Classification of Salmonella enterica serotypes from Australian poultry using repetitive sequence-based PCR


  • Disclaimer: Mention of trade names or commercial products in manuscript are provided for specific information only and does not imply endorsement or recommendation by the authors or their affiliations.

Anthony Pavic, Birling Avian Laboratories, PO Box 111, Bringelly, NSW 2556, Australia. E-mail:


Aims:  To evaluate a semi-automated repetitive extragenic palindromic sequence-based PCR (rep-PCR) system for the classification of Salmonella serotypes from Australian poultry.

Methods and Results:  Using a DNA fingerprint library within the DiversiLab® System, four separate databases were constructed (serogroup B, C, E and Other). These databases contained 483 serologically confirmed (reference laboratory) Salmonella isolates. A blinded set of Salmonella cultures (= 155) were typed by rep-PCR, matched against the internal library and compared with traditional serotyping. The predicted (Kullback–Leibler) serotype of 143 (92·3%) isolates matched traditional typing (< 0·05). Of the 12 (7·7%) remaining isolates, ten (6·5%) resulted in ‘No Match’, one (0·65%) was incorrectly matched to the library (Salm. subsp 1 ser 4,12:-:-), and the other (0·65%) was referenced as Salm. ser. Sofia, whereas rep-PCR and in-house serotyping concurred as Salmonella serovar Typhimurium. Financial analysis showed higher material cost (215%) and a lower labour component (47·5%) for rep-PCR compared with serotyping.

Conclusion:  The DiversiLab® System, with serogroup databases, was successfully implemented as an adjunct for reference serotyping of Salmonella enterica.

Significance and Impact of the Study:  The DiversiLab® System platform is a cost-effective and easy-to-use system, which can putatively determine Salmonella enterica serotypes within a few hours.


Salmonella spp. are zoonotic pathogens of major public health concern that can colonize commercial layer and broiler breeder flocks (FAO/WHO, 2002, 2009; CDC/FoodNet, 2010). The organism may be transmitted vertically or horizontally on-farm and persist throughout processing stages during production (Liljebjelke et al. 2005). Many baseline studies indicate that the postslaughter carriage rates of Salmonella on broiler carcasses varies between different continents, from 36·7% in Australia (FSANZ, 2010), to 15·7% in Europe (EFSA, 2010) and 21·0% in the USA (USDA/FSIS, 2008). The prevalence of the bacterium on postslaughter carcasses results in a link between human salmonellosis and consumption of poultry products (Batz et al. 2005; Braden 2006; Callaway et al. 2008).

Implementation of HACCP-based intervention strategies for the control of Salmonella in poultry production begins on-farm with elite breeder flocks and filters down the production pyramid to the processing plant (Cox and Pavic 2010). While these measures are important for meeting performance standards and reducing carriage rates, fine, strain-level discrimination of Salmonella isolates is required to examine the association between serotypes isolated from the farm, postslaughter and human cases of salmonellosis (Scott et al. 2002; Gan et al. 2010; Kilic et al. 2010).

The conventional technique for typing among the salmonellae has been serotyping using specific antibodies, in accordance with the White–Kauffman–Le Minor scheme. The common antigens targeted are the somatic O-antigen (serogroup) and flagellar H-antigens (serotype), which may have up to three phases present (Brenner et al. 2000; Cox 2000; Grimont and Weill 2007). Further discrimination of prevalent serotypes is performed typically by phage typing (Threlfall and Frost 1990).

As of 2007, there were approximately 2600 serotypes known, with nearly 1530 serotypes identified within Salmonella enterica subsp. enterica (subspecies I), the subspecies primarily responsible for infections in humans and colonization of warm-blooded animals (Grimont and Weill 2007; Guibourdenche et al. 2010). In Australia, Salmonella serovar (ser.) Typhimurium is associated typically with at least 25% of the human cases of salmonellosis annually (Davos 2009). Furthermore, Salmonella Typhimurium (isolated from 13% of carcasses), Salmonella subsp II 1,4,12,27:b:[e,n,x] (referred to hereafter as Sofia; 36·6%), serotypes Kiambu (12·5%), Agona (4·8%) and Infantis (3·3%) represented the top five salmonellae isolated from broiler chicken carcasses in Australia in 2008 (Davos 2009).

