carotovora subsp. atroseptica (Eca), E. carotovora subsp. carotovora (Ecc) and E. chrysanthemi (Ech), are responsible for soft rot diseases in a number of crops world-wide (Pérombelon & Kelman 1980). Ecc and Ech have a wide host range, infecting crops mainly in tropical and subtropical regions, whereas Eca is restricted almost exclusively to potato in temperate regions, causing soft rot of tubers and blackleg of stems. Economically, Eca is the most important pathogen (Pérombelon 1992). In order to improve control measures, detection systems and strategies of breeding for resistance to Eca infection, it is important to improve our understanding of the epidemiology of blackleg disease. A greater awareness of Eca strain diversity in the environment is thus essential.
A number of fingerprinting techniques have previously been used to study diversity within Eca. Serological analyses using Ouchterlony double diffusion (ODD) divide the pathogen into nine serogroups, although 65–95% of strains belong to a single serogroup, serogroup I, limiting the usefulness of this technique (De Boer et al. 1979; De Boer et al. 1987; Bång 1989; Smith & Bartz 1990; Persson 1991; Harju & Kankila 1993). A previous phage typing system, based on only five phages, was used to study the diversity and distribution of Eca strains recovered from the rhizosphere and stems of potatoes grown in the Columbia Basin of the Pacific North-west, USA (Gross et al. 1991). These strains, all of which belonged to serogroup I, were divided into four phage groups. However, 82% of strains typed into a single phage group, thus supporting serological data suggesting that the pathogen is highly homogeneous. Darrasse et al. (1994b), using polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) analysis of a pectate lyase (pelY) gene divided 32 Eca strains into only two PCR-RFLP groups while, in contrast, Maki-Valkama & Karjalainen (1994) divided 10 strains of the organism into nine random amplified polymorphic DNA (RAPD) groups using three RAPD primer pairs. Although the latter study suggested that a high level of diversity could be obtained with Eca using this technique, only 10 strains were tested and, of these, seven belonged to a single serogroup, serogroup I. This analysis thus requires further verification on a larger number of strains.
The aim of this work was to develop techniques or to utilize available techniques to increase the level of discrimination currently possible within Eca, a necessary prerequisite to epidemiological studies and therefore control measures. Biolog, phage typing, RFLP and enterobacterial repetitive intergenic consensus (ERIC) analyses were evaluated for the first time on Eca as part of this work and compared with ODD and RAPD analysis previously tested on Eca.
Materials and Methods
Bacterial strains and phages
Sixty strains of Eca were isolated from a number of sources in eight west European countries (Table 1). In order to fully represent the range of Eca strains present in the environment, several previously typed (S.H. De Boer, personal communication) non-serogroup I strains (obtained from Sweden and Finland) were included in the study. The identity of all strains was confirmed using biochemical tests (Graham 1972; Lelliot & Dickey 1984), supported by fatty acid profiling (unpublished data), before being stab inoculated into potato tuber discs to confirm their pathogenicity on potato according to the method of Klement et al. (1990). Strains were stored on nutrient agar (NA) slopes at room temperature for 1–2 weeks or frozen at − 80 °C in freezing medium for long-term storage (Sambrook et al. 1989). Each Eca strain was tested at least twice by each fingerprinting technique.
Table 1. Erwinia carotovora subsp. atroseptica strains and groupings produced by fingerprinting techniques
The ODD method was used as described by Van der Wolf & Gussenhoven (1992). IPO-DLO antiserum 8898 against serogroup I Eca was produced against whole living cells. Antisera numbers 36, 39 and 43 against serogroups XVIII, XX and XXII, respectively, were produced against glutaraldehyde-fixed cells (De Boer et al. 1979) and provided by S.H. De Boer (Agriculture Canada, Newfoundland). Strains were typed on the basis of a reaction of identity with the homologous type strain (IPO 161) used for the antiserum production.
