A rapid, inexpensive and easy handling method for identification of mastitis pathogens is an essential task in veterinary routine diagnosis and a prerequisite for targeted (antibiotic) treatment and efficient prophylaxes.
Especially, the identification of gram-positive, catalase-negative cocci at species level, in particular, still represents a major challenge. Wyder et al.  and McDonald et al.  clearly showed that an unequivocal identification of Streptococcus spp. and related mastitis associated cocci, especially minor pathogens such as L. garvieae, is difficult by methods currently used in routine diagnosis. We therefore assessed the potential of FTIR spectroscopy and MALDI-TOF MS for identification and differentiation of mastitis relevant gram-positive, catalase-negative cocci. FTIR spectroscopy was already successfully applied for the classification of human pathogenic enterococci and food related lactic acid bacteria ([31, 32]) and MALDI TOF MS is becoming more and more prominent in human clinical diagnosis [20, 22]. The strain panel generated for this study comprised a total of 383 isolates, belonging to 21 species and four genera. In addition to the most important Streptococcus species, other mastitis-associated gram-positive, catalase-negative cocci that occur less frequently but have an intramammary origin, such as L. garvieae or S. parauberis, were included (Table 1).
Since many of these species are phylogenetically closely related, their differentiation and identification is often challenging. For this reason, a spectral reference database and FTIR based ANNs were established and combined to ensure fast and accurate bacterial identification at species level. The reference database, which averages all spectral differences into one distance measure (D-value), is generally a very robust system but it does not provide as much flexibility as ANN . Since the selection of appropriate spectral regions for identification of microbes is known to be species specific [12, 15], care was taken to select spectral regions providing the highest discriminatory power for classification and identification of Streptococci and related species. The spectral regions between 1.500 to 1.200 and 1.200 to 900 cm-1, covering the mixed region and the polysaccharide region, were found to be best suited for the identification and discrimination of gram-positive, catalase-negative cocci.
The ANN detects small differences in biodiversity patterns due to the selection of single wavenumbers in contrast to whole spectral regions as chosen for the library, which explains its higher discriminatory power and robustness compared to the spectral library based identification system. The selection of specific wavenumbers by the ANN ensures that only the most significant information is extracted and put into the training process. The complete net consisted of single nets that are highly discriminatory for the defined classes on the basis of the most discriminative wavenumbers . During the training process, only significant information is extracted, trained and the differences between the spectra used for calibration are stored for achieving high identification accuracy . The final net established in this study was highly discriminatory and provided highly specific and reliable results. For instance, during the external validation, using strains unknown to the ANN, an overall identification correctness of 98.5% at species level was achieved. Although most of the species could be easily separated in three steps (levels) the discrimination of S. dysgalactiae, S. parauberis, S. pyogenes and S. uberis required two additional levels. This might be due to similar spectral patterns that are used for species identification, particularly noticeable for S. uberis and S. dysgalactiae. By combining the ANN with the reference database 100% correct identification was achieved, underpinning the specificity and robustness of FTIR-based identification system developed in this study. It is assumed that FTIR spectroscopy combined with chemometrics could significantly improve the diagnosis of mastitis-associated streptococci and related cocci. Indeed, results from a blind study, including isolates from quarter milk samples sent to routine mastitis diagnostic labs in Austria, confirmed the principal suitability of the established FTIR-systems for routine diagnosis (Table 5).
Another particular advantage of FTIR is its high discriminatory power, allowing discrimination of bacterial subtypes and its use for epidemiological studies and determination of contamination sources [9, 13, 17]. The latter applications could be very important concerning therapy control, prophylaxes and detection of source of infection .
As a second physico-chemical method we included MALDI-TOF MS in this study since it is already established in clinical microbiology but is still rarely used for routine diagnosis in veterinary medicine. For instance, Barreiro et al.  showed that MALDI-TOF is, in principle, suitable for identification of subclinical mastitis pathogens. We therefore used a subset of strains (n = 210) derived from the strain panel, generated for the development of the FTIR identification system, to test the performance of MALDI-TOF MS. Unequivocal results (log(score) values > 2.3) were achieved for 95.2% of the strains and about 2.4% of the strains showed log(score) values between 2.0 and 2.3. Distinguishing S. dysgalactiae from S. canis and S. bovis from S. bovis/equinus complex turned out to be difficult by MALDI-TOF MS but not by FTIR spectroscopy. For instance, one S. dysgalactiae and one S. bovis were even misidentified by MALDI-TOF MS (Table 4). These results might be explained by the close phylogenetic relationship of the aforementioned species to S. pyogenes and S. equinus, respectively [34, 35]. Moreover it must be taken into consideration that the taxonomy of the S. bovis/equinus complex is very complex and has undergone several changes over time, resulting in the reassignment of S. bovis biotype I to S. gallolyticus (for overview see ). Furthermore, the MALDI BioTyper database (V.3) only includes reference MS spectra from six S. lutentiensis, two Streptococus equinus and one Streptococcus infantarius strains but no reference MS spectra from S. bovis strains. Limited resolution of MALDI-TOF MS for closely related Streptococcus spp. was also reported from a recent study of Raemy et al. . Some isolates, especially those belonging to S. dysgalactiae, could only be identified at genus level and isolates belonging to the S. mitis/oralis/pseudopneumoniae group could not be further discriminated by using the SARAMIS software (see , Figure 2). Identification and discrimination of bacterial species by MALDI-TOF MS is mainly based on ribosomal proteins whereas FTIR spectroscopy covered the entire biochemical composition of a cell explaining the higher discriminatory power of the latter technique. In the study presented, the carbohydrate region contributed significantly to the successful identification of streptococci by FTIR. Nevertheless, both methods tested are suitable for the identification of gram-positive, catalase-negative cocci. In contrast to FTIR, MALDI-TOF MS does not require specific culture conditions, which makes it easier to implement it in routine diagnosis, although at the cost of resolution power. The results of the blind study strengthen the assumption that the methods presented in this work are interesting novel tools for routine diagnosis. Notably, A. viridans, L. garvieae and L. lactis represented about 20% of the blind study isolates, suggesting that these so called “minor pathogens” occur more frequently than previously thought. These results are in line with recent findings of Wyder et al. , who also reported similar frequencies of A. viridans and L. garvieae in quarter milk samples from Swiss dairy cows. In order to avoid misclassification and to get validated data on the prevalence of the ‘major mastitis causing’ Streptococcus spp. as well as on other gram-positive, catalase-negative cocci associated with bovine mastitis, reliable and robust identification techniques are of utmost importance and urgent need.