Plant Transformation by Coinoculation with a Disarmed Agrobacterium tumefaciens Strain and an Escherichia coli Strain Carrying Mobilizable Transgenes

Abstract

Transformation of Nicotiana tabacum leaf explants was attempted with Escherichia coli as a DNA donor either alone or in combination with Agrobacterium tumefaciens. We constructed E. coli donor strains harboring either the promiscuous IncP-type or IncN-type conjugal transfer system and second plasmids containing the respective origins of transfer and plant-selectable markers. Neither of these conjugation systems was able to stably transform plant cells at detectable levels, even when VirE2 was expressed in the donor cells. However, when an E. coli strain expressing the IncN-type conjugation system was coinoculated with a disarmed A. tumefaciens strain, plant tumors arose at high frequencies. This was caused by a two-step process in which the IncN transfer system mobilized the entire shuttle plasmid from E. coli to the disarmed A. tumefaciens strain, which in turn processed the T-DNA and transferred it to recipient plant cells. The mobilizable plasmid does not require a broad-host-range replication origin for this process to occur, thus reducing its size and genetic complexity. Tumorigenesis efficiency was further enhanced by incubation of the bacterial strains on medium optimized for bacterial conjugation prior to inoculation of leaf explants. These techniques circumvent the need to construct A. tumefaciens strains containing binary vectors and could simplify the creation of transgenic plants.

Agrobacterium tumefaciens is well known for its ability to mediate transfer of DNA and proteins into the nuclei of plant cells, providing one of the best-studied examples of horizontal DNA transfer and the only known natural instance of interkingdom DNA transfer. In nature, A. tumefaciens detects a variety of chemical signal molecules released from plant wound sites and responds by expressing a battery of dedicated virulence (Vir) proteins, which are encoded by genes residing on the Ti plasmid. These proteins play various roles in the processing and transfer of a portion of the Ti plasmid, called the T-DNA, into the plant cell. A subset of T-DNA-encoded genes direct the production of compounds called opines, which serve as nutrients for the colonizing agrobacteria, while other transferred genes direct the overproduction of phytohormones, leading to the proliferation of infected cells and tumor formation (4, 23, 41, 57).

A. tumefaciens-mediated plant transformation technology has been used extensively in the past decades to introduce genes into dicotyledonous plants and, more recently, into monocots, including rice and maize (20, 24, 37). This process requires the use of so-called disarmed strains, which retain the full complement of vir genes but lack the native T-DNA. The DNA to be transferred is then usually introduced on a separate plasmid, called a binary vector, and is flanked by cis-acting DNA sequences called borders. Binary plasmid vectors must have one or more replication systems that allow replication both in Escherichia coli and in A. tumefaciens, as well as antibiotic resistance genes that allow selection in both bacterial hosts. The transgene is generally linked to a second antibiotic resistance gene that is expressed in plant cells and confers resistance to a plant-specific biocide (1, 21). A. tumefaciens-mediated transformation offers several advantages over alternative methods of plant transformation, since relatively long DNA fragments are stably transferred to plants with few, if any, rearrangements and are integrated into the plant cell nuclear DNA as one copy or a small number of copies. Cotransformation of multiple T-DNAs has also been described, in which one T-DNA contains a biocide resistance-encoding gene while the other contains some gene of scientific or commercial importance. The resulting transgenic plant can be crossed to the wild type, resulting in segregation of the two T-DNAs and generation of marker-free transgenic plants (20, 24).

The process of T-DNA transfer by Agrobacterium is strikingly similar to conjugal DNA transfer by conjugative plasmids. In both phenomena, site-specific endonucleases nick the DNA to be transferred on one strand and remain covalently bound to the 5′ end of the DNA strand to be transferred. In the T-DNA transfer system, this requires the VirD1 and VirD2 proteins, with VirD2 remaining bound to the single-stranded T-DNA, referred to as a T strand. Both types of transfer require a multiprotein membrane-associated transfer apparatus, sometimes called a mating bridge. The T-strand transfer apparatus is encoded by the 11 virB genes and the virD4 gene, which strongly resemble the mating bridges of various conjugation systems (30). This system can transfer several proteins to plant cells in addition to the T-DNA. Two proteins, namely, VirD2 (the conjugal relaxase) and VirE2 (a single-stranded DNA-binding protein), enter the plant cell and contain nuclear localization motifs that mediate nuclear targeting of the transferred DNA (38, 41). Mutations in VirE2 cause greatly impaired tumorigenesis, although virulence can be restored by coinfecting the plant tissue with a T-DNA-deficient virE2⁺ strain or by using transgenic plants that express VirE2 (9, 33).

The VirB pore complex belongs to a gene family called type 4 secretion systems (TFSS) (12, 39). VirB proteins are extremely similar to the mating pair formation operons of the conjugal transfer (tra) systems of IncN and IncW plasmids, such as pKM101 and R388, respectively, as well as the pertussis toxin secretion apparatus of Bordetella pertussis and the toxin export machinery of Brucella spp. (5, 8, 53, 55). It has less, but still significant, similarity to the mating bridge apparatus of IncP plasmids such as RK2 and still less similarity to several other TFSS, including those of Helicobacter pylori and Legionella pneumophila. The VirB system, in conjunction with VirD4, can mobilize derivatives of the nonconjugative broad-host-range IncQ plasmid RSF1010 to agrobacteria or plant cells (3, 6).

