Most plant pathogenic bacteria adopt the type III secretion systems to secrete virulence factors and/or avirulence gene products, which trigger the plant hypersensitive response (HR) and the oxidative burst with hydrogen peroxide (H2O2) as the main component. However, the soil-borne plant pathogen Agrobacterium tumefaciens uses the type IV secretion pathway to deliver its oncogenic T-DNA that causes crown gall tumours on many plant species. A. tumefaciens does not elicit a typical HR on those plants. Here, we report that inactivation of one of A. tumefaciens catalases (which converts H2O2 to H2O and O2) by a transposon insertion highly attenuated the bacterial ability to cause tumours on plants and to tolerate H2O2 toxicity, but not the bacterial viability in the absence of exogenous H2O2. This provides the first genetic evidence that the Agrobacterium–plant interaction involves a plant defence response, such as H2O2 production, and that catalase is a virulence factor for a plant pathogen.
The plant resistance response to bacterial infection is often triggered because of a specific interaction between bacterial avirulence (avr ) and plant resistance (R) gene product (Staskawicz et al., 1995; Van den Ackerveken et al., 1996; Baker et al., 1997). A key component in the plant defence system is the hypersensitive response (HR) recognized as the rapid and localized death of cells in response to an avirulent pathogen (Levine et al., 1994; Staskawicz et al., 1995). Attempts to identify bacterial virulence genes have led to the discovery of hypersensitive response and pathogenicity (hrp) genes, without which pathogenic bacteria lose the ability to cause disease on susceptible plants and the ability to cause the HR on resistant plants. These hrp genes presumably encode bacterial membrane apparatus to deliver avr and virulence gene products into host plant cells (He et al., 1993; Van den Ackerveken et al., 1996). The hrp gene products are homologous to a family of bacterial proteins encoded by both plant and animal pathogenic bacteria as well as non-pathogenic bacteria. These proteins constitute the so-called type III secretion system family (He, 1998; Hueck, 1998).
Catalases catalyse the dismutation of hydrogen peroxide to water and oxygen: 2H2O2→ 2H2O + O2. The primary role for catalases is to protect the cells from damage caused by H2O2. Catalases are not required for bacterial cell growth (Loewen, 1997). Although the AOS have been shown to play an important role in animal defence against bacterial infection, whether catalases play a role as virulence factors for animal pathogenic bacteria varies from case to case. There is a report that catalase has been implicated as a virulence factor for Mycobacterium bovis (Wilson et al., 1995). There are also reports that catalase is not a significant factor for the virulence of Salmonella typhimurium (Papp-Szabo et al., 1994; Buchmeier et al., 1995), Haemophilus influenzae (Bishai et al., 1994) or Bordetella pertussis (Khelef et al., 1996). It is not well established whether catalases play a role in virulence for a plant pathogenic bacterium.
Like most plant pathogenic bacteria, Agrobacterium tumefaciens cannot penetrate intact plant surfaces, but enters through wounds. A. tumefaciens is a soil-borne plant pathogen that causes crown gall tumours on many plant species by transferring a specific segment (T-DNA) of its tumour-inducing (Ti) plasmid into plant cells, where the T-DNA becomes integrated into the plant genome. The T-DNA contains the oncogenes that cause overproduction of plant growth hormones and hence tumours (Kado, 1991; Zupan and Zambryski, 1997). The transfer of T-DNA into plant cells is facilitated by the Ti plasmid-encoded virD2 and virB genes (Zupan and Zambryski, 1997; Zupan et al., 1998). The 11 proteins encoded by virB are homologous to members of the newly recognized type IV secretion systems, some of which are essential for the conjugal transfer of bacterial plasmids and for the secretion of virulence factors for two animal pathogenic bacteria, Bordetella pertussis and Legionella pneumophila (Winans et al., 1996; Christie, 1997; Zupan et al., 1998).