As a reference method, serotyping has been useful during source-attribution investigations (Baggesen et al. 2010). However, the serotyping of isolates in Australia requires submission to a limited number of reference laboratories that stock specific antibodies and employ trained personnel; serotyping can take several days to weeks. This has lead to the adoption of nucleic acid-based methods, such as pulsed-field gel electrophoresis (PFGE) and multilocus sequence typing (MLST), which allow rapid strain typing for source tracking of Salmonella (Foley et al. 2007, 2009). These methods are particularly useful in the poultry industry (Yan et al. 2004; Foley et al. 2007; Ben-Darif et al. 2010; Gan et al. 2010). Currently, many authors advocate restriction fragment length polymorphism as the preferred technique for discrimination of salmonellae (Weigel et al. 2004; Foley et al. 2009; Kilic et al. 2010). An alternative method, repetitive extragenic palindromic sequence-based PCR (rep-PCR) was compared with PFGE and found to be the more discriminatory when applied to a collection of 68 isolates of Salmonella from swine farms (Weigel et al. 2004).

Rep-PCR uses primers that target naturally occurring, noncoding, repetitive DNA sequences interspersed throughout the bacterial and fungal genome (Versalovic et al. 1994). Multiple DNA fragments (amplicons) of different sizes are amplified during PCR and, when separated by electrophoresis, constitute a genomic fingerprint pattern or ‘barcode’ of varying band sizes and intensities, thereby defining a strain (Versalovic et al. 1994; Versalovic and Lupski 2002). The rep-PCR technology has been commercially adapted to an automated format known as the DiversiLab® System, which offers advances in standardization and reproducibility over manual, gel-based rep-PCR analysis (Healy et al. 2005). A modification of the DiversiLab® System is the inclusion of a microfluidics chip-based DNA fragment separation rather than traditional gel electrophoresis, thus yielding distinct and reproducible banding patterns (Anderson et al. 2010). In this chip format, ten samples can be analysed simultaneously with controls, and internal DNA standard markers are added to each well to allow for normalization and efficient chip-to-chip comparisons. Networking of the bioanalyser to web-based software allows for archiving of fingerprint profiles and construction of reference libraries to which unknown (‘query’) isolates can be matched.

Wise et al. (2009) typed 44 poultry Salmonella isolates using the DiversiLab® System; though, only 28 isolates (64%) showed concordance with serotyping. A comparison between MLST and rep-PCR, using the DiversiLab® System, showed the latter provided equivalent or better prediction of serotypes among Salm. enterica isolates, owing to greater discriminatory power (Wise et al. 2009; Ben-Darif et al. 2010; Hyeon et al. 2011). In all of these studies, the query isolates were sourced solely from the United States (Wise et al. 2009; Anderson et al. 2010), the United Kingdom (Ben-Darif et al. 2010) and Korea (Hyeon et al. 2011), which may restrict the diversity of the collection.

It was the necessity for speed and accuracy in serotyping that initiated this study in 2007 (Pavic and Bailey 2009) with final accreditation of this method being granted in 2010. The objectives of this study were to determine whether separating the Salmonella library based upon serogroup (B, C, E and Other) would increase agreement using a blinded set of 155 poultry isolates and whether rep-PCR was more economical to perform than standard serotyping.

Materials and methods

Salmonella strains, culture and DNA extraction

Salmonella strains

To construct the initial libraries, 483 isolates from poultry, representing 41 different serotypes, were grouped using commercial antisera (PROLAB Diagnostic, ON, Canada; Refs TL6002 [O], TKL6001 [H], RL6011-04 [B], PL6013 [C] and PL6017 [E]) according to the manufacturer’s instructions and stored in duplicate on nutrient agar (NA, Oxoid CM3, Basingstoke, UK) slopes at 4°C. For each isolate, one slope was sent to the Australian Salmonella Reference Centre (Institute of Medical and Veterinary Sciences [IMVS], Adelaide, South Australia) for serotyping and the other cultured for purity for DNA extraction.