TFSS are able to convey DNA and proteins into diverse types of recipient cells. For example, the Vir system can mediate DNA transfer to plants, bacteria, yeast, higher fungi, and mammalian cells (3, 11, 28, 36), whereas the IncP tra system can mediate transfer to most eubacteria, yeast, and mammalian cells (15, 19, 51). In this study, we examined whether the promiscuous IncP and IncN group tra systems of plasmids RK2 and pKM101, respectively, can mediate gene transfer to leaf cells of Nicotiana tabacum. These conjugation systems were chosen on the basis of their similarity in gene organization and function to the Ti plasmid vir system (8, 31, 53). In this study, E. coli host strains harboring IncP-type or IncN-type tra systems were tested as donor strains of DNA fragments containing genes that can be selected in plant cells. These strains were unable to stably deliver transgenes to host plants by themselves. However, when a strain containing an IncN-type system was coinoculated with a disarmed A. tumefaciens strain, numerous calli were readily detected that gave rise to stable transgenic plants.

MATERIALS AND METHODS

Bacterial strains, plasmids, and growth conditions.

The A. tumefaciens and E. coli strains and plasmids used in this study are listed in Table 1. A. tumefaciens strains were cultured in AT minimal liquid and solid medium (7) at 28°C. E. coli was cultured in L broth or L agar plates at 37°C (40). For A. tumefaciens, antibiotics were added at the following concentrations: 150 μg/ml for spectinomycin, 100 μg/ml for kanamycin, and 50 μg/ml for carbenicillin. For E. coli, antibiotics were added at the following concentrations: 50 μg/ml for ampicillin, 50 μg/ml for kanamycin, 100 μg/ml for spectinomycin, and 15 μg/ml for tetracycline.

TABLE 1.

Bacterial strains and plasmids used in this study

Strain or plasmid	Relevant features	Source or reference
E. coli
DH5α	endA1 hsdR17(rK− mK+) gyrA thi-1 recA1 relA1 supE44 φ80dlacZΔM15 Δ(lacZYA-argF) U169	40
AB1157	thr-1 leu-6 proA2 his-4 thi-1 argE3 lacY1 galK2 ara-14 xyl-5 mtl-1 tsx-33 rpsL31 supE44	52
JC2926	AB1157 recA13	52
S17-1	thi pro hsdR hsdM+recA RP4(tet::Mu kan::Tn7)	44
A. tumefaciens
EHA101	Strain C58 containing a disarmed derivative of pTiBo542	22
At10002	A348 ΔvirD1 ΔvirD2 ΩnptII	36
Plasmids
pPZP212	Binary cloning vector for Agrobacterium-mediated plant transformation; Spr Smr	18
pMAL-c2X	Protein overexpression vector; Apr	New England Biolabs
RP4	Apr Kmr Tcr IncP-1; P-type tra	47
pKM101	Apr IncN; N-type tra	35
pGW1514	pKM101Ω.925::Tn5 with the bla gene deleted; Kmr	54
pVK228	pTiA6 cosmid clone; Kmr	26
pKP9	348-bp PCR amplicon of the RP4 oriT region cut and inserted into the EcoRI-HindIII site of pPZP212	This study
pKP10	534-bp PCR amplicon of the pKM101 oriT region cut and inserted into the EcoRI-HindIII site of pPZP212	This study
pKP40	4.2-kb NdeI band from cosmid clone pVK228, harboring the virE1-virE3 genes, cloned into NdeI-digested pMAL-c2X	This study
pKP50	535-bp PCR amplicon of the pKM101 oriT region cut with HpaI and inserted into the PmeI site of pPZP212	This study
pKP80	535-bp PCR amplicon of the pKM101 oriT region cut with HpaI and inserted into the second ScaI site (base position 6532) of pPZP212	This study
pKP80ΔN	pKP80 cut with NotI and religated	This study
pKP81	pKP80ΔN cut with ScaI and NdeI end filled at the NdeI end, and religated	This study

Bacterial DNA isolation, recombinant DNA techniques, and PCR amplification.

E. coli plasmids were isolated in accordance with the alkaline lysis method (40). Recombinant DNA techniques and purification of genomic DNA were carried out as described previously (34, 40). PCR amplifications for cloning purposes were carried out with a Hybaid Thermal Reactor and PfxI (Gibco-BRL); reaction conditions were as recommended by the manufacturer, except that 1-min denaturation, annealing, and synthesis steps were used for 35 cycles and primer annealing was carried out at 52°C. DNA sequences of generated fragments and subsequent clones were verified by automated DNA sequencing and analyzed with the LaserGene software package (DNASTAR). PCR amplifications for transgene profiling were carried out with Taq polymerase (Promega) in accordance with the manufacturer's instructions. In both bacterial and plant template PCRs, each reaction contained 50 ng of template DNA, 2 pmol of each primer, and each deoxynucleoside triphosphate at 200 μM. The oligonucleotide primers used are listed in Table 2.

