 |
 |
|
|
|
|
|
|
| |||||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||||
Emended Description of Gardnerella Vaginalis and Description of Gardnerella Leopoldii Sp. Nov., Gardnerella Piotii Sp. Nov. And Gardnerella Swidsinskii Sp. Nov., With Delineation of 13 Genomic Species Within the Genus GardnerellaАвторы: Mario Vaneechoutte1?, Alexander Guschin2?, Leen Van Simaey1?, Yannick Gansemans3?, Filip Van Nieuwerburgh3?, Piet Cools4?Ссылка на первоисточник: INTERNATIONAL JOURNAL OF SYSTEMATIC AND EVOLUTIONARY MICROBIOLOGY Volume 69, Issue 3
AbstractWhole genome sequence analysis (digital DNA–DNA hybridization and average nucleotide identity) was carried out for 81 sequenced full genomes of the genus Gardnerella, including ten determined in this study, and indicated the existence of 13 genomic species, of which five consist of only one strain and of which only five contain more than four sequenced genomes. Furthermore, a collection of ten Gardnerella strains, representing the emended species G. vaginalis and the newly described species Gardnerella leopoldii, Gardnerella piotii and Gardnerella swidsinskii, was studied. Matrix-assisted laser desorption ionization time-of-flight MS analysis of the protein signatures identified specific peaks that can be used to differentiate these four species. Only strains of G. vaginalis produce ?-galactosidase. We emend the description of G. vaginalis (type strain ATCC 14018T=LMG 7832T=CCUG 3717T) and describe the novel species Gardnerella leopoldii sp. nov. (UGent 06.41T=LMG 30814T=CCUG 72425T), Gardnerella piotii sp. nov. (UGent 18.01T=LMG 30818T=CCUG 72427T) and Gardnerella swidsinskii sp. nov. (GS 9838-1T=LMG 30812T=CCUG 72429T).
Keywords Gardnerella vaginalis, Gardnerella piotii sp. nov., Gardnerella leopoldii sp. nov., Gardnerella swidsinskii sp. nov., whole genome sequencing, MALDI-TOF MS
Author Notes The GenBank/EMBL/DDBJ accession numbers for the complete 16S rRNA gene sequences reported here are MH898661 (Gardnerella vaginalis UGent 09.01), MH898662 (Gardnerella vaginalis UGent 09.17), MH898666 (Gardnerella vaginalis UGent 25.49), MH898660 (Gardnerella leopoldii UGent 06.41T), MH898665 (Gardnerella piotii UGent 21.28), MH898658 (Gardnerella swidsinskii GS9838-1T) and MH898659 (Gardnerella swidsinskii GS10234).
One supplementary figure and one supplementary table are available with the online version of this article.
AbbreviationsARDRA, amplified rDNA restriction analysis; MALDI-TOF MS, matrix-assisted laser desorption ionization time-of-flight mass spectrometry; RAPD, randomly amplified polymorphic DNA analysis; ANI, average nucleotide identity; BV, bacterial vaginosis; dDDH, digital DNA–DNA hybridization; PCA, principal components analysis; DA, discriminant analysis; MSP, main spectrum; CDC, cholesterol-dependent cytolysin.
IntroductionGardnerella vaginalis [1] was first described as Haemophilus vaginalis by Gardner and Dukes [2, 3]. Prior to Gardner and Dukes, Leopold [4] had indicated that the organism appeared to be Gram-negative on Gram staining, was Haemophilus-like, haemolytic on defibrinated rabbit blood agar, grew best at low oxygen pressure (micro-aerophilic), and could be isolated from the urine of men and cervix of women, whereby four men of nine couples for which the spouse was cervix-positive were also urine-positive.
