Clostridium difficile infection (CDI) is a major nosocomial disease, placing a considerable burden on healthcare resources. A recent European report estimated the associated per-case cost of CDI at £4577–£8843 [1]. C. difficile has commonly been associated with hospital outbreaks, and most notably the epidemic spread of NAP1/BI/PCR ribotype 027 [2, 3]. Studies of C. difficile epidemiology from diverse geographical locations have described different locally or nationally prevalent strains [4, 5, 6], some of which were associated with increased severity and/or reduced antimicrobial susceptibility [2, 5]. DNA typing systems have been crucial in tracking the emergence and spread of C. difficile strains. Notably, individual surveillance programs have seen significant changes in the prevalence and epidemiology of NAP1/BI/PCR ribotype 027 [7, 8]. This highlights the important role that DNA typing systems play in epidemiological surveillance and in targeting informing infection control interventions.

Most epidemiological studies of C. difficile are local or national in scope, and as such have used different typing techniques. Pulsed-field gel electrophoresis (PFGE) [9], restriction enzyme analysis (REA) [8], PCR-Ribotyping [7], toxinotyping [10], arbitrary-primed PCR (AP-PCR) [11], random amplification of polymorphic DNA (RAPD) [12] and REP-PCR [13] have all been used routinely for C. difficile typing; thus, it is not uncommon for isolates to be referred to by multiple typing designations (eg: NAP1/BI/027; PFGE/REA/PCR-ribotyping). This can be problematic when trying to compare studies or conduct large scale epidemiological analyses. A multi-centre, international study assessed seven typing techniques, and drew attention to the lack of a consensus technique with proven inter-laboratory reproducibility [14]. PCR ribotyping has gained acceptance in Europe as the typing method of choice, with a common nomenclature adopted across most countries (although some local systems have emerged) and different primers can be employed [15, 16, 17]. Agarose gel-based methods offer relatively high sensitivity, specificity and reproducibility relative to other PCR-based techniques, but inter-laboratory variation in banding patterns and interpretation mean that data are difficult to compare between typing centres [14].

The advent of high-resolution, capillary gel electrophoresis-based fragment analysis addressed many of these issues. Reduced PCR cycling and electrophoresis times combined with the ability to increase batch sizes (using multiple fluorescent detection dyes) represents a relatively rapid, high-throughput and low cost means of epidemiological C. difficile investigation [18, 19, 20]. However, CE-ribotyping has yet to gain widespread acceptance, and is hampered by a lack of multi-centre evaluations demonstrating its reliability and the portability of data. Indra et al. proposed and established a web-based repository of electronic C. difficile PCR ribotyping data (WebRibo) [18]. Unfortunately, the lack of protocol standardization and consequent limited capacity to accurately categorize submitted patterns using internationally recognised PCR-ribotype nomenclature has to date diminished the effectiveness of this approach [21]. We describe here the first multi-centre development, validation and evaluation of a standardised protocol for high-resolution PCR-ribotyping using capillary gel electrophoresis.

Increased awareness of CDI is accompanied by the need to develop a consensus typing and analysis method that will allow the establishment of a standardised nomenclature and greater understanding of C. difficile epidemiology. This should be robust, reliable and generate high resolution, portable data, and function at local, national and international levels. Existing reports involving CE-ribotyping have represented only single-centre studies and inter-laboratory variability has not been formally assessed [18, 20]. We report a high level of reproducibility and accuracy associated with a consensus CE-ribotyping protocol. A maximum SD of only ±3.8 bp was recorded in individual fragment sizes, and 98.2% blinded ribotypes were correctly identified at four international ribotyping centres spanning Europe and North America (rising to 98.8% after analysing the discrepancies). Our results demonstrate that this consensus protocol generates sufficiently low level inter-laboratory variation to support accurate PCR-ribotyping for international C. difficile surveillance.

Indra et al. first combined conventional PCR-ribotyping and capillary based electrophoresis techniques (CE-ribotyping) and demonstrated superiority of this approach over the conventional agarose gel-based method [18]. More recently, Xiao et al. reported that the performance of capillary-electrophoresis exceeded that of a commercial high performance electrophoresis system when applied to PCR-ribotyping. [20]. Several epidemiological studies have been conducted using subtly different CE-ribotyping protocols, making future inter-laboratory comparison of results difficult [18, 20, 24, 25]. We hope that the provision of a consensus CE-ribotyping protocol will encourage laboratories to use the technique in the knowledge that data will be directly comparable across centres.

