• Open Access

Optimization of Agrobacterium-mediated transient assays of gene expression in lettuce, tomato and Arabidopsis


  • Tadeusz Wroblewski,

    1. The Genome Center, University of California, Davis, 1 Shiels Ave., Davis, CA 95616, USA
    Search for more papers by this author
  • Anna Tomczak,

    1. The Genome Center, University of California, Davis, 1 Shiels Ave., Davis, CA 95616, USA
    Search for more papers by this author
    • Present address: Laboratory of Nematology, Wageningen University, Wageningen, The Netherlands

  • Richard Michelmore

    Corresponding author
    1. The Genome Center, University of California, Davis, 1 Shiels Ave., Davis, CA 95616, USA
      Correspondence (fax +530-752-9659; e-mail rwmichelmore@ucdavis.edu)
    Search for more papers by this author

Correspondence (fax +530-752-9659; e-mail rwmichelmore@ucdavis.edu)


Agrobacterium-mediated transient assays for gene function are increasingly being used as alternatives to genetic complementation and stable transformation. However, such assays are variable and not equally successful in different plant species. We analysed a range of genetic and physiological factors affecting transient expression following agroinfiltration, and developed a protocol for efficient and routine transient assays in several plant species. Lettuce exhibited high levels of transient expression and was at least as easy to work with as Nicotiana benthamiana. Transient expression occurred in the majority of cells within the infiltrated tissue and approached 100% in some regions. High levels of transient expression were obtained in some ecotypes of Arabidopsis; however, Arabidopsis remains recalcitrant to routine, genotype-independent transient assays. Transient expression levels often exceeded those observed in stably transformed plants. The laboratory Agrobacterium tumefaciens strain C58C1 was the best strain for use in plant species that did not elicit a necrotic response to A. tumefaciens. A wild A. tumefaciens strain, 1D1246, was identified that provided high levels of transient expression in solanaceous plants without background necrosis, enabling routine transient assays in these species.


The identification of thousands of sequences through genome projects has increased the need for quick and simple analyses of gene function. These often involve one or more transgenic approaches to express or silence a gene. However, the transformation and regeneration of most higher plants remain tedious, time-consuming and often costly. Even with species for which these procedures have been greatly simplified, such as Arabidopsis thaliana or Medicago truncatula (Bechtold et al., 1993; Clough and Bent, 1998; Trieu et al., 2000), it can still take several months to produce transgenic plants suitable for analysis. In addition, because the level of transgene expression can vary from one transgenic plant to another as a result of gene silencing phenomena or the gene's position in the chromatin (reviewed by Vaucheret et al., 1998), multiple transgenic individuals are required for the reliable analysis of a single transgene.

Transient assays mediated by agroinfection have been increasingly employed as an alternative to the analysis of stable transformants. Transient expression provides a rapid method for assaying the function of some types of gene; transgenes can often be assayed within a few days of infiltration (Janssen and Gardner, 1989). Its utility has been reported for transgenic complementation (Bendahmane et al., 2000; Van der Hoorn et al., 2000; Johansen and Carrington, 2001; Shao et al., 2003), promoter analysis (Yang et al., 2000) and protein production (Vaquero et al., 1999, 2002). Agrobacterium-mediated transient assays have some limitations, however. They are restricted to species and tissues that are biologically compatible and physically accessible to A. tumefaciens. In addition, because two living organisms participate in the process, the efficiency of transient assays is influenced by experimental variables that affect the virulence of A. tumefaciens and the plant's physiological condition.

The efficiency of transformation is greatly influenced by the compatibility between plant and bacterium. Some strains of Agrobacterium are more virulent than others on a particular plant species and, conversely, some plant species or genotypes are more or less sensitive to particular strains of Agrobacterium (De Cleene and De Ley, 1976; Anderson and Moore, 1979). Such variation may be partly due to differences in the ability of the bacteria to attach to plant cells or differences in either bacterial- or plant-encoded T-DNA transfer machinery (Lippincott et al., 1977; Yanofsky et al., 1985a,b; Nam et al., 1997). In addition, there may be incompatibility between Agrobacterium and the host plant that involves necrosis, which resembles the hypersensitive reaction (HR) elicited as a component of the plant's defence response. Most laboratory strains of A. tumefaciens elicit necrosis in tomato (Van der Hoorn et al., 2000; and this paper). Sufficient numbers of infected and transformed cells can regenerate into transgenic tomato plants, and therefore the necrotic response has not prevented the utilization of A. tumefaciens for the generation of stable transformants of tomato. However, necrosis can be problematic for Agrobacterium-mediated transient assays of Solanaceous species.

Most of the widely used laboratory strains of A. tumefaciens originate from only two wild isolates, C58 (Hamilton and Fall, 1971) and Ach5 (LBA4404; Hoekema et al., 1983). Additional strain designations primarily reflect the introduction of various Ti plasmid derivatives (reviewed by Hellens and Mullineaux, 2000). Thus, the currently available genetic variation amongst laboratory strains of Agrobacterium is extremely limited. The many wild strains of Agrobacterium represent a potential source for novel, more efficient agents for transient assays. The oncogenes present in wild-type T-DNA (Leemans et al., 1981) need not be removed from wild Agrobacterium strains for either compatibility screening purposes or gene expression studies. The several weeks it takes crown-gall tumours to develop precludes their interference with transient assays, which are evaluated within a few days of infiltration with bacteria. Moreover, oncogene expression may possibly delay senescence and actually enhance the efficacy of transient assays in some circumstances.

The studies described in this paper were carried out to increase the efficiency of A. tumefaciens-mediated transient assays (Schob et al., 1997) for lettuce, tomato and Arabidopsis. We examined parameters controlling expression, as monitored using β-glucuronidase (GUS) (Jefferson et al., 1987) and green fluorescent protein (GFP) (Chalfie et al., 1994; Haseloff et al., 1997) activity in planta. We evaluated the role of several chemical, physical and biochemical factors known to affect the efficiency of Agrobacterium-mediated stable transformation in transient expression. The levels of transient expression observed often exceeded the levels of expression in stably transformed tissue. We also investigated genotypic effects of both plant and bacterial components on transient gene expression. As a result, reliable, high levels of transient gene expression in lettuce, tomato and Arabidopsis were achieved and a strain of A. tumefaciens that is compatible with tomato, and therefore particularly useful for transient assays in this species, was identified.


