Comparison of the Vitek MS and Bruker Matrix-Assisted Laser Desorption Ionization–Time of Flight Mass Spectrometry Systems for Identification of Chryseobacterium Isolates from Clinical Specimens and Report of Uncommon Chryseobacterium Infections in Humans

Matrix-assisted laser desorption ionization–time of flight mass spectrometry is becoming more popular and is replacing traditional identification methods in the clinical microbiology laboratory. We aimed to compare the Vitek mass spectrometry (MS) and Bruker Biotyper systems for the identification of Chryseobacterium isolated from clinical specimens and to report uncommon Chryseobacterium infections in humans.

ABSTRACT

Matrix-assisted laser desorption ionization–time of flight mass spectrometry is becoming more popular and is replacing traditional identification methods in the clinical microbiology laboratory. We aimed to compare the Vitek mass spectrometry (MS) and Bruker Biotyper systems for the identification of Chryseobacterium isolated from clinical specimens and to report uncommon Chryseobacterium infections in humans. The microbial database from a hospital was searched for records between 2005 and 2016 to identify cultures that yielded Chryseobacterium. Species identification by the Vitek MS and Bruker Biotyper systems was compared to identification by 16S rRNA gene sequencing. Over the study period, 140 Chryseobacterium isolates were included. Based on 16S rRNA gene sequencing, 78 isolates were C. indologenes, 39 were C. gleum, 12 were uncommon Chryseobacterium species (C. arthrosphaerae, C. culicis, C. cucumeris, C. bernardetii, C. artocarpi, and C. daecheongense), and the remaining 11 isolates were only identified at the genus level. The Vitek MS and Bruker Biotyper systems correctly identified 98.7% and 100% of C. indologenes isolates, respectively. While the Bruker Biotyper accurately identified 100% of C. gleum isolates, the Vitek MS system correctly identified only 2.6% of isolates from this species. None of the uncommon Chryseobacterium species were successfully identified by either of these two systems. The overall accuracies of Chryseobacterium identification at the species level by the Vitek MS and Bruker Biotyper systems were 60.5% and 90.7%, respectively. An upgrade and correction of the Vitek MS and Bruker Biotyper databases is recommended to correctly identify Chryseobacterium species.

INTRODUCTION

Chryseobacterium is a genus of Gram-negative, aerobic, nonmotile, nonfermenting, and non-spore-forming rods that are ubiquitously distributed in natural environments, such as soil, water, and plants (1). Currently, there are more than 100 species with validly published names in the Chryseobacterium genus (see http://www.bacterio.net/chryseobacterium.html). Among these species, C. indologenes is known to be associated with human infections, particularly in immunocompromised patients (2, 3). In contrast, the other Chryseobacterium species are infrequently thought to cause human infections.

The correct identification of pathogens is of paramount importance in clinical practice, epidemiological study, and the basic research of microorganisms. The use of 16S rRNA gene sequencing is considered the reference identification method for many bacteria. Unidentified organisms with clinical significance are often subjected to identification by this method (4). Previous studies have shown that commercial phenotypic identification systems and traditional bacterial identification methods are not reliable for the identification of Chryseobacterium species (5–8). The use of matrix-assisted laser desorption ionization–time of flight mass spectrometry (MALDI-TOF MS) for species identification is becoming more popular and is replacing traditional identification methods in many clinical microbiology laboratories because of its high speed, good accuracy, and cost-effectiveness (9). Currently, there are two MALDI-TOF MS platforms that are commonly used in the clinical microbiology laboratory for microbial identification, namely, the Vitek MS system (bioMérieux, Marcy l'Etoile, France) and the Bruker Biotyper (Bruker Daltonics GmbH, Bremen, Germany).

A previous study reported that the Vitek MS system, using both Knowledge Base v2.0 and Knowledge Base v3.0, identified only 53.8% (35/65) of Chryseobacterium isolates at the species level with a confidence value of ≥85% (5). However, little is known about the successful identification rate of Chryseobacterium species using the Bruker Biotyper. Moreover, there is no research to compare the accuracy of identification of Chryseobacterium between these two common MALDI-TOF MS platforms. In this study, we compared the concordance of the Vitek MS and Bruker Biotyper systems to the 16S rRNA gene sequencing for identification of 140 Chryseobacterium isolates. We also reported the identification of rare Chryseobacterium species that caused human infections.