Salmonella isolation and confirmation

The isolates constituting the blinded set of query strains (n = 155) were isolated from various poultry samples (processed dressed carcasses, by-products, drag swabs and litter) using a modification of the ISO 6579:2002 method (Pavic et al. 2010) at a laboratory accredited with the National Association of Testing Authorities. The following validated modifications were made to the method: samples were emulsified 1:10 in buffered peptone water (BPW; Thermo Fisher Oxoid; CM509, Hampshire, UK); the selective and differential split plate formulated with xylose lysine deoxycholate agar and Hektoen agars (Thermo Fisher Oxoid, PP2027, Adelaide, Australia); confirmation on the chromogenic medium SMID2 (bioMérieux, ref. 43621, Marcy-l’Étoile, France). All confirmed Salmonella were sent to the Australian Salmonella Reference Laboratory.

DNA extraction

For each confirmed Salmonella isolate, one colony from a pure culture was grown on NA for 24 h at 37°C. The pure cultures were then suspended in 2·5 ml of 0·85% sterile saline (bioMérieux ref. 04286, Brisbane, Australia) to obtain a final density of McFarland Standard No. 4 (bioMérieux ref. 70900, Marcy-l’Étoile, France). The DNA from each isolate was extracted from a 1 ml aliquot of each suspension using the NucliSENS® easyMAG® workstation (bioMérieux Inc., Durham, NC) and an eluate volume of 25 μl was used to obtain a final genomic DNA concentration of 25–50 ng μl−1. DNA extracts were stored at −80°C prior to use.

rep-PCR DNA fingerprinting

All DNA samples were amplified using the DiversiLab® Salmonella Kit for DNA fingerprinting (bioMérieux ref 270625) according to the protocol of Wise et al. (2009). Briefly, a total volume of 25 μl/PCR mixture contained 18 μl of the kit-supplied rep-PCR master mix (MM1), 2·5 μl of 10× GeneAMP PCR Buffer I (Qiagen Taq PCR Core Kit 201225; Qiagen, Victoria, Australia), 2 μl kit-supplied primer mix, 0·5 μl (or 2·5 U) of AmpliTaq® polymerase (Qiagen) and 2 μl of DNA extract (concentration approximately 25–50 ng μl−1). DNA amplification was performed using a thermal cycler (Applied Biosystems GeneAmp 2700) under the following conditions: initial denaturation at 94°C for 2 min; followed by 35 cycles of denaturation at 94°C for 30 s, annealing at 50°C for 30 s, extension at 70°C for 90 s; and a final extension at 70°C for 3 min. Separation of rep-PCR products was performed using the Agilent Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA, USA).

Analysis of query strains

The strains were analysed with the Diversilab® System software ver. 3.4, and the resulting DNA fingerprint patterns were viewed initially as electropherograms. The software reporting system includes a dendrogram construction option, as well as an electropherogram of each sample. The Kullback–Leibler distance correlation coefficient, which weighs band presence or absence more heavily than band intensity, was selected to calculate similarity matrices of the virtual gel DNA fingerprint images.

The unknown serotypes were predicted on the basis of rep-PCR fingerprints, using the ‘Top Match’ feature of Diversilab. This feature matches the unknown serotype to the five most similar library serotypes and reports the percentage similarity. If the unknown serotype matched a particular serotype library entry at greater than or equal to 95% similarity, it was considered to be a positive designation. If more than one serotype yielded a ≥95% similarity, the Next Top Match was also reported. Isolates that were less than 95% similar to the library were reported as No Match.

Cost analysis

Salmonella Typhimurium was used for a labour comparison between the traditional full serotyping and molecular typing with all isolates being serogrouped from pure cultures. A stopwatch was used to measure time (in minutes) required to type ten isolates. Traditional serotyping included (i) set-up (labelling slides, emulsifying the colony in serum) (ii) reading by rocking the slide for a maximum of 2 min and (iii) reporting onto worksheets and then transcribing into a computer-based information system. The steps covered using the DiversiLab® System with the DiversiLab®Salmonella DNA fingerprinting kits were (i) DNA extraction, (ii) rep-PCR amplification, (iii) separation and analysis of product and (iv) reporting. The material usage was calculated by the percentage difference in the material cost between the two methods using the traditional method as the denominator.


