Genotypic comparison ofPantoea agglomeransplant and clinical strains
© Rezzonico et al. 2009
Received: 29 April 2009
Accepted: 22 September 2009
Published: 22 September 2009
Skip to main content
© Rezzonico et al. 2009
Received: 29 April 2009
Accepted: 22 September 2009
Published: 22 September 2009
Pantoea agglomeransstrains are among the most promising biocontrol agents for a variety of bacterial and fungal plant diseases, particularly fire blight of apple and pear. However, commercial registration ofP. agglomeransbiocontrol products is hampered because this species is currently listed as a biosafety level 2 (BL2) organism due to clinical reports as an opportunistic human pathogen. This study compares plant-origin and clinical strains in a search for discriminating genotypic/phenotypic markers using multi-locus phylogenetic analysis and fluorescent amplified fragment length polymorphisms (fAFLP) fingerprinting.
Majority of the clinical isolates from culture collections were found to be improperly designated asP. agglomeransafter sequence analysis. The frequent taxonomic rearrangements underwent by theEnterobacter agglomerans/Erwinia herbicolacomplex may be a major problem in assessing clinical associations withinP. agglomerans. In theP. agglomerans sensu stricto(in the stricter sense) group, there was no discrete clustering of clinical/biocontrol strains and no marker was identified that was uniquely associated to clinical strains. A putative biocontrol-specific fAFLP marker was identified only in biocontrol strains. The partial ORF located in this band corresponded to an ABC transporter that was found in allP. agglomeransstrains.
Taxonomic mischaracterization was identified as a major problem withP. agglomerans, and current techniques removed a majority of clinical strains from this species. Although clear discrimination betweenP. agglomeransplant and clinical strains was not obtained with phylogenetic analysis, a single marker characteristic of biocontrol strains was identified which may be of use in strain biosafety determinations. In addition, the lack of Koch's postulate fulfilment, rare retention of clinical strains for subsequent confirmation, and the polymicrobial nature ofP. agglomeransclinical reports should be considered in biosafety assessment of beneficial strains in this species.
Pantoea agglomerans(Beijerinck 1888) comb. nov. , formerlyEnterobacter agglomerans (Beijerinck 1888) Ewing and Fife (1972),Erwinia herbicola(Löhnis 1911) Dye 1964 orErwinia milletiae(Kawakami and Yoshida 1920) Magrou 1937, is a Gram-negative bacterium that belongs to the family of Enterobacteriaceae.P. agglomeransis primarily a plant epiphyte [2–4] commonly found in diverse ecological niches including aquatic environments, soil or sediments [5–7]. Several strains ofP. agglomeransare sold as commercial biological control agents (BCAs) against the fire blight pathogen [Erwinia amylovora(Burrill 1882) Winslow et al. 1999] on apple and pear trees [8,9].P. agglomeransstrains are effective against other bacterioses, such as basal kernel blight of barley  and post-harvest fungal diseases of pome fruits [11–14]. Three commercialP. agglomeransstrains have recently been registered for biocontrol of fire blight in New Zealand (BlossomBless™ strain P10c ), in the United States and in Canada (BlightBan C9-1™ strain C9-1 ; Bloomtime™ strain E325 ). The primary mode of action is competitive exclusion which involves the occupation of sites otherwise colonized by the pathogen, but for some strains reports also indicate the contribution of different antibiotics like herbicolins  pantocins [18–21], putatively phenazine , and other unknown compounds .