Piot et al. [5] were the first to assess the diversity within G. vaginalis and developed a biotyping scheme. Ingianni et al. [6] were the first to establish the presence of at least three genotypes, on the basis of restriction with HpaII and TaqI of the amplified 16S rRNA gene (ARDRA), which was largely confirmed more recently, using ARDRA and rapid amplified polymorphic DNA analysis (RAPD) [7]. However, Pleckaityte et al. [8] and Schellenberg et al. [9] were able to establish only two ARDRA genotypes on the basis of TaqI restriction of the 16S rRNA gene, even though the latter publication included strains belonging to genotype 3, established previously [6, 7]. Interestingly, restriction analysis with eight different restriction enzymes of the (largely non-coding) 16S–23S intergenic spacer region, which is usually polymorphic, did not yield differences between the 34 strains tested by Ingianni et al. [6], indicating (unexpected) homogeneity of this region within the genus Gardnerella. Ahmed et al. [10] recognized the presence of four groups on the basis of neighbour-grouping analyses, using both distributed gene possession data and core gene allelic data. They showed that these groups were non-recombining clades with distinct gene pools, warranting their description as separate species. Balashov et al. [11] developed four PCR assays that could specifically identify representatives of these four groups. Using cpn60 gene sequence analysis, Schellenberg et al. [9] confirmed the presence of the four groups established previously by Ahmed et al. [10]. The existence of different groups (at the species level) within the ‘species’ Gardnerella vaginalis, i.e. within the genus Gardnerella, has been indicated by other studies [8, 12]. Furthermore, comparative genomic studies coupled with functional phenotypic characterization of Gardnerella strains have revealed differences in virulence factors between commensal and pathogenic G. vaginalis strains [13, 14].
Here, we analysed all 81 available full genome sequences by means of digital DNA-DNA hybridization (dDDH) and average nucleotide identity (ANI) and confirm the existence of at least 13 different separate taxa within the species Gardnerella vaginalis, although the taxa that result from our analysis do not correspond one to one to the groups delineated by Ahmed et al. [10], as we found two taxa each within groups 1, 2 and 4 of Ahmed et al. [10] and three within group 3 of Ahmed et al. [10]. Furthermore, we indicate how four of these taxa can be easily and unambiguously distinguished from each other by matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) MS, a routine method for identification of bacterial species in clinical laboratories. We propose emendation to the description of Gardnerella vaginalis (type strain ATCC 14018T) and the description of three new Gardnerella species, i.e. Gardnerella leopoldii sp. nov. (type strain UGent 06.41T), Gardnerella piotii sp. nov. (type strain UGent 18.01T) and Gardnerella swidsinskii sp. nov. (type strain GS 9838-1T).
Isolation and ecologyTable 1 lists the strains used for this study and their origins. All strains were isolated from vaginal swabs, all in Belgium during a longitudinal study [7], except strains GS 9838-1T and GS 10234, which were isolated in Russia, and strain ATCC 14018T, from the USA.
 Genome features
Genomic DNA was obtained for each strain by using the High Pure PCR Template Preparation Kit (Roche). The DNA was fragmented using a Covaris ultrasonicator and fragments ranging from 800 to 1000?bp were recovered on an E-gel (Thermo Fisher Scientific). Library construction was done using the NEBNext Ultra II kit (New England Biolabs). Finally, whole genome sequencing was carried out as a single-index paired-end 300 run on an Illumina MiSeq device.
Contaminating phiX-174 internal standard reads were removed using bowtie2 (v2.3.4). Adapter trimming was done using cutadapt (v1.15). A final quality check was done with FastQC (v0.11.7).
De novo genome assembly was done using SPAdes (v3.11.1) with standard settings for Illumina paired-end reads. Assemblies for strains UGent 21.28, UGent 09.01 and UGent 09.07 were repeated with the CLC Genomics Workbench (v10.0.1) to obtain a less fragmented assembly. This only succeeded for strain UGent 21.28. Scaffolds were removed if they were shorter than 1000 nt or when displaying an abnormal read coverage. Quality assessment of the assemblies was done with Quast (v4.6.2)
To evaluate the assignment of the Gardnerella strains to novel species, dDDH and ANI were calculated pairwise between the ten newly sequenced genomes and the already available genomes (of which QFNV01.1 was excluded as it appeared to be a duplicate of QFNU01.1, and PNGN01.1 that appeared to represent Lactobacillus vaginalis). Pairwise dDDH values were calculated using the online Genome-to-Genome Distance Calculator tool version 2.1 (http://ggdc.gbdp.org) using the recommended formula 2 and were reported with a 95?% confidence interval [15]. Pairwise ANI values were calculated with OAU (v1.2), as recommended at https://www.ezbiocloud.net/tools/ani [16]. A distance matrix of the 16S rRNA gene sequences was obtained using mega7 software.
