Identification of 273 Clinical Aeromonas Strains with a Gold Standard Method and MALDI-TOF: A Review on the Limitations of the Method

Abstract

The genus Aeromonas comprises Gram-negative bacteria widely distributed in aquatic environments, with some species able to cause disease in humans, fish, and other aquatic animals. The dominating species in human infections are Aeromonas caviae, Aeromonas dhakensis, Aeromonas hydrophila, and Aeromonas veronii and the disease presentations gastroenteritis, bacteremia, and wound infections. Matrix-Assisted Laser Desorption Ionization Time of Flight Mass Spectrometry (MALDI-TOF MS) is an extremely rapid method used in clinical microbiology laboratories to identify bacterial isolates at the genus and species level. The present study aimed to evaluate the usefulness of the MALDI-TOF MS to identify 273 clinical isolates of Aeromonas that were also identified by rpoD gene. The latter recognized eight different species, but only 73.6% of the strains of six species were correctly identified with MALDI-TOF MS and results depended upon the species. The higher concordance was with A. veronii (92.8%), A. hydrophila (83.3%), A. caviae (73.1%), and Aeromonas media (60.0%). Our results and those of the reviewed literature corroborate that MALDI-TOF is a promising identifying method being the poorly updated database the main limitation. Improvement requires including a higher diversity of strains from all the described species which should be the responsibility of the provider.

1. Introduction

The genus Aeromonas formed by more than 32 species includes Gram-negative, rod shaped, non-spore forming, facultative aerobes which are considered autochthonous of aquatic systems and widely distributed in soil and foodstuffs [1, 2, 3]. Aeromonas are considered emerging pathogens that can cause a wide spectrum of diseases in humans, mainly gastroenteritis, bacteremia, and wound infections, being able to infect both immunocompromised and immunocompetent patients [2, 3, 4]. Four are the dominating species associated with clinical cases: Aeromonas caviae, Aeromonas dhakensis, Aeromonas veronii and Aeromonas hydrophila [3, 5]. Consumption of contaminated water and food are considered the main routes of transmission and there are few reports of well documented outbreaks that include information about the ingested doses of Aeromonas [6, 7, 8]. In fact, using the latter information the enteropathogenicity of Aeromonas has been considered similar to the one produced by Salmonella or Campylobacter [8].

Phenotypic identification of Aeromonas isolates in clinical microbiological laboratories has been mostly based on different biochemical tests, including commercial identification kits and or automatic or semiautomatic systems [9, 10, 11]. Classic phenotypic characteristics that identify the genus Aeromonas are the positive cytochrome oxidase reaction and the growth in nutritive broth in the presence of the vibriostatic agent O129 [2, 12, 13]. Previous studies reported inconsistent results and misidentification of Aeromonas as members of the genus Vibrio [2, 5, 13, 14, 15], and this is very relevant when confusion occurs with Vibrio cholerae [13, 16, 17, 18]. To avoid this a specific DNA probe and a PCR have been described [17, 18]. Beaz-Hidalgo et al. [12] re-identified 119 strains by molecular methods (16S rRNA-RFLP and rpoD sequences) that previously were identified phenotypically and demonstrated that only 35.5% were correctly identified at the species level. Commercial identification systems like API 20E, Vitek, or MicroScan W/A have demonstrated limitations [9, 11]. In 2010, Lamy et al. [9] compared the accuracy of six commercial systems for Aeromonas identification, using the rpoB sequencing as a reference and concordance was low between both approaches because the commercial systems showed many erroneous identification at the species level.

Correct identification can be achieved by sequencing housekeeping genes (HKG) like gyrB or rpoD that encode proteins with essential functions for the survival of the bacteria [9, 12, 15], and have been very useful in recognizing many species in recent years [19, 20, 21, 22, 23, 24, 25]. However, the latter is costly and time-consuming, being not available for routine diagnosis. Matrix-Assisted Laser Desorption/Ionization Time of Flight Mass Spectrometry (MALDI-TOF MS) is a powerful method increasingly being used in many clinical laboratories to identify bacterial isolates at the species level [26, 27, 28]. This technique is being used as an alternative to conventional identification methods due to its speed in the identification of isolates, which takes only a few minutes [26, 27, 28]. He et al. [29] performed a comparative study of identification by MALDI-TOF MS versus routine phenotypic methods for the identification of suspicious colonies from stool samples. They found that the entire identification procedure from the smear preparation from the culture media to reporting of the final result was completed within 30 min, thus shortening the turnaround time of the test by 2–3 days [29]. The MALDI-TOF MS was first used for the identification of aeromonads by Donohue et al. [30] using 32 strains of 17 genomospecies that included the type and reference strains of the hybridization groups HG1 to HG17 along with some clinical isolates and recognized that the mass spectra showed specific signatures for the genus and the species. Since then MALDI-TOF MS has been used for the identification of Aeromonas spp. in at least a dozen of studies where the technique was evaluated in comparison with a molecular gold standard identification method and results indicated that performance depended directly on the database and on the species [31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42].