Library construction of Salmenterica isolates from poultry

All of the 483 Salmonella isolates that were used to construct the library could be typed with the DiversiLab® System and Salmonella DNA fingerprinting kit. The Kullback–Leibler distance correlation coefficient was selected to automatically compare the rep-PCR profiles and create corresponding dendrograms using diversilab®. Initially, the library was constructed using the most commonly isolated serotypes and, when possible, a minimum of five serologically confirmed isolates of each serotype were used. As less common serotypes were isolated, these were added to the library, which was divided into four separate databases belonging to serogroups B, C, E and Other (Fig. 1). The serotypes included were among those most commonly isolated from commercial broiler chickens.

Figure 1.

 Dendrogram illustrating the internal (a) serogroup B, (b) serogroup C, (c) serogroup E and (d) Other library of 483 poultry-associated Salmonella spp. isolates. The horizontal scale bar indicates the per cent similarity among serotypes. Numbers in parentheses indicates total number of strains included in the cluster.

For most serotypes, there was limited variability among the isolates; for example, all 12 of the serotype Kiambu fingerprints were ≥95% similar to each other (Fig. 1b). However, isolates of some serotypes exhibited more extensive strain-level diversity (i.e. Salm. Infantis had five discrete fingerprint types) when placed in the dendrogram. The cluster represented by isolate SAL180 in Fig. 2 was on average 96·1% similar to the cluster represented by isolate SAL385. Serotype Infantis isolate SAL141 was only 93% similar, on average, to the SAL153 and to the SAL385 clusters.

Figure 2.

 Similarity Matrix illustrating the clustering of Salmonella serovar Infantis represented by five discrete fingerprint types. The group represented by isolate SAL180 was on average 96·1% similar to the cluster represented by isolate SAL385. Serotype Infantis isolate SAL141 was only on average 93% similar to the SAL153 and SAL385 clusters.

During the construction of the library, there were ten instances where isolates of different serotypes showed fingerprints that were ≥95% similar when using the four different libraries based on serogroup (Group B, Group C, Group E and Other): Typhimurium/Abortusovis, Infantis/subsp 1 ser rough:r:1,5, Virchow/subsp 1 ser rough:r:1,2, Anatum/Anatum var 15+, 34+, Orion var 15 + /Orion var 15 + , 34 + , Saintpaul/Typhimurium, Amsterdam/Amsterdam var 15 + , 34 + ; Reading/Saintpaul, Chester/Saintpaul and Chester/Reading.

Using the electropherogram overlay function, DiversiLab® System electropherograms showed that there were slight but consistent peak differences when Salm. Typhimurium was compared with Salm. Abortusovis and Salm. Saintpaul (Fig. 3). Similar slight but critical differences were observed for Reading/Saintpaul, Chester/Saintpaul and Chester/Reading. All the other rep-PCR fingerprints generated from variant and mutant rough strains were indistinguishable from their corresponding ‘parent’ or ‘smooth’ strains (Fig. 4).

Figure 3.

 (a) Electropherogram overlay of rep-PCR amplicons from Salmonella Typhimurium isolate 27133A (grey curve) and Salm. Abortusovis isolate SAL167 (red curve). (b) Electropherogram overlay of rep-PCR amplicons from Salm. Typhimurium isolate SAL388 (grey curve) and Salm. Saintpaul isolate 27845A (red curve). The arrows indicate a fluorescent peak difference between the samples. These peaks were consistently present among all Typhimurium isolates but not among Saintpaul or Abortusovis.

Figure 4.

 Electropherogram overlay of rep-PCR amplicons from Salmonella Anatum var 15 + , 34 +  (SAL410) (grey curve) and Salm. Anatum (SAL409) (red curve). No significant differences in overlays were observed between the two different serotypes.

Classification of unknown Salmonella isolates by comparison to the library

Salmonella isolates (155) obtained from a variety of poultry samples (processed carcasses, by-products, drag swabs and litter), using an internally validated modification of ISO 6579:2002, were typed with the DiversiLab® System and queried against the Salmonella databases previously described. The corresponding duplicate set of isolates was serotyped by conventional methods.