Physiology and chemotaxonomy
MALDI-TOF MS Protein composition profiles of the ten strains were determined by means of MALDI-TOF MS as described previously [17]. Briefly, strains were cultured for 72?h on chocolate agar plates (Becton Dickinson) in 10?% CO2 at 37?°C. Ethanol/formic acid extraction and generation of eight spectra per strain were performed as described previously [17].
The raw spectra were exported as .txt files using the FlexAnalysis software (Bruker) and these files were imported into the BioNumerics software package 7.5 (Applied Biosystems) and processed as recommended in the software manual. After processing the raw spectra, the eight MALDI-TOF mass spectra obtained for each isolate were combined into one newly generated spectrum (composite mass spectrum or main spectrum, MSP) for this strain by calculating the average signal intensity for each data point of the eight technical replicates. Raw spectra with a similarity lower than 95?% of the average of the raw spectra were excluded from contribution to the MSP.
The spectra of the different strains were analysed by using different methods. First, two similar non-hierarchical methods were used to analyse the relationship between the spectra, i.e. a principal component analysis (PCA) and a discriminant analysis (DA). Both PCA and DA are powerful techniques that aim to reduce the complexity and redundancy of the MALDI-TOF spectral data set while simultaneously retaining the information present in the data. To do so, a new set of new variables, so-called principal components (PCs), are generated, which represent the variance in the data set. Spectra from strains are represented as single dots in a three-dimensional plot along these PC axes. As a starting point for PCA and DA, a peak matching table was generated using a constant tolerance of 1.9, a linear tolerance of 300 and a peak detection rate of 10?%.
Second, starting from the same peak matching table, a heat-map was created by generating a tree for both the strains and the peak classes using a UPGMA algorithm based on Spearman ranks correlations. A Mann–Whitney test was used to identify peaks that differed significantly between the different clades.
Third, a hierarchical cluster analysis was performed based on a similarity matrix calculated using the curve-based Pearson product-moment correlation coefficient as an input. The dendrogram was reconstructed using the complete linkage clustering algorithm. Fourth, jackknife analysis was performed using both average and maximum similarities to quantify rates of correct classification of strain spectra into the different taxa and to estimate the internal stability of the different taxa. Biochemical activity Sialidase activity testing was carried out as described previously [7]. ?-Galactosidase activity was tested using ONPG tablets (Sigma Aldrich), as recommended by the manufacturer. Results and discussion
Clinical relevance of the genus Gardnerella
Gardnerella vaginalis can be recovered from the urogenital tract of men and women [2–4, 18] and has been isolated from the rectum of women [19]. It can be present as a commensal as well as the dominant micro-organism in the vaginal microbiome of women suffering from bacterial vaginosis (BV) [10, 20, 21]. BV is asymptomatic in half of cases, but can be associated with malodorous vaginal discharge, increased vaginal pH and the presence of clue cells. Swidsinski et al. [22], using fluorescence in situ hybridization specific for G. vaginalis, were the first to show that this species is able to form biofilms on vaginal epithelium in women with BV, explaining the nature of clue cells, i.e. cells covered with biofilm, predominantly formed by G. vaginalis, and as such providing convincing evidence for the aetiological role of G. vaginalis in this condition.