In the present study, we evaluated the utility of the MALDI-TOF MS system for the identification of a collection of 273 clinical isolates belonging to different species of the genus Aeromonas recognized by their rpoD gene. In addition, we reviewed previous studies that also use this identification method for the genus Aeromonas.

2. Material and methods

2.1 Bacterial isolates

A total of 273 clinical isolates of Aeromonas were recovered from patients affected mainly by gastroenteritis. The latter isolates were recovered mainly from stool samples cultured in Cefsulodin Irgasan Novobiocin (CIN) agar (Yersinia CIN Agar) and the Salmonella-Shigella agar (SS) both from Biomerieux®, Marc l’Etoile, France, after incubation at 37°C for 24 h in aerobic conditions. Aeromonas were identified by MALDI-TOF MS and simultaneously prepared for sequencing the rpoD gene [15].

2.2 Identification by MALDI-TOF MS

Identifications were performed using the direct transfer method following the manufacturer’s instructions (Bruker Daltonics GmbH, Bremen, Germany). In brief fresh colony material was smeared on the holder (MBT Biotarget 96, Bruker Daltonics) using a toothpick, and then 1.0 μl of a saturated solution of MALDI matrix (α-cyano-4-hydroxycinnamic acid) was applied in top of each sample and it was allowed to dry at room temperature. Measurements were performed by a Microflex® LRF MALDI-TOF/TOF equipment (Bruker Daltonics). Spectra ranging from the mass-to-charge ratio (m/z) 2,000–20,000 KDal were analyzed using Biotyper Real-Time Classification software v3.1 (Bruker Daltonics) including database v8 version (the Aeromonas spp. included in the MBT 7854 MSP library are transcribed in Table 1). This software generates a score showing the similarity between the spectra obtained from the clinical sample and the reference spectra and displays the top 10 matching results with the highest scores. As specified by the manufacturer instructions, scores of ≥2.300 indicated highly probable species identification, scores ≥2.000–2.299 probable species identification, scores of ≥1.700–1.999 probable genus identification, and scores of ≤1.699 indicated no reliable identification (Table 2, [31, 33]).

Aeromonas species	No strains	References at culture collectionsa
A. bestiarum	2	CETC 4227^T, DSM 13956^T
A. caviae	3	60PIM, CETC838^T, DSM 7323^T
A. encheleia	2	CECT 4342^T, DSM 11577^T
A. enteropelogenesb	3	DSM 6394^T, DSM 7312, DSM 9381
A. eucrenophila	2	CETC 4224^T, DSM 17534^T
A. hydrophila:	5
A. hydrophila subsp. hydrophila		CECT 839^T
A. hydrophila subsp. anaerogenesc		DSM 30188^T
A. hydrophila subsp. hydrophila		DSM 30187^T
A. hydrophila subsp. hydrophila		LMG 21108
A. hydrophila subsp. ranae		LMG 19707^T
A. icthiosmiad	1	DSM 6393^T
A. jandaei	2	CECT 4228^T, DSM 7311^T
A. media	4	CETC 4228^T, DSM 4881^T, ST_00655_09 ERL, ST_02485_08 ERL
A. molluscorum	2	848^T, DSM 17090^T
A. popoffi	1	LMG 17541^T
A. punctata subsp. caviaec	1	LMG 3773
A. salmonicida:	5
A. salmonicida subsp. masoucida		LMG 3782^T
A. salmonicida subsp. pectinolytica		DSM 12609^T
A. salmonicida subsp. salmonicida		CCM 1307
A. salmonicida subsp. salmonicida		CECT 894^T
A. salmonicida subsp. salmonicida		LMG 3780^T
A. schubertii	2	CECT 4240^T, DSM 4882^T
A. simiae	1	DSM 16559^T
A. sobria	2	CECT 4245^T, LMG 3783^T
Aeromonas sp.	1	VA 1205_09 ERL
A. veronii	7	0807 M090438 IBS, CECT 4199^T,e, CECT 4257^T, CECT 5761^T,f, DSM 11576^T,e, DSM 17676^T,f, DSM 7386^T

Table 1.

References of the different Aeromonas species among 46 strains included in the Bruker MALDI Biotyper spectra database v8 (MBT 7854 MSP).

^a

in bold duplicate strain from different culture collections.

^b

A. trota.

^c

A. caviae.

^d

A. veronii.

^e

Strains CECT 4199^T and DSM 11576^T included in A. veronii are correspond to the type strain of the species A. allosaccharophila.

^f

The CECT 5761^T and DSM 17676^T strains correspond to the species A. culicicola, which was synonymized with A. veronii.

Table constructed from the Biotyper database v8, that is identical to v3.0 and 3.1 elaborated by Latif-Eugenín et al. [35] with the following observations: ^a−fObsolete names from old taxa that correspond now to the following species.

