Antibiotic Production by Erwinia herbicola Eh1087: Its Role in Inhibition of Erwinia amylovora and Partial Characterization of Antibiotic Biosynthesis Genes

Abstract

Mutants of Erwinia herbicola Eh1087 (Ant⁻), which did not produce antibiotic activity against Erwinia amylovora, the fire blight pathogen, were selected after TnphoA mutagenesis. In immature pear fruit Ant⁻ mutants grew at the same rate as wild-type strain Eh1087 but did not suppress development of the disease caused by E. amylovora. These results indicated that antibiosis plays an important role in the suppression of disease by strain Eh1087. All of the Ant⁻ mutations obtained were located in a 2.2-kb region on a 200-kb indigenous plasmid. Sequence analysis of the mutated DNA region resulted in identification of six open reading frames, designated ORF1 through ORF6, four of which were essential to antibiotic expression. One gene was identified as a gene which encodes a translocase protein which is probably involved in antibiotic secretion. A sodium dodecyl sulfate-polyacrylamide gel electrophoresis analysis of plasmid proteins produced in Escherichia coli minicells confirmed the presence of proteins whose sizes corresponded to the sizes of the predicted open reading frame products.

Erwinia herbicola Eh1087, a naturally occurring nonpathogenic epiphyte isolated from apple blossoms in a New Zealand orchard, is inhibitory to Erwinia amylovora, the organism that causes fire blight (31). Fire blight is a necrotic disease of rosaceous plants which is particularly damaging to apple and pear trees and is a major problem in pip fruit production worldwide (54). The incidence of this disease can be sporadic and difficult to predict. Prophylactic control with the antibiotic streptomycin is expensive and has been associated with the emergence of streptomycin-resistant E. amylovora strains (12, 35, 48).

Biological control of fire blight by epiphytic bacteria that are able to inhibit the growth of E. amylovora (16, 28, 55, 56) is currently being investigated as a way to overcome the problems associated with chemical control of the disease. The most frequently isolated epiphytic species that exhibits inhibitory activity against E. amylovora is E. herbicola (Lohnis) Dye (14). E. herbicola has been shown to colonize the stigmas of blossoms in the same way as the pathogen (23) and is thus a good candidate for development of a biological control agent. On the basis of DNA homology studies, E. herbicola has recently been placed in a new taxon, Pantoea agglomerans (21).

Screening of E. herbicola isolates for inhibitory activity in immature pear fruits has revealed that inhibitory strains frequently produce an antibiotic or bacteriocin which may be involved in disease suppression (28, 55, 57–59). The relative role of antibiosis in disease suppression is unclear as not all antibiotic-producing strains are better inhibitors than nonproducing strains (5, 58). However, studies performed with nonproducing strains of E. herbicola (55) and with antibiotic-resistant mutants of E. amylovora (28) have indicated that antibiotics are important in the inhibition of E. amylovora by E. herbicola.

Commonly, E. herbicola antibiotic activity is inhibited by amino acids, especially histidine, suggesting that inhibitory E. herbicola strains typically produce an antibiotic (or a family of related antibiotics) that interferes with histidine biosynthesis (15, 55, 58). Antibiotic activity in cell-free supernatants from Eh1087 broth cultures is not inhibited by amino acids but is inactivated by digestion with β-lactamase, indicating that the antibiotic of Eh1087 is a β-lactam antibiotic (31). The discovery of a strain of E. herbicola which inhibits E. amylovora with an antibiotic different from the antibiotics produced by other inhibitory E. herbicola strains is of interest as biological control strategies are likely to be more successful when there are two or more compatible strains which act by different mechanisms and which may respond differently to climatic conditions.

Eh1087 inhibits a broad spectrum of gram-negative bacterial species in vitro and inhibits the development of fire blight symptoms in immature pear fruits (31) and in excised apple blossoms in glasshouse trials (29). In this study, mutants of Eh1087 that were not able to inhibit E. amylovora in vitro were generated by TnphoA mutagenesis. The aims of this study were to confirm the role of antibiosis in the control of disease by Eh1087 and to identify a genetic locus or loci involved in antibiotic expression.

MATERIALS AND METHODS

Bacterial strains, plasmids, and media.

The bacterial strains and plasmids used in this study are described in Table 1. Bacteria were cultured in or on Luria-Bertani (LB) medium (1% tryptone, 0.5% yeast extract, 0.5% NaCl; solidified with 1.5% Bacto Agar [Difco Laboratories], when required) supplemented with the appropriate antibiotics. Antibiotics and 5-bromo-4-chloro-3-indolyl phosphate (XP) were obtained from Sigma Chemical Co. (St. Louis, Mo.) and were used at the following concentrations: ampicillin, 100 μg · ml⁻¹; rifampin, 50 μg · ml⁻¹; kanamycin, 50 μg · ml⁻¹; tetracycline, 15 μg · ml⁻¹; chloramphenicol, 35 μg · ml⁻¹; and XP, 40 μg · ml⁻¹.