MATERIALS AND METHODS

Study design and collection of isolates.

This study was performed in a 1,000-bed university-affiliated medical center in Taiwan and was approved by the Institutional Review Board of the hospital (EMRP-106-105). The microbial cultures were routinely collected according to the clinical requirements of patients. The isolates were initially identified by the clinical microbiology laboratory using API/ID32 phenotyping kits (bioMérieux) between 2005 to 2013 and using the Vitek MS system between 2014 to 2016, after the upgrade of microbial identification equipment. The isolates were stored as glycerol stocks at −80°C. The microbial database from January 2005 to December 2016 was searched, and 147 nonduplicated Chryseobacterium isolates were identified. Seven isolates were unable to be resuscitated, and 140 surviving Chryseobacterium isolates were reidentified using 16S rRNA gene sequencing, Vitek MS system, and Bruker Biotyper.

16S rRNA gene sequence analysis.

The thawed bacteria were initially subcultured on tryptic soy agar with 5% sheep blood (Becton Dickinson, Sparks, MD). The colony was then inoculated on fresh tryptic soy agar with 5% sheep blood for overnight culture. The total DNA of fresh colonies was prepared using a Wizard Genomic DNA purification kit (Promega, Madison, WI). The primers and protocols for amplification and sequencing of the 16S rRNA gene were as described previously (5). The amplicon size of 16S rRNA gene was 1,498 bp. The assembled sequences of 16S rRNA were submitted to the National Center for Biotechnology Information website for comparison of their nucleotide sequences to GenBank sequence databases using the Basic Local Alignment Search Tool (https://blast.ncbi.nlm.nih.gov/Blast.cgi). The similarity of 16S rRNA sequence was calculated between the submitted sequence and that of the type strain in the GenBank sequence databases. The species was considered to be correctly identified if the isolate shared ≥99.5% sequence identity with the type strain (10). If the sequence identity was 95 to 99.5%, only a genus was assigned (11).

Vitek MS analysis.

The identification by Vitek MS was performed using the direct colony method according to the manufacturer's suggestion (12). In brief, a part of a fresh colony obtained from two additional subcultures was smeared onto a spot of a MALDI disposable target plate and was then covered with 1 μl of Vitek MS-CHCA (bioMérieux), a saturated solution of α-cyano-4-hydroxycinnamic acid matrix in 50% acetonitrile and 2.5% trifluoroacetic acid. Once dry, the target plate was loaded into the Vitek MS platform with the Knowledge Base v3.0 database for target identification according to the default settings. Escherichia coli ATCC 8739 was used for instrument calibration and quality control. The mass spectral fingerprints were analyzed using Knowledge Base v3.0, which contained only C. indologenes and C. gleum in the category of the Chryseobacterium genus. The confidence value of identification was calculated. If the confidence value did not reach ≥99.9%, the experiment was repeated. An identification was considered reliable if an isolate presented with a ≥99.9% confidence value. If the identification did not reach a ≥99.9% confidence value, the identification was defined as no identification (13).

Bruker Biotyper analysis.

Briefly, a portion of a fresh colony obtained from two additional subcultures was applied onto a metallic MALDI-TOF main spectrum profile (MSP) 96 target (Bruker Daltonics) and then overlaid with 1 μl of saturated α-cyano-4-hydroxycinnamic acid matrix solution, as previously described (12). After drying, the plate was loaded into the Bruker Biotyper machine for microbial identification with the manufacturer's recommended settings. The protein spectra were explored with the Bruker Biotyper 3.1 software package. The reference Biotyper library v6.0 6903 MSP was searched to compare the most similar pattern for each isolate. This library included 14 Chryseobacterium species, namely, C. balustinum, C. chaponense, C. ginsenosidimutans, C. gleum, C. hagamense, C. hominis, C. indologenes, C. joostei, C. oncorhynchi, C. oranimense, C. piscicola, C. piscium, C. scophthalmum, and C. tructae. Bacterial test standard (BTS) (E. coli DH5 α strain extract plus two proteins; Bruker Daltonics) was used as a standard for calibration and as a reference for quality control. The cutoffs of identification scores were categorized into ≥2.000, 1.700 to 1.999, and <1.700 according to the manufacturer's recommendation. A score of ≥2.000 represented reliable identification at the species level, and <1.700 was interpreted as no identification. If the identification score failed to achieve ≥2.000, the experiment was repeated.