Identification method (n)						Concordance (%)
rpoD gene sequencing	MALDI-TOF MS; (score ≥ 2.000)	Score values with the database				Correct; species	Global
		0.000–1.699	1.700–1.999	2.000–2.229	2.300–3.000
A. caviae (212)	A. caviae (155)	1	24	129	26	73.1	73.6; (species); 100; (genus)
	A. hydrophila (30)	2	4	23	1
	A. caviae/; A. hydrophila (1)	0	0	1	0
	A.veronii (1)	0	1	0	0
A. veronii (28)	A. veronii (26)	0	0	25	1	92.8
	A. caviae (2)	0	1	1	0
A. hydrophila (18)	A. hydrophila (15)	0	1	15	0	83.3
	A. eucrenophila (1)	0	1	0	0
	A. veronii (1)	0	1	0	0
A. media (5)	A. media (3)	0	0	3	0	60.0
	A. eucrenophila (1)	0	0	1	0
	A. veronii (1)	0	0	1	0
A. rivipollensis (5)	Aeromonas sp. (2)	0	2	0	0	0
	A. hydrophila (1)	0	1	0	0
	A. media (1)	0	1	0	0
	A. veronii (1)	0	1	0	0
A. allosaccharophila (2)	A. veronii (2)	0	0	2	0	0
A salmonicida (2)	A. salmonicida (1)	0	0	1	0	50.0
	A. bestiarum (1)	0	0	1	0
A. bestiarum (1)	A. bestiarum (1)	0	0	1	0	100
Total identifications (score ≥ 2.000)		3	38	174	27 = 201

Table 2.

Concordance between the genetic identification of 273 strains of Aeromonas with the rpoD gene sequencing and that obtained by MALDI-TOF MS.

In bold the species in agreement with rpoD with a score ≥ 2.000. Scores: ≤1.699 indicated no reliable identification, ≥1.700–1.999 probable genus identification, ≥2.000–2.299 probable species identification and ≥ 2.300−3.000 highly probably species identification.

2.3 Classification of the consistency of the MALDI-TOF MS identification in five categories

Consistency of the identification was evaluated following the classification into five categories (categories A to E) that take into consideration the score above and below 2.000 obtained in the first and the second best matches [31, 33, 37]. Category A corresponds to the correct identification: the correct species corresponds to a unique species with a score ≥ 2.000 in the first and second match; Category B is considered inconclusive because the correct species is the first match with a score ≥ 2.000, but the second match possessed the same score but showed a different species; Category C is also considered inconclusive because the first and second matches have a score ≥ 2.000, but the correct species is the second match; Category D embraces the misidentifications because the first and second match corresponds to a wrong species despite the first and second matches showed a score ≥ 2.000 and finally Category E, groups the unidentifiable strains, because the first and second matches showed scores <2.000 [33].

2.4 Preparation of genomic DNA

Clinical isolates were incubated at 37°C for 24 h in aerobic conditions on Tryptic Soy Agar, TSA (bioMerieux, Marcy l’Etoile, France). Genomic DNA for molecular characterization was extracted using the InstaGene™ DNA Purification Matrix (Bio-Rad-Hercules CA, USA) following the manufacturer’s instructions.

2.5 Housekeeping gene sequencing and phylogenetic analysis

All strains analyzed in this study were identified to the species level using the rpoD gene sequence, with the primers and the conditions described by Soler et al. [15]. After sequencing, the raw sequence file (.abi) was processed using DNASTAR SeqMan Pro (DNASTAR, Madison, WI, USA) to obtain a trimmed nucleotide sequence. A phylogenetic tree was generated with the neighbor-joining (NJ) method using the obtained sequences and those from all the known type strains of Aeromonas spp. (http://www.ncbi.nlm.nih.gov/) with the MEGA 7.0 software [43].

3. Results and discussion

Among the 273 strains, eight species were recognized with the rpoD gene as shown in (Table 2) and only three of them embraced 94.5% of the strains i.e., A. caviae 212 (77.6%), A. veronii 28 (10.2%) and A. hydrophila 18 (6.6%). This is in agreement with results obtained in other studies that demonstrated that these are the prevailing species found in association with clinical strains [1, 2, 3, 5]. Interestingly the sequences of the rpoD gene recognized also the species Aeromonas media, Aeromonas rivipollensis, Aeromonas allosaccharophila, Aeromonas salmonicida and Aeromonas bestiarum, which despite being in low abundance, like in our study, have all been isolated from human infections [5, 44].

Rapid identification of pathogens is essential for the diagnosis of human and animal infections. Molecular reference methods like HKG or whole-genome sequencing are considered the gold standard [3, 5, 15, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 45, 46, 47, 48], however, they are costly, time-consuming, and require molecular knowledge to correctly interpret results and therefore are not appropriate for rapid species identification in the clinical diagnostic laboratory. Many studies have demonstrated that biochemical methods for bacterial identification have limited discriminatory power and are cumbersome and time-consuming, something that is counterproductive for the management of infectious diseases [9, 10, 11, 12]. A good alternative is the MALDI-TOF MS technology, a fast and useful tool that is nowadays being used in many clinical laboratories for the identification of bacteria [14, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42]. It has been emphasized that the accuracy of the identification strongly relies upon the robustness of the database, and this has been claimed in many of the reviewed studies Table 3 [31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42].