Despite efficacy trials in commercial orchards demonstrating the potential ofP. agglomeransbiocontrol formulations as an alternative plant protection tool and their approval in the United States by the Environmental Protection Agency (EPA) as microbial pesticideshttp://www.epa.gov/fedrgstr/EPA-PEST/2006/September/Day-20/p8005.htm, registration efforts in Europe are hindered by biosafety concerns stemming from clinical reports that identify strains ofP. agglomeransas opportunistic human pathogens, and have resulted in the current classification of this species as a biosafety level 2 (BL-2) organism in Europe [23–27]. Biosafety classification differs among countries; in the European Union, Directive 2000/54/EC includes "Enterobacterspp." in the list of microorganisms that are currently classified as a biosafety level 2 (BL-2), while the German "Technische Regeln für Biologische Arbeitsstoffe", TRBA 466 and Swiss regulationshttp://www.bafu.admin.ch/publikationen/publikation/00594/index.html?lang=demore explicitly identifyP. agglomeransand its synonyms in BL-2. Several strains maintained in culture collections throughout the world and the type strainP. agglomeransLMG 1286T(= CDC 1461-61T= NCTC 9381T= ICMP 3435T= ATCC 27155T) itself are listed as clinical isolates . Confirmed pathogenicity of this species is difficult to ascertain, since clinical reports involvingP. agglomeransare typically of polymicrobial nature, often involve patients that are already affected by diseases of other origin, lack Koch's postulate fulfillment or any pathogenicity confirmation, and diagnostic isolates are rarely conserved for confirmatory analysis .
There has been insufficient investigations as to whether agriculturally beneficial isolates are distinct from clinical isolates or harbor potential pathogenic determinants that would justify current biosafety restrictions. Comparative analysis of biocontrol and clinical strains, or examination of virulence factors has been performed previously for other 'Jekyll-Hyde' biocontrol species reported to contain opportunistic pathogenic isolates, such asPseudomonas aeruginosa[28–30],Serratia marcescens and theBurkholderia cepaciacomplex [32,33]. Similar comparisons have not been performed forP. agglomerans, leaving a gap in knowledge critical to regulatory authorities.
The aim of our study was to perform a polyphasic genotypic and phenotypic analysis ofP. agglomeransisolates of diverse origin in order to understand whether clinical and biocontrol (environmental) isolates can be distinguished and have undergone a discrete evolution that would indicate specialization towards human pathogenicity or an epiphytic lifestyle. The taxonomy of a collection of clinical and plant isolates was assessed using fluorescent amplified fragment length polymorphism (fAFLP) analysis of total genomic DNA and sequence analyses of specific genes (such as 16S rDNA generrs,gyrBencoding DNA gyrase subunit B, and theP. agglomeransquorum-sensing regulatory genespagRIencoding homoserine lactone receptor and synthase) . The fAFLP analysis was used as well to search for random molecular markers that could serve as a simple and rapid discriminatory marker for clinical and biocontrol strains. Additionally, we examined the distribution of some phenotypic and genotypic traits among strains that may reflect adaptation to the different lifestyles proposed forP. agglomerans, such as growth at 37°C for clinical isolates, presence of pantocin A genes or sorbitol utilization for biocontrol strains, and presence of type III secretion system (T3SS) for plant pathogenic pathovars.
Thirty-two clinical isolates designated asP. agglomerans,E. agglomerans,E. herbicolaorPantoeaspp. were obtained from the American Type Culture Collection (ATCC,http://www.atcc.org/), the Belgian Coordinated Collection of Microorganisms (BCCM/LMG,http://bccm.belspo.be), the Institut Pasteur Collection (CIP,http://www.crbip.pasteur.fr/), the Spanish Type Culture Collection (CECT,http://www.cect.org/) or received from the Hospital de la Santa Crei Sant Pau (Barcelona, Spain) and the Istituto Cantonale di Microbiologia (ICM, Bellinzona, Switzerland). ElevenP. agglomeransstrains with established biocontrol activity obtained from several sources (including the three currently registered commercial strains), twenty environmental isolates and three phytopathogenic strains, together with representative strains of otherPantoeaspecies and closely related genera such asErwinia,PectobacteriumandBrenneria, were included in the study for comparison (see Additional file 1 - Table S1).
PCR primers used for gene amplification and sequencing.