RESULTS

Characteristics of the isolates.

Over the 12-year investigation period, 140 surviving, nonduplicated Chryseobacterium isolates were included in this study. Of these isolates, the most common site of isolation was blood (n = 79; 56.4%), followed by bile (n = 21; 15%), the tip of the central venous catheter (n = 8; 5.7%), abscesses (n = 8; 5.7%), and urine (n = 8; 5.7%) (Table 1).

TABLE 1.

Site of isolation and number of Chryseobacterium isolates in this study

Site of isolationa	No. of isolates of:
	C. indologenes	C. gleum	C. arthrosphaerae	C. culicis	C. cucumeris	C. bernardetii	C. artocarpi	C. daecheongense	Chryseobacterium spp.
Blood	39	26	2	2	0	2	1	0	7
Bile	9	6	0	0	1	0	0	1	4
CVC tip	6	2	0	0	0	0	0	0	0
Abscess	6	0	1	0	1	0	0	0	0
Urine	6	2	0	0	0	0	0	0	0
Respiratory tract	4	1	0	0	0	0	0	0	0
Ascites	3	1	0	0	0	0	0	0	0
Pleural effusion	3	1	1	0	0	0	0	0	0
CSF	2	0	0	0	0	0	0	0	0
Total	78	39	4	2	2	2	1	1	11

^aCVC, central venous catheter; CSF, cerebrospinal fluid.

Species identification by 16S rRNA.

After comparison to the sequences in GenBank, 129 isolates were demonstrated to show ≥99.5%, and 11 isolates had 95 to 99.5% 16S rRNA gene sequence identity to the type strains. Among the 129 isolates identified at the species level, 78 and 39 isolates were identified as C. indologenes (type strain DSM 16777^T; GenBank accession no. LN681561) and C. gleum (type strain ATCC 35910^T; GenBank accession no. GL379781), respectively (Table 2). Twelve isolates were uncommon Chryseobacterium species, including four C. arthrosphaerae (type strain CC-VM-7^T; GenBank accession no. NR_116977) (14), two C. culicis (type strain R4-1A^T; GenBank accession no. FN554975) (15), two C. cucumeris (type strain GSE06^T; GenBank accession no. KX146463) (16), two C. bernardetii (type strain CDC G229^T; GenBank accession no. JX100816) (4), one C. artocarpi (type strain UTM-3^T; GenBank accession no. KF751867) (17), and one C. daecheongense (type strain CPW406^T; GenBank accession no. AJ457206) isolate (18) (Table 2). Of the remaining 11 isolates identified at the genus level, eight were closest to C. indologenes (98.5 to 99.3% 16S rRNA gene sequence identity), and three were closest to C. daecheongense (97.0 to 97.5% 16S rRNA gene sequence identity) (Table 2).

TABLE 2.

Chryseobacterium identified at the species level by the Vitek MS and Bruker Biotyper systems compared to the results of 16S rRNA gene sequencing

16S rRNA sequence-based identification	No. of isolates	Vitek MS			Bruker Biotyper
		Correct identification	Misidentification	No identification	Correct identification	Misidentification	No identification
C. indologenes	78	77	0	1	78	0	0
C. gleum	39	1	38	0	39	0	0
C. arthrosphaerae	4	0	4	0	0	3	1
C. culicis	2	0	1	1	0	0	2
C. cucumeris	2	0	2	0	0	0	2
C. bernardetii	2	0	1	1	0	1	1
C. artocarpi	1	0	1	0	0	0	1
C. daecheongense	1	0	0	1	0	0	1
Chryseobacterium spp.	11	0	7	4	0	2	9

Of the 12 patients with the six uncommon Chryseobacterium species infections, seven (58.3%) were male and five (41.7%) were female, with a median age of 61.5 years (standard deviation, 14.4 years). Malignancy was identified in 66.7% (8/12) of patients. One patient died of C. arthrosphaerae infection (Table 3).