TABLE 2.

Oligonucleotides used in this study

Primer	Sequencea	Amplified region
oriTNF1	5′-ATTGGAATTCAGTTCCTCACAT-3′	pKM101 oriT
oriTNR1	5′-CTTTTAAGCTTCATAGTACCCTCA-3′
oriTPF	5′-CTGGAATTCAGTACACCTTGATAG-3′	RP4 oriT
oriTPR	5′-AGCAAAGCTTTTCCGCTGCATAAC-3′
kanF	5′-GCTGCGAATCGGGAGCGGCGATAC-3′	Part of nptII gene
kanR	5′-CGCTTGGGTGGAGAGGCTATTCGG-3′
actF	5′-TGATGGTGTTAGCCACACTGTCCC-3′	Part of tobacco actin gene
actR	5′-CTCTCAGGTGGAGCTACCACCTTA-3′
repColE1F	5′-TGCCTCGCGCGTTTCGGTGATGAC-3′	Part of ColE1 rep region
repColE1R	5′-ACAGGTATCCGGTAAGCGGCAGGG-3′
fliCF	5′-ACAGGATCCGCGGTAAACGACGAT-3′	Part of E. coli fliC gene
fliCR	5′-TTAAAGCTTGCCAGAAGACAGACG-3′
repANF	5′-CAAAGCGCGTTCTCTGGTTATGTC-3′	Part of pKM101 repA gene
repANR	5′-AGCATGAGTAGTAACCCATAAGCC-3′
27F	5′-AGAGTTTGATCCTGGCTCAG-3′	Universal 16S rRNA primers
1492R	5′-TACGG(T/C)TACCTTGTTACGACTT-3′

^aRestriction sites are in italics. F and R denote forward and reverse primers. All primers were designed in this work, apart from 27F and 1492R (16).

Strain constructions and mating techniques.

Plasmids were introduced into E. coli by transformation (40) and into A. tumefaciens by electroporation of mid-log-phase cells that had been washed and concentrated 20-fold in 15% glycerol. Cell suspensions for inter- and intraspecies bacterial conjugations were concentrated 20-fold by mild centrifugation (7,500 × g for 3 min), deposited on MF-Millipore 0.45 μm filters on Luria-Bertani (LB) agar plates, and incubated for 5 h at 28°C, unless otherwise indicated. The donor-to-recipient cell ratio was 1:10. Following the mating interval, cells were resuspended from the filters in 0.9% NaCl solution, serially diluted, and plated on the appropriate selective medium. E. coli-A. tumefaciens mating mixtures that were to be used directly in N. tabacum infections were prepared similarly, except that a 1:1 cell ratio was used and the cells were resuspended in transformation-specific liquid TTR medium (see below) at the end of the mating interval.

Plant DNA isolations.

DNA was isolated from leaves of young plantlets (fresh or frozen at −80°C) in accordance with the following protocol. A 200- to 300-mg tissue sample was rinsed in distilled water and dipped in 500 μl of lysis buffer containing 175 mM sorbitol, 1 M NaCl, 150 mM Tris-HCl (pH 7.5), 2.5 mM EDTA (pH 8.0), 1% cetyltrimethylammonium bromide, 0.8% N-laurylsarcosine and 0.16% (wt/vol) freshly added sodium bisulfite. Tissue samples were disrupted in lysis buffer with a Dounce homogenizer, an equal volume of chloroform was added, and the mixture was vortexed to complete emulsification. The mixture was clarified by centrifugation (15,000 × g, 5 min), and a second chloroform extraction was carried out. The aqueous phase was transferred to a new tube and nucleic acids were precipitated with a 2/3 volume of isopropanol. The nucleic acid pellet, recovered after centrifugation for 5 min, was washed with 70% ethanol, dried, and resuspended in TE buffer (10 mM Tris-HCl, 1 mM EDTA, pH 7.8) supplemented with 50 μg of RNase per ml). DNA samples were then heated to 65°C for 30 min to inactivate nucleases.

Tobacco transformations.