Apparently, A. tumefaciens has adopted a virulence secretion pathway different from the remainder of the plant pathogenic bacteria. In addition, A. tumefaciens does not elicit a typical HR on most dicots on which it induces tumours, whereas the remainder do (Robinette and Matthysse, 1990; Deng et al., 1995). A. tumefaciens can even inhibit HR elicited by other bacteria (Robinette and Matthysse, 1990). It is not clear whether A. tumefaciens triggers any specific plant defence response or how it copes with the plant general response that may be triggered by wounding.
Here, we report that an A. tumefaciens mutant, lacking one of the catalases because of a transposon insertion, was highly attenuated in tumour formation. The mutant was much more vulnerable to H2O2 toxicity, although it was as viable as the wild type in the absence of exogenous H2O2. This provides the first genetic evidence to support catalase as a virulence factor for a plant pathogen. This also implies that Agrobacterium–plant interaction involves plant defence responses such as the oxidative burst and the corresponding bacterial detoxification of H2O2 by catalase.
An Agrobacterium gene encoding catalase is involved in tumorigenesis
A. tumefaciens A348 was mutagenized with a mini-Tn5 transposon containing a promoterless gene encoding a green fluorescent protein (GFP) variant, which produces bright green fluorescence under UV light. The mini-Tn5 transposon was carried on a plasmid, pAG408 (Suarez et al., 1997). One of the mutants, AG6, contained the transposon insertion at a gene that was differentially induced by pH on a minimal medium (X. Q. Xu and S. Q. Pan, manuscript in preparation). We inoculated the leaves of Kalanchoe plants with this mutant strain AG6 and compared it with the parent strain A348. As shown in Fig. 1, AG6 was highly attenuated in the ability to cause tumours on plants compared with A348. In order to isolate the mini-Tn5-containing DNA fragments, Southern analysis was carried out to estimate their sizes. A 6 kb ClaI DNA fragment containing the sequences downstream from the transposon insertion site and a 5 kb NruI DNA fragment containing the sequences upstream from the insertion site were cloned into the vectors, resulting in plasmids pXQ6 and pXQ7 respectively. Sequence analysis of pXQ6 and pXQ7 revealed that the transposon was inserted at a gene that is homologous to bacterial genes encoding catalases. We designated this gene as katA.
To determine the complete sequence of the gene, we amplified the DNA fragment from A348 using PCR. A fragment of 2.9 kb was obtained that contained both the upstream and downstream sequences of katA. The resulting fragment was cloned into pSW172 (Chen and Winans, 1991) to generate plasmid pXQ9. When pXQ9 was introduced into AG6, it could fully restore the ability of the mutant to cause tumours (Fig. 1), suggesting that pXQ9 carried a full-length katA gene. Sequence analysis indicated that the katA locus carried a single open reading frame (ORF) that encodes a putative protein of 723 amino acids with a molecular weight of 78.7 kDa (accession number AB033631). This putative protein was highly homologous to other bacterial catalases (Fig. 2).
A catalase encoded by katA detoxifies H2O2
To determine whether the katA gene encodes a functional catalase, we analysed the catalase isozyme patterns for the mutant, parent strain and complemented strain, using a catalase activity staining procedure. As shown in Fig. 3A, both the parent strain, A348, and the complemented strain, AG6(pXQ9), had three distinct catalase activity bands (I, II and III), whereas the mutant, AG6, had only one band (I). This demonstrated that the transposon insertion at the katA gene in AG6 knocked out two catalase activity bands. To investigate whether these two catalase activity bands originated from the same katA gene, we loaded different amounts of the bacterial cell extracts onto the polyacrylamide gels for the catalase activity staining. We found that the catalase activity band III disappeared even in A348 when a smaller amount of the cell extract was loaded (Fig. 3B). This suggests that the catalase activity bands II and III originated from the same katA gene product. The catalase activity band III appeared only when a sufficient amount of cell extract was loaded, suggesting that the band III was an aggregated form of catalase activity band II. We tried to determine whether this katA gene encoded a protein that possessed both catalase and peroxidase activities like some of the catalase genes (Loewen, 1997). When we stained for the peroxidase activities with the same cell extracts (Gregory and Fridovich, 1974), we did not find any peroxidase activity associated with the catalase activity bands (data not shown). Taken together, these data suggest that the katA gene encodes a catalase isozyme in the A. tumefaciens cells.
