Introduction

The genus Raoultella, belonging to the Enterobacter-Escherichia clade of the order “Enterobacteriales” [1], includes Gram-negative, oxidase-negative, capsulated, and nonmotile facultatively anaerobic bacilli. Raoultella was recognized as a genus separated from Klebsiella in 2001 by Drancourt et al. [2] on the basis of the sequences of the 16S rRNA and rpoB genes. Raoultella spp. are emerging opportunistic pathogens which are being increasingly associated with infections of different locations in human patients, particularly in the gall bladder and pancreas [3, 4], the urinary tract [5,6,7], or infected wounds [8, 9].

The genus Raoultella includes four species: Raoultella ornithinolytica, Raoultella planticola, Raoultella terrigena, and Raoultella electrica [10]. R. ornithinolytica is considered to be more virulent than the other species, and it is associated with a higher attribute mortality rate [11, 12].

Identification of Raoultella isolates to species level is difficult because of the very close genetic relationship of the four species, and there is no biochemical or phenotypic profile that easily differentiates them from species of the genus Klebsiella, except for the use of L-sorbose and 3-0-methyl-D-glucose as carbon sources and ornithine decarboxylase (ODC) activity, which is always positive for R. ornithinolytica and variably positive for R. terrigena [2, 13, 14]. It has been reported that matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) shows high accuracy for Raoultella identification to the genus level, although there is no agreement on this regard [13, 15]. In the present study, we carried out a comparative evaluation of the performance of the two more extensively used MALDI-TOF MS platforms (Vitek MS and Bruker Biotyper) with Vitek2 and API20E systems, for identification at the species level of 97 clinical isolates originally phenotypically identified as Raoultella spp. by Vitek2, taking identification by molecular methods as the gold standard.

Materials and methods

Ninety-seven isolates, from clinical samples, of Raoultella spp. (63 R. ornithinolytica, 30 R. planticola, and 4 R. terrigena), were included in this study. First, isolates were identified with Vitek MS, in parallel with Vitek2 and API, and finally with Bruker Biotyper. By first using MALDI TOF, a more sensitive a priori system, we try to minimize a selection bias as much as possible.

Identification using phenotypic methods

The organisms were identified using two different methods: the Vitek2 automated system with Vitek2 GN cards (bioMérieux, Marcy-I’Étoile, France) and the API20E system (bioMérieux).

MALDI-TOF MS identification

Identification was performed with both Vitek MS V2.0 Knowledge Base-Clinical Use (bioMérieux, Marcy-I’Étoile, France) and MALDI Biotyper (Bruker Daltonics GmbH, Bremen, Germany). Both libraries are for in vitro diagnostic (IVD) and not for research use only (RUO). The identifications are gold standard (molecular methods) based. To be called correctly to species, it should achieve a score > 2 in the case of Bruker and a score of 99.9% in the case of Vitek MS.

Identification using molecular methods

All the strains were subjected to PCR amplification of the Klebsiella oxytoca pehX gene to distinguish between this species and Raoultella [16]. In addition, the bla gene was amplified and digested with the NotI enzyme for identification at the species level of R. ornithinolytica and R. planticola, as described elsewhere [17]. Lastly, in those strains that did not amplify the bla gene, the housekeeping genes rpoB, gyrA, and parC were sequenced for the identification of R. terrigena [18].

Data were analyzed with SPSS 20.0 software (IBM Corp, Armonk, NY, USA). Two-by-two concordances were estimated by Cohen’s kappa coefficient of agreement (k) and concordance index (CI).