BV, whether or not symptomatic, is associated with increased risks for preterm delivery, intrauterine growth retardation, pelvic inflammatory disease, postpartum endometritis and human immunodeficiency virus infection [10, 20, 21]. The potentially pathogenic role of G. vaginalis has been disputed, possibly also due to limited taxonomic fine-tuning, whereby different species that may confer distinct ecological or pathological properties have been considered as a single species. Indeed, recent findings have shown that Gardnerella groups 1 and 2 (as defined by Ahmed et al. [10]) were significantly associated with BV as defined by Nugent score, whereas Gardnerella groups 3 and 4 were not [23]. Furthermore, Gardnerella groups (as defined by Ahmed et al. [10]) could best be characterized on the basis of sialidase activity and vaginolysin production [14]. Taxonomic position of the genus Gardnerella
Gardnerella was initially considered as a Gram-negative organism, as reflected by its original designation Haemophilus vaginalis [3]. This was still the case at the time of its reclassification as Gardnerella vaginalis [1], although its Gram-positive characteristics had been outlined partially and the name Corynebacterium vaginale had been proposed [e.g. 24]. However, its taxonomic status as belonging to Gram-positive bacteria is clear, based on electron microscopy analysis of the cell wall structure [25, 26], the absence of diaminopimelic acid (DAP) from the peptidoglycan layer [27, 28] and of lipopolysaccharide from the cell wall [26]. Also, fructose 6-phosphate phosphoketolase activity, assumed to be specific for the genus Bifidobacterium [29], indicates a relationship to the Gram-positive bifidobacteria, which is further substantiated by 16S rRNA gene sequencing, indicating highest sequence similarity of 93.1?% to Bifidobacterium bifidum, the type species of the genus [30], or to B. coryneforme and B. minimum [31].
However, Gardnerella is clearly different from bifidobacteria, due to major differences in the G+C content of the genomic DNA, i.e. 38.0–43.4?mol% (Fig. 1), as compared to 55–67?mol% for Bifidobacterium [29]. On the other hand, Gardnerella vaginalis has been found to consistently branch in between different Bifidobacterium species [32, 33]. Finally, vaginolysin, a haemolysin that belongs to the cholesterol-dependent cytolysins (CDCs) which thus far have been described mostly from Gram-positive organisms [34, 35], is encoded by most strains of the genus Gardnerella [36]. Interestingly, the Lactobacillus iners genome codes for inerolysin, a CDC with a significant primary sequence similarity to vaginolysin (68.4?%) [37].
 It has been recognized that the unusual thinness of the cell walls of G. vaginalis (8–12?nm) accounts for the Gram-variable staining of Gardnerella [26, 27].
16S rRNA gene phylogeny
Pairwise comparison of the 16S rRNA genes of the 81 genomes indicated that none of these sequences showed less than 98.5?% similarity (data not shown). Furthermore, we could not detect signature sequences that could be used to identify individual species. As such, the Gardnerella species cannot be delineated on the basis of the 16S rRNA gene.
PCR assays specific for the groups described by Ahmed et al. [10] Ahmed et al. [10] carried out full genome sequence analysis and established four groups within the genus Gardnerella. Subsequently, Balashov et al. [11] developed four different primer/probe quantitative PCR assays, each one being specific for one of the groups described by Ahmed et al. [10], based on the comparison of full genomes (one genome per group). Finally, Schellenberg et al. [9] were able to amplify the specific gene for each group, using the same primers as used by Balashov et al. [11] (without probes) in a simple PCR format. We found that Gardnerella genomic species 1 and 2 (this study; numbering: see Fig. 1) were recognized by the group 1 assay of Balashov et al. [11], Gardnerella genomic species 3 and 4 (this study) by the group 2 assay, Gardnerella genomic species 8, 9 and 10 (this study) by the group 3 assay, and Gardnerella genomic species 5 and 6 (this study) by the group 4 assay. Whole genome sequence studies
Whole genome sequence analysis (dDDH and ANI) of 81 full genome sequences, including ten sequences that were determined in this study, is outlined in Fig. 1.
According to the recommended cut-offs for species delineation of 70 % for dDDH [38] and 96 % for ANI [39], we established a total of 13 species, of which five consist of only one strain (genome), whereby dDDH results were in excellent correspondence with the delineation obtained with ANI.