Infiltration of plant tissues with suspensions of A. tumefaciens

To identify the efficient methods for Agrobacterium infection, we tested different infiltration media as well as several techniques for tissue infiltration. For lettuce, pure distilled-deionized (dd) water gave results as good as any infiltration medium reported previously (Rossi et al., 1993; Kapila et al., 1997; Schob et al., 1997; Llave et al., 2000). dd water was therefore routinely used for most of the work presented in this paper, the exceptions being experiments in which the effects of various chemicals were evaluated, as described below.

Vacuum infiltration of detached leaves (Rossi et al., 1993; Kapila et al., 1997) or seedlings gave variable results and usually weak expression, probably due to uneven tissue penetration by the bacterial suspension. Prolonged exposure to the vacuum rapidly decreased the temperature of the suspension of A. tumefaciens, which may have additionally reduced the expression. We therefore utilized agroinfiltration of attached leaves (Schob et al., 1997; see ‘Experimental procedures’) as our method of agroinfection. This method also allowed multiple assays to be conducted on a single leaf (Figure 1).

Figure 1.

Transient gene expression in the leaves of lettuce and tomato after infiltration with Agrobacterium tumefaciens containing pTFS40 (Table 3). (a) GUS activity in a leaf of Lactuca serriola (LS102). Of all the plants analysed in this study, the highest level of transient gene expression was observed in the leaves of this genotype. The pattern of GUS staining reflects the restriction of the infiltrated suspension by the major leaf veins. (b) Multiple infiltrations in a single leaf, in this case 12 in L. sativa cv. Mariska. (c) Necrosis (arrowed) in tomato (cv. Rio Grande 76R) induced by different strains of A. tumefaciens.

Leaf structure affected infiltration greatly. In Nicotiana benthamiana, Petunia hybrida (data not presented), tomato and Arabidopsis, the bacterial suspension penetrated the tissue easily and the infiltrated regions were usually circular or irregular in shape. In lettuce, pepper, N. tabacum and cotton (data not presented), the suspension of A. tumefaciens spread through the leaf lamina, but was delineated by prominent veins. The density of the bacterial suspension was also important for infiltration. Suspensions less dense than OD600 = 0.1 (OD, optical density) resulted in weak transgene expression. Infiltrations of bacterial suspensions with densities above OD600 = 1.0 often resulted in tissue yellowing or wilting. Tests with lettuce indicated that the optimal bacterial density for infiltration was between 0.3 and 0.8. We therefore used suspensions of A. tumefaciens diluted to densities of OD600 = 0.4–0.5 for the rest of the experiments presented in this paper. For all the plant species evaluated, the intensity of GUS staining reached its maximum within 4–5 days following agroinfiltration and decreased slowly after that time. Therefore, most of the data presented here were collected 4 days post-infiltration (dpi).

As a result of the > 6000 infiltrations performed during the collection of data for this paper, we noted differences in the level of transient gene expression within single plants, in tissues of different developmental stages and in plants of different ages. Infiltration of young leaves before expansion, as well as older leaves, usually resulted in weak expression. For all plant species tested, higher levels of expression were often observed in the first true leaves of the seedlings when compared with leaves produced later. For Arabidopsis thaliana plants with rosettes composed of about 30 leaves, the fourth and fifth leaves from the centre usually exhibited the highest levels of expression. For lettuce, N. benthamiana and tomato, when multiple infiltrations into single leaves were made, the level of transient expression varied depending on the region of leaf infiltrated for the assay. Depending on the age of the leaf, the expression on the tip of the leaf was often different in comparison with that on the basal part. The basal regions were composed of ontogenetically younger cells. In general, tissues that had recently undergone rapid cell expansion exhibited the highest levels of transient expression. In order to take ontogenetic variability into account, all data presented in this paper were generated by averaging the results from at least 10 replications.

Proportion of cells exhibiting transient expression and levels of transient expression

For many experiments, it is pertinent to know what proportion of cells is expressing a transgene, as well as the level of expression averaged over the whole tissue. Therefore, we assessed the transient expression of GFP at the cellular level utilizing confocal laser scanning microscopy (CLSM) and biochemical assays of GUS activity. Strain C58C1 containing the GFP gene controlled by the cauliflower mosaic virus (CaMV) 35S promoter (pBINm-gfp5-ER; Haseloff et al., 1997) was infiltrated into the leaves of wild lettuce LS102. In this particular GFP gene construct, the basic chitinase signal sequence immobilizes the GFP protein in the endoplasmic reticulum (ER), thereby preventing its cell to cell movement (Haseloff et al., 1997).

A high percentage of infiltrated cells of lettuce expressed GFP; however, the level of expression varied somewhat from cell to cell. On average, more than 80% of the mesophyll cells within infiltrated areas exhibited GFP expression (data not shown). In some areas, nearly 100% of the cells expressed GFP. Groups of cells exhibiting stronger expression were often juxtaposed with clusters of cells exhibiting weaker expression. These differences in expression level may be due to uneven penetration of the parenchymal tissue by the suspension of A. tumefaciens during infiltration, as previously suggested by Kapila et al. (1997).

In lettuce, transient expression of the 35S-GUS gene introduced by strain C58C1 was more than an order of magnitude higher than the expression of the same transgene in stably transformed plants. The average transient activity of GUS in 10 infiltrated non-transgenic leaves of lettuce cv. Mariska was 287.3 ± 88.2 nmol/mg/min, compared with an activity of 12.3 ± 7.9 nmol/mg/min measured in 10 independent transgenic individuals of the same cultivar.

Screening of wild strains of A. tumefaciens as agents for transient assays

To increase the diversity of strains of A. tumefaciens available for transient assays, we initially evaluated 42 wild strains. These strains had been previously isolated from crown-gall tumours formed on various hosts in nature [C. Kado, L. Epstein and A. Dandekar, University of California, Davis (UC Davis), personal communication, 2003]. To confirm their identity and pathogenicity, as well as their potential utility for tomato, we inoculated stems of tomato cv. Moneymaker with each wild strain. Twenty-four strains induced crown-gall tumours in the infected stems within 6 weeks after inoculation and were studied further (Table 1). All were identified as Biovar I (Moore et al., 1988). The size and morphology of the tumours differed slightly between the strains, but detailed evaluations of tumorigenicity were not performed.