The clustering of the DiversiLab® System dendrogram and the Top Match function (Fig. 5) of the classification report was subsequently utilized as a guide to putatively assign serotype. The data summarized in Table 1 show the agreement between traditional serotyping and DiversiLab® System library matches using 95% similarity as a threshold. Of the 155 isolates tested, 145 had a putative serotype match obtained using the library, a significant (P < 0·0001, sign and binomial test, with probability at 0·5) 143 were in concordance with the traditional serotyping result (98·6%).

Figure 5.

 Top Match reports generated by the DiversiLab® System software. Unknown Salmonella isolates (a) 27838A and (b) 26513A, previously serogrouped to [C], were queried against the poultry-associated Group C database and the five most similar entries along with serotype designation and virtual gel images are displayed.

Table 1.   List of Salmonella isolates studied with predicted serotype by rep-PCR and serological results. Numbers in parentheses denotes total number of strains or matches, assuming the traditional serological result as correct
Traditional serological result*DiversiLab® Top Match†DiversiLab® Next Top Match†% Agreement‡Comments
  1. *Serological result from Australian Salmonella Reference Centre

  2. †A similarity value of ≥95% by Kullback–Leibler distance correlation coefficient considered to be significant for Top Match and Next Top Match.

  3. ‡Agreement calculated on the basis if one of the Top Match or Next Top Match results is in concordance with serotyping.

Abortusovis (3)Abortusovis (3)Typhimurium (3)100One band difference between Typhimurium and Abortusovis
Agona (4)Agona (4) 100 
Amsterdam (3)Amsterdam var 15 + , 34 +  (2), No Match (1)Amsterdam (2)66Variant strains
Anatum (6)Anatum (2), Anatum var 15 + , 34 +  (4)Anatum var 15 + , 34 +  (2), Anatum (4)100Variant strains
Cubana (3)Cubana (3) 100 
Enteritidis (4)Enteritidis (2), No Match (2) 50Phage-type 26 (IMVS)
Hvittingfoss (1)Hvittingfoss (1)   
Infantis (2)Infantis (1), subsp 1 ser rough:r:1,5 (1)Infantis (1), subsp 1 ser rough:r:1,5 (1)100Infantis ‘smooth’ isolate
Isangi (1)Isangi (1) 100 
Kiambu (10)Kiambu (8), No Match (2) 80One unmatched isolate incorrectly grouped as E+
Liverpool (4)Liverpool (4) 100 
Mbandaka (1)Mbandaka (1) 100 
Orion var 15 + , 34 +  (2)Orion var 15 + , 34 +  (1), Orion var 15 +  (1)Orion var 15 + , 34 +  (1), Orion var 15 +  (1)100Variant strains
Saintpaul (1)Saintpaul (1)Typhimurium (1)100Band difference
Singapore (1)Singapore (1) 100 
Sofia (66)Sofia (63), Typhimurium (1), No Match (2)Abortusovis (1)95Typhimurium/Abortusovis lab suspect, reference error
Subsp 1 ser 4,12:d:- (1)Subsp 1 ser 4,12:d:- (1) 100 
Subsp 1 ser 4,12:-:- (1)Typhimurium (1) 0 
Tennessee (12)Tennessee (12) 100 
Typhimurium (21)Typhimurium (20), No Match (1)Abortusovis (15)100One band difference, four different phage types (135, 135a, 6 var, 9).
Virchow (8)Virchow (4), subsp 1 ser rough:r:1,2 (2), No Match (2)Virchow (2), subsp 1 ser rough:r:1,2 (3)75Virchow ‘smooth’ isolate
Total 155    

Two isolates gave discrepant traditional serotyping and rep-PCR results. One of these outliers was Salm. subsp 1 ser 4,12:-:-, which was not included in the library but was incorrectly matched as Salm. Typhimurium (97·5% similarity). The other isolate was one of the Salm. Sofia isolates typed (= 66), which was matched as Salm. Typhimurium/Abortusovis using the DiversiLab® System rep-PCR fingerprinting kit. The DiversiLab® System result concurred with the in-house Typhimurium result using anti-Typhimurium antibodies. In this instance, the reference laboratory may have serotyped the incorrect isolate, which was reported as Salm. Sofia.