N. tabacum cv. Samsun NN plants were used for transformation. Leaves from young plants were collected and surface sterilized (46). Bacterial cultures grown to mid-log phase were centrifuged and resuspended in tobacco transformation and regeneration medium (TTR medium) containing 1× Murashige and Skoog salts medium (Gibco BRL) supplemented with 3% sucrose, 1 mg of thiamine per liter, 100 mg of myo-inositol per liter, 1 mg of benzyladenine per liter, and 0.1 mg of naphthylene acetic acid per liter (pH 5.8). Explant leaves were cut aseptically into 1-cm² strips and immediately dipped into the bacterial cell suspensions (1 × 10⁸ to 5 × 10⁸ CFU/ml). In E. coli-A. tumefaciens coinfection assays, suspensions of similar cell numbers were combined. Leaf strips, after 3 to 5 min of treatment in the bacterial suspensions, were transferred and incubated on TTR solid medium (liquid medium plus 0.6% phytagar) for 36 to 48 h at 24°C under constant light. After the bacterial-explant tissue cocultivation period, the strips were transferred and incubated on TTR selective plates, supplemented with 100 μg of kanamycin per ml and 200 μg of timentin per ml. Tumors were counted after 10 to 20 days. Regenerated shoots were cultivated and rooted in TTR selective medium without the phytohormones until ready for planting in soil. Induction of cloned virE genes in E. coli cultures harboring plasmid pKP40 was done by addition of 1 mM isopropyl-β-d-thiogalactopyranoside (IPTG) in early log phase cultures and further incubation for 3 to 4 h, until cell harvest. In these experiments, IPTG (0.5 mM) was also added to TTR medium at all N. tabacum infection stages prior to the application of antibiotics.

RESULTS

Plant-directed shuttle vectors mobilizable by the IncP and IncN transfer systems.

In order to examine whether the IncP and IncN conjugal transfer systems could be used to mobilize plasmid DNA directly into plants, the origins of transfer (oriT) from plasmids RP4 and pKM101 (17, 35) were PCR amplified and cloned into the multicloning site of plasmid pPZP212 (18). The resulting plasmids, pKP9 and pKP10 (harboring the IncP and IncN transfer origins, respectively), retained all of the features of pPZP212 (Fig. 1). Plasmid pKP9 was introduced into E. coli strain S17-1, which contains the IncP conjugation system, while pKP10 was introduced into strain DH5α(pKM101), which harbors the IncN conjugation system. Filter matings between these donor strains and E. coli recipients (Table 3) showed that both vectors were mobilized with extremely high efficiency.

FIG. 1.

Plasmids carrying the IncN-type or IncP-type origin of conjugal transfer region (oriT) designed for plant transformation. All plasmids are derived from pPZP212 (18), whose functional map is shown at the top. pVS1 ori denotes the broad-host-range replication origin of plasmid pVS1, while pBR322 ori denotes the high-copy replication origin of plasmid ColE1. The aadA gene product confers resistance to spectinomycin and streptomycin. LB and RB are the left and right T-DNA borders, respectively, nptII is the neomycin phosphotransferase gene conferring resistance to kanamycin, P_35S is the cauliflower mosaic virus 35S promoter, lacZ is the LacZ α-peptide gene, and MCS is the multiple cloning site (18). The oriT insertion sites are indicated by triangles. Vector pKP9 has an IncP-type oriT site, while all of the other plasmid have IncN-type oriT sites. Plasmids pKP80ΔN and pKP81 were derived from pKP80 by deletion of the indicated DNA sequences.

TABLE 3.

Mobilization frequencies of vectors pKP9 and pKP10 in conjugal transfer to E. coli and A. tumefaciens recipients^a

Recipient	Donor S17-1(pKP9)		Donor DH5α(pKM101)(pKP10)b
	pKP9 (no. of transconjugants/donor)	pKP9 (no. of transconjugants/recipient)	pKP10 (no. of transconjugants/donor)	pKP10 (no. of transconjugants/recipient)
AB1157	>1	1.2 × 10−1	>1	6.4 × 10−1
EHA101	3.8 × 10−4	1 × 10−5	2.8 × 10−1	5.7 × 10−3

^aAs counted at the end of the filter-mating interval, 1 × 10⁸ to 5 × 10⁸ recipients per ml) were recovered. The donor-to-recipient cell ratio was 1:10. Filter matings were carried out at 28°C for 5h on LB medium. The data shown are representative of at least three experiments.

^bPlasmid pKM101 exhibited frequencies of transfer to E. coli recipients identical to those reported for pKP10 mobilization, whereas it did not detectably transfer or stabilize in A. tumefaciens recipients.

Similar conjugation experiments were done with A. tumefaciens strain EHA101 as the recipient. Both pKP9 and pKP10 were mobilized from their respective donors into this strain, although at lower frequencies than those observed with the E. coli recipient (Table 3). Notably, the IncN conjugation system was more efficient than the IncP system by approximately 3 orders of magnitude (Table 3).

IncP-mediated and IncN-mediated transformation of tobacco leaf explants.

Plasmids pKP9 and pKP10 contain the nptII gene expressed from a strong plant promoter, and the integration of this gene into plant genomic DNA confers resistance to aminoglycoside antibiotics such as kanamycin. In control experiments, both plasmids were introduced into disarmed A. tumefaciens strain EHA101. The resulting strains were used to inoculate surface-sterilized leaf strips of tobacco by the leaf disk infection method (46), and transformation of plant cells was selected with solid plant tissue culture medium containing kanamycin. EHA101(pKP9) and EHA101(pKP10) transferred the nptII gene at high efficiency (more than 70 tumors per leaf strip). These efficiencies were comparable to that of control strain EHA101(pPZP212). Transformation in this experiment was presumably mediated by vir-encoded proteins of the Ti plasmid, acting upon the border sequences of these plasmids.