Subsequently, we determined whether H2O2 was toxic to A. tumefaciens cells and whether the katA gene product could detoxify H2O2. Our plate zone-of-inhibition tests showed that AG6 was more vulnerable to killing by H2O2 than A348 and AG6(pXQ9) (Fig. 4). The vulnerability of AG6 to H2O2 was observed on both MG/L and IB plates (Fig. 4). It appeared that the inhibition zones on MG/L plates were smaller than those on IB plates. We speculate that this was because A. tumefaciens cells grew better on MG/L than on IB and they could generically tolerate H2O2 better on MG/L than on IB.
We then conducted liquid killing assays to determine the lethal dosage for H2O2 toxicity. We exposed A. tumefaciens cells to various concentrations of H2O2 in liquid IB medium and then determined colony-forming units (cfu) on MG/L plates. As shown in Fig. 5, 0.75 mM H2O2 could kill 100% of the mutant AG6 cells, whereas 2.0 mM H2O2 was required to kill 100% of the wild-type strain A348 cells. The LD50 for AG6 was 0.2 mM H2O2 and the LD50 for A348 was 0.8 mM H2O2. Consistent with the inhibition zone tests on plates (Fig. 4), this indicated that the mutant AG6, lacking one of the catalases, was considerably more vulnerable to H2O2 toxicity than the wild-type A348, suggesting that catalase played a role in detoxifying H2O2 during tumorigenesis.
The katA gene does not affect the bacterial viability
As the katA gene was required by A. tumefaciens to detoxify H2O2 (Figs 4 and 5), we wondered whether the mutation at katA could affect the bacterial viability because H2O2 could also be produced inside bacterial cells during bacterial metabolism. We monitored the growth behaviour for both the katA mutant AG6 and the wild-type A348. Both strains could form similar sizes of single colonies 2 days after incubation on MG/L and IB agar plates (data not shown). There was no significant difference in growth rate between the two strains grown in MG/L or IB liquid media (Fig. 6). The viable cell counts were also similar between the two strains grown in the liquid media (data not shown). These suggest that the mutation at katA did not affect the bacterial ability to grow in the absence of exogenous H2O2. Previous experiments have demonstrated that in the presence of plant signal molecules IB medium can induce expression of the A. tumefaciens virulence genes, which are directly responsible for causing tumours on plants (Rogowsky et al., 1987; Winans, 1990; Cangelosi et al., 1991; Kado, 1991; Zupan and Zambryski, 1997) and thus should be induced by the plant environment. Our recent experiments have confirmed that IB is representative of the growth conditions that the bacteria encounter inside plant tissues during the infection process (Li et al., 1999). Therefore, the katA gene should not affect the bacterial viability if H2O2 is not produced during the Agrobacterium–plant interaction. The attenuated ability of AG6 to cause tumours on plants should not be owing to the bacterial viability per se in the absence of exogenous H2O2. This suggests that Agrobacterium–plant interaction involves the plant production of H2O2 and the bacterial detoxification of H2O2 by catalase.
