Bmc Infectious Diseases Accuracies of Leuconostoc Phenotypic Identification: a Comparison of Api Systems and Conventional Phenotypic Assays

Background: Commercial diagnostics are commonly used to identify gram-positive bacteria. Errors have been reported mostly at the species level. We have found certain phenotypic criteria used in API systems which significantly misidentify Leuconostoc, an emerging human pathogen, at the genus level. We also attempt to find practical, conventional phenotypic assays for accurate identification of this group of bacteria.


Background
Leuconostoc is a gram-positive coccoid or coccobacillary emerging human pathogen found in environment, foods and food products [1]. Risk factors of infection include antibiotic pressure, foreign device, or underlying immune defects. The organism is naturally highly resistant to vancomycin with MIC ≥ 256 µg/ml but could be successfully treated with penicillin with MIC ranging from 0.25 to 1.0 unit/ml. Commercial diagnostics are commonly used to identify gram-positive bacteria, with errors mostly at the species level [2,3]. Here we report inaccuracies of the Analytical Profile Index systems (API 20 STREP and API 50 CHL, Biomérieux, Inc., Lyon, France) in identifying Leuconostoc at the genus level. We also propose practical methods for clinical bacteriology laboratories to identify this organism.

Methods
Clinical isolates of catalase negative gram-positive coccoid or coccobacillary pairs and chains with αor γ-hemolysis in our institute during 1997-2004 were subject to an identification aid by API 20 STREP (bioMérieux, Inc., Lyon, France), following the instruction manual. Those identified as Leuconostoc by API 20 STREP were re-examined by the same kit and by API 50 CHL (bioMérieux, Inc., Lyon, France) according to the instruction manuals, by Leuconostoc conventional phenotypic assays, by Leuconostoc-and Lactobacillus-specific PCR's, and by 16S rDNA sequence analysis as previously described [4]. The 800-bp 16S rDNA fragment corresponds to Escherichia coli positions 10 to 806. The sequencing results were compared with those available in the GenBank, using BLASTN. Criteria for our conventional phenotypic assays for Leuconostoc are catalase-negative gram-positive coccoid or coccobacillary bacteria evaluated after growth in thioglycolate broth at 35°C for 24-48 hours [5], vancomycin MIC ≥ 256 µg/ml by Etest (AB BIODISK, Solna, Sweden), CO2 production from glucose in de Man, Sharp, Rogosa (MRS) broth (Difco, Detroit, MI, USA) with Durham tubes and negative pyrrolidonyl arylamidase (PYR), leucine arylamidase (LAP), and arginine dihydrolase (ADH) [6]. Leuconostoc-specific PCR was performed on all isolates as described [7], with slight primer modifications, as stated below. These modifications were to make the primer sequences most complementary and specific to Leuconostoc strains in GenBank. Forward and reverse primer sequences were 5'-CACAGCGAAAGGTGCTT-GCAC-3' and 5'-GATCCATCTCTAGGTGACGCC-3', respectively. To further assess accuracy of the API 20 STREP kit, additional catalase-negative gram-positive coccoid or coccobacillary isolates during 2005-2006 with vancomycin MIC ≥ 256 µg/ml were also evaluated by the same phenotypic and genotypic assays (isolates 31-38 in Table 1). Our gold standard for Leuconostoc identification is that the organisms fulfill both conventional phenotypic criteria and either or both of the genotypic assays (PCR and 16S rDNA sequence analysis). As we suspected that some isolates might have been Lactobacillus, also a lactic acid bacteria with overlapping phenotypes, Lactobacillusspecific PCR was also performed on all isolates as described [8]. PCR using universal primers targeting bacterial 16S rRNA conserved sequences was also performed to ensure template quality. The forward primer Y1 corresponds to positions 20 to 43 in the E. coli 16S rRNA sequence and the reverse primer Y2 corresponds to E. coli positions 361 to 338 [9]; this protocol gave positive results for all isolates in the study. Leuconostoc mesenteroides ATCC 8293, Pediococcus pentosaceus ATCC 33316, Lactobacillus pentosus ATCC 8041 and Lactobacillus plantarum ATCC 14917 served as controls for all assays. This study has been approved by The Institutional Review Board of The Faculty of Medicine, Chulalongkorn University.

Results
Our clinical bacteriology laboratory has a busy service, serving a 1,500-bed university hospital. Out of several thousands of gram-positive bacteria isolated during 1997-2004, 26 catalase-negative gram-positive isolates (isolates 1-26) were initially identified as Leuconostoc by API 20 STREP. 7 catalase-negative gram-positive strains with vancomycin MIC ≥ 256 µg/ml were isolated during 2005-2006 (isolates 31-33 and 35-38). Thus, 33 clinical isolates entered the study. As 16S rDNA sequencing analysis was performed after the other tests, the results were not complete. Some isolates could not be retrieved, and some could not be amplified. 11 isolates of isolates 1-26 were reproducibly identified by API 20 STREP as Leuconostoc (Table 1). Only 3 of the 11 isolates, however, were confirmed by both genotypic and all defined phenotypic criteria (Table 1). 7 catalase-negative gram-positive isolates with vancomycin MIC ≥ 256 µg/ml (isolates 31-33 and 35-38) were all identified as Leuconostoc by API 20 STREP, only 2 of which were confirmed genotypically. API 20 STREP identified Lactobacillus pentosus ATCC 8041 and Pediococcus pentosaceus ATCC 33316 as Leuconostoc with 81.1% and 39.3% identity, respectively. Regarding all 31 non-leuconostoc strains (including reference strains and clinical isolates), API 20 STREP identified 10 of them as Leuconostoc with over 90% identity. 16S rDNA sequencing data were available in 7 of the 10 isolates and all were closely-related Weissella spp. 6 isolates were read as Leuconostoc with 50-90% identity. Two of these were Lactobacillus pentosus ATCC 8041 and Lactobacillus salivarius (isolate 26) and two were Pediococcus (isolates 35 and 38). API 50 CHL identified almost all Leuconostoc correctly, at least to the genus level, except for isolate 3. The kit, however, identified Pediococcus pentosaceus ATCC 33316, 7 out of 8 Weissella spp., all 6 strep-     (Table 1). All these belonged to non-leuconostoc isolates and do not affect conventional phenotypic interpretation of these isolates as non-leuconostoc. Identity percentages given by API 20 STREP and API 50 CHL had poor correlations with PCR or phenotypes.
With regard to the 7 isolates of catalase-negative grampositive bacteria with high-level vancomycin resistance which were all identified as Leuconostoc by API 20 STREP, only 2 of them were confirmed as such by genotypic assays. API 50 CHL correctly identified both isolates as Leuconostoc, one of which was correct at the species level.