TABLE 1.

Bacterial strains and plasmids

Strain(s) or plasmid	Characteristicsa	Source or reference
E. herbicola strains
Eh1087	Wild-type strain isolated from Malus × domestica, Rfr	This study
EhA11, EhA12, EhA17, EhA19, EhA46	Kmr TnphoA-induced mutants of Eh1087 which do not produce antibiotic in vitro	This study
E. amylovora Ea8862	Wild-type strain isolated from Malus × domestica	ICMPb
E. coli strains
DH5α	endA hsdR supE thi gyrAU169 φ80d relA lacZΔM15 Δ(argf-lacZYA)	Bethesda Research Laboratories
P678-54	Minicell-producing strain, F−thr leu supE lacY tonA galB mal xyl ara mtl min	1
SM10	thi thr leu tonA lacY supE RecA::RP4-2-Tc::Mu (λpir+)	49
Plasmids
pRT733:TnphoA	oriR6K Apr Kmr	53
pUC18/19	Apr LacZα+	60
pBR322	Apr Tcr pMB1 replicon	6
pACYC184	Cmr Tcr p15A replicon	9
pLAFR3	Tcrmob+ Tra− LacZα λcos pRK2 replicon	51
pRK2013	Kmr Tra+ helper plasmid pRK2 replicon	13
pDelta1	Apr Kmr Tc strA sacB lacZ′	Life Technologies
pLK2	Mutant fragment clone from EhA17 in pUC19, containing left-hand portion of TnphoA plus 12 kb of Eh1087 DNA adjacent to the site of TnphoA insertion, Kmr	This technology
pLA17, pLA255, pLA272	Complementing cosmids from Eh1087 genomic library	This study
pBH3.8	HindIII 3.8-kb subclone of pLA255 in pBR322, Tcr	This study
pBE5	EcoRI 5-kb subclone of pLA255 in pBR322, Tcr	This study
pAH8	HindIII 8-kb subclone of pLA255, Cmr	This study

^aAp^r, Km^r, Cm^r, and Rf^r, resistance to ampicillin, kanamycin, chloramphenicol, tetracycline, and rifampin, respectively. 

^bICMP, International Collection of Micro-organisms from Plants, Landcare Research/Manaaki Whenua New Zealand Ltd., Auckland, New Zealand. 

TnphoA mutagenesis of Eh1087.

TnphoA was mobilized into Eh1087 with a suicidal vector plasmid, pRT733 (Ap^r), which cannot replicate without the λpir gene product (33). Equal volumes (0.5 ml) of overnight LB broth cultures of the donor strain, Escherichia coli SM10(pRT733::TnphoA), and Eh1087 (Rf^r) were mixed, and the cells were sedimented by centrifugation. The cells were gently resuspended in 200 μl of LB broth, deposited on a sterile 0.22-μm-pore-size filter (Millipore) on a prewarmed LB agar plate, and incubated for 7 to 8 h at 25°C. After incubation, the cells were resuspended in 5 ml of sterile 0.85% (wt/vol) saline, and 100-μl portions were plated onto LB agar supplemented with rifampin, kanamycin, and XP. The plates were incubated overnight at 30°C. Control plates containing E. coli and Eh1087 grown on the selective medium were also included. TnphoA insertion mutants were screened for in vitro inhibition of E. amylovora as described below and for auxotrophy by resuspending individual colonies in 10 mM MgSO₄ and then transferring the preparations to M63 minimal medium supplemented with vitamin B₁ (41).

In vitro inhibition of E. amylovora.

The in vitro inhibition assay was performed on a minimal medium (27) supplemented with niacin (Sigma) at a concentration of 50 mg · liter⁻¹ (HSN medium). Individual colonies of exconjugants growing on the selective medium were transferred with toothpicks onto HSN agar plates seeded with a soft agar overlay lawn of Ea8862 and incubated overnight at 30°C. Colonies that failed to produce inhibition zones on Ea8862 lawns were also tested for inhibition of Ea8862 in immature pear fruits and for production of antibiotic activity in cell-free culture supernatants.

Antibiotic production in broth cultures.

Overnight HSN broth cultures that had been incubated at 30°C were centrifuged to sediment the cells. The pH of the broth supernatants was adjusted to 6.8, and the supernatants were sterilized by filtration. Drops (20 μl) of each sterile broth supernatant were dropped onto an Ea8862 soft agar overlay lawn freshly prepared on HSN agar. The resulting plates were incubated overnight at 30°C, and the presence of zones of inhibition was recorded.

Inhibition of E. amylovora in immature pear fruits.

