Expression of a crown gall biological control phenotype in an avirulent strain of Agrobacterium vitis by addition of the trifolitoxin production and resistance genes

Background Agrobacterium vitis is a causal agent of crown-gall disease. Trifolitoxin (TFX) is a peptide antibiotic active only against members of a specific group of α-proteobacteria that includes Agrobacterium and its close relatives. The ability of TFX production by an avirulent strain of Agrobacterium to reduce crown gall disease is examined here. Results TFX was shown to be inhibitory in vitro against several A. vitis strains. TFX production, expressed from the stable plasmid pT2TFXK, conferred biological control activity to an avirulent strain of A. vitis. F2/5, against three virulent, TFX-sensitive strains of A. vitis tested on Nicotiana glauca. F2/5(pT2TFXK) is significantly reduces number and size of galls when co-inoculated with tumorigenic strain CG78 at a 10:1 ratio, but is ineffective at 1:1 or 1:10 ratios. F2/5(pT2TFXK) is effective when co-inoculated with tumorigenic strain CG435 at 10:1 and 1:1 ratios, but not at a 1:10 ratio. When F2/5(pT2TFXK) is co-inoculated with CG49 at a 10:1 ratio, the incidence of gall formation does not decline but gall size decreases by more than 70%. A 24 h pre-inoculation with F2/5(pT2TFXK) does not improve biological control at the 1:10 ratio. Conclusions TFX production by an avirulent strain of Agrobacterium does confer in that strain the ability to control crown gall disease on Nicotiana glauca. This is the first demonstration that the production of a ribosomally synthesized, post-translationally modified peptide antibiotic can confer reduction in plant disease incidence from a bacterial pathogen.


Background
Agrobacterium vitis strains are causative agents of crown gall, an economically important disease [1,2]. A. vitis F2/ 5 is an effective biological control agent against many A. vitis tumorigenic strains [3]. Strain F2/5 produces an antibiotic toxic to many A. vitis strains in vitro. However, two lines of evidence suggest that this antibiotic plays a minor role in disease suppression. Strains that are susceptible to the antibiotic in vitro, such as A. vitis strain CG78, are able to infect the plant in the presence of F2/5 [3], and Tn5 mutants of F2/5 lacking F2/5 antibiotic production appear to be unaffected in crown gall biological control [4]. Biological control by F2/5 is grape-specific, as F2/5 is not effective on non-grapevine host plants such as Nicotiana glauca. Furthermore, F2/5 is not effective against all A. vitis strains [3]. Thus, enhancing the F2/5 biological control phenotype and extending the host range of the efficacy of F2/5 beyond grape would be beneficial for disease control in field applications.
Agrobacterium rhizogenes strain K84 is the most studied crown gall biological control strain and is commercially utilized for crown gall disease control worldwide. Strain K84 biological control is primarily due to production of two plasmid-encoded antibiotics, agrocins 84 and 484, encoded by genes on pAgK84 and pAgK434 respectively [5]. Agrocin 84, an adenosine analog [6], is effective against tumorigenic strains carrying nopaline/agrocinopine tumor-inducing plasmids, and requires the acc system in the target strain for activity [7]. Agrocin 434, a di-substituted cytidine analog, is effective against, and specific for, a broad range of A. rhizogenes strains [8]. Curing of either agrocin-encoding plasmid results in reduction of biological control activity [9]. Thus, K84 demonstrates the efficacy of antibiosis for crown gall biological control. However, the commercial application of the K84 biological control system is limited where Agrobacterium strains of certain crops are not inhibited by K84. As a result, alternative biological control systems for crown gall disease are needed.
Trifolitoxin (TFX) is a peptide antibiotic produced by Rhizobium leguminosarum bv. trifolii T24. Antibiotic production and resistance functions are encoded by the seven-gene tfx operon and the unlinked tfuA gene [10,11]. TFX is derived from post-translational cleavage and modification of the tfxA gene product. TFX effectively inhibits growth of members of the α-proteobacteria including strains of Brucella, Ochrobactrum, Rhodobacter, Rhodospeudomonas, Rhizobium, and the etiological agent of crown gall disease, Agrobacterium [12]. TFX is highly specific for this group, as demonstrated by a lack of observable effect on the majority of the bacterial population in the bean rhizosphere [13]. This suggests that TFX could be used to control crown gall diseases of various plants with limited effects on non-target bacteria outside of that very specific group of TFX-sensitive α-proteobacteria. TFX is rapidly degraded in nonsterile soil but is readily extractable from sterile soil inoculated with a TFX-producing strain, suggesting that TFX is sensitive to in situ proteolysis [14]. Despite this rapid turnover, TFX production confers enhanced nodulation efficiency upon TFX-producing Rhizobium strains under field conditions [15].
Degradation of antimicrobial peptides such as cecropin B and attacin E also has been observed in plant apoplastic fluids [5,16,17]. This is likely due to apoplastic proteinases [5,16]. Expression of antimicrobial peptides in plants has had mixed results for enhancing disease resistance. For example, cecropin expression in transgenic tobacco did not confer resistance to P. syringae pv tabaci [18], likely due to low apoplastic peptide concentrations due to pro-teolysis [19]. Therefore, prior to this work it was not clear whether a peptide antibiotic could play a role in the inhibition of galling by tumorigenic Agrobacterium in planta.
Agrobacterium rhizogenes is capable of producing TFX by addition of the stable plasmid, pT2TFXK, which contains the tfx operon but not tfuA [11,12,20]. This suggests that TFX production by crown gall biological control strains of Agrobacterium, such as A. rhizogenes K84 and A. vitis F2/5, may be enhanced by TFX production. These strains would be excellent delivery vehicles for TFX to the infection court.
Here we present experiments demonstrating the TFX-sensitivity of a range of A. vitis strains and the effect of the TFX production and resistance phenotypes on the expression of biological control of crown gall using the model plant Nicotiana glauca. Strain F2/5 was chosen as the recipient of the TFX genes in this work because it is an avirulent strain that expresses a biological control phenotype on only one host. As a result, F2/5 is an ideal strain in which to test the ability of the TFX system to confer biological control of crown gall disease and to broaden the host-range efficacy of a known biological control strain.