We next tested for transfer of nptII to plant cells from E. coli strains containing either pKP9 or pKP10 and expressing the IncP or IncN conjugation system, respectively. No stably transformed plant calli were detected. This was not entirely surprising, since plant transformation by A. tumefaciens requires the transfer of at least one protein, VirE2, which was not present in these assays. We therefore introduced an additional plasmid (pKP40) that expresses the virE operon (Table 1) from the Ptac promoter. High-level expression of VirE proteins on this plasmid was confirmed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (data not shown). This plasmid was introduced into E. coli IncP donor strain S17-1(pKP9) and IncN donor strain JC2926(pGW1514)(pKP10) (Table 4; pGW1514 is a Kn^r Ap^s derivative of pKM101 used in place of pKM101 because pKP40 encodes Ap^r). In the case of the IncN system, no calli were detected. In the case of the IncP system, slow-growing calli were occasionally detected. However, these calli invariably failed to proliferate stably when placed on fresh selective medium. The significance of these slow-growing calli is not known.

TABLE 4.

N. tabacum transformation with tra(N)-mobilizable vector pKP10

Infecting strain(s)	No. of tumors/no. of stripsa in expt:						Total no. of tumors/no. of stripsa
	I	II	III	IV	V	VI
DH5α(pKM101)(pKP10)	1b/105	0/60	NTc	0/52	0/35	NT	1b/252 (0.004)
JC2926(pGW1514)(pKP10)	NT	0/60	0/100	0/54	0/35	0/46	0/295
JC2926(pGW1514)(pKP10)(pKP40)	NT	0/120	0/100	0/50	0/36	0/42	0/348
JC2926(pGW1514)(pKP10) + JC2926(pGW1514) (pKP40)	NT	NT	0/50	NT	NT	NT	0/50
DH5α(pKM101)(pKP10) + JC2926(pGW1514)(pKP40)	NT	NT	NT	0/52	NT	NT	0/52
DH5α(pKM101)(pKP10) + EHA101	142/60 (2.3)	NT	NT	NT	155/60 (2.58)	NT	297/120 (2.47)
JC2926(pGW1514)(pKP10) + EHA101	NT	NT	NT	NT	160/85 (1.88)	119/56 (2.12)	279/141 (1.97)
EHA101	0/48	0/60	0/40	0/36	0/27	0/32	0/243
EHA101(pKP10)	1,030/26 (39.6)	417/6 (69.5)	512/12 (46)	647/10 (64.7)	780/10 (78)	1,008/12 (84)	4,394/76 (57.8)

^aAverages are in parentheses.

^bTumor failing to differentiate.

^cNT, not tested.

As described above, the VirE2 protein can be transferred to plant cells independently of T-DNA (33, 50). It has become routine in several research laboratories to coinoculate plant wounds with two strains, one of which transfers T-DNA and the other of which transfers VirE2. Furthermore, particular conjugal transfer intermediates or gene products of IncQ and IncW plasmids are thought to compete with and hinder VirE2 export at the VirB transport apparatus (29, 45). We therefore tested whether two E. coli strains could be used for extracellular complementation of protein VirE2. Leaf strips were coinoculated with strains S17-1(pKP9) and S17-1(pKP40), both of which express the IncP transfer genes. The first of these contains the nptII gene, while the second expresses VirE2. In some experiments, very small numbers of slow-growing calli were detected, but as observed above, none of these calli proliferated on fresh selective media. Similar experiments carried out with the IncN system did not yield any calli (Table 4, rows 4 and 5).

In an effort to use a transformation medium more nutrient rich and accommodating to the nutritional needs of the E. coli strains, a minimal medium was used (M9 [see reference 40] supplemented with yeast extract, Casamino Acids, and glucose at 0.5% each) in place of TTR medium in all infection steps prior to leaf strip transfer to selective TTR plates. Use of this medium resulted in bacterial overgrowth and subsequent plant tissue maceration, and this approach was therefore abandoned.

Plant transformation by coinoculation with E. coli donor strains and a disarmed A. tumefaciens strain.

It seemed possible that the IncN and IncP conjugation systems might be able to transfer T-DNA but not VirE2. To test this, we attempted to transform leaf strips by coinoculation with an E. coli donor of the nptII gene and disarmed A. tumefaciens strain EHA101 as a donor of VirE2. When the IncP system was used as a DNA donor, we did not observe any transformed calli in more than 500 infected strips. However, when the IncN system was tested, we observed abundant numbers of calli that proliferated at normal rates and differentiated readily into kanamycin-resistant transgenic shoots and plantlets (Table 4, infections 6 and 7). E. coli strains containing pKM101 or pGW1514 were tested, and in both cases, an average of two calli per leaf strip were detected. This transformation frequency is 30 to 40 times lower than that of the control transfer of the nptII gene directly from A. tumefaciens (Table 4, row 9). Nevertheless, transformed plant cells were readily detectable.