Immature pears (Pyrus communis L. cv. Bartlett) were surface sterilized in a 0.5% (wt/vol) sodium hypochlorite solution for 10 to 15 min and then washed for 20 min with running water. The method of Erskine and Lopatecki (16) was modified so that the pears were aseptically sliced and 3-mm-thick slices were placed in sterile petri dishes containing water-saturated filter paper discs. Each treatment included 60 slices from at least 20 pears. Overnight LB broth cultures of wild-type strain Eh1087, Ant⁻ mutants, and Ea8862 were centrifuged to sediment the cells, and the pellets were resuspended in sterile 0.85% (wt/vol) saline to an optical density at 600 nm of 0.2 (approximately 5 × 10⁸ CFU · ml⁻¹). For each treatment pear slices were inoculated with 50 μl of either Eh1087 or an Ant⁻ mutant and 50 μl of Ea8862. Controls containing only saline and controls containing only pathogen were included in each assay. The pear slices were incubated at 25°C for 4 to 6 days and were considered positive for infection when water soaking and/or ooze production was observed. Levels of infection were determined on the third day after inoculation as follows: percent infection = (number of infected pear slices/total number of pear slices) × 100.

Screening Ant⁻ mutants for loss of immunity to Eh1087 antibiotic activity.

Ant⁻ mutants of Eh1087 were screened for the loss of immunity to the Eh1087 antibiotic activity on HSN agar by using a streak plate method. A plate containing a single central streak of Eh1087 that was cross-streaked with Ant⁻ mutant strains and with Eh1087 and Ea8862 controls was prepared. Zones of inhibition were observed after overnight incubation at 30°C.

Rates of growth of Ant⁻ mutants in immature pear fruits.

The rates of growth of Ant⁻ mutant strains in immature pear fruits were compared with the rate of growth of wild-type strain Eh1087. Overnight cultures in LB broth containing appropriate antibiotics were diluted to a density of approximately 5 × 10⁸ CFU · ml⁻¹, and 50-μl portions of the diluted broth preparations were used to inoculate immature pear slices as described above for the immature pear fruit assay. At different times over a 24-h period three pear slices were removed from each treatment. The pear slices were washed for 1 min with vigorous shaking in 10 ml of 0.85% (wt/vol) saline plus 1.5% peptone. Aliquots (20 μl) were plated in triplicate onto selective media and incubated overnight to obtain viable bacterial counts.

General DNA manipulations.

Restriction endonucleases and bacterial alkaline phosphatase (Bethesda Research Laboratories) were used as recommended by the manufacturer. T4 DNA ligase was obtained from Boehringer Mannheim. Total DNA was purified by the method of Chesney et al. (10), and plasmid DNA was purified by the method of Sambrook et al. (46). BamHI digestion of total DNAs from Ant⁻ mutants yielded fragments containing the Km^r gene of TnphoA and IS50L, in addition to the left-hand flanking Eh1087 DNA; these fragments were designated mutant fragments. Total BamHI-digested DNAs from Ant⁻ mutants were cloned into pUC19 and used to transform competent cells of E. coli DH5α by the method of Sambrook et al. (46). DH5α colonies containing cloned mutant fragments were selected on LB agar supplemented with kanamycin. Plasmid DNAs from Km^r colonies (mutant fragment clones) were purified and restriction mapped. Mutant fragment clone pLK2 was derived from strain EhA17 and contained 12 kb of Eh1087 DNA adjacent to the site of the TnphoA insertion. A 2.4-kb EcoRI-SalI fragment was isolated from this insert and used as a probe in hybridizations with both a genomic library and a plasmid visualization gel containing Eh1087. When DNA fragments were used as probes, they were purified from agarose gel slices by the method of Heery et al. (24) and were labelled with ³²P by using the Bethesda Research Laboratories random primer system as recommended by the manufacturer. DNA was transferred onto Hybond N⁺ membranes (Amersham) according to the manufacturer’s instructions by using a Vacublot system (Bio-Rad). Southern-blotted DNA was hybridized by standard methods. The membranes were dried and autoradiographed for 6 to 12 h at −80°C with Hyperfilm-MP film (Amersham) in Cronex intensifying cassettes (DuPont).

Eh1087 plasmid visualization and hybridization.

Large plasmid preparations were made by the method of Comai and Kosuge (11). Electrophoresis was carried out in 0.3% agarose for 36 h at 1 V · cm⁻¹ at 4°C. A strain of Agrobacterium tumefaciens carrying two plasmids (200 and >300 kb) was included as a control (this strain was supplied by B. Palmer, University of Canterbury). Southern-blotted Eh1087 DNA was hybridized with a fragment probe derived from pLK2 to determine whether the mutated site was chromosomal or plasmid borne.

Construction and screening of a genomic library of Eh1087.

