Precise Species Identification for Acinetobacter: a Genome-Based Study with Description of Two Novel Acinetobacter Species

ABSTRACT The genus Acinetobacter comprises species with ecological significance and opportunistic pathogens and has a complicated taxonomy. Precise species identification is a foundation for understanding bacteria. In this study, we found and characterized two novel Acinetobacter species, namely, Acinetobacter tianfuensis sp. nov. and Acinetobacter rongchengensis sp. nov., based on phenotype examinations and genome analyses of the two strains WCHAc060012T and WCHAc060115T. The two strains had ≤89.69% (mean, 79.28% or 79.72%) average nucleotide identity (ANI) and ≤36.4% (mean, 20.89% or 22.19%) in silico DNA-DNA hybridization (isDDH) values compared with each other and all known Acinetobacter species. Both species can be differentiated from all hitherto known Acinetobacter species by a combination of phenotypic characteristics. We found that Acinetobacter pullorum B301T and Acinetobacter portensis AC 877T are actually the same species with 98.59% ANI and 90.4% isDDH values. We then applied the updated taxonomy to curate 3,956 Acinetobacter genomes in GenBank and found that 6% of Acinetobacter genomes (n = 234) are required to be corrected or updated. We identified 56 novel tentative Acinetobacter species, extending the number of Acinetobacter species to 144, including 68 with species names and 76 unnamed taxa. We also found that ANI and the average amino acid identity (AAI) values among type or reference strains of all Acinetobacter species and taxa are ≥76.97% and ≥66.5%, respectively, which are higher than the proposed cutoffs to define the genus boundary. This study highlights the complex taxonomy of Acinetobacter as a single genus and the paramount importance of precise species identification. The newly identified unnamed taxa warrant further studies. IMPORTANCE Acinetobacter species are widely distributed in nature and are of important ecological significance and clinical relevance. In this study, first, we significantly update the taxonomy of Acinetobacter by reporting two novel Acinetobacter species, namely, Acinetobacter tianfuensis and Acinetobacter rongchengensis, and by identifying Acinetobacter portensis as a synonym of Acinetobacter pullorum. Second, we curated Acinetobacter genome sequences deposited in GenBank (n = 3,956) using the updated taxonomy by correcting species assignations for 6% (n = 234) genomes and by assigning 94 (2.4%) to 56 previously unknown tentative species (taxa). Therefore, after curation, we further update the genus Acinetobacter to comprise 144 species, including 68 with species names and 76 unnamed taxa. Third, we addressed the question of whether such a large number of species should be divided in different genera and found that Acinetobacter is indeed a single genus. Our study significantly advanced the taxonomy of Acinetobacter, an important genus with science and health implications.

IMPORTANCE Acinetobacter species are widely distributed in nature and are of important ecological significance and clinical relevance. In this study, first, we significantly update the taxonomy of Acinetobacter by reporting two novel Acinetobacter species, namely, Acinetobacter tianfuensis and Acinetobacter rongchengensis, and by identifying Acinetobacter portensis as a synonym of Acinetobacter pullorum. Second, we curated Acinetobacter genome sequences deposited in GenBank (n = 3,956) using the updated taxonomy by correcting species assignations for 6% (n = 234) genomes and by assigning 94 (2.4%) to 56 previously unknown tentative species (taxa). Therefore, after curation, we further update the genus Acinetobacter to comprise 144 species, including 68 with species names and 76 unnamed taxa. Third, we addressed the question of whether such a large number of species should be divided in different genera and found that Acinetobacter is indeed a single genus. Our study significantly advanced the taxonomy of Acinetobacter, an important genus with science and health implications.