We wondered when H2O2 was produced during Agrobacterium–plant interactions. Because mechanical wounding is known to cause accumulation of H2O2 at the wound sites (Bradley et al., 1992; Wojtaszek, 1997) where A. tumefaciens causes infection, we wondered whether the toxicity imposed on A. tumefaciens by plants was simply due to the H2O2 produced at the wound sites where the bacteria were inoculated. As the oxidative burst is known to be transient for a few hours (< 5 h) (Baker and Orlandi, 1995; Bradley et al., 1992), we reasoned that we could allow A. tumefaciens cells to avoid the oxidative burst generated during wounding by inoculating the bacterial cells at a lag time after the wounds were inflicted. We inoculated the bacterial cells on the wounds at various time intervals after wounding. However, we found that the mutant AG6 lacking one of the catalases still caused highly attenuated tumour formation, even when it was inoculated onto the wounds inflicted 24 h earlier, just as it was when inoculated onto the fresh wounds (data not shown). As a control, the wild-type bacterial cells inoculated onto the fresh wounds were as tumorigenic as those inoculated onto the wounds inflicted 24 h earlier (data not shown). This suggested that H2O2 at the wounds did not impose a chemical barrier for A. tumefaciens, even if it might be produced because of wounding, and that plant production of H2O2 was triggered by the bacterial infection process.
In nature, the soil-borne plant pathogen A. tumefaciens causes crown gall tumours on many plant species, including woody and herbaceous plants belonging to 140 genera of more than 60 families. This Gram-negative bacterium can transfer oncogenic T-DNA, which subsequently becomes integrated into plant chromosomes, into plant cells. Under laboratory conditions, the bacterial ability to transfer T-DNA has been extended to many non-natural hosts, including monocot plants (Hiei et al., 1994), yeasts (Bundock et al., 1995; Piers et al., 1996) and even fungi (de Groot et al., 1998). These indicate that A. tumefaciens possesses an unusual transfer apparatus that is functional with exceptionally different recipient cells.
However, it is unknown how the hosts respond to and defend against the infection of this promiscuous plant pathogen. In this paper, we demonstrate that a transposon insertion into the A. tumefaciens katA gene homologous to catalases highly attenuated the bacterial ability to cause tumours on plants and to tolerate H2O2 toxicity, but not the bacterial viability. Our experiments indicated that the gene product encoded by the katA gene possessed the catalase activity and not peroxidase activity, indicating that H2O2 is the sole substrate for this catalase activity. This suggests that the attenuated virulence of this transposon mutant was a result of the reduced ability to detoxify H2O2. Because the mutant was as viable as the wild-type in the absence of exogenous H2O2, we reason that H2O2 toxicity must have come from plant tissues. This suggests that during the Agrobacterium–plant interaction plants produce H2O2 and A. tumefaciens produces catalases to detoxify H2O2. We speculate that H2O2 production is a plant defence response to A. tumefaciens infection. However, it is unknown whether H2O2 production is the only plant response to A. tumefaciens infection or part of plant defence response that can be triggered both by A. tumefaciens and by other pathogenic bacteria. It would be of interest to determine whether A. tumefaciens triggers other plant defence responses.
The antimicrobial role of active oxygen species (AOS) including H2O2 has long been recognized in animal systems. The engulfment of bacteria during phagocytosis allows the production of AOS to occur in the enclosed areas around the bacteria so that the concentrations of AOS can reach antimicrobial levels. However, plant cells lack the mobility to surround or engulf invading pathogens; bacterial colonization occurs in plant intercellular spaces that are not clearly compartmentalized. In addition, H2O2 is neutral and is able to pass easily through biological membranes. It is not clear whether AOS produced in plants reach sufficient levels for the antimicrobial effects to play a significant role in the plant defence response (Baker and Orlandi, 1995; Wojtaszek, 1997).
As a major component of the oxidative burst, H2O2 has been observed to accumulate in plants challenged with pathogenic micro-organisms, including fungi, bacteria and viruses, as well as in cultured plant cells treated with elicitor preparations and in response to mechanical stress (Baker and Orlandi, 1995; Wojtaszek, 1997). During bacterial–plant interactions, the oxidative burst occurs in two distinct phases (Baker and Orlandi, 1995). Phase I is a relatively short-lived non-specific response that occurs immediately after inoculation with either a compatible or incompatible pathogen. Phase II is a relatively long-lived response occurring 1.5–3 h after inoculation and appears to be specific to incompatible interactions. A major question in the studies of the oxidative burst during plant–pathogen interactions is whether the oxidative burst plays a causal role in pathogen growth restriction or whether AOS are merely produced as a consequence of pathogen infection.