Eh1087 genomic DNA was partially digested with Sau3A, and 25- to 30-kb DNA fragments were purified on a linear sucrose gradient, dephosphorylated, and used as insert DNA. Cosmid pLAFR3 individual vector arms were prepared by the method of Staskawicz et al. (51). Insert DNA was ligated to pLAFR3 vector arms and packaged into λ phage heads in vitro. Test and preparative libraries were made by the method of Fleischmann et al. (17). The genomic library was screened by colony hybridization to a DNA probe prepared from plasmid pLK2. Colony blots were prepared on Hybond N⁺ membranes (Amersham) and hybridized as recommended by the manufacturer.

Complementation of Ant⁻ mutants of Eh1087.

pLAFR3 cosmid clones that positively hybridized to the pLK2 fragment probe were mobilized from E. coli DH5α to Eh1087 TnphoA insertion mutants by triparental mating performed by using helper plasmid pRK2013 (13) and the conjugation conditions used for TnphoA mutagenesis. Plasmids pBR322, pACYC184, pBH3.8, pBE5, and pAH8 were introduced into Ant⁻ strains for complementation by electroporation with a Bio-Rad Gene Pulser apparatus used as recommended by the manufacturer.

Construction of pAH8.

Plasmid pAH8 consisted of two contiguous HindIII fragments that came from within the region of DNA common to all of the complementing cosmids and were cloned into vector pACYC184. To construct pAH8, the two HindIII fragments (3.8 and 4.2 kb) were cloned into pACYC184 in a shotgun cloning experiment and transformed into DH5α. Cm^r transformants were separately hybridized to ³²P-labelled probes prepared from each HindIII fragment. The transformants that hybridized to both HindIII fragment probes, which indicated that both fragments were inserted into pACYC184, were restriction mapped to confirm that each fragment was oriented correctly.

E. coli minicell protein analysis.

E. coli minicell-producing strain P678-54 was electroporated with the appropriate plasmids. Plasmid-containing minicells were purified by differential centrifugation and differential rate sedimentation in sucrose gradients (18). Frozen minicells (100-μl aliquots) were thawed on ice, and 900 μl of M63 labelling buffer (M63 medium supplemented with all of the amino acids except methionine) was added to each 100-μl aliquot. The minicells were preincubated for 20 to 40 min at 37°C with orbital shaking, and then [³⁵S]methionine (20 μCi · ml⁻¹) was added to each flask and the cells were incubated for an additional 30 to 45 min. Labelled cells were transferred to an Eppendorf tube and collected by centrifugation for 2 min at 12,000 × g. The cell pellet was resuspended in 60 μl of storage buffer (7 g of Na₂HPO₄ per liter, 3 g of KH₂PO₄ per liter, 4 g of NaCl per liter, 0.1 g of MgSO₄ per liter) and stored frozen at −20°C. The labelled minicells were thawed, diluted with an equal volume of 2× Laemmli sample buffer (34), and boiled for 3 min. The proteins were analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) in 10 to 20% polyacrylamide gradient gels by using the Laemmli system. The gels were stained with Coomassie blue to visualize molecular weight markers and then soaked in an Amplify (Amersham) solution for 30 min and vacuum dried onto filter paper for autoradiography with a model 443 slab dryer (Bio-Rad) for 5 h at 60°C.

DNA sequencing.

DNA sequencing was performed by the dideoxynucleotide method (47) by using Taq DNA polymerase. Either restriction fragments were cloned into pUC vectors and sequenced directly from the plasmids by using forward and reverse sequencing primers for pUC templates or nested deletion clones were generated in plasmid pDelta with the Deletion Factory system (Life Technologies) and then sequenced directly from plasmid pDelta by using SP6 or T7 primers. In addition, several custom-made DNA oligonucleotide primers were used to fill in gaps. To exactly map the site of each TnphoA insertion, the DNA sequence adjacent to the inserted transposon in each mutant fragment clone (see above) was read directly from the transposon by using a TnphoA primer (TACTTGTGTATAAGAGTCAG) (37). DNA and amino acid sequences were analyzed with the DNASIS (Hitachi Software Engineering Co.), FASTA (45), BlastX (2), and PROSITE (3) software packages.

Nucleotide sequence accession number.

The DNA sequence data described in this paper have been deposited in the GenBank database under accession no. AF006625.

RESULTS

Selection of Eh1087 mutants deficient in antibiotic production.

TnphoA mutagenesis was used to create 12 TnphoA insertion mutants (Ant⁻) of Eh1087 which failed to inhibit Ea8862 in vitro. The Ant⁻ mutants were prototrophic, indicating that the mutated genes were directly involved in antibiotic synthesis, secretion, and/or regulation. No auxotrophic mutants were obtained from the 800 Km^r colonies, indicating that transposon insertion was not random. On minimal agar, Ant⁻ colonies had white or pink-orange pigmentation, in contrast to the yellow pigmentation of wild-type strain Eh1087 colonies. The Ant⁻ mutants remained resistant to the Eh1087 antibiotic in vitro.