RESULTS
Identification of two novel Acinetobacter species, namely, Acinetobacter tianfuensis and Acinetobacter rongchengensis. Two Acinetobacter strains, namely, WCHAc060012 T and WCHAc060115 T , were recovered from hospital sewage using an Acinetobacter chromogenic agar plate in 2018. We obtained the nearly complete 16S rRNA gene sequences (1,352 bp) of the two strains using PCR with the universal primers 27F and 1492R (25) and Sanger sequencing as described previously (26) for preliminary species identification. Comparison of the 16S rRNA gene sequences in the EzBioCloud database (27) and the 16S rRNA gene sequence-based phylogenetic tree (see Fig. S1 in the supplemental material) revealed that the two strains indeed belonged to the genus Acinetobacter. Strains WCHAc060012 T and WCHAc060115 T had the highest identity of 16S rRNA gene sequences with Acinetobacter chengduensis WCHAc060005 T (98.96%; accession no. MK796535) and Acinetobacter chinensis WCHAc010005 T (98.05%; accession no. NR_165666), respectively. However, it is well known that analysis based on the 16S rRNA sequence is insufficient for accurate taxonomic assignment (28). We then compared partial rpoB sequences (861 bp) of the two strains with those of Acinetobacter type strains. The two strains were also distinct from all known Acinetobacter species and formed two evolutionary clades in the phylogenetic tree based on partial rpoB gene sequences (see Fig. S2 in the supplemental material). Strain WCHAc060012 T had the highest identity of the partial rpoB sequence with Acinetobacter wanghuae dk386 T (89.08%), while WCHAc060115 T had the highest identity with Acinetobacter piscicola KCTC 62134 T (95.23%).
To further explore their precise species assignations, the two strains were subjected to whole-genome sequencing using Illumina HiSeq X10 platform. For strain WCHAc060012 T , 6,401,206 reads and 1.92 giga-bases (Gb) were generated with an actual 549.9Â coverage, which were assembled into a 3.5-Mb draft genome sequence containing 116 contigs (N 50 , 77,978 bp) with a G1C content of 42.3% . For strain WCHAc060115 T , 5,479,547 reads and 1.64 Gb were generated with an actual 396.8 Â coverage, which were assembled into a 4.1-Mb draft genome sequence containing 248 contigs (N 50 , 68,539 bp) with a G1C content of 37.7%. We determined the average nucleotide identity (ANI) values between WCHAc060012 T and WCHAc060115 T and between the two strains and type strains of all Acinetobacter species. Compared with type strains of all Acinetobacter species, ANI values of WCHAc060012 T ranged from 77.09% (Acinetobacter puyangensis ANC 4466 T ) to 82.70% (Acinetobacter cumulans WCHAc060092 T ), while those of WCHAc060115 T ranged from 77.71% (A. puyangensis ANC 4466 T ) to 89.69% (Acinetobacter piscicola LW15 T ) ( Table 1). The ANI value between WCHAc060012 T and WCHAc060115 T was 79.46% (Table 1). These ANI values are well below the 95% to 96% threshold used to define bacterial species (29). We then performed in silico DNA-DNA hybridization (isDDH) analyses for WCHAc060012 T , WCHAc060115 T , and type strains of all Acinetobacter species. The isDDH values of WCHAc060012 T and type strains of all Acinetobacter species were 19.2% to 23.4%, while those of WCHAc060115 T and type strains of all Acinetobacter species were 20.0% to 36.4% (Table 1), which are far below the 70% cutoff used to define a species (30,31). The isDDH value between WCHAc060012 T and WCHAc060115 T was 21.7% (Table 1). Both ANI and isDDH analyses clearly indicate that the two strains represent two novel Acinetobacter species. In the phylogenomic tree based on core genes ( Fig. 1), WCHAc060012 T and WCHAc060115 T are most closely related to A. cumulans WCHAc060092 T and A. piscicola LW15 T , respectively.
The two novel Acinetobacter species may be able to be differentiated from other Acinetobacter species by a combination of phenotypic characteristics. The phenotypic characteristics tested using the genus-targeted set of physiological and metabolic tests are presented in the standard way used in previous nomenclatural proposals (32,33). The phenotypes for the two novel Acinetobacter species, together with those for all known Acinetobacter species with validly published names, are summarized in Data Set S1 in the supplemental material. For both strains, growth occurs at various pHs from 7 to 8 and the temperatures range 20 to 35°C. Strain WCHAc060012 T grows at 30°C in the presence of 0% to 3% (wt/vol) NaCl in tryptic soy broth (TSB), while WCHAc060115 T grows in 0% to 4% (wt/vol) NaCl. Both strains were positive for the catalase test but negative for the oxidase activity. Cells of the two strains are Gramnegative coccobacilli; strictly aerobic; nonsporogenous; incapable of swimming motility; and capable of growing on media such as tryptic soy agar (TSA), Luria-Bertani (LB) agar, BHI agar, and Müller-Hinton agar (all from Hopebio). Colonies are light yellow, circular, opaque, smooth, convex, with entire margins, and approximately 1.0 to 2.0 mm in diameter after 24 h of incubation at 30°C on BHI agar plates.