Results

The enzyme NofI digested the amplification of the bla gene in 63 isolates, identifying them as R. ornithinolytica while the 30 undigested ones were identified as R. planticola. In the four isolates in which the gene bla was not amplified, the rpoB, gyrA, and parC genes were sequenced and those four isolates were definitely identified as R. terrigena. Vitek MS, Bruker Biotyper, and Vitek2 assigned a correct genus identification to 96 (98.9%), 97 (100%), and 96 (98.9%) isolates, respectively. On the other hand, API galleries allowed genus identification for only 18 (18.5%) isolates. Vitek MS, Bruker Biotyper, Vitek2, and API globally assigned a correct species identification in 88 (90.7%), 80 (82.4%), 39 (40.2%), and 12 (12.3%) cases. Vitek MS correctly identified 62 (98.4%) R. ornithinolytica, 22 (73.3%) R. planticola, and 4 (100%) R. terrigena. Bruker Biotyper correctly identified 63 (100%) R. ornithinolytica, 17 (56.6%) R. planticola, and none of (0%) R. terrigena. Vitek2 correctly identified 10 (10.3%) R. ornithinolytica, 29 (96.6%) R. planticola, and none of (0%) R. terrigena, and it did not assign any identification in one isolate. Finally, API galleries correctly identified 10 (10.3%) R. ornithinolytica, 0 (0%) R. planticola, and 2 (50%) R. terrigena and identified 2 additional isolates as R. ornithinolytica when the strips were incubated for 48 h; despite these results, the API20E developer system indicated in many cases “possibility of Raoultella” (Table 1).

Table 1 Sensitivity, specificity, and predictive values of four methods for the identification of Raoultella at the species and genus level
Full size table

In comparison with molecular methods, the sensitivities and specificities of Vitek MS, Bruker Biotyper, Vitek2, and API20E results were 98.8/57.8% (95%CI 93.9–99.9%/33.5–79.7%), 98.7/37.0% (95%CI 93.3–99.9%/19.4–57.6%), 95.1/14.9% (95%CI 83.4–99.4%), and 13.0/62.5% (95%CI 6.9–21.6%/35.4–84.8%), respectively. We have also calculated kappa coefficient (ĸ) and concordance index (CI) between Vitek MS with molecular techniques and Bruker Biotyper with molecular techniques showing a result of ĸ 0.664 (95%CI 48.2–84.6%) with an appropriate degree of agreement for the first system and ĸ 0.446 (95%CI 28.2–61.0%) with a moderate degree of agreement for the second system (Table 2).

Table 2 Identification results for R. ornithinolytica (n = 63), R. planticola (n = 30), and R. terrigena (n = 4) by Vitek MS, Bruker Biotyper, Vitek®2, and API20E
Full size table

Of note the 10 (15.8%) R. ornithinolytica correctly identified by Vitek2 were the same isolates as those correctly identified by API20E in the first 24 h, and they were the only isolates showing ODC activity. Only in two additional strains the ODC was positivized by incubating the API20E gallery for 48 h. Malonate became positive by all R. ornithinolytica and R. planticola isolates. The 2-nitrophenyl-β-D-galactopyranoside (ONPG) test was positive in all R. ornithinolytica, R. planticola and R. terrigena isolates. The Voges-Proskauer reaction was positive in all strains of R. planticola and R. terrigena, and in 60 (95.2%) of R. ornithinolytica [19].

Discussion

Both Vitek2 and API20E only rely on the ODC to differentiate K. oxytoca (ODC negative) from R. ornithinolytica (ODC positive). Moreover, the Vitek2 system does not include R. terrigena in its database.

In our study, the two MALDI-TOF MS systems that have been evaluated are more efficient than the Vitek2 system or API20E galleries for a reliable identification of Raoultella at a genus level, so this proteomics-based identification is an appropriate alternative to the phenotypic tests. Although Vitek2 is almost as reliable as MALDI-TOF for identification of the genus Raoultella, with a sensitivity of 95.1% (but with the limited specificity of 14.9), it needs 24 h until results are available. Within the two MALDI-TOF systems studied, Vitek MS is slightly better than the Bruker Biotyper one, with a sensitivity and specificity of 98.8/57.8% from the first to 98.7/37.0% of the second. Finally, API galleries did not assign correct identification at species level of most clinical isolates of R. ornithinolytica and R. planticola and identified about 50% of the uncommon species R. terrigena, being aware of the limited number of strains studied.

The finding of low correlation between R. ornithinolytica isolates identified by both molecular methods and MALDI-TOF with the positivity of ODC is justified by the existence of ODC-negative R. ornithinolytica isolates, as has been outlined by Walckenaer et al. [17].

Vitek2 and API require additional biochemical tests to identify the ODC-negative R. ornithinolytica isolates. The current commercial phenotypic identification systems are not optimized for the identification of Raoultella species.

In summary, although more information is required about the reliability of MALDI-TOF MS for the identification of Raoultella species, it would be expected that more cases of infections by these microorganisms could be diagnosed with these identification approaches [20].