Table 1.  Description of 24 tumorigenic strains of Agrobacterium tumefaciens
StrainHost of originGeographical originKanamycinTetracyclinepTFS40*Reaction on tomatoReference or source
25 mg/L50 mg/L2.5 mg/L5.0 mg/L
  • R, resistant, growth equal to no selection; r, partially resistant, decreased growth relative to no selection; s, partially sensitive, minimal but detectable growth; S, sensitive, no growth; N, necrosis 4 days after infiltration; x, no transformants obtained; –, no observable reaction.

  • *

    Introduction of pTFS40 (Table 3) into the strain confirmed by polymerase chain reaction (PCR).

1D1108EuonymusDeitrick, MD, USARRSS+NC. Kado, UC Davis
1D1249DahliaTubingen, GermanyRrSS+C. Kado, UC Davis
1D1299CherryTokyo, JapanrssS+NC. Kado, UC Davis
1D132CherryContra Costa, CA, USAssrs+NC. Kado, UC Davis
1D135PeachYolo, CA, USAssSS+NC. Kado, UC Davis
1D1405CherrySacramento, CA, USAssSS+NC. Kado, UC Davis
1D1460CaneSanta Cruz, CA, USARRsS+NC. Kado, UC Davis
1D1479LippiaYam Hill, OR, USARrSSxNC. Kado, UC Davis
1D1480EuonymusRosenberg, OR, USASSSSxNC. Kado, UC Davis
1D1487AppleMedford, OR, USAsSSS+C. Kado, UC Davis
53-2AWalnutCA, USAssrs+NL. Epstein, UC Davis
Ec1??rsss+NA. Dandekar, UC Davis
85-1AWalnutCA, USARRsS+NL. Epstein, UC Davis
173AWalnutCA, USARRsSxNL. Epstein, UC Davis
132AWalnutCA, USARrssxNL. Epstein, UC Davis
A281Engineered sSRs+NSciaky et al. (1978)
136-1AWalnutCA, USAsSrsxNL. Epstein, UC Davis
Ach5??SSsS+NOoms et al. (1981)
15955??SSSS+Barker et al. (1983)
K12Engineered sSsS+NDandekar et al. (1987)
99WalnutCA, USAsSssxNL. Epstein, UC Davis
92AWalnutCA, USAsSssxNL. Epstein, UC Davis
143-1AWalnutCA, USASSssxNL. Epstein, UC Davis
69-2B1WalnutCA, USASSsS+NL. Epstein, UC Davis

To determine the levels of antibiotics that would be useful for the selection of bacteria transformed with binary plasmids, all 24 tumorigenic strains were tested for endogenous resistance to kanamycin and tetracycline. Six were resistant to kanamycin at 25 and 50 mg/L, and three to tetracycline at 2.5 and 5.0 mg/L. None of the strains was resistant to both antibiotics (Table 1). We also tested six selected wild strains and the laboratory strain C58C1 for their ability to grow in the presence of six antibiotics (Table 2). For all six strains tested, tetracycline and gentamycin were more effective than kanamycin. The resistance of C58C1 to chloramphenicol concurs with the previously reported presence of chloramphenicol acetyltransferase in this strain (Tennigkeit and Matzura, 1991). Small percentages (less than 0.1%) of resistant colonies were observed for many of these strains and were probably due to spontaneous mutations conferring resistance. Spontaneous mutation to tetracycline resistance has been well characterized for strain C58 (Luo and Farrand, 1999).

Table 2.  Sensitivity of one laboratory and six tumorigenic strains of Agrobacterium tumefaciens to six antibiotics
50 mg/L100 mg/L2.5 mg/L5 mg/L10 mg/L20 mg/L50 mg/L100 mg/L200 mg/L10 mg/L20 mg/L100 mg/L
  • Numbers indicate the percentage of colonies that formed in the presence of the antibiotic relative to non-selective conditions.

  • *

    The laboratory strain C58C1 contained plasmid pCH32, which carries TetR.

C58C1*  0  0100100< 0.0010 10 < 0.001  0100 10  0.005
1D1108100  0.1  0  0  0.10100100  0 10  1100
1D1249 10 10  0  0  0.0150  0.01  0.005  0  0.01  0.01  0
53-2A < 0.001< 0.001100 < 0.001< 0.0010100 10  0100100100
85-1A100  1 < 0.001 < 0.001  0.0010100100< 0.001 < 0.001 < 0.001 10
K12  0  0 < 0.001 < 0.001  0.0010 10 < 0.001  0  0.1  0.001 10
69-2B1 < 0.001< 0.001  0  0< 0.0010 10  5  0 10  1  5

The binary plasmid pTFS40 (British Sugar Corporation, Norwich, UK; Table 3), carrying the CaMV 35S promoter-controlled GUS-intron (Vancanneyt et al., 1990) reporter gene in its T-DNA, was then successfully introduced into 16 of the 24 tumorigenic strains (Table 1). For comparative purposes, the same plasmid was also introduced into three laboratory strains: C58C1, GV3101 (Holsters et al., 1980) and LBA4404 (Hoekema et al., 1983). Strain C58C1 carried pCH32, an additional helper plasmid (Hamilton, 1997; Table 3). Each of the 19 strains carrying pTFS40 successfully mediated transient expression of GUS as detailed below.

Table 3.  Vectors and constructs used
Construct or vectorHost plasmidReference for the host plasmidGene expressed in planta from CaMV 35S promoterReference or source
pTFS40pSLJ1006Jones et al. (1992)GUSBritish Sugar Corp, Norwich, UK
pDraGON-GpCB301Xiang et al. (1999)GUSUnpubl. authors own assemble
pDraGON-G:GFPpCB301Xiang et al. (1999)GUS, GFPUnpubl. authors own assemble
pMD1:AvrPTOpBIN19Bevan (1984)AvrPTOScofield et al. (1996)
P19N/AN/AP19 from TBSVA. Jackson, UC Berkeley, pers. comm.
TEV-P1/HcPropSLJ75515Jones et al. (1992)P1/HcPro from TEVJohansen and Carrington (2001)
TuMV-P1/HcPropCB301Xiang et al. (1999)P1/HcPro from TuMVJ. Carrington OSU (Sanchez et al., 1998)
pCH32pCC113Chen et al. (1991)NoneHamilton (1997)
pBIN m-gfp5-ERpBI121Jefferson et al. (1987)GFPHaseloff et al. (1997)