Our experiments demonstrated that the ability of A. tumefaciens to tolerate the H2O2 toxicity was important for tumorigenesis. This provides the genetic evidence that H2O2 plays an important role in restricting the bacterial infection. However, the exact role of H2O2 in Agrobacterium–plant interaction is still unknown as H2O2 has been implicated in other roles, such as (i) cross-linking of the plant cell wall (Bradley et al., 1992), (ii) induction of plant genes involved in cellular protection and defence (Levine et al., 1994; Jabs et al., 1997), and (iii) initiation of plant cell death in the hypersensitive response leading to disease resistance (Chen et al., 1993; Levine et al., 1994; Alvarez et al., 1998). Nevertheless, this report provides the first genetic evidence that the Agrobacterium–plant interaction involves a plant defence response, such as H2O2 production.
Strains, plasmids and growth conditions
Agrobacterium tumefaciens strains were grown in MG/L, IB or AB medium (Cangelosi et al., 1991) at 28°C, supplemented with 100 μg ml−1 kanamycin, 5 μg ml−1 tetracycline or 100 μg ml−1 carbenicillin as required. Escherichia coli strains were grown on Luria–Bertani (LB) medium (Sambrook et al., 1989) at 37°C, supplemented with 50 μg ml−1 kanamycin, 10 μg ml−1 tetracycline or 50 μg ml−1 ampicillin as required. Mini-Tn5 transposon mutagenesis of A. tumefaciens strain A348 was carried out by pAG408 as has been described previously (Suarez et al., 1997).
Total DNA was extracted as has been described previously (Charles and Nester, 1993) from the A. tumefaciens mutant, AG6. Approximately 1 μg of total DNA was digested with ClaI or NruI and then electrophoresed on a 0.9% agarose gel. The DNA fragments were then transferred onto nylon membrane Zeta-Probe GT (Bio-Rad), using transfer apparatus PosiBlot (Stratagene). The plasmid pAG408 was labelled as the probe by random priming with the enhanced chemiluminescence kit (Amersham). The labelling, hybridization and signal detection were conducted according to the manufacturer's instructions.
Cloning and sequencing of the katA gene
The total DNA was extracted from AG6, which contains a mini-Tn5 insertion at the katA gene. Southern analysis revealed a 6 kb ClaI DNA fragment, containing the sequences downstream of the mini-Tn5 insertion at the katA gene, and a 5 kb NruI DNA fragment, containing the sequences upstream of the insertion. These DNA fragments were extracted from the agarose gels and were cleaned using the Geneclean II Kit (BIO 101). The ClaI DNA fragment was cloned into pBluescript-II KS– at the ClaI site, and the NruI DNA fragment was cloned into pTZ19R at the SmaI site. The resulting plasmids were designated pXQ6 and pXQ7 respectively. Sequencing of pXQ6 and pXQ7 was carried out using a mini-Tn5 specific primer and the M13 reverse and −40 universal primers. The resulting sequence data were then used to generate primers for further sequencing. DNA sequencing was carried out using the ABI Prism 377 DNA Sequencer.
In order to clone the full-length katA gene, we designed primers p83 (5′-GGTGCGCTAGCCAAATTCGTCACCAAGC-3′) and n84 (5′-CAATCGCTAGCGTTCGGCCCTCTG-3′) that can, respectively, reanneal to the upstream and downstream sequences of the katA gene. Both primers had a NheI site to facilitate subsequent cloning. The total DNA from A. tumefaciens strain A348 was used as the template for PCR to amplify a 2.9 kb DNA fragment. The PCR product was digested with NheI and ligated into pSW172 (Chen and Winans, 1991) that had been digested with XbaI. The resulting plasmid, pXQ9, was sequenced in both directions independently to obtain unambiguous sequence data. Plasmid pXQ9 was introduced into the mini-Tn5 mutant AG6 to create AG6(pXQ9) by triparental mating (Ditta et al., 1980), based on selection on MG/L medium supplemented with 100 μg ml−1 of kanamycin and 5 μg ml−1 of tetracycline.