Agrobacterium acts during coinoculation experiments as a transfer intermediate.

As described above, plant transformation was readily detectable when an E. coli strain expressing the IncN conjugation system and containing a plant-selectable gene was coinoculated with a disarmed A. tumefaciens strain. There are several possible explanations for this. First, as originally intended, the plant cell could have received the nptII gene directly from the E. coli strain and received the VirE2 protein from the A. tumefaciens strain. Alternatively, plasmid pKP10 could have been transferred from the E. coli strain to the A. tumefaciens strain via the IncN conjugation system. The A. tumefaciens strain could then have processed the T-DNA at its T-DNA border sequences and transferred it, as well as VirE2, to the plant cells.

These alternative hypotheses can be tested by using an A. tumefaciens strain that has a virD2 mutation and is unable to process T-DNA. If the A. tumefaciens strain is needed only to transfer VirE2 and is not needed to transfer T-DNA, then a virD2 mutation should not affect transformation (36). Alternatively, if the A. tumefaciens strain acts as a recipient for pKP10 and processes and transfers the T-DNA, then such a virD2 mutation would be expected to abolish plant transformation. Coinoculation assays were therefore performed with a strain bearing a nonpolar deletion of virD1 and virD2. This strain, At10002, is unable to transform plant cells because of its inability to process T-DNA at the border sequences but retains the ability to transfer VirE2 in coinoculations (36). Coinoculations with this strain in combination with DH5α(pKM101)(pKP10) yielded no transformants in the more than 300 strips tested. This indicates that the tumorigenesis detected as described above required T-DNA-processing functions and therefore required transfer of pKP10 from the E. coli donor into A. tumefaciens.

Further support for this conclusion was obtained by PCR amplification of the transferred DNA in the transformed plant cells. If the T-DNA was transferred by the IncN oriT site, then this site should be disrupted in transformed plant cells. If transfer occurred via the T-DNA borders, then the IncN oriT site should be intact. A total of 18 transgenic plants were tested. In control experiments, the nptII and actin genes were detected in all 18 plants, while two bacterial genes (fliC and repA) were not detected (Table 5), indicating the absence of bacterial DNA in these plant tissues. Significantly, an intact copy of the IncN oriT region was also detected in all 18 plants (Table 5), indicating that this site had not been processed during transfer. Of the 18 lines tested, 5 contained DNA from the ColE1 rep region, suggesting that these lines may have inherited all or most of the pKP10 sequence. The finding of vector sequences in similar experiments has been described previously (27). However, the fact that most transgenic plants did not contain these sequences provides further support that the nptII gene was transferred via the T-DNA borders.

TABLE 5.

PCR amplification results from 18 different N. tabacum transgenes for pKP10 shuttle vector DNA, and from infecting bacterial strains, with various bacterium- or plant-related primers^a

DNA template	Amplified region	No. of plants containing each gene/ no. of plants tested
pKP10	nptII	18/18
pKP10	oriT(N)	18/18
pKP10	oriT ColE1	5/18
N. tabacum	Actin gene	18/18
E. coli	fliC	0/18
pKM101	repA	0/18

^aPrimers were designed so as to amplify gene segments specific for (i) vector pKP10 parts [the nptII kanamycin resistance gene, the tra(N) origin of transfer (oriT_N), and the ColE1 rep region], (ii) the N. tabacum chromosomal actin gene (positive control), and (iii) E. coli donor strain-specific genes (fliC chromosomal flagellar gene, plasmid pKM101 repA gene).

Construction of improved vectors mobilizable by the IncN conjugation system.

The coinoculation assay described above could, in principle, be useful for rapid construction of transgenic plants. If vectors containing the IncN oriT region were to be useful, it seemed important to restore the multicloning site that was originally present on plasmid pPZP212. The 534-nucleotide IncN oriT region was introduced into the unique PmeI site and one of the ScaI sites of pPZP212 (creating plasmids pKP50 and pKP80, respectively, Fig. 1). Both plasmids closely resemble pKP10 but bear an intact multiple cloning site. These plasmids were introduced into A. tumefaciens strain EHA101, and the resulting strains were tested for transformation of tobacco leaf tissue. Both transformed tobacco at frequencies similar to those of the previous plasmids (pKP10 and pPZP212). These plasmids were also introduced into E. coli strain DH5α(pKM101), and the resulting strains were coinoculated with EHA101 onto tobacco leaf strips. The strain containing pKP80 transferred the nptII gene with efficiencies similar to that of the strain containing pKP10, while the strain containing pKP50 was less efficient.

It seemed plausible that these coinoculation experiments might not require replication of the plasmid in A. tumefaciens, since transfer of the plasmid from E. coli to A. tumefaciens could be followed immediately by T-DNA transfer to the plant without plasmid replication. To test this, plasmid pKP80 was digested with NotI, which releases DNA spanning the pVS1 replication and stability genes, and the remaining backbone of the plasmid was ligated, thus creating plasmid pKP80ΔN (Fig. 1). In plant transformation assays, an E. coli strain containing pKM101 and pKP80ΔN, when coinoculated with EHA101, transformed tobacco at frequencies about fivefold lower than that of the control strain (Table 6). This confirms our prediction that replication of the shuttle replicon in A. tumefaciens is not essential for plant transformation.