Discussion
Commercial diagnostics have been widely used in bacteriology laboratories to identify common organisms, such as Streptococcus, to the species level, or to identify unusual gram-positive organisms in clinical specimens, e.g. Aerococcus, Lactobacillus, and Leuconostoc, among others. Studies illustrating inaccurate identification of various grampositive pathogens have been published [3,[10][11][12][13][14]. In this study, our purpose is to raise an awareness that Leuconostoc, an emerging human pathogen, can be overdiagnosed by certain commercial diagnostics.
Occasional discrepant results among the same biochemical tests obtained from API 20 STREP and from manual conventional assays are not unexpected, as incomplete agreement of various automated and manual systems have been reported [15][16][17]. Reproducibility of API 20 STREP for Leuconostoc identification is only moderate in our study. Previous studies have shown higher consistency of bacterial identification by commercial diagnostics [18,19]. Clinical isolates in our study were initially identified during the time spanning from 1997-2006 and thus repeated API 20 STREP testing was done months or years thereafter. Our lower reproducibility could be, at least partly, due to loss or change in some characteristics by repeated subculture [20,21].
All 6 standard and clinical Leuconostoc strains were correctly identified by API 20 STREP and 5 by API 50 CHL, at least at the genus level. On the contrary, specificity of Leuconostoc identification by these API kits were only moderate at best. As evidenced by 16S rDNA sequence analysis, most of the isolates misidentified as Leuconostoc by API systems were in the genus Lactobacillus and Weissella, which are closely-related bacteria, followed by Pediococcus, also one of the lactic acid bacteria [6]. API 20 STREP failed to identify these isolates obviously because Lactobacillus and Weissella are not listed in the Identification Table [22]. Even some strains of streptococci and enterococci initially were identified as Leuconostoc. Streptococcus constellatus isolate 19 was misidentified at the species level. API 50 CHL misidentified most Weissella in this study obviously because only one species, Weissella viridescens, is included in the Identification Table of the kit.
Our conventional phenotypic criteria correlated well with Leuconostoc-specific PCR and 16S rDNA sequence analysis in almost all isolates, except for isolates 7 and 36 (Table  1). Isolate 7 was phenotypically compatible with Leuconostoc but negative by PCR. This could be closely-related bacteria which are certain lactobacilli such as L. sanfrancisco, or L. fructosus, or Weissella [23], or some rare Leuconostoc not detected by our PCR protocol. Weisella is a recently-described genus found in a variety of foods. Some of its members used to be Leuconostoc paramesenteroides and heterofermentative lactobacilli. Reliability of the conventional phenotypic criteria in this study is evidenced by the fact that only 1 of the 8 Weissella isolates and none of the 2 Lactobacillus isolates (one identified by 16S rDNA sequencing and both by PCR) was misidentified as Leuconostoc.
The importance of accurate identification of Leuconostoc also needs to be emphasized in the clinical arena. Case reports based on incomplete and/or inappropriate phenotypic criteria with or without assistance of commercial diagnostics are subject to potential errors [24][25][26][27], given the fact that Leuconostoc and related bacteria possess overlapping phenotypes. Flawed clinical reports include an incorrect argument that heterofermentative Lactobacillus must hydrolyze arginine [25], while in fact L. sanfrancisco and L. fructosus do not [23], and labeling the organism as Leuconostoc even though the organism was LAP positive [26].  [22]. Appropriate modifications of the kit criteria for Leuconostoc would significantly enhance its accuracy. For clinical laboratories, we propose that all catalase-negative grampositive coccoid or coccobacillary bacteria with high level of vancomycin resistance (MIC ≥ 256 µg/ml) be tested with the manual phenotypic assays listed in Table 1: Gram's staining of the isolate grown in thioglycolate broth, arginine dihydrolase (ADH), leucine arylamidase (LAP), gas production in MRS broth, and pyrrolidonyl arylamidase (PYR) test. Users of this method are to accept that, even though more accurate than the current API systems, these conventional assays could still occasionally misidentify certain lactobacilli and Weissella as Leuconostoc. This practical guide should minimize avoidable inaccurate identification of this emerging pathogen.

Conclusion
The current API systems, similar to some other commercial identification systems for microorganisms, still need improvement before they can reliably identify certain unusual gram-positive pathogens. They lack specificity in Leuconostoc identification. We propose that, for accuracy and reliability, identification of vancomycin-resistant, catalase-negative gram-positive bacteria be performed by practical, conventional phenotypic assays, with assistance of a genotypic confirmation where available.