PCR amplicons were purified from the PCR mix by washing twice with 50 μl of double-distilled water (ddH2O) on a MultiScreen PCR Plate (Millipore, Molsheim, France), resuspended in 30 μl of ddH2O, and quantified spectrophotometrically as described above. The cycle-sequencing reaction was performed with 20-40 ng of purified PCR product using the ABI PRISM BigDye Terminators v1.1 Cycle Sequencing Kit (Applied Biosystems, Foster City, CA, U.S.A.) according to the manufacturer instructions employing the same primers used for PCR amplification. For 16S rRNA gene sequencing, additional primers 16S-609R and 16S-533R (Table1) were used to obtain complete coverage of the amplicon. Cycle sequencing products were cleaned through water-swelled Sephadex G-50 columns (Amersham Biosciences, Uppsala, Sweden) on MultiScreen HV plates (Millipore) and sequenced on an ABI PRISM 3100 Genetic Analyzer. Obtained sequences were assembled using the Sequencher software (version 4.0.5; Gene Codes Corporation, Ann Arbor, MI, U.S.A.).
Phylogenetic trees were generated on the basis of partial 16S rDNA,gyrBandpagRIsequences without choosing any outgroup. DNA sequences were aligned with ClustalW . Sites presenting alignment gaps were excluded from analysis. The Molecular Evolutionary Genetics Analysis (MEGA) program version 4.0  was used to calculate evolutionary distances and to infer trees based on the Minimum Evolution (ME) method using the Maximum Composite Likelihood (MCL) formula. Nodal robustness of the inferred trees was assessed by 1000-bootstrap replicates.
For those strains received asE. agglomerans,P. agglomeransorPantoeaspp. from international culture collections but not clustering withP. agglomeransin the 16S rDNA andgyrBtrees, identification was sought by blasting the obtained nucleotide sequences in the NCBI database. Since the best hits often led to poorly characterized or obviously misdentified bacteria only the best match with a secure identification was retained. Confidence of secure identifications was based either on relatedness to theP. agglomeranstype strain or position in the BLAST distance tree. In order to be considered trustworthy, obtained hits were required to be flanked by sequences of representatives of the same species and not be part of a clade containing strains from related species with dissimilar identification.
Primers combinations and number of different peak positions generated used in the selective amplification step.
Raw data collected from the ABI 3130XL sequencer were analyzed using the GeneMapper v4.0 software (Applied Biosystems). To remove noise, only peaks with an absolute intensity greater than 200 (combinations CGC, TG) or 300 (combinations GG, AT) were retained for final analysis. The fAFLP profiles were converted into a binary matrix of presence/absence of each peak and this data was used to construct a UPGMA (Unweighted Pair Group Method with Arithmetic mean) dendrogram using the MEGA software. Nodal robustness of the inferred trees was assessed by 1000-bootstrap replicates. The fAFLP patterns of all strains were analyzed in order to identify peaks that may be distinctive of either biocontrol or clinical isolates in agreement with the classification shown in Additional file 1 - Table S1.
Growth ability ofP. agglomeransstrains on sorbitol was studied using 200-μl microcultures in 100-well Bioscreen C MBR system honeycomb plates (well volume 400 μl) at 24°C with regular shaking at 15-min intervals in M9 minimal medium containing 10 mM sorbitol as sole carbon source. All strains were grown overnight in LB, collected by centrifugation, and washed twice with sterile 0.9% NaCl before being inoculated in M9 at an initial OD600of about 0.02. Growth curves were measured in triplicates by periodically quantifying the absorbance through a 420- to 580-nm wide band filter (OD420-580 nm) using a Bioscreen C MBR system (Growth Curves Oy, Helsinki, Finland).
whereN 0andN t represent absorbance measured at two consecutive time points and Δtis the time interval (i.e., 1 h) between the two measurements. The highest optical density, the maximal growth rate, as well as the time needed to reach the latter value were recorded for each strain. A comparison of these parameters was performed among the average values obtained for clinical, biocontrol or plant-pathogenicP. agglomeransstrains. Correlations between OD420-580 nmmeasured in the Bioscreen C MBR system and number of colony forming unit (CFU) was estimated for representative strains by dilution plating on LB agar.
The accession numbers for the sequences produced for this study are: 16S rRNA gene [GenBank: FJ611802-FJ611887];gyrBgene [GenBank: FJ617346-FJ617427];hrcNgene [GenBank: FJ617428-FJ617436];pagRIgenes [GenBank: FJ656221-FJ656252].