Reference	No	Origin	HKG reference method	Results from MALDI-TOF MS				Conclusions MALDI-TOF MS
				Equipment/ Software	No. of correct ID (%)
					Global concordance		Detailed species
					Genus	Species
Lamy et al. [37]	31	Type and reference strains	rpoB	Bruker/Biotyper 2.0 (Bruker Daltonics, Bremen, Germany)	31 (100)	29 (90.6)	A. aquariorum, A. bivalvium and A. tecta) absent in the data base were incorrectly identified	It will probably be the most accurate phenotypic method for identification to the species level
	49	Environmental (not specified)			49 (100)	46 (93.9)	A. caviae 4/5 (80.0); A. hydrophila 13/13 (100); A. veronii 23/24 (95.8); A. media 6/7 (85.7)
	90	Human clinical isolates (not specified)			90 (100)	81 (90.0)	A. caviae 20/23 (87.0); A. hydrophila 29/32 (90.6); A. veronii 27/30 (90.0); A. media 2/2 (100); A. jandaei 1/1 (100); A. allosaccharophila 1/1 (100); A. salmonicida 1/1 (100)
Benagli et al. [38]	92	Well characterized strains from 21 species	gyrB	AXIMA Confidence™ (Shimadzu Biotech, Kyoto, Japan) MS/ SARAMIS™ software (BioMérieux, France)	NA	NA	Created 11 new SuperSpectra™ analyzing 92 Aeromonas strains belonging to the 21 known species of the genus	52/741 (7%) strains could not be identified due to the absence of 10 species (23 strains) or poor coverage of the SuperSpectra™ for 29 strains.; With the new database with well-characterized strains, it is a rapid and economic method for identification of the species.
	741	Human and environmental (not specified)		AXIMA and SARAMIS supplemented with 11; SuperSpectra™ of the listed species*	741 (100)	689 (93.0)	A. caviae 189/197 (95.9); A. hydrophila 165/167 (98.8); A. veronii 95/103 (92.2); A. media 88/91 (96.7); A. salmonicida 43/44 (97.7); A. bestiarum 30/30 (100); A. eucrenophila 24/25 (96.0); A. sobria 21/21 (100); A. popofii 13/19 (68.4); A. tecta 12/12 (100); A. encheleia 9/9 (100)
Chen et al. [39]	217	Human clinical isolates (not specified)	rpoB	Bruker/ Biotyper 3.1 (Bruker, Daltonics, Bremen, Germany)	217 (100)	147 (67.7)	A. veronii 57/61 (93.4); A. caviae 56/61 (91.8); A. dhakensis 0/58 (0); A. hydrophila 34/35 (97.1); A. sanarelli 0/1 (0); A. taiwanensis 0/1 (0)	The 217 strains of the listed species were used to standardized the method creating their own database.; It could be incorporated in routine clinical laboratories after completing the database with new spectra.
	100	Human clinical isolates (not specified)	rpoB	After 123 new spectra were included in the database	100 (100)	97 (97.0)	A. caviae 30/30 (100); A. veronii 29/30 (96.7); A. dhakensis 29/30 (96.7); A. hydrophila 9/10 (90.0)
Vávrová et al. [40]	64	Czech Collection of Microorganisms (CCM)	Type and reference strains	Bruker/Biotyper 3.0 (Bruker Daltonics, Bremen, Germany)	64 (100)	41 (64.0)	NA	It is not a reliable method for identification of the species.
Shin et al. [41]	65	Human (blood, body fluid, feces)	gyrB and rpoB (not concatenated)	Bruker/Biotyper 2.0 (Bruker Daltonics, Bremen, Germany)	64 (98.5)	60 (92.3)	A. caviae 19/23 (82.6); A. hydrophila 35/35 (100); A. veronii 6/6 (100); A. aquariorum 0/1 (0)	Could be a good alternative to HKG sequencing in clinical laboratories for identification of the species.
Pérez-Sancho et al. [31]	151	Environmental (diseased fish)	rpoD	Bruker/Biotyper 3.0 (Bruker Daltonics, Bremen, Germany)*	151 (100)	62; (41.0)	A. veronii 10/31 (32.3); A. salmonicida 19/29 (65.5); A. sobria 6/27 (22.2); A. hydrophila 20/22 (91.0); A. bestiarum 6/17 (35.3); A. popoffii 0/13 (0); A. media 0/4 (0); A. dhakensis 0/3 (0); A. piscicola 0/2 (0); A. allosaccharophila 0/1 (0); A. encheleia 1/1 (100); A. caviae 0/1 (0)	The added spectra corresponded to one isolate of A. dhakensis and A. piscicola and two isolates of the most common species in fish diseases: A. hydrophila, A. bestiarum, A. salmonicida and A. piscicola.