Of the ten isolates that returned No Match to the library (similarity value threshold <95%), nine matched with the library, though with similarity values marginally below the threshold. The other isolate belonged to a serotype (Salm. Kiambu [B]) included in the original library but the isolate was incorrectly serogrouped as group E and, therefore, was not queried against the correct database.

Financial comparison

The time and motion analysis was broken up into three stages: (i) set-up, (ii) reading and (iii) reporting. The time measured for these stages showed that for set-up, molecular typing required 19 min as compared to 10 min for serotyping, which is a 90% increase in labour. However, when the interpretation of the results was measured it only required 2 min to Top Match the ten isolates, whereas for serotyping the time was 20 min. Thus, it takes 10 times longer to interpret the results by the standard serotyping compared with the DiversiLab® System. The reporting of the results was also quicker with the DiversiLab® System (5 min) as compared to 10 min for serotyping, a 50% reduction in time. When the total labour component was compared, the traditional serotyping method required a total of 40 min (4 min/sample) as compared to the NucliSENS® EasyMAG plus DiversiLab® System of 21 min (2·1 min/sample) a difference of 47·5%. The percentage saving in labour costs could be offset by a larger material cost (Australian Dollars) between the two techniques. The material cost of rep-PCR was 215% (2·15 times) more expensive than the traditional serotyping using commercially available antibodies for Salm. Typhimurium. Overall, both methods had the same cost (Australian Dollars).


Currently, in Australia, it can take up to 4 weeks for a Salmonella isolate from poultry to be serotyped by a reference laboratory. In urgent cases, where notifiable serotypes such as Salm. Enteritidis (Arzey 2005) are suspected, a serotyping result can be obtained in 5 days. Given the short rearing period of broiler flocks (approximately 42 days) and short shelf-life of raw chicken meat, this can be problematic for the poultry industry. While serotyping kits are available to identify suspect Salm. Enteritidis, the DiversiLab® System offers a rapid approach to verify in-house serotyping of suspects with high confidence and allows more time to implement corrective actions.

In this study, an automated nucleic acid extraction platform (NucliSENS® EasyMAG®), which results in a clean and consistent nucleic acid extract, was used as an alternative to the manufacturer’s recommended manual DNA extraction protocol (UltraClean™ Microbial DNA Isolation Kit (Mo Bio Laboratories, Solana Beach, CA, USA). The NucliSENS® EasyMAG® was used because of inconsistencies in DNA extraction and subsequent band intensity observed in fingerprints generated in preliminary studies (Pavic and Bailey 2009). Purity of the DNA was not checked for each isolate in the present study; though, preliminary investigations comparing the NucliSENS® EasyMAG® and the manual extraction protocol showed the DNA to be of high purity (A260/A280 values of 1·7–1·9; unpublished data). The protocol for the NucliSENS® EasyMAG® requires few handling steps; therefore, the potential for (cross-) contamination is greatly reduced, when compared to manual extraction procedures. The other advantage of using the automated platform for nucleic acid extraction is that the technician is available to perform other tasks in the laboratory. One final modification to the manufacturer’s protocol was the initial cell density, which was accurately measured using a McFarland Standard No. 4 of a third-generation subculture, rather than following the manufacturer’s suggestion of obtaining a ‘loopful of inoculum’. This modification was made to standardize the final intensity of the DNA by controlling initial cell density and the DNA extraction process (Loens et al. 2007; Dundas et al. 2008).

An internal library consisting of four separate databases based on serogroup (Group B, Group C, Group E and Other) was constructed as an adjunct to the Salmonella DiversiLab® System reference library housed in the USA (Healy et al. 2005; Wise et al. 2009). In this study, the creation and extension of internal databases allows for the addition of local isolates. Separation of the library fingerprints into respective serogroups allows for greater discrimination and a possible reduction in false matches as reported by Wise et al. (2009). In regard to human illness, the serotypes archived in the library account for over 57% of poultry serotypes that are also isolated from human clinical samples (Davos 2009). Serotype Sofia is predominant, accounting for at least a third of the Salmonella strains isolated annually from Australian chickens and chicken products, but with a low incidence in humans (<0·2–0·3%) (Harrington et al. 1991; Davos 2009; Gan et al. 2010; Mellor et al. 2010).