We have shown that a basic crude DNA extraction method is sufficient to demonstrate good reproducibility of the system. However, many laboratories have more sophisticated higher-throughput DNA extraction platforms, which invariably yield good quality DNA, and these could be expected to be viable alternatives. Several different sets of PCR primers have been reported for use in PCR-ribotyping in recent years [15, 17, 26]. The primers originally described by Stubbs et al. unfortunately had a 4 bp mismatch in the 23S reverse primer [15] and were subsequently redesigned [17]. It is generally accepted that these primer sets offer equivalent levels of performance and discrimination, but generate different relative fragment lengths. Bidet et al. primers generate fragments ~34 bp shorter than those designed by Stubbs et al., and ~53 bp shorter than those designed by Sadeghifard et al. The loss of accuracy associated with sizing larger DNA fragments when using electrophoresis should ideally be minimised, and the selection of primers that generate smaller fragments would therefore be preferable. In order to avoid potential primer:template mismatches and minimise inter-laboratory variation, the primers designed by Bidet et al., were selected for use in this consensus method [17]. It is important that standardized protocols for multi-centre surveillance can be easily adopted by participating laboratories. PCR consumables can be acquired or changed relatively easily, but access to specific DNA sequencing instrumentation can present difficulties. It is therefore essential that a standardised protocol can operate on more than one instrument type without significant variation in ribotype profiles. Our study has incorporated the use of two different sequencing instruments with little effect on PCR-ribotype recognition or fragment-size variation. These results suggest that laboratories can adopt the consensus protocol for use on different automated sequencers; further validation is required to confirm this issue.

Previous reports using CE-ribotyping have called for validation of the technique using internationally accepted reference strains [18, 20]. The standardised consensus method used here was tested on a well characterised collection of 70 different PCR-ribotypes [22, 23] to evaluate the performance of the protocol against prevalent and clinically relevant PCR-ribotypes. Data on this set of strains are available on-line in a NCBI BioProject database to facilitate working in this field (http://www.ncbi.nlm.nih.gov/bioproject/248340). Several examples of profiles that shared a very high level of similarity were incorporated in this study (eg. 027 and 081; 015 and 046; 106 and 174; 014 and 020) (Fig. 2). CE ribotyping was able to consistently discriminate between all these isolates. It is generally accepted that such discrimination is difficult with conventional agarose gel-based PCR-ribotyping unless gel quality is particularly high. The number of unique PCR-ribotypes has expanded rapidly in recent years and now exceeds 650 (Fawley, Wilcox, unpublished). As this library expands, the number of ribotypes with similar profile is likely to increase. In turn, the interpretation of PCR-ribotype profiles will become more complex. CE-ribotyping offers a solution for future ribotyping strategies by offering accurate fragment sizing and increased discrimination between ribotypes, including those with similar profiles that can be confused using conventional agarose gel-based PCR-ribotyping. CE-ribotyping has already confirmed reproducible differences in DNA profile of isolates previously assigned to the same ribotype using agarose gel electrophoresis, most notably those associated with ribotypes 014, 001 and 027 [18, 20, 27]. Multi-locus sequence typing studies have further shown that individual ribotypes tend to be associated with a single ST-type, but that a small number are associated with multiple ST-types [22, 27, 28].

CE-ribotyping represents a cost-effective technique to further discriminate between ribotypes of very similar DNA profile in order to further understand phylogenetic relationships. The consumables price per test for PCR-ribotyping by CE-method and by conventional agarose gel-based method are comparable (under EUR 10) [19]. Current opinion would indicate that the largest barrier to widespread adoption of CE-ribotyping remains the acquisition of an automated sequencing instrument. However, such instruments are now in widespread use; most medical and academic institutions use automated sequencing technology and many offer access to their instruments (on or off site) on a service provision basis. In this way CE-ribotyping could reasonably be performed in any laboratory and need not be reserved for larger reference centres.

Our results suggest that data generated using the consensus method are fully portable; 98.8% (rising to 100% after analysing the discrepancies) of ribotype profiles were correctly identified when data were exchanged electronically between participating laboratories. Hitherto, surveillance programs have been impeded due to the inability of laboratories to exchange typing data efficiently. Notably, understanding the spread of NAP1/BI/PCR ribotype 027 strains across N. America and Europe was affected by inefficient inter-laboratory exchange of PCR-ribotyping data, involving the laborious exchange of bacterial isolates to establish significant epidemiological links [2, 6]. We have demonstrated that a standardised CE-ribotyping protocol allows easy electronic exchange of typing data, which could streamline communication and so reduce the time for countries/regions to become aware of new and emerging ribotypes.

There have been calls for improvements to PCR-ribotype nomenclature [19, 29]. Stubbs et al. first assembled a library of PCR-ribotypes [16] but, while this collection has expanded to >650 types, lack of direct access to this library has led some to use local ribotype nomenclatures. These have only local immediate applicability. The UK library was developed using agarose gel-based ribotyping, and further development is hindered by interpretation issues associated with this technique. Standardised and fully portable ribotyping data would be a significant step forwards to consolidating a single library of C. difficile reference strains, which would bring clarity for researchers and those involved in surveillance of this important pathogen.