With the exception ofpagRI, for which they are shown directly in the corresponding figure, accession numbers and other sources of reference sequences not obtained in this work are indicated below.Complete genomes:C. koseriATCC BAA-895 [NCBI: NC_009792],E. amylovoraEa273http://www.sanger.ac.uk/Projects/E_amylovora/,E. coliK-12 MG1655 [NCBI: NC_000913],Enterobactersp. 638 [NCBI: NC_009436],E. tasmaniensisEt1/99 [NCBI: NC_010694],K. pneumoniae342 [NCBI: NC_011283],P. stewartiisubsp.indologenesDC283http://www.hgsc.bcm.tmc.edu/microbial-detail.xsp?project_id=125.16S rRNA gene:E. cloacaeATCC 13047T[GenBank: AJ251469],E. sakazakiiATCC 51329 [GenBank: AY752937],Pantoea sp.LMG 2558 [GenBank: EF688010],Pantoea sp.LMG 2781 [GenBank: EU216736],Pantoea sp.LMG 24198 [GenBank: EF688009],Pantoea sp.LMG 24199 [GenBank: EF688012],Pantoea sp.LMG 24200 [GenBank: EF688011],Pantoea sp.LMG 24534 [GenBank: EU216737],P. terreaLMG 22051T[GenBank: EF688007],S. entericasvtyphiCT18 [NCBI: NC_003198].gyrB gene:E. cloacaeATCC 13047T[GenBank:EU643470],E. sakazakiiATCC 51329 [GenBank:AY370844],Pantoeasp. BD502 [GenBank: EF988786],Pantoeasp. BCC757 [GenBank: EF988776],Pantoea sp.LMG 2558 [GenBank: EF988812],Pantoea sp.LMG 2781 [GenBank:EU145271],Pantoea sp.LMG 24196 [GenBank: EF988758],Pantoea sp.LMG 24199 [GenBank: EF988768],Pantoea sp.LMG 24200 [GenBank: EF988770],Pantoea sp.LMG 24202 [GenBank: EF988778],Pantoea sp.LMG 24534 [GenBank: EU145269],P. terreaLMG 22051T[GenBank:EF988804],S. entericasvtyphiCT18 [NCBI: NC_003198].
Sequence analysis ofgyrBrevealed that 26 of the 32 clinical isolates obtained from international culture collections asP. agglomerans,E. agglomeransorPantoeaspp. did not justifiably belong toP. agglomerans, but clustered distant from type strain LMG 1286T. Based on genotypic similarity, these strains belonged either to otherPantoeaspp. or other Enterobacteriaceae genera. In contrast, classification of biocontrol strains was more precise than for presumptively clinical strains, and all of these could be identified unequivocally asP. agglomerans sensu stricto(Figure2).
Congruence between phylogenies derived fromrrs(Figure1) andgyrB(Figure2) gene sequences was imperfect. Analysis using 16S rDNA enabled only limited separation of strains within eachPantoeaspp., whereas analysis usinggyrBsequences revealed higher variability and enabled finer resolution of distinct branches with some strains clustering alongsideP. agglomeransLMG 1286Tin therrstree. ThegyrBclades corresponded largely to the MLST-groups recently defined by Brady et al.  forPantoeaspp. (Figure2).
Four strains (EM13cb, EM17cb, ATCC 29001 and SC-1) that grouped with representative strains ofPantoeaMLST-groups C, D and F in therrstree clearly diverged usinggyrBsequences. Clinical isolate EM13cb and cotton pathogen SC-1 clustered with LMG 2558 (MLST-group C), while two other clinical isolates, EM17cb and ATCC 29001, clustered with LMG 24534 (MLST-group F) and LMG 5343 (MLST-group E) using eitherrrsorgyrB. In contrast, LMG 5343, LMG 24198 (MLST-group B) and LMG 24199 (MLST-group A), all clustered unexpectedly withP. agglomeransin therrstree (Figure1) but were clearly divergent usinggyrB. This demonstrated the resolution limits of 16S rDNA sequence analysis amongPantoeaspp. BothrrsandgyrBsequences assigned two additional presumptive-clinical strains (ATCC 27995 and ATCC 27996) to the related speciesPantoea ananatis(Serrano 1928) Mergaert et al. 1993, while most of the other human isolates (including representatives from Brenner's biotypes VII-XII ) clustered far from theP. agglomerans sensu strictogroup and could be roughly assigned toErwiniaorEnterobacterspp. (see Additional file 2 - Table S2) based on BLAST comparison. Indicative of the uncertainty surrounding identification of this species, the BLAST best-hits list often included isolates clearly misidentified asP. agglomeransorPantoeaspp. Specifically, strains with extremely low sequence similarity with theP. agglomeranstype strain LMG 1286T(well below 90%) were interspersed among better characterized Enterobacteriaceae.