Utilization of wild Agrobacterium strains

To test the utility of the wild A. tumefaciens strains in transient assays, all 19 strains carrying pTFS40 (16 tumorigenic and three laboratory) were infiltrated into the leaves of six genotypes of four plant species: wild lettuce (Lactuca serriola, LS102), cultivated lettuce (L. sativa, cv. Valmaine and cv. Mariska), tomato (Lycopersicon esculentum cv. Rio Grande 76R), N. benthamiana and Arabidopsis thaliana (Col-0). Leaves were evaluated histochemically for GUS activity at 4 dpi (Jefferson et al., 1987). Significant differences between strains and plant genotypes were observed (e.g. Figure 2); A. tumefaciens strains were ranked by the level of GUS activity for each plant genotype. Of the three laboratory strains, C58C1 induced the highest expression in every plant species tested. Ten of the wild strains induced high levels of expression similar to the laboratory strain C58C1. To investigate the genotype or species by strain specificity, the patterns of activity were compared pairwise. For all three lettuce accessions (LS102, cv. Mariska, cv. Valmaine), the differences in expression between A. tumefaciens strains were consistent and therefore did not indicate genotype by strain specificity. Comparisons between lettuce, Nicotiana and Arabidopsis indicated some degree of genus by strain specificity. Four wild strains, 15955, 1D1249, A281 and K12, exhibited different patterns for each genus; for example, infiltration with 1D1249 and 15955 resulted in weak GUS expression in lettuce and Arabidopsis thaliana, but mediated strong expression in N. benthamiana. A281 also induced weak expression in lettuce, but strong expression in both Arabidopsis thaliana and N. benthamiana. K12 induced an intermediate level of expression in lettuce and Arabidopsis, but high levels in N. benthamiana (Figure 2).

Figure 2.

Comparison of transient gene expression induced by laboratory strains of Agrobacterium tumefaciens (C58C1 and GV3101) or six tumorigenic strains of A. tumefaciens harbouring binary plasmid pTFS40 in the leaves of wild lettuce Lactuca serriola (LS102), cultivated lettuce L. sativa (cv. Valmaine and cv. Mariska), cultivated tomato (cv. Rio Grande 57R), Nicotiana benthamiana and Arabidopsis thaliana (Col-0). Each panel is representative of at least 15 independent infiltrations.

In tomato, many of the tested strains induced high levels of GUS expression, but the results were variable due to severe necrosis in the infiltrated areas (Figure 2). The expression mediated by strains 69-2B1, 53-2A and 85-1A in Arabidopsis appeared to be slightly higher than that observed after infiltration with C58C1. Therefore, wild strains of A. tumefaciens can be used as efficient mediators of transient expression without additional engineering. However, laboratory strain C58C1 remained the most convenient agent for use with lettuce, N. benthamiana and Arabidopsis.

Evaluation of wild Agrobacterium strains on tomato and other solanaceous species

Because agroinfiltration with the currently available laboratory strains is often associated with necrotic responses in tomato and pepper (Van der Hoorn et al., 2000; and this paper), we screened the 19 A. tumefaciens strains to assay for strains that induced minimal necrotic background in tomato. Sixteen of the 19 strains utilized (Table 1), including, as expected, the laboratory strain C58C1 and its derivatives GV3101 and GV2260, induced necrosis in the infiltrated area in tomato cv. Rio Grande 76R. The amount of necrosis induced varied between the strains. Laboratory strains C58C1, GV3101 and LBA4404 caused necrosis at bacterial densities as low as OD600 = 0.1–0.2. Necrotic areas were often surrounded by a blue perimeter of cells exhibiting GUS activity (Figure 2).

Three strains, 1D1249, 1D1487 and 15955, did not elicit necrosis. Transient expression was much higher with 1D1249 than with either of the other two strains. On further testing, we found that 1D1249 did not elicit a background necrotic response in tomato, even when the density of the bacterial suspension used for infiltration was as high as OD600 = 2.0. We then tested the utility of 1D1249 on other solanaceous plants for which C58C1 cannot be used due to background necrosis. 1D1249 induced no necrosis in all genotypes tested; these included 21 varieties of cultivated tomato (Lycopersicon esculentum), two wild tomato species (Lycopersicon cheesmanii and Lycopersicon pimpinellifolium) and three accessions of pepper (Capsicum annuum and C. frutescens) (data not shown). Strain 1D1249 can therefore be used to conduct transient assays in tomato and other solanaceous species without the necrosis inherent in the use of current laboratory strains.

Evaluation of Arabidopsis ecotypes

Because transient assays in Arabidopsis are not as facile as in N. benthamiana and lettuce, we evaluated the utility of 12 different Agrobacterium strains on 10 different ecotypes of Arabidopsis thaliana. Leaves of Ag-0, Col-0, Col-4, Est, Gie-0, Mt-0, Tsu-1, Sei-0, Van-0 and WS were infiltrated with each strain that elicited the highest levels of transient GUS expression on the Col-0 ecotype during our initial screen. These strains included C58C1, GV3101, 1D1108, 1D1249, 1D1405, 1D1487, 53-2A, 85-1A, A281, 15955, K12 and 69-2B1. Plants were cultivated under different conditions from those used for the initial screens (see ‘Experimental procedures’). Ten leaves were evaluated for each ecotype by strain combination and assigned a numerical score relative to each other (data not shown).

A background response to A. tumefaciens was only evident in two ecotypes, Mr-0 and Ag-0. The infiltrated areas of Mr-0 leaves remained green; however, expansion of the infiltrated spot was clearly inhibited with all of the strains tested, and resulted in leaf deformation a few days after infiltration. Ecotype Ag-0 exhibited yellowing and necrosis, the extent of which was somewhat strain dependent.

Although relative differences in GUS expression were usually consistent from ecotype to ecotype (Figure 3a), some differences in the intensity of staining were observed amongst the strains tested. For the ecotype Columbia (Col-0), the level of GUS expression was remarkably higher in this experiment, when compared with the initial screens, presumably due to the altered growth conditions (Figure 2; see ‘Experimental procedures’). Similar increased GUS intensity was also observed in several other ecotypes, most notably Gie-0 and Tsu-1 (data not shown). The highest levels of expression were observed in the ecotype WS (Wassilewskija; WT-08A-10, Lehle Seeds, Round Rock, TX, USA); however, this accession also exhibited obvious plant to plant variability. This variability may have been due to genetic heterogeneity, as WS plants also differed slightly in their morphology and time to flower.