Pérez-Sancho et al. [31]	137	Environmental (diseased fish)	rpoD	Bruker/Biotyper 3.0 (Bruker Daltonics, Bremen, Germany)* after the database was supplemented with 14 mass spectra.	137 (100)	76 (55.5)	A. veronii 21/29 (72.4); A. sobria 22/25 (88.0); A. salmonicida 13/27 (48.1); A. hydrophila 14/20 (70.0); A. bestiarum 3/15 (20.0); A. popoffii 2/11 (18.2); A. media 0/4 (0); A. dhakensis 0/2 (0); A. allosaccharophila 0/1 (0); A. caviae 0/1 (0); A. encheleia 1/1 (100); A. piscicola 0/1 (0)	It is a useful system for the identification to the genus level, but not for identification of the most prevalent species implicated in fish disease.; The evaluation of 137 additional strains showed some improvements for some species.
Latif Eugenín [35]	179	Human (feces, blood, ascites fluid, wound, sputum, bile, urine, faringeal)	rpoD	Bruker/Biotyper 3.0 & 3.1 (Bruker Daltonics, Bremen, Germany)*	176 (98.3)	163 (91.1)	A. caviae 113/117 (96.6); A. veronii 30/32 (93.8); A. hydrophila 15/15 (100); A. media 3/6 (50.0); A. salmonicida 1/2 (50.0); A. allosaccharophila 0/2 (0); A. dhakensis 0/1 (0); A. bestiarum 1/1 (100)	It is a useful and reliable system for the identification of species. With an updated database would identify all the species.
Zhou et al. [36]	37	Human (feces)	MLPA	VITEK MS (BioMérieux, France)*	37 (100)	6 (16.2)	A. caviae 0/6 (0); A. hydrophila 0/6 (0); A. veronii 0/6 (0); A. dhakensis 0/6 (0); A. enteropelogenes 5/6 (83.3); A. media 1/2 (50.0); A. sanarelli 0/1 (0); A. taiwanensis 0/2 (0); A. bivalvium 0/1 (0); A. allosaccharophila 0/1 (0)	The system is better than the VITEK2 biochemical tests in identifying the genus and Aeromonas spp.
Du et al. [42]	62	China State Key Laboratory of Infectious Disease Prevention and Control (CDC) Collection Bacterial Bank	WGS/MLPA	Microfelx/Autof MS1000 database (Antobio, China)*	62 (100)	49 (79.0)**	A. caviae 18/21 (85.7); A. hydrophila 7/7 (100); A. veronii 18/18 (100); A. media 1/1 (100); A. jandaei 3/5 (60.0); A. enteropelogenes 2/2 (100); A. dhakensis 0/8 (0)	A. caviae misidentified with A. hydrophila.; With an updated, mass spectrometry could be recommended for routine laboratories
Sun et al. [32]	58	Human clinical isolates (blood)	gyrB	Bruker/Biotyper (Bruker Daltonics, Bremen, Germany)*	NA	31 (53.4)	A. dhakensis 0/26 (0); A. jandaei 0/2 (0); No information from the results from the other 4 species	Showed poor coincidence with gyrB sequencing analysis
Kitagawa et al. [33]	58	Human clinical isolates (not specified)	rpoB	Bruker/Biotyper 3.4 (Bruker Daltonics, Bremen, Germany)*	53(91.4)	37(63.8)***	A. caviae 15/19 (78.9); A. hydrophila 14/15 (93.3); A. veronii 8/11 (72.7); A. jandaei 0/1 (0); A. dhakensis 0/12 (0)	It is a good method, but an upgrade of the database is required to differentiate the species
				VITEK MS ver. KB3.2 (BioMérieux, France)*	58 (100)	34 (58.6)	A. caviae 19/19 (100); A. hydrophila 15/15 (100); A. veronii 0/11 (0); A. jandaei 0/1 (0); A. dhakensis 0/12 (0)
Zhang et al. [34]	188	Human clinical isolates (blood, feces, wound, bile, urine, other sites no especified)	rpoD, gyrB and gyrA (not concatenated)	VITEK MS (BioMérieux, France)*	188 (100)	75 (39.9)	A. caviae 43/54 (79.6); A. hydrophila 31/49 (63.2); A. veronii 0/47 (0); A. jandaei 1/2 (50.0); A. dhakensis 0/34 (0); A. sobria 0/1 (0); A. media 0/1 (0)	The system was unreliable for species identification
This study	273	Human clinical isolates (mainly feces)	rpoD	Biotyper 3.1 (Bruker Daltonics, Bremen, Germany)*	273 (100)	201 (73.6)	A. caviae 155/212 (73.1); A. veronii 26/28 (92.8); A. hydrophila 15/18 (83.3); A. media 3/5 (60.0); A. rivipollensis 0/5 (0); A. allosaccharophila 2/2 (0); A. salmonicida 1/2 (50.0); A. bestiarum 1/1 (100)	It is not currently possible to confidently distinguish all known species. Database should be corrected for errors and updated