In vitro TFX antibiosis against A. vitis
Agrobacterium vitis strains (Table 1) were tested for sensitivity to trifolitoxin. As expected based on previous results [12], the tested Agrobacterium vitis strains were sensitive to TFX-producing strains. However, the level of susceptibility was lower than predicted based upon previous TFX sensitivity measurements with CG48 and CG74 [12]. No zones of A. vitis growth inhibition were observed around R. leguminosarum T24 colonies, and only relatively small zones were observed around R. etli CE3(pT2TFXK), which produces more TFX than T24. Furthermore, one A. vitis strain, F2/5, was TFX-resistant.
Because A. vitis F2/5 produces an antibiotic to which most of the tumorigenic strains are sensitive [3], the effect of TFX on A. vitis was assessed against TFX-producing and non-producing Rhizobium strains. None of the tested strains were sensitive to R. leguminosarum T24, which produces relatively low amounts of TFX. All of the A. vitis strains except for F2/5 were sensitive to R. etli CE3(pT2TFXK) as evidenced by zones of growth inhibition around the CE3(pT2TFXK) colonies. A. vitis growth was not inhibited by a non-TFX metabolite or nutrient competition by CE3(pT2TFXK) as evidenced by the lack of a zone around the near-isogenic tfxA mutant CE3(pT2TX3K) colony ( Fig. 1). Wild type F2/5 produces a zone of inhibition versus the other agrobacteria used in this work. This occurred because of the antibiotic production previously observed by F2/5. These zones of inhibi-tion were much larger when pT2TFXK was added to F2/5 and unchanged with the addition of pT2TX3K.

In planta biological control of crown gall conferred by TFX production
As expected, F2/5 did not inhibit galling by tumorigenic A. vitis strains on N. glauca. However, a 10:1 ratio of A. vitis F2/5(pT2TFXK):pathogen caused a statistically significant reduction in mean gall size relative to the TFX nonproducing controls on N. glauca stems for all three tested tumorigenic strains (Figs. 3 and 4). High concentrations of F2/5(pT2TFXK) also reduced gall incidence for CG435 and CG78, but not for CG49 (Table 2). A 1:1 ratio of F2/ 5(pT2TFXK):CG435 also resulted in a significant reduction in gall size and in gall incidence compared to controls. A 1:1 ratio of F2/5(pT2TFXK):CG49 or F2/5 Table 1: Bacterial Strains used in this work and their TFX phenotype (production, resistance, and sensitivity). Overproduction of TFX occurs when the TFX production is conferred using a multi-copy, broad host range plasmid such as pT2TFXK.

Strain
Characteristics Reference

Figure 1
In vitro TFX antibiosis assay against A vitis CG78. Legend: The indicated TFX producing strains and their non-producing mutant derivatives cultured as a single colony near the centers of the plates and allowed to grow 4 days at 27°C. The plates were then misted with a dilute suspension of A. vitis CG78. The plates were photographed two days following misting. A zone of growth inhibition is present around the R. etli CE3(pT2TFXK) colony, but not around any of the other colonies.
(pT2TFXK): CG78 did not affect either incidence of galls or reduce gall size. Similarly, an excess (a 1:10 ratio) of any of the virulent strains to F2/5(pT2TFXK) resulted in a high incidence of disease and large gall size.