Figure 3.

Transient gene expression in the leaves of different Arabidopsis thaliana ecotypes after infiltration with eight strains of Agrobacterium tumefaciens harbouring the binary plasmid pTFS40 (Table 3). All plants were cultivated in a growth chamber illuminated by high-pressure sodium lamps (see ‘Experimental procedures’). (a) GUS activity in three Arabidopsis ecotypes mediated by six tumorigenic and two laboratory (C58C1 and GV3101) strains. No strain vs. ecotype specificity was observed, and ecotype Ws displayed the strongest expression of all the analysed accessions. The expression for Col-0 was stronger in comparison with the previous experiments, which used plants grown under different conditions (Figure 2). (b) GUS expression induced after infiltration with C58C1 containing pTFS40 in the leaves of five Arabidopsis ecotypes, reported by Nam et al. (1997) to be more amenable to transformation by A. tumefaciens: 1–4, Ms-0 (CS1376, CS1377, CS905 and CS6797, respectively); 5, M7884S (CS3113); 6, Wei-0 (CS3110); 7–10, Aa-0 (CS900, CS934, CS935 and CS6600, respectively); 11, Ws (CS915); 12, Be-0 (CS964).

To identify Arabidopsis accessions most suitable for transient expression, we evaluated an additional 12 lines originating from six ecotypes. These six ecotypes had been previously reported to be highly amenable to Agrobacterium-mediated stable transformation (Nam et al., 1997). Ecotypes Ms-0 (CS1376, CS1377, CS905, CS6797), M7884S (CS3113), Wei-0 (CS3110), Be-0 (CS964), Aa-0 (CS900, CS934, CS935, CS6600) and Ws (CS915) were all infiltrated with strain C58C1, carrying pTFS40, and assayed for GUS activity. All showed strong transient expression, particularly accessions CS6600 (Aa-0) and CS915 [Ws – Wassilewskija; Arabidopsis Biological Resource Center (ABRC), OH, USA] (Figure 3b). This indicates that the plant factors responsible for high efficiency of stable transformation also affect the efficiency of transient expression. In this experiment, the expression was uniformly high amongst the ecotype Ws plants represented by the line CS915 from the ABRC.

The influence of Agrobacterium on AvrPto-dependent HR

Because we wished to use transient assays in studies of disease resistance, we investigated whether wild A. tumefaciens affected the HR elicited in plant cells in response to a bacterial avirulence factor. Transient expression of AvrPto induces a clear HR in lettuce (data not shown). Therefore, pMD1:AvrPto (a binary plasmid carrying the AvrPto gene under the control of the CaMV 35S promoter; Table 3) was introduced into six strains: 1D1108, 1D1249, Ach5, A281, K12 and C58C1. Each strain was then used to infiltrate the leaves of wild lettuce (LS102). As a control for the expression level, the same panel of Agrobacterium strains, but carrying pTFS40 instead of pMD1:AvrPto, was infiltrated side by side into the same leaf. In each case, the intensity of the HR paralleled the intensity of the GUS stain. There was no case in which we observed high levels of GUS expression but reduced HR, or vice versa. In a second experiment, strain C58C1 carrying pMD1:AvrPto was mixed (1 : 1) with each of the 16 wild A. tumefaciens strains, or with C58C1 lacking pMD1:AvrPto as a control. Again, there was no indication of either an inhibitory or enhancing effect of any of the A. tumefaciens strains on AvrPto-induced HR. Therefore, there was no evidence that the use of A. tumefaciens would interfere with the evaluation of HR-inducing agents.

Effect of culture conditions and chemical components

Because many chemicals and physicochemical conditions have been reported to enhance transformation efficiency (or transient expression) in various plant species, we tested the effects of several of these compounds on transient expression. These included acetosyringone (Stachel et al., 1986; Rogowsky et al., 1987), a compound commonly added to cultures of Agrobacterium and to the medium used for agroinfiltration (Van der Hoorn et al., 2000), different ranges of pH (pH 5.5–8.5) during the infiltration, or the addition of l-cysteine (Olhoft and Somers, 2001; Olhoft et al., 2001) and azaserine (Roberts et al., 2003). We also tested the effects of supplemental energy sources, amino acids and mineral ions. These tests of transient gene expression were carried out using A. tumefaciens strain C58C1 (pTFS40) in leaves of lettuce (LS102) and N. benthamiana.

We observed no improvement in transient expression, when compared with our standard conditions, on addition of any of these compounds (data not shown). Similarly, no change in the level of transient gene expression was observed over the pH range tested in the infiltration solution. However, an inhibitory effect on gene expression was observed with concentrations of l-cysteine above 25 mm, l-glutamine above 100–150 mm and azaserine above 0.5 mm. This may indicate that many factors previously reported to improve the efficiency of stable transformation or transient expression may be specific to particular experimental situations, and their efficacy may not be general.

The effect of virE and virG genes

Because the over-expression of the virG gene of A. tumefaciens can enhance the transformation efficiency of celery, carrot and rice (Liu et al., 1992), and the over-expression of virE and virG can improve tomato transformation (Hamilton, 1997), we evaluated the effect of over-expression of these genes in our transient assays. We introduced pCH32, a helper plasmid designed to over-express both virE and virG (Table 3), into two wild A. tumefaciens strains, 1D1108 and 1D1249. In addition, strain C58C1, with and without pCH32, was used as a control. All three strains with and without pCH32 were also transformed with pDraGON-G (Table 3), a binary vector carrying a chimeric GUS transgene.

Slightly higher GUS activity was observed in leaves infiltrated with A. tumefaciens strains 1D1108 and C58C1 carrying pCH32, compared with the isogenic strain lacking the plasmid (Figure 4). However, no difference in GUS activity was observed in leaves infiltrated with strain 1D1249 with and without pCH32. Thus, it appears that over-expression of virE and virG genes can enhance transient expression mediated by some strains, but not in all cases.

Figure 4.

Effect of VirE and VirG genes on the transient expression induced by two strains of Agrobacterium tumefaciens in the leaves of wild lettuce Lactuca serriola. VirE and VirG genes were carried on the helper plasmid pCH32 (Hamilton, 1997). The addition of pCH32 slightly enhanced the GUS activity mediated by C58C1, but not by 1D1249. Average patterns of expression from six replicated infiltrations are shown.