Table 3.

Comparison of different studies where MALDI-TOF has been used to identify Aeromonas spp. recovered mainly from clinical cases.

*

Database do not include all known Aeromonas species.

**

This data is not the same (83.87% 52/62) as in Du et al. [42] abstract but corresponds to the correct identifications shown in their Table 1 (MS).

***

This data do not derives from the abstract but from information provided in Table 1 [33].

HKG: house keeping genes. NA: data not available. WGS: Whole-genome sequencing. MLPA: Multilocus Phylogenetic Analysis.

3.1 Relationship between the score and the identification

In relation to the identification obtained from the first match with the MALDI-TOF and applying the cut-offs established by the manufacturer (Table 3) only 27/273 (9.9%) of the rpoD identified strains had a score ≥ 2.300 that corresponded to a highly probable species identification and 174/273 (63.7%) as a secure genus identification and a probable species identification (score ≥ 2.000–2.299), while 13.9% (38/273) of the strains presented scores (≥1.700–1.999) that indicated that they probably belonged to the genus. There were also 1.1% of the isolates that matched with Aeromonas but with scores ≤1.699 indicating that they did not obtain a reliable identification (Table 3). MALDI-TOF was able to identify correctly at the genus level 100% of the strains but only 73.6% (201/273) at the species level. The percentage of concordance between both methods of identification was 92.8% (26/28) for A. veronii, 83.3% (15/18) for A. hydrophila, 73.1% (155/212) for A. caviae, 60.0% (3/5) for A. media, 50.0% (1/2) for A. salmonicida and the single strain of A. bestiarum was correctly identified. The strains of A. rivipollensis obtained scores that indicated they probably belonged to the genus and A. allosaccharophila got a probable species identification with A. veronii (Table 3). None of the strains of A. rivipollensis, and A. allosaccharophila were identified correctly with MALDI-TOF MS. All the strains of these two species, 5 and 2, respectively were misidentified as Aeromonas sp., A. hydrophila, A. media and A. veronii (Table 3). Regarding incorrect identifications, 30 A. caviae isolates were misidentified as A. hydrophila and 1 as A. veronii, and 2 A. hydrophila as A. eucrenophila and A. veronii.

3.2 Limitations and errors of the biotyper database that may explain the results

Notice that as shown in Table 1, the majority of the species are represented by duplicate strains from two different type collections, for instance in the case of A. bestiarum strains CECT 4227^T is the same strain as DSM 13956^T or for A. hydrophila CETC838^T is the same strain as DSM 7323^T and this happens for fifteen species therefore, this database only has one strain of many species. In other cases, the species name is obsolete like the case of A. enteropelogenes that currently corresponds to A. trota, A. hydrophila subsp. anaerogenes that now corresponds to A. caviae and A. icthiosmia to A. veronii. In addition, the strains CECT 4199^T and DSM 11576^T are mislabeled and correspond to the type strain of the species Aeromonas allosaccharophila. The same for strains CECT 5761^T and DSM 17676^T which correspond to the species A. culicicola, which was synonymized with A. veronii, this being, nowadays the accepted species name.

The limitations of the Biotyper database for not including species can be expected as occurs with results of A. rivipollensis (Table 2). Until 2016, A. rivipollensis was considered as A. media due to its similar phylogenetic features [49]. However, only 1/5 strains of A. rivipollensis was identified by MALDI-TOF as A. media, which is in agreement with the recognition of a strain identified as A. media from a patient with gastroenteritis that was finally recognized to belong to A. rivipollensis through genome sequencing [50]. The present study together with the one of Bertran et al. [50] represents the first clinical records of this species.

3.3 Review of the literature

Several studies have also evaluated the usefulness of the MALDI-TOF MS to identify clinical strains of Aeromonas using in parallel the sequences of HKGs as reviewed in Table 3 [31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42]. Concerning identification at the genus level, our percentages are in line with around 100% of previous published studies. Regarding the identification at the species level, there is a variable percentage of concordance between studies. Few authors obtained a good percentage of concordance between MALDI-TOF MS and a reference method. Lamy et al. [37] studied 32 type and reference strains, 49 environmental and 90 clinical strains, and obtained a concordance of 90.6%, 93.9%, and 90.0%, respectively, in comparison with rpoB gene sequencing. Shin et al. [41] reidentified 65 clinical strains previously characterized by gyrB sequencing and showed a 92.3% concordance. Similar results were presented by Latif-Eugenín [35] in a comparison of 179 clinical strains with rpoD. In the study by Du et al. [42], 62 CDC bacterial bank strains were identified based on the whole-genome sequencing, MLPA of seven HKGs, and concordance with Autof software was 79.0% (Antobio, China). However, other authors obtained lower concordance, in line with our results, or generated additional mass spectra to complete the available database. In this sense, Benagli et al. [38] characterized the first 92 strains of 21 species and generated 11 SuperSpectra™ from 11 species, reaching a concordance of 93.0%. Chen et al. [34] characterized 217 clinical isolates previously identified by rpoB sequencing and found that only 67.7% were correctly identified at the species level. However, when they completed the database with new spectra and they used it to validate again the method using 100 additional genetically identified strains a concordance of 97.0% was obtained. Vávrová et al. [40] obtained a 64.0% of concordance by analyzing 64 isolates from Czech Collection of Microorganisms, 50 of the 64 strains belonged to species included in the Biotyper v3.0 database and in this group, 41 were correctly identified (82.0%). Worse results were reported by Pérez-Sancho et al. [31] 41.0% comparing with rpoD, Sun et al. [32] 53.4% comparing with gyrB, and Kitagawa et al. [33] 63.8% comparing with rpoB, all of them using the Biotyper software [31, 32, 33]. The worst concordances the 16.2% and 39.9% were obtained by Zhou et al. [36] comparing to MLPA and Zhang et al. [34] comparing to rpoD, gyrB, and gyrA, in their respective studies performed with VITEK MS [38, 39].