Physiological profiling (Biolog) analysis
Single colonies of Eca strains (Table 1), grown on NA at 27 °C for 48 h, were streaked onto tryptone soya agar (TSA; Oxoid, Basingstoke, UK) plates and incubated at 27 °C for a further 24 h. Using sterile cotton swabs, bacteria were removed from the TSA plates, suspended in sterile saline (0·85% NaCl) and the optical density (O.D.) adjusted to between 0·3 and 0·5 at 595 nm. The suspension (150 μl) was added to the 96 wells of both a Gram-negative and Gram-positive Biolog plate (Biolog, Hayward, CA, USA), incubated at 27 °C for 24 h, and O.D. (595 nm) readings taken on a microtitre plate reader (Multiskan MS; Labsystems, Cambridge, UK). Wells giving values of at least twice the control well were considered positive. To improve the level of discrimination, only results from the 47 tests (from a total of 128 individual tests) showing a differential response to the different Eca strains were analysed.
Phage host range determination
Twenty-two different Eca-specific phages, previously isolated from sewage (Toth et al. 1989), were purified and a high titre lysate made according to the method of Toth et al. (1993) using serogroup I Eca strain SCRI 1043. High titre lysates of the 22 phages were diluted to a routine test dilution (RTD; the lowest dilution to give confluent lysis on Eca strain SCRI 1043) and the phages then stored in phage buffer (Sambrook et al. 1989) over chloroform at 4 °C. Soft agar overlays (3 ml distilled water agar (0·7%) containing 300 μl bacterial overnight culture in Luria Bertani (LB) broth) were made on LB plates and 5 μl RTD of each lysate spotted onto the overlays. After overnight incubation at 25 °C, clear zones, turbid zones or individual plaques were considered positive for phage sensitivity.
To ensure the phenotype stability of strains when using Biolog, ODD and phage typing techniques, five Eca strains (Z147, IPO 856, H1, 161 and 1040) were inoculated into potato tuber discs according to the method of Klement et al. (1990). Following re-isolation of the strains, the re-inoculation step was repeated and strains from this infection were re-isolated and re-typed using the above techniques.
Total genomic DNA (gDNA) was extracted from Eca cells grown in LB broth at 27 °C for 18 h according to the method of Sambrook et al. (1989) and tested as follows. Enterobacterial repetitive intergenic consensus analysis using primers ERIC2 (5′-GCGAGTGGGGTCAGTGAATGAA-3′) and ERIC1R (5′-ATGTAAGCTCCTGGGGATTCAC-3′) (Versalovic et al. 1991) was carried out in a PCR reaction mix containing: 2 μl DNA (15 ng μl−1), 0·3 μl Taq polymerase (5 U μl−1; Ampligene, Rockville, MD, USA), 5 μl 10× PCR buffer (Ampligene), 5 μl dNTPs (1·25 mmol l−1), 1 μmol l−1 of each primer ERIC2 and ERIC1R and HPLC-grade water to a total volume of 50 μl. DNA amplification was performed on a GeneATAQ thermocycler (Pharmacia, Biotech, St Albans, UK) under the following conditions: 96 °C for 6 min for the first cycle, 94 °C for 1 min, 52 °C for 1 min and 72 °C for 2 min repeated for 30 cycles, and a final cycle of 72 °C for 6 min. Following PCR amplification with the ERIC primers, 10 μl from each reaction was electrophoresed through a 1·8% (w/v) agarose gel as described by Sambrook et al. (1989). Reproducibility was assessed by comparing two independent DNA extractions from each of the 60 strains.
Restriction fragment length polymorphism analysis of gDNA was carried out as follows: gDNA (1 μg) from each Eca strain (Table 1) was digested with restriction endonucleases HindIII or EcoRI (Boehringer Mannheim, Lewes, UK) and treated with RNase (Boehringer Mannheim), both according to the manufacturer's instructions. Digested samples (20 μl) were electrophoresed through a 1·5% (w/v) agarose gel and DNA bands visualized with ethidium bromide as described by Sambrook et al. (1989). The probe (B-probe) (Darrasse et al. 1994a) was labelled with digoxigenin (Boehringer) according to the manufacturer's instructions and used for Southern blot analysis of the gel (Sambrook et al. 1989).