The effect of viral silencing suppressors

We then investigated whether we could increase the level of transient expression by adding inhibitors of post-transcriptional gene silencing (PTGS). In N. benthamiana, high levels of expression following agroinfiltration have been reported to trigger PTGS; the co-expression of viral silencing suppressors interferes with PTGS, resulting in enhanced expression of GFP (Voinnet et al., 2003). To test whether such suppressors would be efficacious in our system, we performed parallel experiments with lettuce (LS102) and N. benthamiana using two marker genes simultaneously. GUS and GFP were introduced using C58C1 harbouring pDraGON-G:GFP (Table 3), a binary vector carrying both of these reporter genes, each controlled by separate but identical CaMV 35S promoters in its T-DNA. Prior to infiltration, a suspension of C58C1 (pDraGON-G:GFP) was mixed, in a 1 : 1 ratio, with a suspension of C58C1 harbouring one of the three following viral silencing suppressors in the binary plasmid: P19 (from tomato bushy stunt virus, TBSV), P1/HcPro (from turnip mosaic virus, TuMV) or P1/HcPro (from tobacco etch virus, TEV) (Table 3). Expression of all three silencing suppressors was controlled by the CaMV 35S promoter. A 1 : 1 mixture of C58C1(pDraGON-G:GFP) with a suspension of C58C1 lacking additional plasmids was used as a control. The levels of GFP and GUS gene expression were then evaluated at 4 dpi.

In N. benthamiana, the activity of each of the three silencing suppressors clearly resulted in stronger expression of GFP (Figure 5). P19 seemed to have the greatest effect, in agreement with the observations of Voinnet et al. (2003). When the same leaves were assayed for GUS activity, no obvious enhancement was observed in conjunction with any of the three suppressors (Figure 5). The greater stability of GUS relative to GFP protein probably accounted for the differences observed between these two marker genes. Therefore, co-expression of viral silencing suppressors may enhance the level and/or durability of transient expression in N. benthamiana, especially of less stable proteins.

Figure 5.

Effect of the co-expression of viral silencing suppressors on the transient expression of GFP and GUS in the leaves of Nicotiana benthamiana and Lactuca serriola (LS102). C58C1 (pDraGON-G:GFP; Table 3) was mixed 1 : 1 with C58C1 containing three different viral silencing suppressors under the control of the 35S promoter (see Table 3) in the following binary vectors: c, control: pCB301‘empty’; TE, TEV-P1/HcPro; Tu, TuMV-P1/HcPro; P19, P19 from TBSV. Transient expression assays were carried out 4 days post-infiltration. (a) Enhancement of expression of GFP in N. bethamiana leaf due to the activity of P19 and, to a lesser extent, by TEV-P1/HcPro and TuMV-P1/HcPro. (b) Lack of an observable effect of the silencing suppressors on GUS activity in N. bethamiana. (c) Lack of an observable effect of the silencing suppressors on GFP activity in wild lettuce L. serriola (LS102). (d) Lack of an observable effect of viral silencing suppressors on GUS activity in wild lettuce L. serriola (LS102). All leaves are representative of at least 10 replications.

In lettuce, none of the three silencing suppressors had any obvious effect on GFP or GUS expression (Figure 5). Even at 8 dpi (data not presented), when the effects of the silencing suppressors were particularly obvious with GFP in N. benthamiana, the patterns of transient gene expression in lettuce were similar with and without the silencing suppressors. The ability of these silencing suppressors to prevent PTGS of transiently expressed genes therefore appears to be dependent on plant genotype, and to some extent may reflect the host specificity of the viral sources of the suppressors themselves.

None of the three silencing suppressors had an observable effect on transient GUS expression in Col-0 (data not presented). One suppressor was derived from a virus that is pathogenic in Arabidopsis (TuMV). Thus, there was no evidence that silencing was responsible for the relatively weak transient expression in Arabidopsis.

Transient vs. stable gene expression in Arabidopsis

We also investigated whether PTGS could be responsible for the lower level of GUS expression in stably transformed plants, compared with the high level of expression observed in transient assays with Arabidopsis thaliana plants (ecotype Ws-0; CS915) from the same T-DNA (Figure 6). We transiently expressed the viral silencing suppressor P1/HcPro from TuMV (Table 3) in the leaves of Ws-0 plants that had been stably transformed with 35S-GUS and were expressing the GUS gene. Much higher GUS activity was often observed in the regions of the leaves in which the silencing suppressor had been infiltrated, compared with the regions infiltrated with a control strain that lacked the silencing suppressor (Figure 6). Therefore, PTGS was indeed at least partially responsible for the lower levels of GUS activity observed in the transgenic plants.

Figure 6.

Effect of transient expression of the TuMV-P1/HcPro silencing suppressor on the level of GUS activity in transgenic Arabidopsis thaliana (ecotype Ws, CS915). The detached leaves above illustrate the typical levels of GUS staining obtained in transient assays utilizing Agrobacterium C58C1 carrying pTFS40 with 35S-GUS in wild-type plants. The leaves of the intact plant stably transformed with the same T-DNA (35S-GUS) were infiltrated with Agrobacterium C58C1 carrying TuMV-P1/HcPro and, across the leaf midvein, with C58C1 carrying the empty vector as a control. Clear enhancement of GUS expression indicates that the lower level of expression in the transgenic plant is due to post-transcriptional gene silencing (PTGS).

The differences between the effects of PTGS on transient vs. stable transgene expression may reflect the timing of the onset of PTGS relative to the timing of the gene expression assays. In transgenic plants, the level of expression is the net result of long-term transcription and translation limited by PTGS. In transient assays, high levels of gene product may accumulate prior to the initiation of PTGS.


We studied a range of genetic and physiological factors affecting transient expression and developed a protocol for efficient and routine transient assays in several plant species. Lettuce proved to be amenable to high levels of transient expression and was at least as easy to work with as the more widely used N. benthamiana; transient expression occurred in the majority of cells and approached 100% in some regions of the leaf. The level of transient expression often exceeded that observed in stable transgenics. The laboratory strain C58C1 remained the best strain for use in transient assays of gene expression in plant species that did not elicit a necrotic response to A. tumefaciens.