To confirm that there were single TnphoA insertions in the Ant⁻ mutants, Southern-blotted DNAs from Ant⁻ mutants were hybridized with a TnphoA probe consisting of a 1.4-kb BamHI-HindIII fragment that contained part of the central TnphoA region and part of IS50R. Hybridization with this probe revealed that single TnphoA insertions occurred in 6 of the 12 Ant⁻ mutants and double insertions occurred in the remaining mutants (data not shown).

Five of the single-insertion mutants, EhA11, EhA12, EhA17, EhA19, and EhA46, failed to suppress fire blight disease in immature pear fruits (Fig. 1A) and to exhibit antibiotic activity against Ea8862 in cell-free HSN culture supernatants. At the high inoculum levels used in the immature pear fruit assay, the rates of growth of the Ant⁻ mutants and wild-type strain Eh1087 were the same (data not shown), indicating that the lack of inhibition by Ant⁻ strains was not due to reduced growth.

FIG. 1.

(A) Levels of infection in immature pear fruits 3 days after infection, as determined with pear slices inoculated with pathogenic strain Ea8862 (Ea) and wild-type strain Eh1087 or Ant⁻ mutants (EhA strains). Pear slices inoculated with Ea8862 and saline were included as positive controls; 95% confidence intervals are indicated. (B) Complementation of mutant strain EhA17 with cosmids pLA15, pLA255, and pLA272. Levels of infection in immature pear fruits 3 days after inoculation are shown; 95% confidence intervals are indicated.

BamHI digestion of Ant⁻ mutant DNAs released fragments containing the left-hand portion of TnphoA, including the kanamycin resistance gene, and target DNA flanking the insertion site. These fragments were cloned into pUC19 and restriction mapped. Overlapping restriction maps of these mutant fragment clones allowed the TnphoA insertion sites for the corresponding mutants to be tentatively mapped in a 2.2-kb region of DNA.

Complementation of Ant⁻ mutants.

An Eh1087 total DNA library was constructed in cosmid vector pLAFR3 and was screened by colony hybridization with a 2.4-kb EcoRI-SalI fragment probe derived from mutant fragment clone pLK2. Three cosmids hybridized with the probe and restored wild-type in vitro antibiotic activity to each of the five Ant⁻ mutants by trans complementation. EhA17 transformed with any of the three cosmids was indistinguishable from wild-type strain Eh1087 as determined by the immature pear fruit bioassay (Fig. 1B). A 10- to 11-kb region common to all three of the cosmids was identified by a combination of restriction mapping and Southern hybridization (Fig. 2A).

FIG. 2.

(A) Restriction map of the DNA region common to complementing cosmids pLA15, pLA255, and pLA272. The restriction enzymes used were HindIII (H), EcoRI (E), SalI (S), and BamHI (B). The triangles indicate the sites of TnphoA insertion for EhA strains. The numbers of the strains corresponding to the insertion sites are indicated above the triangles. (B) Predicted ORFs identified by sequence analysis of the 5-kb EcoRI fragment. (C) Restriction fragments used for complementation studies. The name of each construct is indicated on the left. pAH8 was a pACYC184 construct; pBE5 and pBH3.8 were pBR322 constructs. Data for complementation of Ant⁻ strain EhA12 for each construct are shown on the right.

Three overlapping restriction fragments that were in the common region were subcloned and used to test for complementation of the Ant⁻ mutants (Fig. 2C). Only one mutant strain was complemented by any of the fragments tested. This strain, EhA12, was fully complemented to wild-type antibiotic production in vitro by the introduction of a 3.8-kb HindIII fragment cloned into pBR322 (pBH3.8). When plasmid constructs containing larger overlapping fragments which extended the left-hand margin of the 3.8-kb HindIII complementing fragment (pBE5 and pAH8) were introduced into EhA12, complementation was reduced, and smaller inhibition zones were observed. Plasmid pAH8 also partially restored antibiotic production in vitro to mutant strain EhA17, although this partial complementation was not stable. The inhibition zones produced were turbid rather than clear, and inhibition zones were not produced after the organism was subcultured once.

DNA sequence analysis.

The 5-kb EcoRI fragment insert of pBE5, which completely overlapped the complementing 3.8-kb HindIII fragment and which contained all of the TnphoA insertion sites, was sequenced in both directions. Sequencing each mutant fragment clone directly from the TnphoA primer allowed us to confirm the TnphoA insertion sites, which initially had been indicated by restriction mapping. A sequence analysis in which DNASIS was used revealed six major open reading frames (ORFs) (designated ORF1 through ORF6) (Fig. 2B and Table 2). For each of the ORFs except ORF3, putative promoter sequences and potential ribosome-binding sites were found; no ribosome-binding site was found for ORF3. In two cases there was potential translational coupling of pairs of ORFs, with the extreme 3′ end of one ORF overlapping the 5′ end of the next ORF downstream; this was true at the ORF2-ORF3 and ORF5-ORF6 boundaries. All of the ORFs had the same orientation and were closely grouped together. There were 50 nucleotides between the stop codon of ORF3 and the start codon of ORF4 and 26 nucleotides between the stop codon of ORF4 and the start codon of ORF5.