Two plant epiphytic isolates (EPS486 and EPS595) from apple and pear trees and two wheat root isolates (P8SAA and P10QLC) formed a clearly distinct group in both trees which is not embraced by any of the MLST-groups defined forPantoea, and are considered representative of a single newPantoeaspp.
Utilization of sorbitol byP. agglomeransas a sole carbon source was restricted to only a few biocontrol isolates, indicating this as an important feature for phytopathogen antagonism. In addition to the commercial biocontrol strain C9-1, which has two plasmid-encoded sorbitol-utilization operons , only the biocontrol strains Eh252 and P10c were able to efficiently metabolize sorbitol. StrainP. ananatisLMG 2665T, included as a positive control for sorbitol utilization, andP. agglomeransstrains C9-1 and Eh252 gave absorbance readings that indicated a growth after 6-8 h from inoculation, while the lag-phase of P10c was protracted up to 24 h, suggesting that a certain signal may be required for this strain before C6-sugar metabolism is triggered.
Pantocin A biosynthetic genes were amplified in just four biocontrol isolates (i.e., C9-1, Eh252, Eh318 and CPA-2) and one clinical strain LMG 5343. Genome sequence analysis of C9-1 has revealed that in this strain the gene cluster coding for pantocine production is situated on a low-GC genomic island of about 29 kbp inserted between themutSandnarLgenes, which was probably acquired by horizontal gene transfer . Using primers narL-rev and mutS-rev, designed on the flanking regions of the insertion site, a 554-bp fragment could be amplified in allP. agglomeransstrains that were negative forpaaABC(i.e., with no genomic island insertion). The large size of the genomic island inpaaABC-positive strains prevented recovery of a PCR-product amplicon. This indicates that the insertion site of the pantocin genomic island is the same as in C9-1 for allPantoeastrains carrying the pantocin A genes. The origin of the pantocin genes remains unknown, and no near or distant homologues have been identified in any other organism after an extensive BLAST search.
Discrimination of clinical and plant-associated isolates ofP. agglomeranshas important implications for the registration of biocontrol products for plant protection. We conducted a polyphasic genotypic analysis of a collection of strains from different ecological niches. Our first observation was that a majority of clinical strains were in fact not trueP. agglomeransas defined by Gavini et al.  based on taxonomic discrepencies revealed by sequence analysis of the 16S rDNA andgyrBgenes. All biocontrol strains in the collection were found to be correctly identified asP. agglomerans. The reason for this discrepancy is ascribed to the fact that bacteria selected for their biological control properties are typically better characterized, including DNA sequencing, in comparison to those obtained in clinical diagnostics where rapid identification for implementation of therapeutic treatment is the primary concern and relies on less precise biochemical identification methods (e.g., API20E and Vitek-2 from bioMerieux or Phoenix from BD Diagnostic Systems). Biochemical methods have previously been shown to misidentifyP. agglomeransandEnterobacterspp. [43,46–49], which our results confirm. Additionally, many archival strains were deposited in culture collections more than 30 years ago when the genusPantoeawas not yet taxonomically established and biochemical identification was less accurate.
We identified a single discriminatory marker for biocontrol strains using fAFLP which may be of use in biosafety decisions for registration of beneficial isolates. Only biocontrol isolates had this fAFLP band, eventhough all strains ofP. agglomerans sensu strictohave indication of the gene found within the band. For differentiation purposes this is irrelevant since the purpose is to identify a genomic marker, not a specific gene.