A348, AG6 and AG6(pXQ9) were grown on MG/L plates at 28°C for 2 days and were resuspended in MG/L liquid at OD600 = 0.5. Small cuts were made on the leaves of Kalanchoe plants with a hypodermic needle. Then 5 μl of bacterial cell suspension was inoculated onto each area of wounds. Inoculation was repeated at least three times on leaves of different plants. Photographs were taken 6 weeks after inoculation.
Catalase isozyme assay
A. tumefaciens strains A348, AG6 and AG6(pXQ9) were grown overnight at 28°C in 100 ml of MG/L liquid medium to OD600 = 1.4. The cells were harvested by centrifugation at 4000 r.p.m. for 10 min at 4°C. The cell pellets were washed and resuspended in 5 ml of extraction buffer containing 0.05 mM phosphate and 0.4 mM EDTA (pH 7.8). The cells were sonicated for 30 s six times on ice with 2 min cooling on ice between sonications. The cell debris was removed by centrifugation at 1100 r.p.m. for 10 min at 4°C. The cell-free supernatant was diluted 2× with the extraction buffer, and 20 μl of each diluted extract was electrophoresed on 7.5% native polyacrylamide gels. The resolving gel buffer was prepared at pH 8.1 instead of pH 8.9. Electrophoresis was performed at 150 V for 3 h. Catalase isozymes were visualized using an activity staining procedure as described by Clare et al. (1984).
Hydrogen peroxide killing assays
A. tumefaciens strains were grown overnight at 28°C in MG/L or IB liquid media. The bacterial cells were harvested by centrifugation. After washing, the cells were resuspended in the corresponding liquid medium at OD600 = 0.7, and 1 ml of the cell suspension was added into 9 ml of top agar medium, containing the corresponding medium and 1% agar at 50°C. The top agar was immediately plated onto the corresponding solid medium (the plates were 145 mm in diameter). After solidification, sterile Whatman 3MM disks (0.6 cm in diameter) containing 10 μl of 2 mM (0.02 μmol), 20 mM (0.2 μmol) or 200 mM (2 μmol) H2O2 were placed on the surfaces. Zones of inhibition were visualized after incubation overnight at 28°C. To determine the survival rates of A. tumefaciens cells in the presence of H2O2, the bacteria were grown in MG/L liquid medium overnight at 28°C. The cells were harvested by centrifugation, washed and resuspended in IB liquid medium to a final concentration of OD600 = 0.2. Aliquots (2 ml) of the cultures were transferred into sterile tubes and H2O2 was added to the tubes to final concentrations of 0.25 mM, 0.5 mM, 0.75 mM, 1 mM, 1.5 mM or 2 mM. The cell suspensions were incubated at 28°C for 3 h. They were appropriately diluted and then plated on MG/L agar medium to determine the number of cfu. The survival rates of H2O2-treated bacteria were expressed as the percentage of cfu recovered from H2O2-containing media compared with untreated control samples.
A. tumefaciens A348 and AG6 were grown in MG/L liquid medium overnight at 28°C. The cells were harvested by centrifugation, washed and resuspended in MG/L or IB liquid media to an initial concentration of around OD600 = 0.2. Aliquots of the culture were removed after 6, 12 and 24 h for measurement of OD600, and plated on MG/L or IB plates for determination of viable counts.
We thank C. A. Guzman for providing the plasmid pAG408, H. M. Soo and L. W. Tan for technical support and L. P. Li, X. Tang, L. M. Chang, Y. H. Jia, B. F. Lu and Q. M. Hou for discussion. This work was supported by the National University of Singapore Academic Research Grants RP970323 and RP3972375 (to S.Q.P.). We also thank S. M. Wong for critical reading of the manuscript.