The template DNA concentration was not carefully controlled, and this may affect band intensity rather than the presence or absence of bands in the profile (Healy et al. 2005). The Kullback–Leibler distance correlation coefficient was selected when comparing similarity because this statistical approach weighs band presence/absence more heavily than band intensity; this negates any potential changes in banding because of differences in template DNA concentration and offers greater differentiation between the fingerprints of different serotypes.

Although there was limited variability in the fingerprint profiles for each serotype (Fig. 1), some serotypes exhibited more extensive strain-level discrimination. For example, Salm. Infantis had five discrete fingerprints. Divergent genetic lineages and heterogeneity among Salm. Infantis isolates has also been demonstrated using multilocus enzyme electrophoresis (Beltran et al. 1988), highlighting allelic variation in conserved chromosomal structural genes.

Wise et al. (2009), using the DiversiLab® System as an alternative to serotyping Salmonella (n = 44), were unable to distinguish between Hadar and Istanbul, and Typhimurium and Infantis with a consolidated library. In the current study, the initial serogrouping of the isolates and classification using a library based on serogroup avoids mismatching of Typhimurium and Infantis as these serotypes are from serogroup B and serogroup C, respectively. However, Hadar and Istanbul would likely remain indistinguishable because they both belong to serogroup C, although neither of these serotypes were isolated in this study.

As mentioned previously, ten cases were observed where isolates of different serotypes showed fingerprints that were ≥95% similar when using the four different libraries based on serogroup, but most of these were owing to closely related variant and rough strains. Rough strains may lack phenotype but probably represent the same genotypes as the respective parent strains; therefore, similar fingerprints would be expected. However, variant serotypes, which are distinguishable by antibody-based serotyping as a result of lysogenization by specific phage factors (Grimont and Weill 2007), have inserts, so different fingerprint patterns may be expected. The instance where variant strains of serotypes (e.g. Anatum/Anatum var 15+, 34+; Fig. 4) were not differentiated could be a limitation of the DiversiLab® System. Although it is widely accepted that the position of the converted variants is expressed after conversion by lysogenic phage (e.g. phage ɛ15 + ɛ34) (Grimont and Weill 2007), one explanation for the high degree of similarity between the electropherograms could be that the variant strains are the same at a genetic level but the genes encoding the variant factors are not expressed or are silent in the parent strain.

Salmonella isolates that do not express their O antigens are termed ‘rough’ and this morphological change may be due to low osmolarity, nutrient limitation and temperatures below 30°C, which are commonly encountered in the environment (e.g. processing plant and wood shavings) (Gerstel and Romling 2001). In this study, subsp 1 ser rough:r:1,5 is likely the rough strain of Salm. Infantis (6,7,14:r:1,5), and subsp 1 ser rough:r:1,2 is likely the rough strain of Salm. Virchow (6,7,14:r:1,2). The similarity of fingerprints suggests the ‘rough’ isolates are derived from the ‘smooth’ isolates. This hypothesis was considered with respect to Salm. ser. rough:r:1,5 and Salm. Infantis, which have identical flagellar antigens (Grimont and Weill 2007), suggesting these isolates are variants of that serotype. According to the complete White–Kauffman–Le Minor scheme there are 11 other serotypes that have the same flagellar structure: Bradford (serogroup [B]); Czernyring [O54]; Abertbanju [V]; Lubumbashi [S]; Hindmarsh [C2-C3]; Linde [P]; Jamaica [D1]; Ughelli [E1]; Senegal [F]; Tennenlohe [K] and Gege [N] (Grimont and Weill 2007). However, none of these serotypes have been isolated in Australia (Davos 2009), supporting the initial hypothesis that the rough isolate is a variant of Salm. Infantis.