We used a genetic approach to overcome the incompatibility between tomato and laboratory strains of A. tumefaciens by screening a collection of tumorigenic strains. Although the majority of tumorigenic strains surveyed also elicited a necrotic response in tomato, three strains did not. Of these, 1D1249 mediated high levels of transient expression in tomato and other solanaceous plant species. This has increased the diversity of strains available for transient assays. Transient assays of tomato with strain 1D1249 are now routine without the risk of background necrosis. We attempted to disarm 1D1249; however, extended periods of subculture at elevated temperatures (Hamilton and Fall, 1971) failed to cure 1D1249 of its Ti plasmid (data not shown). 1D1249 is therefore not currently useful for generating stable transgenic plants.

Necrotic responses to Agrobacterium have been reported previously for grapes (Yanofsky et al., 1985a, 1985b), tomato, potato, cucumber and pepper (Van der Hoorn et al., 2000). The reaction in grapes was explained as a sensitivity to elevated levels of auxin produced by the wild-type T-DNA (Deng et al., 1995); however, in our studies, necrosis also occurred when disarmed strains were used. Avirulence factors of bacterial phytopathogens are delivered to the plant cell via a Type III secretion system (reviewed by Salmond, 1994). Although Agrobacterium lacks Type III secretion (Goodner et al., 2001; Wood et al., 2001), it is possible that bacterial proteins are transferred to the plant cell by the Type IV secretion system; these may then be recognized and trigger necrosis in some species. Another candidate for the induction of necrosis is flagellin. Tomato seems to be particularly sensitive to flagellin (Felix et al., 1999), which is consistent with its sensitivity to many analysed strains of A. tumefaciens.

Transient assays of Arabidopsis are still not as routine as for lettuce, tomato and N. benthamiana. We increased expression by changing the plant growth conditions and selecting the best ecotypes to assay; however, the results with Arabidopsis were still variable and fluctuated from experiment to experiment more than in the other species tested. Obtaining reliable data in Arabidopsis required multiple repetitions. Further optimization of the experimental conditions may yet provide more reliable transient assays in Arabidopsis.

Important variables appeared to be the compatibility between A. tumefaciens and the plant, as well as the physiological condition of both organisms. Empirical observations made over the course of these experiments suggested that plant age and fitness were critical. Young tissue composed of newly expanded cells from vigorously growing plants often exhibited the highest levels of transient expression; this may reflect the elevated physiological activity of such cells. In addition, potential temporal and spatial differences in the induction of PTGS may generate further variation. The durability of expression could be enhanced by the co-expression of viral silencing suppressors (Voinnet et al., 1999, 2003). However, this may require the use of silencing suppressors that are specifically selected to be effective in the targeted plant species (reviewed by Voinnet, 2001). Interestingly, many of the factors reported to be important in transformation efficiency did not have obvious effects under our conditions in the species tested. Future enhancement of assays for transient gene expression will probably involve optimization of the plant's physiological condition, manipulation of factors promoting T-DNA transfer, optimization of transcription and the application of silencing suppressors.

The presence of A. tumefaciens seemed to have little influence on transient expression in our assays. Changes in gene expression patterns are known to be induced by A. tumefaciens (Ditt et al. 2001; Veena et al., 2003). However, the presence of wild-type T-DNA neither enhanced nor decreased GUS activity or the degree of necrosis induced by AvrPto. This was surprising because the transfer of wild-type T-DNA would be expected to result in elevated production of IAA (indole-3-acetic acid) and zeatin (Gheysen et al., 1985). Furthermore, the presence of the tms gene (involved in the IAA synthesis pathway) in the T-DNA has been reported to inhibit HR elicited by Pseudomonas in tobacco (Robinette and Matthysse, 1990). Although the wild-type T-DNA did not interfere with our assays, the use of wild Agrobacterium strains may not be appropriate for functional analyses of certain types of gene, particularly those involving plant hormones.

In comparison with the other systems of assessment of in planta transgene expression, Agrobacterium-mediated transient assay has several advantages. Unlike the viral vectors, it is not limited to small proteins (Porta et al., 1996). It is much simpler to perform than particle bombardment and protoplast transformation. It is rapid and, unlike stable transformation, allows the analysis of genes that have deleterious effects on growth and development. In addition, analysis involves the simultaneous assay of many independently transformed cells and is not confounded by somaclonal variation that can occur in stable transformation experiments. The amount of tissue transiently expressing a gene is also sufficient to perform both RNA and protein analysis, and is potentially useful for laboratory-scale in planta protein production (Vaquero et al., 1999). Earlier reports have indicated that the level of transient expression in tobacco is up to a maximum of five times higher than that in stably transformed plants (Vaquero et al., 2002). In our experiments, the activity of GUS in lettuce leaves after agroinfiltration was more than 20 times higher than that in stably transformed lettuce plants. Furthermore, ‘in spot gene silencing’ offers the possibility for transient RNAi experiments (Johansen and Carrington, 2001), and therefore transient assays will have an increasing role in functional genomics. However, transient assays have some limitations. One constraint is that they are only highly effective in a subset of plant species. More importantly, only those genes that provide a measurable phenotype within a few days in the targeted tissue (usually leaves) are amenable to transient analysis. This still includes a large number of genes, however, and as methods of analysis become more sophisticated, the numbers of assessable, and therefore accessible, genes will increase.

Experimental procedures

Agrobacterium culture

Competent cells of all strains of A. tumefaciens were prepared according to Sambrook et al. (1989) and the Cell-Porator Manual (Life Technologies-Invitrogen, Carlsbad, CA, USA) with the following minor modifications. The concentrations of NaCl and KCl in SOC medium (Sambrook et al., 1989) were reduced by half and the bacteria were washed and resuspended for storage in 5% glycerol. YEP medium (Bacto-Trypton, 10 g/L; yeast extract, 10 g/L; NaCl, 5 g/L; pH 7.5) was used for liquid and solid (10 g/L of agar) bacterial cultures. Tetracycline (5 mg/L) and kanamycin (50 mg/L) were used to maintain pTFS40 in A. tumefaciens. Kanamycin-resistant strains (1D1108, 1D1249) containing pDraGON or pMD1, for which nptII was the sole selectable marker gene, were maintained by elevating the level of kanamycin to 150 mg/L. (Routine use of 1D1249 with various plasmids carrying nptII as a sole selection marker required an increase in the kanamycin concentration to 250–300 mg/L – data not presented.)