3.4 Relationship between the score and the identification

In our study, some correctly identified species had a contradictory score, for instance, 24 A. caviae and one A. hydrophila correctly identified (9.1%) presented scores of 1.700–1.999. One (0.4%) of the correctly identified A. caviae presented a score of ≤1.699. In the cases of misidentified strains, all of them presented scores ≤1.999, except 23 A. caviae isolates that presented scores of ≥2.000. In addition, considering not only the score of the first-best match but also the score of the second-best match the consistency of the identification was classified in a subgroup of 110 strains into one of the five categories (A to E) following the criteria described in the materials and methods section [33].

Among the 110 strains, it was included an important number of strains of A. caviae that were randomly selected among the 212 strains of this species because it was one of the species with higher errors, and the group of A. rivipollensis because it is a species recently described and from which a poor knowledge exist in relation to MALDI-TOF MS behavior (Table 4). Category A (accurate identification) was obtained for 23.7% (23/101) and 25.0% (1/4) of the A. caviae and A. hydrophila isolates, respectively. Category B and C, considered inconclusive results, were found in 46.5% (47/101) of A. caviae. Also, 28.7% (29/101) of the A. caviae strains were classified as unidentifiable as occurred for example with all the isolates of A. rivipollensis.

Species	n	First match	Score	Second match	Score	Category*
A. caviae (101)	23	A. caviae	2.038–2.337	A. caviae	2.001–2.315	A
	24	A. caviae	2.031–2.260	A. hydrophila	1.933–2.222	B, E
	3	A. caviae	2.016–2.178	A. hydrophila	1.941–1.994	B
	20	A. hydrophila	2.029–2.348	A. caviae	2.013–2.268	C
	2	A. hydrophila	2.119–2.170	A. hydrophila	2.079–2.111	D
	9	A. caviae	1.805–1.991	A. caviae	1.724–1.931	E
	1	A. caviae	1.976	A. veronii	1.918	E
	12	A. caviae	1.570–1.977	A. hydrophila	1.566–1.926	E
	1	A. caviae	1.844	A. jandaei	1.701	E
	2	A. hydrophila	1.759–1.897	A. caviae	1.688–1.883	E
	2	A. hydrophila	1.950–1.957	A. hydrophila	1.914–1.941	E
	1	A. hydrophila	1.680	A. veronii	1.631	E
	1	A. veronii	1.717	A. hydrophila	1.712	E
A. hydrophila (4)	1	A. hydrophila	2.218	A. hydrophila	2.052	A
	1	A. hydrophila	1.889	A. veronii	1.640	E
	1	A. eucrenophila	1.911	A. hydrophila	1.846	E
	1	A. veronii	1.794	A. eucrenophila	1.777	E
A. rivipollensis (5)	1	A. hydrophila	1.942	A. jandaei	1.726	E
	1	A. media	1.800	A. veronii	1.625	E
	1	A. veronii	1.719	A. molluscorum	1.678	E
	1	Aeromonas sp.	1.907	A. bestiarium	1.866	E
	1	Aeromonas sp.	1.768	A. hydrophila	1.749	E

Table 4.

Consistency of MALDI-TOF MS identification results of 110 selected strains.

*

Consistency categories represent: A, Accurate identification (First and Second match correct species scores ≥2.000); B, inconclusive (First and Second match scores ≥2.000, First correct species and Second a different species); C, inconclusive (First and Second match scores ≥2.000, First species incorrect and Second correct species); D, misidentified (First and Second match wrong species despite scores are ≥2.000); E, unidentifiable (First and Second match showed scores <2.000).

In bold the species in agreement with rpoD identification.

3.5 Identification of Aeromonas caviae

Regarding the identification of A. caviae, our data shows a 73.1% concordance between both methods. This percentage is lower than in previous studies Table 3: Latif-Eugenín et al. [35] obtained 96.6% global concordance, Chen et al. [39] 94.5%, Lamy et al. [37] 85.7%, Shin et al. [41] 82.6%, Kitagawa et al. [33] 78.9% (all of them with Biotyper software), Du et al. [42] 85.7% (with Autof), Kitagawa et al. [33] 100% and Zhang et al. [34] 79.6% (both using VITEK MS). Nevertheless, other authors such as Pérez-Sancho et al. [31] (analyzing with Biotyper) or Zhou et al. [36] (with VITEK MS) reported 0% of concordance (Table 3). In our study, the lower concordance could be partly explained because a non-negligible number of A. caviae strains 14.1% (30/212) were misidentified as A. hydrophila. The database for this species is limited because only had two spectra from the same strain from two different culture collections (CECT 838^T is the same strain as DSM 7323^T). Notice that looking deeper at the MALDI-TOF MS results of the 101 randomly selected strains of A. caviae the consistency of the first and second matches resulted in 51 of the strains belonging to categories C, D and E. The 24 A. caviae misidentified as A. hydrophila also presented a score of ≥2.000. Our results differ from the studies of Latif-Eugenín [35], Lamy et al. [37], Chen et al. [39], and Kitagawa et al. [33] in which 91.4%, 90.6%, 100%, and 100% respectively of the correctly identified strains presented a score of ≥2.000.