Random amplified polymorphic DNA analysis was carried out using primers OPB-07, OPB-11 and 91300 as described by Maki-Valkama & Karjalainen (1994) but with a modified master mix containing: 2 mmol l−1 MgCl2, 1 mmol l−1 dNTPs, 0·4 μmol l−1 primer, 2 U Stoffel Taq DNA polymerase (Perkin Elmer, Warrington, UK) and 10× PCR buffer (Perkin Elmer). Since primer 91300 offered no additional differentiation of strains following analysis with OPB-07 and OPB-11, it was not used further.
Analyses of data
The phage typing and RAPD data were converted to binary form (0 = absence or negative; 1 = presence or positive), similarity matrices constructed using the simple matching coefficient and cluster analysis performed on the similarity matrices using the unweighted average pair group method (Jackson et al. 1989). To compare typing techniques directly, Simpson's index of diversity (Simpson's D) was used (Hunter & Gaston 1988) to assess both the richness (the number of types in a sample) and evenness (the relative distribution of individual strains among the different types) of strain groupings. Simpson's D is termed ‘diversity index’ throughout this text. Similarity between strains refers to groupings obtained from analysis of dendrograms.
Ouchterlony double diffusion analysis
Forty-six of the 60 Eca strains tested belonged to serogroup I, three were tentatively grouped into serogroup I based on a reaction of identity, although only a weak precipitation line was formed (data not shown), eight belonged to four other serogroups and three could not be typed using ODD (Table 1). Even with the inclusion of eight selected non-serogroup I strains, a low diversity index was obtained for serogrouping (0·26), reflecting an uneven distribution between the five serogroups present, i.e. 85% serogroup I.
Cluster analysis, based on the differential response to 47 carbon sources, showed considerable variation between the Eca strains (data not shown). However, when strains Z141, IPO 856, H1, 161 and 1040 were re-tested several weeks later, either directly or following re-isolation from tubers, the replicates often failed to cluster alongside the original sample. For three strains tested (H1, Z141 and 161), replicates grouped together but showed 5–20% variation in similarity. In other cases, samples failed to cluster with the other replicates and showed up to 35% variation in similarity (data not shown).
Forty-nine of the 60 strains tested showed sensitivity to the phages (Table 2). Following cluster analysis, two groups (A and B) were defined at a similarity of approximately 38% (Table 1 and Fig. 1). At 70%, group A divided into 2 subgroups (A1 and A2), although within subgroup A1 three clearly distinct clusters were present (A1i, A1ii and A1iii). The majority of non-serogroup I strains grouped into cluster A1ii and in all but one case these strains failed to react with any of the phages. At 70%, group B divided into four subgroups (B1–B4). With the exception of cluster A1ii, which contained the phage-resistant strains, all subgroups/clusters showed considerable heterogeneity between member strains and produced a diversity index of 0·83 (Table 1).
Table 2. Phage type of Erwinia carotovora subsp. atroseptica strains
For RFLP analysis, following digestion of gDNA with EcoRI, the B-probe hybridized to a single band of > 25 kbp in all strains, offering no differentiation between these strains (data not shown). However, when EcoRI was replaced by HindIII, hybridization patterns divided the strains into two groups with either a 6·5 kbp (group 1) or a 21 kbp (group 2) hybridization product, representing 65% and 35% of strains, respectively. This grouping produced a diversity index of 0·46 (Table 1).
Amplification of gDNA using primers ERIC2 and ERIC1R generated four to six bands of between approximately 2 kbp and 400 bp for each strain (data not shown) and divided the strains into three groups representing 77% (group 1), 15% (group 2) and 8% (group 3) of strains. The groupings produced a diversity index of 0·39 (Table 1).