In experiments to test the efficacy of additives, acetosyringone was added to the liquid culture to a final concentration of 200 µm at pH 5.6, as described by Kapila et al. (1997). The pH in the infiltration solution was stabilized by adding 10 mm 2-(N-morpholino)ethanesulphonic acid (MES) (pH 5.5, 6.0, 6.5) or 10 mm tris(hydroxymethyl)aminomethane (Tris) (pH 7.0, 7.5, 8.0, 8.5). l-Cysteine was tested at concentrations of 1, 3, 10, 25, 50, 100 and 200 mm, azaserine at 0.05, 0.1, 0.2, 0.5, 1.0, 2.0 and 5.0 mm and l-glutamine at 1, 3, 10, 25, 50, 100 and 200 mm in the infiltration solution. Glucose was tested at 0.5% and magnesium chloride at 10 mm.

Plant material and growth conditions

Tomato seeds were obtained from the Tomato Genetics Resource Center (UC Davis; tgrc.ucdavis.edu) and pepper seeds were provided by M. Jahn (Cornell, Ithaca, NY, USA). Seeds of Arabidopsis ecotype WS (WT-08A-10) were purchased from Lehle Seeds, Round Rock, TX, USA. Ecotypes Ag-0 (CS22630), Col-4 (CS933), Sei-0 (CS1504) and Gie-0 (N1192) were obtained from the ABRC (http://www.biosci.ohio-state.edu/~plantbio/Facilities/abrc/abrchome.htm). Ecotypes Col-0, Est, Mt-0, Tsu-1 and Van-0 were supplied by J. Chory (Salk Institute, CA, USA). The other ecotypes, with their respective accession numbers, listed in Figure 3 were also obtained from the ABRC.

All of the lettuce, tomato, pepper and N. benthamiana plants used in these experiments were grown in a glasshouse. The temperature varied between 22 and 24 °C during the day and decreased during the night to approximately 18–20 °C. Most of the experiments were performed during the autumn and spring when the day and night lengths were approximately even. The light intensity varied between 400 and 1000 µE. During the winter, we provided additional light generated by high-pressure sodium lamps (300 W per 2 m2, ∼1 m above the plants) from 15.00 h to midnight. After infiltrations, plants were maintained under the same conditions.

For the initial experiments, in which wild strains of A. tumefaciens were screened, Arabidopsis plants were grown in a growth chamber at 20 °C with a 10 h day/14 h night photoperiod produced using fluorescent light (120 W/m2, 0.5 m above the plants, corresponding to approximately 160 µE) and a relative humidity of 70%−80%. For later experiments, evaluating the various Arabidopsis ecotypes, plants were also cultivated in a growth chamber, but growth was enhanced by using two 300 W high-pressure sodium lamps. Lamps were placed 1 m above the plants and produced an intensity of 270 µE at the plant level. The temperature was 22 °C during the 16 h day and 18 °C during the 8 h night, and the relative humidity was kept at 60% during these later experiments.

Crown-gall test for virulence of A. tumefaciens

The epicotyls of 2-week-old tomato seedlings (cv. Moneymaker) were pierced with a syringe equipped with a 21G1 needle. One drop of the A. tumefaciens bacterial suspension contained in the syringe (prepared as for leaf infiltrations, see below) was then ejected on to the stem where it was rapidly absorbed. The first signs of tumour formation were visible after 2 weeks. Evaluations for the presence or absence of tumours were made after 6 weeks.

Leaf infiltration

The day prior to infiltration, liquid cultures of A. tumefaciens were initiated from bacteria growing on agar plates. The cultures were grown in YEP liquid medium overnight at 29 °C on a shaker. A new culture was started the next morning by inoculating fresh medium with the overnight culture (1 : 10 ratio, v : v). These cultures were grown under the same conditions for an additional 5–7 h. Bacteria were then harvested by centrifugation at 1000 g and resuspended in dd water. Bacterial densities were adjusted to OD600 = 0.4–0.5 prior to infiltration. Bacteria were maintained at room temperature rather than on ice until use, which was always within 3 h of harvesting.

Infiltrations were performed as described by Schob et al. (1997). This involved infusion of the bacterial suspension by applying pressure against the lower side of a leaf lamina with a syringe containing the bacterial suspension but lacking a needle. Young leaves were used for the majority of these experiments: second to sixth true leaves prior to full expansion (about 1/2–2/3 of the full size) for lettuce, second and older true leaves (about 2/3–3/4 of their full size) for tomato, fourth and older true leaves (about 4/5–1/1 of their full size) for N. benthamiana, and tenth and older true leaves (about 1/3–1/2 of their final size) for Arabidopsis. Only Arabidopsis plants in the vegetative stage were used, as leaf expansion was greatly reduced in flowering plants.

Evaluation of gene expression

Quantitative evaluations of GUS expression were carried out following Jefferson et al. (1987) using 1 mm MUG (4-methylumbelliferyl-β-d-glucuronide) solution in the extraction buffer. GUS activity was evaluated after 15 min of incubation at 37 °C, at which time the number of nanomoles of 4MU (4-methylumbelliferone) generated by 1 mg of total protein in the extract during 1 min of incubation time (nmol/mg/min) was calculated. Histochemical staining was performed following Jefferson et al. (1987). During staining, the leaves were incubated at 37 °C with gentle agitation for 6 h. The ratio of the staining solution to the plant tissue was always 10 : 1. Each Arabidopsis photograph in Figure 3 is representative of at least 20 infiltrated leaves from two replicated experiments. Leaves were collected at 4 dpi.

GFP activity was visualized using a FluoroImager (Molecular Dynamics, Sunnyvale, CA, USA) or a Confocal Laser Scanning Microscope (Leica TCS-4D).


We thank C. Kado, Shu-Yi Wang, L. Epstein, A. Dandekar and M. Escobar (all from UC Davis) for the wild strains of A. tumefaciens, Jung-Youn Lee (W.J. Lucas Laboratory, UC Davis) for help with confocal microscopy, A. Jackson (UC Berkeley) and J. Carrington (Oregon State University) for viral silencing suppressors, O. Ochoa for glasshouse assistance, M. Shetab and K. Lahre for technical help and B. Martineau for critical reading of the manuscript. This research was supported by the NSF Plant Genome Program, award #0211923.