3.6 Identification of Aeromonas hydrophila and Aeromonas veronii

Concerning to A. hydrophila and A. veronii, the percentage of correct identifications obtained in our study were 83.3% and 92.8% respectively (Table 2), in agreement with most of the previous studies mentioned, with percentages above 90%. As show in Table 3 lower concordances were also reported for A. hydrophila by Zhang et al. [34] (63.3%) and Zhou et al. [36] (0%), and for A. veronii by Pérez-Sancho et al. [31] (32.3%), Kitagawa et al. [33] (72.3% with Biotyper and 0% with VITEK MS), Zhou et al. [36] and Zhang et al. [34] (both 0% of concordance). It is striking that all studies using VITEK MS did not recognize any A. veronii strain [33, 34, 36].

3.7 Limitations of the MALDI-TOF MS deviations

As commented an important limitation of MALDI-TOF MS is that the database has just a single representative or a reduced number of strains of each species. In most cases, the spectra of a species have been created with two strains (specifically in the case of the type strains) and these correspond to the same, what changes is the reference given in different bacteria culture collections. Moreover, as commented, it suffers from the use of outdated names because they have been synonymized, such as “Aeromonas punctata” for A. caviae, “Aeromonas trota” for Aeromonas enteropelogenes, and “Aeromonas ichthiosmia” for A. veronii [40]. The MALDI-TOF library used in the study (Biotyper v.8 MBT 7854 MSP) included spectra from 46 strains belonging to 20 Aeromonas species and 8 subspecies and a strain identified at the genus level only (Table 1). Therefore, this database does not have 46 strains, but only 27 because nearly all the diversity within a taxon was represented, as commented, by the same strain from two or three different culture collections. Another big limitation is that database does not contain all the species described in this genus such as the species described from clinical cases A. dhakensis (formerly known as Aeromonas aquariorum), A. rivipollensis, A. intestinalis, A. enterica, A. taiwanensis, A. tecta, A. diversa and A. sanarelli, as well from the environment such as A. crassostreae, A. aquatilis, A. australiensis, Aeromonas bivalvium, A. cavernicola, A. fluvialis, A. piscicola. A. lusitana and A. rivuli [5, 20, 21, 22, 23, 24, 25, 51]. This is especially relevant for species that have shown to be highly relevant in clinical cases, such as A. dhakensis that showed to be relatively abundant among the clinical strains studied by Chen et al. [39] n = 88, Sun et al. [32] n = 26, Kitawaga et al. [33] n = 12 and Zang et al. [34] n = 34, and that could not be identified correctly as shown in Table 3, neither in a recent study that compared several methods for the identification of Aeromonas spp. [42]. In fact, a higher sepsis-related mortality rate has been described for infections produced by A. dhakensis when compared with other species [32] and it is also the principal species causing soft tissue infections especially if patients have liver cirrhosis or malignancy [51]. Furthermore, this species showed relatively high antimicrobial resistance to carbapenems and beta-lactams, so early awareness of the presence of this species is important to avoid treatment failure [51, 52]. In conclusion, this is an important pathogenic species and the company supplying the software should be sensitive to these discoveries. In agreement with our results other researchers that investigates clinical strains recovered A. allosaccharophila and A. salmonicida both species poorly known from the clinical perspective [9, 36, 47]. Despite this, a high virulence profile was demonstrated for A. salmonicida [5, 44].

It is clear that the identification of Aeromonas to species level has clinical significance since resistance to beta-lactams is species dependent and therefore accurate identification is important [53, 54]. Errors in the MALDI-TOF database should be corrected and considering the changes in the taxonomy of Aeromonas be regularly updated to incorporate spectra for newly discovered species [5, 20, 21, 22, 23, 24, 25]. As we commented previously, several studies demonstrated the utility of improving the database composition like occurred in the studies performed by Benagli et al. [38] that developed a reference library system using the SARAMIS™ software; Chen et al. [39] that generated a new model of database [34], or the one of Pérez-Sancho et al. [31] where the number of correct identifications increased after the addition of 14 new spectra. These results showed that with an improvement and update of the database, MALDI-TOF MS could be a promising method for the identification of Aeromonas species and highlight the potential of this system to be part of the routine microbiology laboratory workflow. However, the update of the database should be the responsibility of the company supplying the equipment and not of the user that does not have the resources or the time to assume this duty. The Biotyper software database is now in version 13, but includes the same 46 Aeromonas strains that versions i.e., v 3.0 and 3.1 used by Latif et al. [35] in 2015, which means that almost after 10 years and after many publications claiming for improvement, the providing companies are not reacting and this looks unacceptable.

4. Conclusions

MALDI-TOF MS is a fast and economical tool used in most clinical routine laboratories for the identification of bacteria. However, it is currently not possible to confidently distinguish all known Aeromonas species with this system. The database for this species is limited and redundant and, despite all the periodic updates, there have been no changes for aeromonads to include all the species described in this genus. These results call for an update of the commercial database considering the actual changes in the taxonomy of this genus and the discovery of new pathogenic species, and this should be done regularly by the provider incorporating spectra for newly characterized species.