Discussion
As all strains of A. vitis tested were sensitive to TFX both in this work and in a previous study [12], experiments were conducted to determine the effectiveness of a TFX-producing, avirulent strain of A. vitis in the prevention of crown gall caused a three strains of A. vitis. Although A. vitis F2/ 5 is an effective crown-gall biological control agent against most tumorigenic A. vitis strains when co-inoculated with a tumorigenic strain, it is not an effective against all strains of A. vitis.
Where A. vitis F2/5 is effective as a biological control agent, its control is only effective when numbers of F2/5 are equal to, or greater than, the number of cells of the virulent strain [3]. Strains that are resistant to F2/5 biological control are known. For example, strain CG78 is not controlled when co-inoculated with F2/5 [1].
Two principle benefits of TFX production by F2/5 are demonstrated here. Biological control is extended to a new host, N. glauca. Furthermore, biological control is extended to a strain that F2/5 fails to control (CG78). These effects are due to TFX production as demonstrated by the lack of efficacy of F2/5 against CG49, CG435, and CG78, and by the lack of efficacy of the near-isogenic TFX-nonproducing F2/5(pT2TX3K) against CG78. TFX confers biological control when the TFX-producing strain is present in excess of the tumorigenic strain. Thus, F2/5(pT2TFXK) effectively inhibited gall incidence by two of the three tested tumorigenic strains when co-inoculated in approximately 10-fold excess. F2/5(pT2TFXK) inhibited gall size by more than 70% with third tumorigenic strain tested. At 1:1 or 1:10 inoculum ratios of F2/5(pT2TFXK):pathogen biological control was reduced or lost. High ratios of biological control: pathogen strain should be easily achieved in field situations by dipping the roots of planting stock in suspensions of the biological control strain, or by directly applying the bacterial suspension to the planting bed. Furthermore, TFX is inhibitory towards all tested species of Agrobacterium [12]. These results suggest that TFX production would enhance crown gall biological control for other biological control strains, such as A. rhizogenes K84, and on other host plants, especially where a mixed inoculum of different tumorigenic Agrobacterium species occurs.
K84 biological control is thought to be primarily due to production of two plasmid-encoded antibiotics, each of which accounts for a portion of the observed disease control [9]. This naturally occurring example suggests that

Figure 2
Ability of Agrobacterium to produce TFX with addition of pT2TFXK. Legend: Ability of F2/5(pT2TFXK) to inhibit ANU794. This inhibition is reversed by the addition of the TFX resistance genes to ANU794. Assay was performed as described in Figure 1 with F2/5(pT2TFXK) cultured in a single colony in the center of the plate. One day after growth at 28°C, the plates were sprayed with a dilute suspension of ANU794(pT2TX3K) (A) or ANU794 (B).

A B
pyramiding biological control mechanisms within one strain can enhance disease control. Strain F2/5 also produces an A. vitis-specific antibiotic in vitro, but this antibiotic is not involved in disease control [4]. Thus, the addition of TFX to F2/5 is a successful example of pyramiding unrelated disease control mechanisms and clearly demonstrates that individual antibiotics can be effectively developed as a mechanism of disease control.
Although plasmid-borne traits are frequently unstable, pT2TFXK contains the RK2 plasmid-partitioning locus that confers a high degree of stability both in vitro and under field conditions [15]. Stable TFX expression would be beneficial under field conditions when the biological control agent is inoculated on seeds or on roots dipped in a bacterial suspension prior to planting.
Trifolitoxin production enhances rhizosphere competitiveness of Rhizobium etli CE3 (pT2TFXK) in sterile soil and enhances nodulation efficiency in non-sterile soil when compared to a TFX-sensitive strain [20]. Trifolitoxin production also increases bean nodulation competitiveness of Rhizobium etli CE3 in the presence of indigenous rhizobia under agricultural conditions [15]. Field application of R. etli CE3 (pT2TFXK) dramatically reduced the di-versity of indigenous α-proteobacteria in the bean rhizosphere without affecting unrelated species [13]. Thus, TFX production by A. vitis F2/5 should enhance the competitiveness and aid in establishment of this beneficial strain in the rhizosphere. The ability to displace indigenous tumorigenic Agrobacterium strains would be beneficial in those areas already infested with the pathogen.
Compared to other treatments available for pathogenic agrobacteria, the effects of TFX on non-target species is very low. Copper, a broad spectrum bacteriocide, is commonly used as a treatment of bacterial diseases. Although no comprehensive studies have been published on the taxonomic range of bacteria inhibited by K84, it is known that K84 produces multiple antibiotics and that K84 can inhibit strains of Erwinia and Pseudomonas that are unaffected by TFX [23][24][25]. TFX-producing strains inhibit only a very specific group within the α-proteobacteria in culture and in the field [13,26]. These observations suggest that a K84-producing strain may have a more serious im-