TABLE 3.

Clinical information and outcome of 12 patients with uncommon Chryseobacterium species infection

Case no.	Species	Yr of isolation	Site of isolation	Age (yrs)/gendera	Clinical manifestation(s) and underlying illness(es)	Outcome
1	C. arthrosphaerae	2008	Blood	74/M	Primary bacteremia, gastric ulcer, bladder tuberculosis	Survived
2	C. arthrosphaerae	2012	Pleural effusion	42/M	Thoracic empyema, esophageal carcinoma, liver cirrhosis	Died
3	C. arthrosphaerae	2014	Abscess	46/M	Septic arthritis	Survived
4	C. arthrosphaerae	2014	Blood	54/M	Biliary tract infection, cholangiocarcinoma	Survived
5	C. culicis	2009	Blood	60/F	Biliary tract infection, hepatocellular carcinoma, liver cirrhosis	Survived
6	C. culicis	2012	Blood	64/M	Central line-associated bloodstream infection, Ewing sarcoma	Survived
7	C. cucumeris	2005	Bile	80/F	Biliary tract infection, cholangiocarcinoma, diabetes mellitus, hypertension	Survived
8	C. cucumeris	2013	Abscess	42/F	Necrotizing fasciitis	Survived
9	C. bernardetii	2009	Blood	67/F	Primary bacteremia, gastric ulcer	Survived
10	C. bernardetii	2015	Blood	75/M	Primary bacteremia, tonsil cancer	Survived
11	C. artocarpi	2016	Blood	52/M	Primary bacteremia, hypopharyngeal cancer	Survived
12	C. daecheongense	2016	Bile	82/F	Biliary tract infection, carcinoma of the ampulla of Vater, diabetes mellitus, hypertension	Survived

^aM, male; F, female.

Species identification by Vitek MS.

The Vitek MS accurately identified 98.7% (77/78) of C. indologenes isolates (Table 2). However, only 2.6% (1/39) of C. gleum isolates were successfully identified; the remaining 97.4% (38/39) of isolates were misidentified as C. indologenes (Table 4). For the uncommon Chryseobacterium species, none was correctly identified by the Vitek MS; instead, seven and two isolates were misidentified as C. indologenes and C. gleum, respectively (Table 4). Regarding the 11 isolates that were identified only at the genus level by 16S rRNA gene sequencing, 63.6% (7/11) were misidentified as C. indologenes, and four were given “no identification” (Tables 2 and 4). The overall correct identification rate of Chryseobacterium at the species level by Vitek MS was 60.5% (78/129).

TABLE 4.

Chryseobacterium species misidentified by the Vitek MS and Bruker Biotyper systems compared to the results of 16S rRNA gene sequencing

16S rRNA sequence-based identification	No. of isolates	Vitek MS		Bruker Biotyper
		Misidentification	No. of isolates	Misidentification	No. of isolates
C. gleum	39	C. indologenes	38
C. arthrosphaerae	4	C. indologenes	4	C. gleum	3
C. culicis	2	C. indologenes	1
C. cucumeris	2	C. gleum	2
C. bernardetii	2	C. indologenes	1	C. gleum	1
C. artocarpi	1	C. indologenes	1
Chryseobacterium spp.	11	C. indologenes	7	C. gleum	1
				C. tructae	1

Species identification by Bruker Biotyper.

All 78 C. indologenes and 39 C. gleum isolates were accurately identified by Bruker Biotyper (Table 2). Of the six uncommon Chryseobacterium species, none was accurately identified by the Bruker Biotyper. Three C. arthrosphaerae and one C. bernardetii isolate were misidentified as C. gleum instead (Table 4). Of the 11 isolates that were identified at the genus level by 16S rRNA gene sequencing, nine isolates were given “no identification,” one was misidentified as C. gleum, and one was misidentified as C. tructae (Tables 2 and 4). The overall accuracy rate of microbial identification for Chryseobacterium at the species level by Bruker Biotyper was 90.7% (117/129).