Our polyphasic analysis indicated that clinical and biocontrol strains co-cluster withinP. agglomerans sensu stricto. This suggests that both isolates from clinical and environmental habitats have undergone indistinguishable evolutionary changes and that there is no discernable specialization of clinical isolates toward human pathogenicity or biocontrol isolates toward a plant-associated life-style. We observed no evidence, but can not exclude, the possibility that clinical isolates may have acquired specific pathogenicity factors beyond T3SS on plasmids or other mobile elements, as has been reported for phytopathogenic strains [44,45]. The T3SS discovered in some strains, however, was found to be more closely related to that in biocontrolPseudomonasspp. indicating a non-pathogenic function . Furthermore, only one clinical isolate had a T3SS gene compared to six environmental isolates. Comparison between the completed genome of biocontrol strain C9-1 and the in progress genome sequencing of the clinical type strain ofP. agglomeransLMG 1286T(T.H.M. Smits, B. Duffy et al., unpublished data) indicates that several features including antibiotic production (revealed by the presence ofpaaABCgenes ), and nectar sugar utilization as a sole carbon source are generally associated with antagonistic activity. Our results demonstrate, however, that while many biocontrol strains have such traits, not all do and thus these are not universal features of biocontrol potential. Also, we have demonstrated for the first time the presence of the antibiotic biosynthetic genespaaABCin clinical strains, indicating that these may not be unique signatures of biocontrol isolates. What if any role pantocin may contribute to animal associations remains to be determined. There was no difference in growth at 37°C between clinical and biocontrol isolates, with both types of strains growing poorly at this temperature compared to growth at 27°C, and reinforcing the weakness of this criteria to determine pathogenicity.
Returning to the fundamental problem of insufficient confidence in identification procedures, we have shown that specific gene sequences (such asgyrBrather than 16S rDNA) are more robust than biochemical identification regardingP. agglomerans. The several reports ofP. agglomeransfrom clinical literature upon which biosafety decisions have been based all lack a clear establishment of this species as a primary and singular cause of disease. With rare exception such isolates are not available for precise taxonomic confirmation and detailed clinical histories are typically absent for individual strains. We conducted a small survey of three clinical diagnostic laboratories in Switzerland and found thatP. agglomeransis infrequently recovered.P. agglomeranswas identified, predominantly as a polymicrobial co-isolate in patients, 21 times in the last four years at the ICM in Bellinzona (M. Tonolla, personal communication) and six times in the last three years at the Kantonsspital Lucerne (M. Hombach, personal communication). At the Institute for Medical Microbiology of Zürich University several strains identified asPantoeawere reported to be present in the collection of clinical samples, but no genetic sequence related toP. agglomeranswas retrieved from their sequence database (G. Bloemberg, personal communication). Thus,P. agglomeranscorrectly characterized appears to be a more infrequent clinical organism than literature indicates.
Our study indicates that current restrictions on registration of microbial pesticides based onP. agglomeransbiocontrol strains in Europe warrant review. The primary argument for biosafety concerns is not supported by the fact that a majority of clinical strains are currently misclassified asP. agglomeransas determined by sequence analysis of 16S rDNA andgyrB. Further analysis of specific genes and fAFLP patterns also distinguish beneficial from clinical strains withinP. agglomerans sensu stricto. Moreover, the lack of pathogenicity confirmatory tests with clinical strains (i.e., Koch's postulates) and the polymicrobial nature in clinical reports, which is probably just a reflection of the natural abundance of this species in the environment, draws into question the biosafety concerns with plant beneficial isolates.
The authors are grateful to P. Coll (Hospital de la Santa Crei Sant Pau, Barcelona, Spain), A. Bonaterra (University of Girona, Spain) and M. Tonolla (ICM Bellinzona, Switzerland) for providing some of the strains used in this study, S. Barnett for providing DNA of Australian strains, and C. Pelludat (ACW) for helpful discussion. Financial support was provided by the Swiss Federal Secretariat for Education and Research (SBF C06.0069), the Swiss Federal Office of the Environment (BAFU), and the Swiss Federal Office of Agriculture (BLW Fire Blight Control Project). This work was conducted within the European Science Foundation funded research network COST Action 873 'Bacterial diseases of stone fruits and nuts'.
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.