TABLE 2.

Characteristics of ORFs identified on pBE5

ORF	Nucleotide positions	No. of amino acids	Mol wt (103)	Homology(ies) of gene product	Necessary for antibiotic biosynthesis
ORF1	327-602	92	10.4	None	Not known
ORF4	928-1263	112	11.6	Aldehyde dehydrogenases	Yes
ORF3	1260-1736	159	17.5	RNase E	Yes
ORF4	1781-2464	228	27.6	Methyltransferases, ornithine aminotransferase	Yes
ORF5	2488-3924	479	50.8	Drug resistance translocases	Yes
ORF6	3924-4712	263	29.3	Dehydrogenases	Not known

Two of the mutations, the mutations in strains EhA46 and EhA11, were in a 1,437-bp ORF whose predicted protein product had a molecular mass of 50.8 kDa. When translated, this ORF displayed approximately 30% amino acid identity and 50% amino acid similarity to the cmcT gene product of Nocardia lactamdurans and the products of other genes of the drug resistance translocase family. A hydropathy profile of the encoded protein (data not shown) indicated that there were multiple hydrophobic domains that were characteristic of channel-forming proteins, and this profile was similar to the profiles of other gene products of this family. Because of the similarities in amino acid composition, overall size, and hydropathy profiles, we propose that ORF5 encodes a transmembrane protein that is functionally related to the drug resistance translocases. This gene product is essential for antibiotic expression in Eh1087 and probably is involved in antibiotic secretion.

The remaining three mutations (in strains EhA17, EhA12, and EhA19) each occurred in a separate ORF (ORF2, ORF3, and ORF4, respectively). Sequence homology searches were performed with other proteins in the database for each of these translated ORFs. ORF2 encoded a putative 11.6-kDa protein with 50 to 66% amino acid similarity and 30% amino acid identity (four gaps) to various aldehyde dehydrogenases. The sequence contained a GXGX₂AX₃A consensus motif found in the βαβ dinucleotide-binding fold of NADP⁺-linked enzymes. Similarity to aldehyde dehydrogenases also was observed in the sequence extending 190 nucleotides upstream of the putative start codon of ORF2, although a stop codon at nucleotide 706 prevented extension of this ORF. ORF3 encoded a putative 17.5-kDa protein; the amino-terminal region of this protein (amino acids 6 to 86) is 28% identical and 45% similar to the carboxyl-terminal sequence of the ams gene product of E. coli (8). As reported for the carboxyl-terminal residues of the E. coli ams gene product (8), the amino-terminal residues (100 amino acids) of the ORF3 protein were also highly hydrophilic (hydropathy profile data are not shown). ORF4 encoded a putative 27.6-kDa protein whose amino-terminal region (amino acids 27 to 112) exhibited significant similarity to methyl transferases. This region included another NADP⁺-binding site motif (amino acids 61 to 70). No typical ATP-binding site sequence motif was present, although the carboxyl-terminal region of the ORF4 protein (amino acids 181 to 213) exhibited 42% identity to a probable ATP-binding protein (accession no. pir S28007) and 33% identity to a putative ATPase (pir S40525).

Two of the other ORFs which were identified, ORF1 and ORF6, flanked the region in which the TnphoA insertions occurred. ORF1 encoded a 10.4-kDa protein with no obvious similarities and ORF6 encoded a 29.3-kDa protein with approximately 50% amino acid similarity to dehydrogenases and reductases over the entire ORF, although this similarity did not include any binding motifs found in enzymes known to use flavin adenine dinucleotide or NAD(P) cofactors. Because antibiotic biosynthesis genes often are found in clusters, it is possible that these two putative genes also may be involved in antibiotic expression in Eh1087.

Minicell protein analysis.

Proteins produced by E. coli minicells carrying the three constructs used for complementation studies (pBH3.8, pBE5, and pAH8) were analyzed by SDS-PAGE (Fig. 3). In minicells containing pBH3.8, plasmid-encoded proteins whose sizes corresponded to the sizes of the predicted ORF3 product (17 kDa) and the ORF4 product (27 kDa) were identified. These two protein bands also were found in minicells containing pBE5, along with a 50-kDa protein whose size corresponded to the size of the predicted ORF5 product (50.8 kDa). In both of these constructs, additional bands were observed at 61 and 42 kDa, along with smaller bands at 10 and 11 kDa. In minicells containing pAH8, proteins whose sizes corresponded to the sizes of the ORF4, ORF5, and ORF6 (33-kDa) products were observed in addition to bands at 48, 42, and 26 kDa. The 17-kDa protein found in minicells containing pBH3.8 and pBE5 was not present in minicells containing pAH8. If the 17-kDa protein is the ORF3 protein, then its expression in plasmids pBH3.8 and pBE5 but not in the larger plasmid pAH8 can be correlated with complementation of mutant strain EhA12, which carried a mutation in ORF3. EhA12 was complemented by pBH3.8 and pBE5 but not by pAH8.