Phenotypic differences between the two novel Acinetobacter species and each of the known species with validly published names are indicated in Data Set S1. When considering only clearly positive or clearly negative results, the most useful combinations of characteristics for differentiating WCHAc060012 T from all known Acinetobacter species include growth on L-glutamate, D-malate, malonate, and phenylacetate but no growth on (Continued on next page) L-arabinose, L-arginine, azelate, and glutarate (Data Set S1). Strain WCHAc060115 T could be differentiated from all known Acinetobacter species by the combination of assimilation trans-aconitate, citrate (Simmons'), and L-tartrate but not b-alanine and 4-aminobutyrate (Data Set S1). We also identified antimicrobial resistance genes from genome sequences of the two strains (see Table S1 in the supplemental material). Both strains had genes mediating resistance to aminoglycosides, sulfonamides, and macrolides, while WCHAc060115 T also harbored two carbapenemase genes, namely, bla NDM-1 and bla OXA-58 , and WCHAc060012 T carried a tetracycline-resistant gene tet (39).
Acinetobacter pullorum Elnar et al. 2020 and Acinetobacter portensis Ana et al. 2020 are the same species. During the process of studying WCHAc060012 T and WCHAc060115 T , we also found a pair of synonyms, namely, A. pullorum and A. portensis. A. pullorum B301 T was isolated from raw chicken meat at a local market in Korea (14). It has been shown that A. pullorum B301 T is closely related to Acinetobacter celticus ANC 4603 T (14). Four A. portensis strains were isolated from raw meat samples in supermarkets in Porto, Portugal. A. portensis is also closely related to A. celticus ANC 4603 T (15). A comparison of the 16S rRNA gene sequences for the two type strains showed a 99.70% similarity. The draft genome sequence of A. pullorum B301 T (GenBank accession no. JAAARQ000000000) and that of A. portensis AC 877 T (GenBank accession no. LWRV00000000) have a 90.4% isDDH value and a 98.59% ANI value. Both ANI and isDDH analyses clearly indicate that the two species are actually the same species. In the phylogenomic tree, A. pullorum B301 T and A. portensis AC 877 T indeed cluster together ( Fig. 1, highlighted in red).
A comparison of the physiological and biochemical features of the two type strains shows phenotype coherence, which is summarized in Table S2 in the supplemental material. According to previous reports (14), A. pullorum and A. portensis are different in the acidification of D-glucose and utilization of b-alanine and D-glucose, which is likely due to intraspecies variability or assay conditions. Based on principles by the International Code of Nomenclature of Bacteria (12), A. pullorum has the priority of species name over A. portensis. We therefore propose that A. portensis (15) is a later heterotypic synonym of A. pullorum (14).
