Introduction

Over the last decades, the numbers of prosthetic joint replacements have continuously increased and are forecast to increase even further with the ageing population [1]. This in turn correlates with a higher number of associated prosthetic joint infections (PJIs). Implant-associated infections have significant consequences on morbidity and mortality and place a high economic burden on the healthcare system [2].

For focused therapy of PIJ, identification of pathogens from synovial fluid is highly relevant. Consequently, aspiration of synovial fluid with microbiological analysis is recommended by the AAOS guide-lines for joint infections [3]. However, several challenges exist in the diagnostic procedures of PJI. Most importantly, samples in association with PJI or bone infections after fracture fixation generally have low bacterial densities with concomitant cellular debris, so that direct bacterial identification out of these samples is difficult [4]. Enrichment steps, e.g. in liquid medium and prolonged incubation (up to 2 weeks), increase sensitivity and specificity [5,6]. They allow observance of slow growing small colony variants [7] or fastidious bacteria, and lead to improved diagnosis [8]. Any possible reduction in processing time will likely have a positive effect on treatment.

For microbes cultured on solid media, Matrix-Assisted Laser Desorption/Ionization Time-Of-Flight Mass Spectrometry (MALDI-TOF MS) has become the gold standard for species identification. This technology has brought significant changes in processing of microbiological probes and replaced most cost- and time intensive phenotypic biochemical assays. For identifying microbial species out of liquid culture, MALDI-TOF is already regularly used for blood specimen (reviewed in [9,10]) and is highly standardized during routine diagnostics through commercialized assays such as the Sepsityper workflow [11]. These procedures circumvent the need for an additional overnight cultivation step. A first study has demonstrated the feasibility of such assays also for orthopedic samples using the BACTEC system [12].

Here, we examined whether BacT/Alert blood culture vials inoculated with synovial fluid were suitable for the Sepsityper workflow, if this yielded microbial identification concordant with the standard culture procedure, and if different levels of host residue influence these results.

Results and discussion

Several experimental factors influence MALDI-TOF mass spectrum quality when bacteria are analyzed from clinical samples. Residual host material ionizes along with microbial markers. If the host material is in excess, this can lead to spectra representing the host rather than the microbe. Consequently, a main goal of sample pre-processing is eliminating host material, and enriching for microbial cells. A possible option is amplification through liquid culture. This may however introduce several components into the sample interfering with the MALDI-TOF process [13]. In older blood culture systems, charcoal and salt ions from the culture media were shown to interfere with the results [14]. Even after successful amplification, liquid cultures will still contain remnants of host materials, such as blood or tissue cells, soluble proteins, or mucus. These materials generate background noise mass peaks, overlapping with microbial spectra and interfere with the interpretation of bacterial proteome profiles [15]. Therefore, amplicon pretreatment removing host cells and proteins, while further concentrating the microbes, improves spectrum quality.

To test if the residues contained in synovial inoculum had an influence of species identification, we used a total of 23 native synovial samples (Table 1) over a course of 5 weeks, which contained a visible degree of host residue (blood, pus), or were highly viscous. The samples had been analyzed by culture on solid media and were additionally inoculated into pediatric BacT/Alert blood culture bottles. All bottles automatically flagged positive by the next morning (~16h). Using the Sepsityper workflow, 21/23 (91%) of the samples gave concordant results with culture on solid media. Out of these, only two missed the manufacturer-recommended threshold of 2.000 for species-level identifications. Acceptance of lower score values down to 1.700 when using the Sepsityper protocol, however, appear feasible [9]. The two remaining samples (9%) were from mixed-species specimens. In one case (#11) the initial mixed result could not be confirmed by reanalysis, and the Sepsityper identification was concordant with this second culture result (Table 1). Most importantly, there was no apparent hindrance through blood, pus, or high viscosity present in the inoculum.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Results of spiked culture Sepsityper identification.

https://doi.org/10.1371/journal.pone.0243790.t001

In order to test if this procedure also worked outside artificial laboratory conditions, we prospectively applied it on positively flagged specimen from our routine diagnostics. There, specimen had been inoculated on ward by the treating physician, and the specific condition of the inoculum was unknown (Fig 1). A total of 468 blood culture bottles inoculated with synovia from 355 clinical specimens were received during the study period, out of which 87 were flagged positive. Fifty bottles were excluded because they were not made immediately available for the study, mainly because they flagged positive during the night during the night leading to unrepresentatively high microbial yield, leaving 37 bottles to be tested by the Sepsityper workflow.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Prospective study scheme.

https://doi.org/10.1371/journal.pone.0243790.g001

Pathogen identification concordant with results from isolates cultured on solid agar was achieved in 30/37 (81%) samples at the first try, however 7/37 (19%) again not at levels required for species-level identifications. One sample gave discordant species-level matches for the bottle (Streptococcus mitis) and solid agar (Streptococcus pneumoniae). Reliable distinction of S. pneumoniae from viridans group streptococci is important because of the different pathogenic properties of these organisms, and this has also previously been achieved by MALDI-TOF MS by others [16].

Five out of 37 samples (14%) contained mixed-species cultures of either Staphylococcus aureus combined with different enterobacteria, or Pseudomonas aeruginosa combined with Entercoccus faecieum (4 samples of the same patient). Mixed cultures have previously been shown to be problematic, where the ration of the organisms is unequal [17,18]. Among our samples, only in one the result indicated a mixed culture with both species, in the other four only a single species was found with high score values, however, plating of the culture indicated no more than 4-fold differences in cfu between each of the two species present. In the six remaining samples, identification was not achieved due to the complete lack of mass signals, potentially attributable to the loss of the bacterial pellet or too low cell counts in the culture (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Results of prospective Sepsityper identification.

https://doi.org/10.1371/journal.pone.0243790.t002

Conclusion

In summary, the Sepsityper workflow also worked well on mono-species samples derived from BacT/Alert blood culture bottles inoculated with synovia. The major contributor to failed detection was the absence of mass signals. The level of concordant identifications in both our specimen groups was comparable to those achieved by others for synovia [12], or blood cultures (summarized in [9,10]). In the case of mixed cultures, the test was not able to reliably detect the mixture, despite apparent similar cfu numbers. Host remnant material in the inoculum, such as blood or pus, had no detrimental effect on identification score values of the MALDI BioTyper system, and neither had the initial viscosity of the synovial sample.

Materials and methods

Acknowledgments

The authors acknowledge contributions of clinical and routine diagnostic laboratory staff for acquisition and processing of diagnostic specimen. Specifically, the authors thank Sabine Ceramella and Yvonne Laukat for expert technical assistance.