Figure 3
Effect of TFX production and inoculum ratio on gall formation. Legend: Nicotiana glauca stems were wounded with a dissecting needle, and 5 ul of inoculum (mixtures noted at left) was placed on the wounds. The top stem was inoculated with CG435 as a positive control. Stems 2 and 3 were inoculated with mixtures of biological control test strains and CG435 at either a 1:1 (left three inoculation sites on each stem) or 10:1 (right three inoculation sites) biological control:pathogen ratios. The photograph was taken approximately 2 months post-inoculation. pT2TFXK confers biological control on F2/5 at the 10:1 ratio, but not at the 1:1 ratio.

Figure 4
Effect of TFX production and inoculum ratio on pact on non-target organisms than does a TFX-producing strain. However, there has been no study of the effects of K84 on non-target soil bacteria by culture-independent means to compare the effects of K84 versus the TFX system in a natural system.

Conclusions
TFX production effectively enhances crown gall biological control by Agrobacterium vitis F2/5. The host-plant range and range of tumorigenic A. vitis strains controlled by F2/ 5 are both broadened by TFX production. In addition, virulent strains of A. vitis not previously controlled by F2/5 are controlled following the addition of the trifolitoxin production and resistance genes to F2/5. The plasmid used to confer trifolitoxin production and resistance, pT2TFXK, is stably maintained in the absence of selection pressure [20]. This plasmid is not self-transmissible but is mobilizable. With the TFX system, the biological control of Agrobacterium pathogens may be extended to many crops far beyond what is possible today with currently available commercial products.

Bacterial and plant growth conditions, strain construction
Bacterial strains are listed in Table 1. The A. vitis strains (without pT2TFXK or pT2TX3K) were obtained from Dr. T.J. Burr, Cornell University. Bacteria were grown on BSM agar [21] media at 27°C. A. vitis F2/5(pT2TFXK) and F2/ 5(pT2TX3K) were constructed by triparental mating using standard procedures. Transconjugants were selected on BSM media amended with 50 µg/ml kanamycin. Trimethoprim (10 µg/ml) was added to counterselect the E. coli DH5α donor and helper strains. The helper strain was E. coli DH5a pRK2013 [27]. Strains containing the plasmids pT2TFXK and pT2TX3K were grown for routine propagation on BSM amended with 50 µg/ml kanamycin. Prior to use in making inoculum suspensions for biological control assays these strains were grown overnight on BSM agar without kanamycin. In vitro antibiosis assay TFX antibiosis assays were performed as described previously [11]. The effect of TFX on various Agrobacterium vitis strains (Table 1) was assessed using Rhizobium leguminosarum bv. trifolii strain T24 and R. etli strain CE3 (pT2TFXK) as producing strains. Rhizobium leguminosarum bv. trifolii T24::Tn5-1 and R. etli CE3 (pT2TX3K) were used as TFX non-producing negative controls.

In planta biological control of crown gall by TFX-producing strains
Agrobacterium vitis strains were suspended in sterile distilled water prior to the determination of colony forming units (CFU) per ml. These suspensions were adjusted to OD 650 0.10 (approximately 10 8 CFU/ml) using a Shimadzu UV-160 spectrophotometer and sterile distilled water, and stored until inoculation at 4°C. Actual inoculum viability and cell density were measured by dilution plating on BSM agar medium on the day that plants were inoculated [22]. Prior to inoculation, tumorigenic strains were diluted 10-fold with sterile distilled water to approximately 10 7 CFU/ml. Strain F2/5 and its derivatives were left undiluted, or diluted 10-fold (for CG49 and CG435 experiments) or 100-fold (for CG78 experiments). Thus, the CFU ratios were approximately 10:1, and 1:1 or 1:10 avirulent:tumorigenic strain. Immediately prior to plant inoculation tumorigenic strains were mixed 1:1 (vol:vol) with the appropriate biological control test strain. Positive controls were diluted 1:1 (vol:vol) with water. Thus, all plant inoculum contained approximately 5 × 10 6 CFU/ml of the tumorigenic strains. Plants (Nicotiana glauca) were inoculated by wounding the stem with a dissecting needle. Three or four inoculations were made per inoculum mixture on each of two plants. Thus, each of the two experiments included 6 to 8 repetitions per treatment. A 5 µl drop of bacterial suspension was placed on the wound and allowed to air dry. Inoculation sites were wrapped loosely with Parafilm (American National Can) for 1 week post-inoculation. Gall diameter perpendicular to the long axis of the stem was measured 4 to 7 weeks post-inoculation using a caliper, and all measurements were included for statistical analysis. Results were analyzed using ANOVA at the α = 0.05 level of significance.