Curation of Acinetobacter genomes with the updated taxonomy. Based on the above findings, the valid species names of Acinetobacter should be updated to comprise 68 species at present ( Table 2). In addition, there are 20 tentative species designations of Acinetobacter (www.szu.cz/anemec/Classification.pdf) ( Table 2). We then applied the updated Acinetobacter taxonomy to curate the 5,997 Acinetobacter strains with genome sequences deposited in GenBank (accessed by 1 August 2020). Before curation, we performed a quality-control check for all of the genomes. Among the 5,997 genomes, 2,041 were discarded due to low quality defined by .300 contigs for individual genomes (n = 444), a ,50-kb N 50 value (n = 458), ,90% genome completeness (n = 20), genome contamination (n = 206), or genome heterogeneity (n = 913). We then used the remaining 3,956 genomes for precise species identification by both ANI and isDDH. Among the 3,956 Acinetobacter genomes, 3,777 were labeled with a known Acinetobacter species (n = 175), Acinetobacter genomosp. (n = 2), Acinetobacter calcoaceticus/Acinetobacter baumannii complex (n = 1), or uncultured Acinetobacter (n = 1) (Data Set S2), which were updated by our curation. Species were misidentified for 55 Acinetobacter genomes   Table S3 in the supplemental material). The 55 misidentified genomes include 13 labeled with A. baumannii but actually belonging to other Acinetobacter species and four of non-A. baumannii Acinetobacter species actually belonging to other closely related species (Table S3), while the remaining 38 genomes should be assigned to novel taxa (see below for details). Therefore, there were 234 genomes whose species identification needs to be corrected (n = 55) or updated (n = 179) according to the findings in this study (Data Set S2). After precise species identification, among the 3,956 Acinetobacter strains with genome sequences available, most (n = 3,124, 79.0%) belonged to A. baumannii, followed by Acinetobacter pittii (n = 174, 4.4%), Acinetobacter nosocomialis (n = 103, 2.6%), and Acinetobacter indicus (n = 68, 1.7%; Table 3). However, 94 (2.4%) strains could not be assigned to any known Acinetobacter species nor to any known tentative species designations (Data Set S3). Instead, the 94 strains could be assigned to 56 potentially novel Acinetobacter species, which are named taxon 40 to 95 here (Table 4 and Fig. 2), as Acinetobacter taxon 39 has been used before. Characterization of taxon 40 to 95 by phenotype methods is warranted to further establish their species status with proper species names under current International Code of Nomenclature of Prokaryotes (12).
Acinetobacter is indeed a single genus comprising 144 species at present. The identification of the 56 taxa also extends the number of Acinetobacter species to 144, including 68 with species names and 76 unnamed taxa. The large number of species raises the question whether Acinetobacter is indeed a single genus or actually should be divided into different genera. ANI values among type strains of all species and reference strains of all taxa of the genus Acinetobacter are $76.97% (76.97% to 95.98%) a Acinetobacter pakistanensis is a later synonym of Acinetobacter bohemicus (57). b Acinetobacter guangdongensis is a later synonym of Acinetobacter indicus (20). c Acinetobacter grimontii is a later synonym of Acinetobacter junii (21). d Acinetobacter dijkshoorniae is a later synonym of Acinetobacter lactucae (22). e Acinetobacter mesopotamicus is a later synonym of Acinetobacter lwoffii (19). f Acinetobacter portensis is a later synonym of Acinetobacter pullorum (this study). (see Data Set S4 in the supplemental material). The ANI values are higher than 72.50% to 73.70%, which has been proposed as the 95% confidence interval of the boundary to define a bacterial genus (34). To further verify the genus Acinetobacter, the average amino acid identity (AAI) values among type strains of all species and reference strains of all taxa of the genus Acinetobacter were also calculated, which are .66% (66.5% to 97.4%) (see Data Set S5 in the supplemental material). This is higher than the proposed cutoff of 65% AAI used to define a bacterial genus (34,35). Both ANI and AAI analyses suggest that all Acinetobacter species and unnamed taxa identified so far indeed belong to a single genus.

DISCUSSION
In this study, we first found and characterized two novel Acinetobacter species. We also found that A. pullorum and A. portensis are synonyms and then updated the taxonomy of the genus Acinetobacter. We applied the updated taxonomic assignments to curate genome sequences deposited in GenBank with the label of Acinetobacter and found that 6% (n = 234) of the 3,956 genomes with good quality need to be corrected or updated for species identification. We also identified 56 previously unknown tentative species designations, which further update the genus Acinetobacter to comprise 144 species, including 68 with species names and 76 taxa without species names. Such a large number of species raises the question whether Acinetobacter should be divided into multiple genera. Although the boundary of bacterial genera based on genome sequences is less established than that of species and requires more studies (9), our ANI and AAI analyses suggest that all Acinetobacter species indeed belong to a single genus. The mechanisms and factors driving the divergence of Acinetobacter to form the evolutionary trajectory and generate the remarkable species diversity and form have not been understood (36,37).