FIG. 3.

SDS-PAGE of proteins in E. coli minicells carrying pBR322 (lane A), pBH3.8 (lane B), pBE5 (lane C), pAH8 (lane D), and pACYC184 (lane E). Molecular weights (in kilodaltons) are indicated on the left. The molecular weights of proteins whose sizes correspond to the sizes predicted for the products of ORF3 through ORF6 are indicated between lanes C and D.

Ant⁻ mutations are located on a plasmid.

Eh1087 carries a large indigenous 200-kb plasmid (Fig. 4). A Southern blot of a plasmid visualization gel was hybridized with the 2.4-kb fragment from pLK2 used as a probe to screen the genomic library. The probe hybridized to the plasmid band, indicating that the mutations in antibiotic activity were plasmid borne, not chromosomal.

FIG. 4.

Agarose gel electrophoresis of plasmid DNAs from A. tumefaciens (lane A) and E. herbicola Eh1087 (lane B) and corresponding autoradiograph of DNAs from A. tumefaciens (lane C) and Eh1087 (lane D) when they were hybridized with a DNA probe derived from mutant fragment pLK2 (see text). Note that the >300-kb plasmid of A. tumefaciens is not visible on the gel.

DISCUSSION

Twelve Ant⁻ mutants of Eh1087 which were not able to inhibit E. amylovora in vitro were selected after TnphoA mutagenesis. Antibiotic biosynthesis generally involves multienzyme pathways, and the genes for these pathways tend to be grouped in 20- to 30-kb regions of DNA (39). The tight clustering of TnphoA insertion points in a 2.2-kb region of DNA and the lack of auxotrophic mutations suggest that hot spot insertions of TnphoA may have occurred. Similarly, Gantotti et al. (20) reported high ratios of Tn5-induced mutants of E. herbicola defective in bacteriocin production, and McGowan et al. (40) also observed hot spot insertion of Tn5 into the Erwinia carotovora carR gene, a regulatory gene involved in carbapenem production in this species.

Five of the Ant⁻ mutants with single TnphoA insertions were investigated further. These Ant⁻ mutants failed to produce antibiotic activity in vitro and failed to inhibit Ea8862 in immature pear fruits. The immature pear fruit assay is widely used to assess the efficacy of potential biological control agents, and the results of the immature pear fruit assay generally correlate with protection in orchard trials (4, 56). The inability of Eh1087 Ant⁻ mutant strains to inhibit E. amylovora was not due to impaired competitive ability resulting from a reduced growth rate, as the growth rates in immature pear fruits were similar for the mutants and wild-type strain Eh1087. Mutants of Eh1087 deficient in antibiotic production failed to protect pear fruits from fire blight disease, even when the bacterial populations were high. This is in contrast to the results of previous studies in which antibiotic-deficient mutants of inhibitory E. herbicola strains Eh252 (55) and Eh318 (59) provided some degree of disease protection for pear fruits, which suggested that competition may also play a role in the inhibition of E. amylovora by E. herbicola. The sizes of natural E. herbicola populations in apple (30) and pear (36) blossoms were found to rapidly increase during the late blossom period, possibly indicating that E. amylovora could be displaced by competitive exclusion. However, when immature pear fruits were separately inoculated with Ea8862 or Eh1087, the sizes of the Ea8862 populations were approximately 10-fold higher than the sizes of the Eh1087 populations after 24 h (data not shown), indicating that competitive exclusion in immature pear fruits by Eh1087 probably does not contribute to the antagonism observed. The results of this study suggest that the contribution of antibiosis in Eh1087 is more important to the inhibitory potential of this strain than is the contribution of antibiosis in other inhibitory E. herbicola strains.

Southern hybridization with a radiolabelled probe isolated from the 2.2-kb region of TnphoA insertions showed that the mutations obtained were localized on a 200-kb indigenous plasmid, indicating that at least some of the genes for antibiotic biosynthesis in Eh1087 are plasmid borne. E. herbicola strains carry numerous cryptic plasmids (19, 20, 22, 55). In inhibitory strain Eh112Y the determinants of bacteriocinogenicity are carried on a 96-MDa (150-kb) plasmid (20), but in Eh252 the genes for antibiotic production are chromosomally encoded (55).