TABLE 6.

N. tabacum transformation with different tra(N)-mobilizable vectors^a

Donor strain(s)	Total no. of tumors/total no. of leaf strips (avg)
Direct coinoculation
DH5α(pKM101, pKP10) + EHA101	119/49 (2.4)
DH5α(pKM101, pKP50) + EHA101	7/64 (0.13)
DH5α(pKM101, pKP80) + EHA101	82/40 (2.05)
DH5α(pKM101, pKP80ΔN) + EHA101	22/55 (0.4)
DH5α(pKM101, pKP81) + EHA101	336/52 (0.7)
Strains filter mated prior to plant inoculationb
DH5α(pKM101, pKP10) + EHA101	228/35 (6.5)
DH5α(pKM101, pKP80ΔN) + EHA101	58/42 (1.3)
DH5α(pKM101, pKP81) + EHA101	67/38 (1.7)
A. tumefaciens control donor strains
EHA101(pKP10)	1,440/21 (68.5)
EHA101(pKP50)	1,172/15 (78.1)
EHA101(pKP80)	880/12 (73.3)

^aAll vectors were mobilized from the E. coli donor strains to E. coli or A. tumefaciens recipients with similar frequencies in filter matings (data not shown). E. coli DH5α(pKM101) strains harboring each of the vectors on the left were used to transform tobacco strips in the presence of helper A. tumefaciens strain EHA101.

^bLeaf strip infection followed filter matings of the DH5α(pKM101) donors harboring the relevant vectors and EHA101 on LB agar plates for 5 h (see Materials and Methods).

Additional DNA was deleted from plasmid pKP80, removing all pVS1 rep sequences and the ColE1 rop gene region, creating plasmid pKP81 (Fig. 1). As anticipated, deletion of the rop gene elevated the copy number of this plasmid about fourfold, as estimated by agarose gel electrophoresis (data not shown). In coinoculation assays, an E. coli strain containing plasmid pKP81 incited tumors with the same efficiency as pKP80ΔN (Table 6). The smaller size and higher copy number of this plasmid could enhance its utility in genetic manipulations and in plant transformation.

Interbacterial conjugation prior to plant inoculation increases transformation efficiency.

Genetic and biochemical evidence described above indicates that plant transformation via coinoculation occurs by transfer of the donor plasmid from E. coli harboring the IncN transfer system to A. tumefaciens, followed by transfer of T-DNA from A. tumefaciens to the plant. Since the two bacterial strains are in contact only during plant inoculation, the IncN-mediated plasmid transfer event must have occurred on leaf surfaces. We tested for conjugal transfer of pKP10 from the E. coli donor to EHA101 in the cell suspension used for leaf strip infection at the end of the leaf-dipping process (after 20 min). Approximately 10⁴ A. tumefaciens transconjugants were detected per 10⁸ E. coli donor cells. This indicates that conjugal transfer of pKP10 can occur on leaf explants, despite the fact that the conditions are probably suboptimal for interbacterial conjugation.

Conjugation between E. coli and A. tumefaciens was also tested with TTR agar medium, which is optimized for leaf explants rather than for bacteria. Mobilization of pKP9 and pKP10 from their respective hosts, S17-1(pKP9) and DH5α(pKM101, pKP10), was readily detected, although efficiencies were 20-fold and 50-fold lower, respectively, than when matings were carried out on LB medium.

Since mobilization of these plasmids from E. coli to A. tumefaciens is a critical step in the process we describe, it seemed worthwhile to optimize this process. We therefore mixed E. coli donor strains with A. tumefaciens strain EHA101 on filters. These matings were incubated on LB agar medium at 37°C to allow mobilization of their plasmids from E. coli to A. tumefaciens, and the cells were then resuspended and used to inoculate leaf strips. Tumorigenesis efficiency was elevated approximately threefold (Table 6 and Fig. 2). This stimulation was the same for a plasmid that is able to replicate in A. tumefaciens (pKP10) as it was for plasmids that cannot (pKP80ΔN and pKP81), confirming that replication in A. tumefaciens is not essential for T-DNA processing. Similar experiments with an IncP conjugation system rather than an IncN system failed to yield useful numbers of tobacco transformants (data not shown), probably because of the low efficiency of conjugation of the IncP system into A. tumefaciens (Table 3).

FIG. 2.

Comparison of N. tabacum explants developing calli and shoots after leaf disk infection with strain EHA101(pKP10) (A) and with strain EHA101 conjugated with DH5α(pKM101)(pKP10) for 5 h prior to infection (B). In panel B, explant strips harboring 5 to 10 transformation events each are shown.