Along with many Acinetobacter species identified recently (11,(13)(14)(15)(16)(17)(18), the above findings highlight that Acinetobacter is a highly diverse and complex group (38). The species status of two novel Acinetobacter species, namely, A. tianfuensis and A. rongchengensis, was established by both genome-and phenotype-based methods. In addition to known species, there were 76 tentative novel Acinetobacter taxa, including 56 identified in this study. The identification of new taxa invites more studies on these tentative species by both genome-and phenotype-based methods to establish their species status and to propose appropriate species names under the current code for prokaryotes (12). Alternatively, it has also been proposed to create a new code that would use DNA sequences as the type material to rule the nomenclature of prokaryotes (39) or to establish placeholder species names using genome-based taxonomy (10). Indeed, there is an urgent need to find a solution to deal with the exploration of new taxonomic findings generated by genome sequencing (40).
In conclusion, we characterized and reported two novel Acinetobacter species, namely, A. tianfuensis and A. rongchengensis. A. tianfuensis may be distinguished from all other Acinetobacter species by its ability to grow on L-glutamate, D-malate, malonate, and phenylacetate but not grow on L-arabinose, L-arginine, azelate, and glutarate.
A. rongchengensis may be differentiated from all other Acinetobacter species by the combination of assimilation trans-aconitate, citrate (Simmons'), and L-tartrate but not  b-alanine and 4-aminobutyrate. We also found that A. portensis is a later heterotypic synonym of A. pullorum. We demonstrated that some Acinetobacter genome sequences deposited in GenBank are required to be corrected and identified 56 novel tentative Acinetobacter taxa, which warrant further phenotype-based characterizations.

MATERIALS AND METHODS
Strains and preliminary species identification. Hospital sewage (1 ml) was collected from the influx mainstream of the wastewater treatment plant at West China Hospital in June 2018, which was added in 10 ml nutrient broth (Oxoid, Basingstoke, UK) and was incubated overnight at 30°C with shaking. The culture suspension was diluted to 0.5 McFarland standard and was then further diluted to 1:100 with saline. A 100-ml aliquot was then streaked onto an Acinetobacter chromogenic agar plate (CHROMagar, Paris, France). The plate was then incubated at 30°C overnight. All isolates recovered from the plate were subjected to preliminary species identification by partial sequencing of the RNA polymerase b subunit-encoding rpoB gene using PCR and Sanger sequencing as described previously (5). Isolates with #98% identity of the 861-bp partial rpoB sequence (corresponding to nucleotide positions 2915 to 3775 of A. baumannii CIP 70.34 T ; accession no. DQ207471) to type strains of all known Acinetobacter species may belong to novel species and were characterized as described below. Two Acinetobacter isolates, namely, WCHAc060012 T and WCHAc060115 T , were recovered from the plate and had #98% identity of the 861-bp partial rpoB sequence to type strains of all known Acinetobacter species.
Analysis based on 16S rRNA and rpoB genes. Boiled lysates were used as the PCR template, and PCR amplicons were sequenced using the Sanger method (26). The nearly complete 16S rRNA gene sequences of WCHAc060012 T and WCHAc060115 T were obtained using PCR with universal primers 27F and 1492R (25). The 16S rRNA gene sequences of type strains of each Acinetobacter species were retrieved from their depositions in GenBank or from their whole-genome sequences. The longest common fragments of the 16S rRNA gene sequences (1,352 bp) were aligned using MAFFT v7.471 (41), and a maximum likelihood phylogenetic tree (42) based on the 1,352-bp sequences was inferred using RAxML v8.2.12 (43) with the general time reversible (GTR) model.
To further investigate the taxonomic position of WCHAc060012 T and WCHAc060115 T , 861-bp partial rpoB sequences of type strains of each Acinetobacter species were retrieved from their depositions in GenBank or from their whole-genome sequences. Sequence alignment and the construction of a maximum-likelihood phylogenetic tree were performed as described above.