The 5-kb EcoRI fragment that contained all of the TnphoA insertion sites was sequenced. A computer analysis of the sequence resulted in identification of six ORFs, four of which (ORF2 through ORF5) were essential for antibiotic expression. The putative protein product of ORF5 was a transmembrane protein that exhibited significant homology to the family containing drug resistance translocases, which are thought to confer antibiotic resistance to organisms by mediating the removal of the antibiotic by using energy derived from transmembrane proton gradients (25). It is proposed that the ORF5 product is involved in the export of the Eh1087 antibiotic. The potential translational coupling of ORF5 with ORF6 is characteristic of many bacterial genes whose products are required in equimolar quantities (42) and suggests that there is a possible connection between the functions of the products of the two ORFs. Two other ORFs, ORF2 and ORF3, also overlapped, suggesting that the products of these ORFs may also be related functionally. ORF2 probably encodes a dehydrogenase. The deduced features of the products of ORF3 and ORF4 were insufficient to allow us to predict the functions of the genes.

Antibiotic activity was fully restored to Ant⁻ mutants by cloned 25- to 30-kb wild-type DNA regions from a cosmid library of Eh1087. Attempts to complement Ant⁻ mutants with restriction fragments subcloned from within the 10- to 11-kb region that was common to all of the complementing cosmids were usually unsuccessful. The only mutant in which antibiotic production was restored was EhA12, which was fully complemented by pBH3.8 and partially complemented by pBE5 and pAH8. As the pBH3.8 insert starts only 35 nucleotides upstream from the ORF3 amino terminus, it is possible that expression of this ORF in pBH3.8 is driven by a promoter in pBR322 and that the larger inserts of constructs pBE5 and pAH8 include greater lengths of DNA between the vector promoter and the ORF3 start site, which could weaken expression of ORF3. Alternatively, the DNA upstream of the HindIII site may contain negative regulators of ORF3 expression. Proteins whose sizes correspond to the sizes predicted for the ORF3 and ORF4 products were present in E. coli minicells carrying plasmids pBH3.8 and pBE5 but not in minicells carrying plasmid pAH8, although pAH8 contained the relevant sequences. These results are consistent either with the occurrence of vector-driven expression in constructs pBH3.8 and pBE5, which is much weaker in pAH8, or with the presence of negative regulators for these ORFs in pAH8. The presence of regulatory sequences in pAH8 may also explain the appearance of proteins whose sizes correspond to the sizes predicted for ORF5 and ORF6 products in E. coli minicells carrying plasmids pBE5 and pAH8 but not the smaller plasmid, pBH3.8, despite the fact that the appropriate sequence is present on this plasmid. These results could be explained by the presence of sequences upstream of the HindIII site which exert a positive regulatory effect on expression of ORF5 and ORF6 and a negative regulatory effect on expression of ORF3 and ORF4. In the mutant strains that were not complemented, TnphoA insertion may have exerted polar effects on genes downstream of the insertion sites (32), indicating that genes essential for antibiotic expression in Eh1087 may exist beyond the right-hand margin of the DNA region used in our complementation experiments.

Eh1087 is the first reported strain of E. herbicola that inhibits E. amylovora by means of an antibiotic that appears to be a β-lactam (31). Bacterially produced β-lactams were first discovered in the early 1980s and include the monocyclic β-lactams (monobactams) (52) as well as the bicyclic clavulanic acids (26) and carbapenems (44) and the related cephem antibiotics cepalosporins and cephamycins (38, 50). With the exception of the cephalosporins and cephamycins, whose biosynthesis has been well-characterized (38), very little is understood about the biosynthesis of bacterial β-lactam antibiotics.

There was no evidence which indicated that the antibiotic biosynthesis genes of Eh1087 were related to the cephalosporin or cephamycin biosynthesis genes, which are highly conserved (38).

It is tempting to speculate that the Eh1087 antibiotic might be a carbapenem. Carbapenems have a wide spectrum of potent antimicrobial activity against both gram-positive and gram-negative bacteria, in contrast to the monobactams and clavulanic acids, which tend to have weak antibacterial activities. Erwinia spp. produce a simple carbapenem, carbapen-2-em-3-carboxylic acid, which is susceptible to the same β-lactamase as the Eh1087 antibiotic (31, 40, 44). Studies performed with radiolabelled biosynthetic precursors in Erwinia and Serratia (7) strains have led to proposed hypothetical biosynthetic schemes for the carbapenem antibiotics. These interrelated pathways, which can be experimentally tested, include dehydrogenation and methyl transfer steps which are not incompatible with the predicted functions of some of the ORFs identified in Eh1087.

It is still possible that the Eh1087 antibiotic is a non-β-lactam antibiotic that has β-lactam-like properties. Nozaki et al. (43) described the discovery of a novel antibiotic, lactivicin, produced by Empedobacter lactamgenus YK-258, which has various biological properties commonly observed in β-lactam antibiotics, including susceptibility to hydrolysis by β-lactamase, but which does not have a β-lactam ring in its structure.

In this study we demonstrated that Eh1087 inhibition of E. amylovora is mediated by the production of an antibiotic. The antibiotic of Eh1087 is of particular interest as it is unlike all previously characterized antibiotics produced by E. herbicola. Molecular biological investigations to date have revealed four putative genes that are essential for antibiotic expression.