Ahmed et al. [10] concluded that G. vaginalis appears to include four non-recombining groups of organisms with distinct gene pools and genomic properties and with greatly expanded core gene content. They also concluded that these four different groups had their own characteristic genome size and G+C content, which could not be confirmed in our study.
Other studies largely confirmed the existence of these four groups [9, 11, 12, 40], although the designation of groups/clades and of strains used differs sometimes substantially between different research groups.
Here, we studied a collection of ten Gardnerella strains by comparing their fully sequenced genomes with the already published genomes, by means of dDDH and ANI (Fig. 1). We found that group 1 as described by Ahmed et al. [10] contains two species, of which we describe one as G. vaginalis. Their group 2 contains two species as well, of which we describe one as G. piotii. Their group 3 contains three species, and finally their group 4 contains two species, which we describe here as G. leopoldii and G. swidsinskii. The different conclusions obtained in comparison with those reached by Ahmed et al. [10] might be explained by the fact that their analysis was based upon visual inspection of a phylogenetic tree reconstructed from core gene allelic data, whereas we based our conclusions on full genome similarity statistics (ANI and DDH), which are widely used and accepted as the gold standard for prokaryotic species delineation [15, 39]. Based on comparative genomics of the different Gardnerella (genomic) species, specific primers and probes should be designed that allow the unambiguous identification of the different (genomic) species.
Phenotypic characterization
The biochemical activity of the genus Gardnerella, as has been determined in different studies [1, 30, 41–44], is summarized in Table S1 (available in the online version of this article). We found β-galactosidase activity to be present only in strains of G. vaginalis. All strains of G. piotii tested (data presented only for the two strains included in this study) as well one out of four strains of G. vaginalis, were positive for sialidase activity. Janulaitiene et al. [14] showed that all strains of group 4 (as defined by Ahmed et al. [10]) were negative for sialidase activity. This is in line with our finding that all strains of G. leopoldii and G. swidsinskii are negative for sialidase activity, because, according to the system proposed by Ahmed et al. [10], both species are classified as belonging to group 4.
MALDI-TOF protein profiling
Protein profiling was carried out for the ten strains included in this study, on a Bruker MALDI-TOF device. A heatmap of the MALDI-TOF peak clustering is shown in Fig. S1. Peaks that are significantly different between the four species (P<0.05) and that belong to the 10 % highest peak intensities are shown in Table 2.
 DA (Fig. 2) and PCA (data not shown) clearly delineate the four proposed species. Furthermore, jackknife group analysis confirmed the internal stability of the different taxa (100 % correct group assignments, data not shown). All analyses of the MALDI-TOF protein profiles clearly support the existence of the four species.
 The description of the genus is in agreement with that provided by Greenwood & Pickett 1980 [1] for the genus Gardnerella, emended as follows.
DNA G+C content ranges between 38.0 and 43.4 mol%. dDDH and ANI indicate that Gardnerella consists of at least 13 genomic species, of which four are described in this study. The mean fatty acid composition consists of 6.9±1.5 % of 14 : 0, 3.0±0.9 % of 16 : 1 ω7c, 36.6±2.9 % of 16 : 0, 4.8±1.3 % of 18 : 2 ω6,9c, 35.2±2.2 % of 18 : 1 ω9c, 2.5±2.0 % of summed feature 7 (comprising 18 : 1 ω7c, 18 : 1 ω9t or 18 : 1 ω12t, or any combination of these isomers) and 9.2±1.4 % of 18 : 0 [27].
Emended description of Gardnerella vaginalis (Gardner and Dukes 1955) Greenwood and Pickett 1980
The description remains as given by Gardner and Dukes 1955 [3] and Greenwood and Pickett 1980 [1] with the following emendation. Cells are Gram-negative to Gram-variable coccobacilli, 0.5×1.5 µm in size. Growth occurs on chocolate agar under CO2-enriched conditions (5 %) after 2–5 days. Colonies are pinpoint white to greyish with a smooth surface. β-Galactosidase-positive. Can be unambiguously differentiated from other species described here by means of MALDI-TOF, dDDH and ANI. The DNA G+C content ranges from 41.0 to 42.8 mol%.