DISCUSSION

In this study, we have described efforts to use E. coli strains expressing two different conjugation systems to transfer a selectable marker gene directly to tobacco cells. Our efforts were stimulated by a variety of reports indicating that broad-host-range plasmid conjugation systems in general are rather promiscuous in transferring genes to heterologous recipient cells. We were unable to detect stably transformed N. tabacum cells with the IncP- and IncN-type transfer systems. In the case of the IncP system, slow-growing calli were occasionally detected, but these invariably failed to thrive when subcultured on fresh selective medium. The significance of these abortive calli is not known but might be attributable to unstable transfer of the transgene. Abortive calli were barely detectable with the IncN conjugation system. Our failure to detect stable plant transformation could have several explanations. First, the E. coli strains may have failed to bind to the plant cells in a productive fashion. Binding of A. tumefaciens to plant cells requires the products of several att loci, the chvA and chvB genes, and the cel genes, whose products synthesize cellulose fibrils (32). E. coli lacks these genes and might therefore fail in productive cell-cell contact. Equally likely, the foreign cell envelope of the plant cell might block transfer, or there could be a block in transporting the transgene to the nucleus or across the nuclear pore. A. tumefaciens T strands contain VirD2 covalently bound at their 5′ ends, and this protein contains nuclear targeting sites that interact with importin alpha and with cyclophilins (2, 13). The corresponding relaxases of the IncN or IncP transfer systems would not be expected to contain such motifs. On the other hand, the nuclear targeting motif of VirD2 is dispensable for tumorigenesis (43), possibly because similar motifs are found on VirE2 (10), which is thought to coat the T strand. An additional potential problem was the lack of the VirE2 protein, which is required for transformation. Finally, the integration of the transgene could have been blocked, although there is only suggestive evidence that Vir proteins play any direct role in T-strand integration (56).

If the block in transfer was attributed to the lack of VirE2, then we reasoned that this block would be circumvented by providing VirE2 via a disarmed A. tumefaciens strain. There is a long literature describing this extracellular complementation, and it is used routinely in several laboratories. Coinoculation of an E. coli strain expressing the IncP transfer system with EHA101 did not lead to stable transformation. In contrast, similar experiments with the IncN transfer system gave large numbers of stable transformants. While our original intent in these experiments was to have EHA101 provide VirE2 directly to plant cells, it turned out that the mechanism of delivery involved two transfer steps, in which the entire plasmid was transferred from E. coli to A. tumefaciens and the T-DNA was then transferred to the plant cells. The first step was presumably carried out by the IncN tra genes and the IncN-type oriT site, while the second step is predicted to be Vir protein mediated, acting upon the T-DNA borders. The greater efficiency of the IncN system than the IncP system could have at least two explanations. First, IncN-mediated plasmid transfer from E. coli to A. tumefaciens is approximately 1,000-fold more efficient than IncP-mediated transfer. The IncN tra system was previously found to transfer promiscuously between proteobacteria (25). Another possible factor is that in the IncN system, the mobilizing plasmid could itself be transferred to A. tumefaciens and, despite its inability to replicate in this host, could transiently play some role in the second mobilization step. In contrast, the IncP system is encoded on the chromosome and should not itself be transferred.

Our finding that the mobilized plasmid does not require a broad-host-range replication origin indicates that this plasmid does not need to replicate in A. tumefaciens. This should not be surprising, since the conjugation of this plasmid from E. coli to A. tumefaciens should result in a covalently closed circular plasmid. The T-DNA processing step also should not require vegetative replication of the plasmid. From a practical standpoint, this finding can simplify the construction of plasmid vectors and reduce their size, since an E. coli origin and selectable markers will suffice. Our use of plasmids that fail to replicate in A. tumefaciens could be useful in enhanced biocontainment. In the limited number of N. tabacum transgenic samples tested, no E. coli DNA was detected (Table 5). Nevertheless, in future work, it might be preferable to use E. coli donors that are blocked in the transfer of the mobilizing plasmid.

To date, the efficiency of DNA transfer to plants in the two-step process described here is lower than that of conventional single-step T-DNA transfer from A. tumefaciens. It might be possible to increase the efficiencies in several ways. First, it might be possible to elevate the efficiency of E. coli-to-A. tumefaciens conjugation, possibly by finding and eliminating any restriction barrier or by optimizing the conditions of conjugation. However, the efficiencies reported here are sufficient for routine creation of transgenic plants in all but the most recalcitrant species. Further increases in transfer efficiency might be attainable by preconjugating the donor strains prior to infection or by using approaches that are useful for standard Agrobacterium-mediated plant transformation, such as overexpression of VirG and VirE in the disarmed Agrobacterium strain (42, 48), adding acetosyringone to artificially stimulate expression of the vir regulon (49), or placing genes of interest under stronger or early-expressed promoters (14).

The coinoculation method described here offers the convenience of conducting all DNA manipulation steps and strain construction in E. coli rather than in A. tumefaciens. This makes it unnecessary to introduce binary plasmid DNA into disarmed Agrobacterium strains, and the use of recombination-deficient E. coli strains could facilitate the experimental manipulation of unstable DNA regions.