Amplification of gDNAs using both RAPD primers OPB-07 and OPB-11 produced a total of up to eight bands for each strain (Fig. 2). Cluster analysis, based on the presence or absence of these bands, grouped the strains into two clearly distinguishable groups (1 and 2) at a similarity of approximately 70% (Table 1 and Fig. 3). These groups were further divided into six subgroups (1a (consisting of three clusters 1ai, 1aii and 1aiii), 1b and 2a−2d) at approximately 90% similarity, producing a diversity index of 0·82.
Common groupings between techniques
The results of each fingerprinting technique were cross-tabulated in an attempt to identify common groupings between them (data not shown). A clear relationship emerged between serogroups and phage types (Fig. 1) where all non-serogroup I strains clustering into subgroup A1 (cluster A1ii) were phage-resistant, with the exception of strain 495 (serogroup XXXV). This suggests that phages used in the study were serogroup I-specific (Table 1, Fig. 1).
No clear relationships were observed between RFLP and ERIC groupings (Table 1 and Fig. 3), although groups produced by each of these techniques were related to groups produced by RAPD analysis (Fig. 3). For example, RAPD group 1 contained 16 of the 21 strains belonging to RFLP group 2. Moreover, RAPD subgroup 2b and cluster 1aii contained all ERIC group 3 strains and five of the nine group 2 strains, respectively.
Until the late 1980s, ODD was the only technique available for studying diversity within Eca. The high level of serological homogeneity within the subspecies, however, has severely limited the application of this technique for Eca strain typing. When ODD was used to type the 60 Eca strains in this study, little differentiation was observed between strains and, of all the techniques tested, this gave the lowest diversity index (0·26). The low diversity index was due to the predominance of serogroup I strains used in the study, which also reflects a similar predominance in nature (De Boer et al. 1979; De Boer et al. 1987; De Boer & McNaughton 1987; Bång 1989; Hyman & Pérombelon 1989; Persson 1991; Harju & Kankila 1993; Perminow 1997).
Biolog is traditionally used for bacterial identification (Klingler et al. 1992; Jones et al. 1993) and has been used with members of the Erwinia genus (Lacroix et al. 1995). However, it has also been used to differentiate between bacterial strains, e.g. Xanthomonas campestris, relating these strains to their different hosts and to DNA homology groups (Chase et al. 1992; Hildebrand et al. 1993; Verniere et al. 1993). Due to the success of this method with X. campestris, Biolog was used in this study in an attempt to differentiate within Eca, but was unable to provide clearly distinct groups. This came to light when replicate strain samples failed to cluster together. Following these results, a more detailed analysis was carried out using strains inoculated and re-isolated from tubers. Whether replicates were tested on the same day or several weeks apart, the variation in similarities between replicates was too great to produce meaningful groupings. This variation may have been due to small differences in the level of inoculum used in the Biolog plates, leading to different rates of substrate utilization, changes in optical densities and therefore changes in the overall substrate utilization profiles. Overall, these data suggest that, compared with X. campestris pathovars, Eca is nutritionally homogeneous.
The phage typing system used in this study was based on the sensitivity of the Eca strains to 22 Eca-specific phages, in contrast to only five phages used in the system developed by Gross et al. (1991). While these workers found little variation between Eca strains, the new system showed a high level of diversity. Although this may have been the result of the different Eca strains used, it is more likely to be due to either the use of more phages in this system or to more diverse phage populations being isolated from sewage than from plant and rhizosphere soil. Phage typing was the most discriminatory of all the techniques used, with all replicate strains producing identical phage types. Using phage typing, the 60 Eca strains clustered into two groups although, besides differences in phage sensitivity, the reason for this clear division is not known. Subgroup A1 (cluster A1ii) contained phage-resistant strains and, by cross-tabulating the results of both ODD and phage typing, a clear relationship emerged between the techniques, i.e. only serogroup I strains were sensitive to the phages (Fig. 1). This is not surprising, however, since all the phages used in the system were isolated on a serogroup I strain of Eca, SCRI 1043 (Toth et al. 1997). In the case of Eca, lipopolysaccharide O-chains are known to be involved in serogroup determination (De Boer & McNaughton 1987; Murray et al. 1990), and may also be a receptor for phage attachment (Ackermann & DuBow 1987). Due to the high predominance of serogroup I in countries such as Scotland and the Netherlands (De Boer et al. 1979; De Boer et al. 1987; De Boer & McNaughton 1987; Hyman & Pérombelon 1989), this phage typing system will provide a useful means of differentiating between strains. Our results also suggest that there may be considerable value in isolating phages capable of plaquing on non-serogroup I strains to extend the phage typing system to other serogroups.