Whole-genome sequencing of the two strains. Genomic DNA from an overnight culture of each of the two strains was prepared using the QIAamp DNA minikit (Qiagen, Hilden, Germany) and was then subjected to whole-genome sequencing using the HiSeq X10 sequencing platform (Illumina, San Diego, CA, USA) with an approximate 250Â coverage. Reads were de novo assembled into contigs using the program SPAdes v3.15.1 (44). Potential contaminations of WCHAc060012 T and WCHAc060115 T genomes were checked using CheckM v1.1.3 (45). Antimicrobial resistance genes were identified from genome sequences using the ABRicate program (https://github.com/tseemann/abricate) to query the ResFinder database 4.1 (https://cge.cbs.dtu.dk/services/ResFinder/).
Precise species identification and phylogenomic analysis of the two strains. Whole-genome sequences of type strains of all Acinetobacter species (Data Set S2) were retrieved from the NCBI database. Genome sequences of WCHAc060012 T and WCHAc060115 T were compared with those of type strains of Acinetobacter species using the average nucleotide identity based on BLAST (ANI) and in silico DNA-DNA hybridization (isDDH). ANI and isDDH values were calculated using the fastANI v1.32 (46) and genome-to-genome distance calculator (formula 2) (47) with the recommended parameters and/or default settings, respectively. A $96% ANI (31) or $70.0% isDDH (31,47) was used as the cutoff to define a bacterial species.
Phenotypic characterization for strains of two novel species. The metabolic and physiological properties were assessed using the standardized genus-targeted set of metabolic/physiological tests as described previously (5,32,53). The two strains were grown on brain heart infusion (BHI) agar (Oxoid) plates at 30°C overnight, and the colony morphology was observed by naked eyes. Cell morphology was visualized by light microscopy (CX21 microscope; Olympus, Japan). The Gram staining was carried out with a Gram staining kit (bioMérieux, Marcy l'Etoile, France). The cultivation temperature was 30°C unless indicated otherwise. Cell motility was tested in LB medium with 0.4% agar. Growth at various temperatures (20,25,32,35,37,41, and 44°C) was tested in 5-ml aliquots of BHI broth dispensed into tubes (16mm inner diameter) as described previously (5). Salt tolerance tests at different NaCl concentrations (0% to 10%, wt/vol, in increments of 1.0%) were performed in tryptic soy broth (TSB; Hopebio, Qingdao, China) after incubation for 2 days. Growth at pH 4.0 to 11.0 (at intervals of 1 pH unit, adjusted by adding HCl or NaOH) was examined in TSB for 2 days. The anaerobic growth was examined on a BHI agar plate, which was placed in an anaerobic bag (bioMérieux) at 30°C for 7 day (26). Aerobic acid production from glucose and gelatin hydrolysis was performed using the API 20NE system (bioMérieux), and the results were observed after 48 h. Hemolysis of sheep blood and utilization of citrate (Simmons') were examined according to methods described previously (5). The characteristics for the assimilation of the other carbon sources were determined using the basal mineral medium (54) supplemented with 0.1% (wt/vol) carbon source as described previously (5).
Curation of all available Acinetobacter genomes for precise species identification. All genome sequences labeled as Acinetobacter species in GenBank (n = 5,997, accessed by 1 August 2020) were retrieved. The assemblies, completeness, contamination, and heterogeneity of the genomes were evaluated using QUAST v5.0.2 (55) and CheckM v1.1.3 (45). Genome assemblies were discarded due to low quality defined by .300 contigs, a ,50-kb N 50 value, ,90% genome completeness, genome contamination indicated by $2 in CheckM, or none-zero genome heterogeneity value for individual genomes. ANI and isDDH values between each of the genomes and type strains of Acinetobacter genomes were calculated, using the fastANI v1.32 (46) and genome-to-genome distance calculator (formula 2) (47), respectively. A $96% ANI (31) or $70.0% isDDH (31,47) was used as the cutoff to define a bacterial species. AAI was calculated between each pair of genome sequences using CompareM v0.1.2 (56) with the recommended parameters.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. DATA SET S1, XLSX file, 0.03 MB.