Comparison of species identification by Vitek MS and Bruker Biotyper systems.

Overall, the concordance of the Vitek MS and Bruker Biotyper MALDI-TOF MS systems compared to 16S rRNA gene sequencing for identification of Chryseobacterium was 60.5% and 90.7%, respectively. Both the Vitek MS and Bruker Biotyper systems showed a high rate of correct identification of C. indologenes isolates (98.7% versus 100%). For C. gleum, the Vitek MS system accurately identified 2.6% of isolates, and the Bruker Biotyper correctly identified 100%. None of the uncommon Chryseobacterium species were successful identified by these two MALDI-TOF MS systems.

DISCUSSION

Chryseobacterium does not commonly cause human infections. Among more than 100 species, C. indologenes accounted for the majority of infections in humans (2, 3). However, the identification of Chryseobacterium is usually dependent on commercial automatic phenotyping systems, such as Vitek 2 (bioMérieux) and Phoenix 100 ID/AST (Becton Dickinson Co., Sparks, MD), in previous research (2, 3). One previous study revealed the low accuracy of these traditional phenotyping methods for identification of Chryseobacterium (5). Therefore, the epidemiology of Chryseobacterium infections in the previous studies that relied on these methods could have substantial bias.

We used 16S rRNA gene sequencing as a standard for the identification of Chryseobacterium species. In our study, C. indologenes and C. gleum accounted for 55.7% and 27.9% of Chryseobacterium isolates, respectively. We also identified six uncommon Chryseobacterium species isolated from clinical specimens. C. arthrosphaerae was initially isolated from the feces of the pill millipede (14), C. culicis was from the midgut of the mosquito (15), C. cucumeris was from the cucumber root (16), C. bernardetii was isolated from human sputum (4), C. artocarpi was from the rhizosphere soil of an Artocarpus integer (17), and C. daecheongense was found in freshwater lake sediments (18). Of these species, C. arthrosphaerae, C. cucumeris, C. artocarpi, and C. daecheongense were never reported as having been isolated from human sources. However, C. bernardetii and C. culicis were found in the dental plaque and sputum of human (4, 19). Our study identified 12 patients with these rare Chryseobacterium species infections, including one fatal case. Malignancy is a common comorbidity in these patients. These six uncommon Chryseobacterium species accounted for 8.6% of Chryseobacterium isolates isolated from the clinical specimens. This finding suggests that these Chryseobacterium species are actually not rare in the clinical setting.

In the present study, we compared the accuracies of Vitek MS and Bruker Biotyper microbial identification platforms to those of 16S rRNA gene sequencing for identifying Chryseobacterium species. Our previous small-scale study compared the correct identification rate of four commonly used microbial identification platforms, the API/ID32, Phoenix 100, Vitek 2, and Vitek MS, for identification of 65 Chryseobacterium isolates (5). In that study, the Vitek MS system successfully identified 97.2% (35/36) of C. indologenes and 0% (0/22) of C. gleum isolates. Lo et al. (6) used the Bruker Biotyper to analyze the species of 15 C. gleum isolates confirmed by 16S rRNA gene sequencing, and all of these isolates were correctly identified by Bruker Biotyper. In the present study, we compared the identification of 140 Chryseobacterium isolates using both the Vitek MS and Bruker Biotyper systems. These two platforms accurately identified almost all C. indologenes isolates. However, only 2.6% of C. gleum isolates were correctly identified by Vitek MS system. In contrast, the Bruker Biotyper successfully identified all C. gleum isolates. The low identification rate of C. gleum by the Vitek MS system may lead to substantial underestimation of prevalence if the microbial identification depends on this platform.

Conclusions.

To understand Chryseobacterium infection in humans, correct identification of the species is particularly important for both clinical practice and microbiological research. Although MALDI-TOF MS has been shown to reliably identify many microorganisms, our study revealed the weakness of MALDI-TOF MS in the identification of Chryseobacterium. An upgrade of the databases is recommended to accurately identify the species of Chryseobacterium.