The type strain is ATCC 14018T (=Dukes 594T=DSM 4944T=CCUG 3717T=JCM 11026T=NCTC 10915T=NCTC 10287T). The genome of the type strain has a size of 1.66 Mb and a DNA G+C content of 41.72 mol%. The whole genome sequence of the type strain is available from NCBI under accession number QJUZ00000000.
Other strains, not included in this study: ATCC 14019, 41V.
Gardnerella leopoldii (le.o.pol’di.i. N.L. gen. masc. n. leopoldii of Leopold, named after Sidney Leopold, who described the first strains of the species).
Cells are Gram-negative to Gram-variable coccobacilli, 0.5×1.5 µm in size. Growth occurs on chocolate agar under CO2-enriched conditions (5 %) after 2–5 days. Colonies are pinpoint white to greyish with a smooth surface. Negative for sialidase activity and β-galactosidase. Can be unambiguously differentiated from other Gardnerella species described here by means of MALDI-TOF, DDH and ANI. The DNA G+C content ranges from 41.9 to 43.2 mol%.
The type strain is UGent 06.41T (=LMG 30814T=CCUG 72425T), which was isolated from a human vagina in Ghent, Belgium, in October 2012. The G+C content of the type strain is 41.9 mol%. The whole genome sequence of the type strain is available from NCBI under accession number CP029984.
Other strains, not included in this study: AMD, 409-05.
Gardnerella piotii (pi.o’ti.i. N.L. gen. masc. n. piotii of Piot, named after Peter Piot, a Belgian microbiologist who first attempted to distinguish different groups within G. vaginalis).
Cells are Gram-negative to Gram-variable coccobacilli, 0.5×1.5 µm in size. Growth occurs on chocolate agar under CO2-enriched conditions (5 %) after 2–5 days. Colonies are pinpoint white to greyish with a smooth surface. Postive for sialidase activity and negative for β-galactosidase activity. Can be unambiguously differentiated from other Gardnerella species described here by means of MALDI-TOF MS, DDH and ANI. The DNA G+C content ranges from 41.1 to 42.3 mol%.
The type strain is UGent 18.01T (=LMG 30818T=CCUG 72427T), which was isolated from a human vagina in Ghent, Belgium, in October 2012. The G+C content of the type strain is 42.3 mol%. The whole genome sequence of the type strain is available from NCBI under accession number QJUV00000000.
Other strains, not included in this study: N153/00703C2mash.
Gardnerella swidsinskii (swid.sin’ski.i. N.L. gen. masc. n. swidsinskii of Swidsinski, named after Alexander Swidsinski, a German microbiologist, who recognized that clue cells in Gram stains of vaginal swabs of women with BV are indicative for biofilm formation by Gardnerella).
Cells are Gram-negative to Gram-variable coccobacilli, 0.5×1.5 µm in size. Growth occurs on chocolate agar under CO2-enriched conditions (5 %) after 2–5 days. Colonies are pinpoint white to greyish with a smooth surface. Negative for sialidase activity and β-galactosidase. Can be unambiguously differentiated from other Gardnerella species described here by means of MALDI-TOF, DDH and ANI. The DNA G+C content ranges from 41.4 to 42.3 mol%.
The type strain is GS 9838-1T (=LMG 30812T=CCUG 72429T), which was isolated from a human vagina in St.-Petersburg, Russia in October 2008. The G+C content of the type strain is 41.0 mol%. The whole genome sequence of the type strain is available from NCBI under accession number QJVB00000000.
Funding information
The authors received no specific grant from any funding agency.
Acknowledgements
We thank Anna Krysanova, MD, and Alevtina Savicheva, MD, of the Laboratory of Microbiology of the D. O. Ott Research Institute of Obstetrics, Gynecology and Reproductology, St. Petersburg, Russia, for providing the G. swidsinskii strains.
Conflicts of interest
The authors declare that there are no conflicts of interest.
© 2019 IUMS
| |||||||||||||||||||||||||||||||||||||