The failure to adequately differentiate Salm. Typhimurium from Salm. Abortusovis and Salm. Saintpaul (all group O:4 [B]) using the Kullback–Leibler statistical function suggested that these serotypes are very similar; though, closer examination of the electropherograms showed peak differences (Fig. 3). For determination of S. Typhimurium and S. Abortusovis serotypes, the serotyping result of the unknown isolates matched the DiversiLab® System result with the higher similarity value from the Top Match. This was owing to the presence of an extra band or fluorescent peak in the Typhimurium profile compared with the Abortusovis electropherogram (Fig. 3a). Salmonella Typhimurium (1,4,[5],12:i:1,2) and Salm. Abortusovis (1,4,12:c:1,6) have similar O antigenic structures (H antigenic formulae differ) (Grimont and Weill 2007), and they might share a high degree of DNA sequence homology (Popoff et al. 1984), which may be reflected in their rep-PCR fingerprint patterns. Furthermore, if the size and position in the genome of the genes responsible for antigenic differences are similar, with respect to the priming site for rep-PCR, then banding patterns could be very similar. Salm. Chester, Salm. Reading and Salm. Saintpaul all showed a high degree of similarity but these serotypes are less frequently isolated from Australian poultry and can also be differentiated by differences in the electropherogram overlays.

Of the two cases where the DiversiLab® System result did not concord with serotyping, one of the serotypes, Salm. subsp 1 ser 4,12:-:-, was not included in the library and the other, Salm. Sofia, incorrectly matched as Salm. Typhimurium/Abortusovis but also reacted with the anti-Typhimurium antibodies during initial serogrouping (laboratory suspect Salm. Typhimurium). This result may indicate that there may have been a serotyping error at the reference laboratory. Salm. subsp 1 ser 4,12:-:- was not included in the library but was incorrectly matched at 97·5% similarity to Salm. Typhimurium. The cell surface antigenic formula of Salm. subsp 1 ser 4,12:-:- and Salm. Typhimurium (1,4,[5],12:i:1,2) are similar and, because the somatic antigen of any given Salmonella serotype is an expression of its genomic content (Wise et al. 2009), it is not surprising that the rep-PCR fingerprints between these two serotypes are also similar.

The financial analysis did not account for capital cost, yearly service fees and depreciation, which often differ between countries and are dependent upon the usage of the equipment. Although the capital costs associated with the rep-PCR technology are substantial, financial analysis showed that predicting serotype ‘in-house’ reduces real or direct costs such as freight and increases in efficiencies such as a significant reduction in time to result. The initial capital cost is a major concern for any laboratory; however, the DiversiLab® System platform allows analysis of a range of micro-organisms that includes (but is not limited to) Escherichia, Campylobacter, Listeria and Enterococcus. In addition, the microfluidics chip can be used to view any PCR product and the software used to archive results. There was a 47·5% reduction in labour when ten Salm. Typhimurium isolates were tested simultaneously. Despite a 215% increase in material costs, these are more than offset by labour savings, and ongoing material costs may be decreased based upon volume purchased.

The advantages of the protocol are the ease of DNA extraction using the NucliSENS® EasyMAG® platform, standardized easy-to-use kits on an open-ended system, and user-friendly software that offers inter-laboratory comparisons through the web-based master library. One of the disadvantages of this system is the manual syringe priming station, which is used to load the chip with the gel–dye mix. Occasionally, improper use of the syringe will result in air bubbles during loading and any loading errors are not identified until chip analysis. This problem would be resolved by the implementation of an automated injection system, similar to those used in high-performance liquid chromatography machines. Another drawback of the DiversiLab® System is the run size limit of 13 chip spaces.

Using Salmonella databases distinguished by serogroup, the semi-automated rep-PCR-based method, evaluated for putatively assigning serotypes of Salmonella, correctly classified 143 of 155 (92%) unknowns. The internal libraries are continuously updated with new serotypes and the DiversiLab® System is currently applied in-house as a screening method to limit the number of confirmed positive Salmonella isolates sent to reference laboratories for serotyping. In conclusion, this study has shown that rep-PCR can be utilized to reliably and rapidly predict Salmonella serotype as an adjunct or substitute for standard serological typing.


The authors thank The Australian Salmonella Reference Centre, Institute of Medical and Veterinary Science, Adelaide, Australia for the serotyping of Salmonella isolates. Gratitude is also expressed to Gavin Bailey (bioMérieux Australia), Wylie Armstrong, Taha Harris, Jarrod Tinker and Sheren To (all from Birling Avian Laboratories) for their technical assistance. This project was funded solely by an Australian Poultry company.