An alternative approach to studying phenotypic differences between strains was to compare differences at the molecular level. RFLP differentiated the 60 Eca strains into only two groups, producing a low diversity index (0·46). The B-probe, which has no significant homology to known genes or to other sequence data held in the EMBL and GenBank databases (Darrasse et al. 1994a), produced a single hybridization product within each group. This suggests that the sequence may be represented within the Eca genome as a single copy and is thus of limited value for studying diversity within Eca.
Enterobacterial repetitive intergenic consensus analysis, which monitors the distribution of repetitive sequences around the genome using sequence-specific primers, has been used successfully to type a number of Gram-negative bacteria (Lupski & Weinstock 1992). Moreover, ERIC analysis can successfully differentiate within Ecc and between Ecc and Eca (Laplaze, unpublished). However, when applied to Eca, only three groups were produced, giving a low diversity index (0·39) and suggesting that, like RFLP, ERIC analysis is of limited value for differentiating within Eca (Table 1).
In contrast, a higher level of diversity was observed within Eca using RAPD analysis, which gave a higher diversity index (0·83) (Fig. 3). The difference in sensitivity between RAPD and ERIC analyses may be due to the specificity of their respective primers, i.e. RAPD analysis uses non-specific primers which bind randomly to regions over the entire genome, while ERIC analysis uses primers specific to a repetitive sequence. Although both methods would detect large sequence changes, e.g. insertions/deletions over much of the genome, only RAPD analysis would detect smaller changes, e.g. point mutations, outside a repetitive sequence, thus offering a higher degree of sensitivity.
Maki-Valkama & Karjalainen (1994) previously differentiated 10 strains of Eca into nine groups based on RAPD analysis. Although nine subgroups were also defined in this study, banding patterns, produced on amplification of gDNA using primers OBP-07 and OBP-11, varied between laboratories. The most plausible explanation for these differences in banding patterns is the slight differences in protocols used by each laboratory, necessary for optimizing conditions in each laboratory. Once conditions were optimized, however, RAPD results were reproducible (Maki-Valkama & Karjalainen 1994).
In conclusion, a number of phenotypic and molecular fingerprinting techniques have been compared for their ability to differentiate between strains of Eca. Contrary to previous findings, considerable diversity has been demonstrated within Eca. Of the molecular techniques tested, RAPD analysis was the most discriminatory, as it was most sensitive in detecting small changes in sequence over the entire genome and, in part, encompassed the lesser variation observed with both ERIC and RFLP analyses. However, phage typing was both the most reproducible and the most sensitive of the techniques, providing most strains with unique phage types. In addition, it is also the technically most straightforward of all the techniques used. We are currently attempting to isolate phages to non-serogroup I strains to extend the phage typing system to these strains. Phage typing will now form the foundation for new studies into the epidemiology of blackleg disease.
The authors acknowledge the support of EU grants AIR 3CT93–0876 and VALUE CTT-477, in addition to the different national research agencies of the authors. The valuable contributions made by P.M. de Vries, M. van den Brink-Egberts, Susanne Zock, Wetlesen Akselsen, Sophie Rosentaub, Nicole Payet and Pascal Exbrayat are gratefully acknowledged. The authors also thank S.H. De Boer, Agriculture Canada, Newfoundland, for supplying strains.