Introduction

To date, the biotechnological improvement of plants has frequently been limited to the introduction or manipulation of single genes. However, most agronomic characteristics are polygenic in nature and rely on complex metabolic and regulatory pathways. Therefore, genetic modifications of traits or introduction of new pathways often require transfer of multiple genes into the plant genome (Lyznik and Dress 2008; Naqvi et al. 2010). Examples are the golden rice (Oryza sativa) from Ye et al. (2000) or the red corn (Zea mays) from Zhu et al. (2008). Several strategies like cotransformation (Zhu et al. 2008), retransformation (Li et al. 2003) and sexual crossing (Zhao et al. 2003) can be used for the introduction of multiple genes into plant cells. Through repeated transformation or crossing an essentially unlimited number of transgenes can be combined in a single plant. But these techniques are very time-consuming, labor-intensive and demand distinct selection markers for every round of transformation or crossing. Furthermore the multiple integration sites strongly impede the generation of homozygous lines and exacerbate commercial use because multiple integration sites are not compatible with current legal requirements for the release of transgenic plants (Taverniers et al. 2008). Improvements provide the cotransformation via cobombardment of several plasmids. With this approach numerous genes can be simultaneously transformed using a single selection marker (Francois et al. 2002). Chen et al. (1998) demonstrated the integration of up to 13 different genes and plasmids, respectively, in rice. Surprisingly, cobombardment of several plasmids often results in transgenic plants that integrated all the transgenes at a single genomic locus (Agrawal et al. 2005; Zhu et al. 2008). However, cotransformation often leads to the integration of multiple copies and complex integration patterns (Pawlowski and Somers 1996; Birch 1997), again complicating commercial use. The delivery of the transgenes as a single multigene construct offers considerable benefits. Although multigene constructs were used successfully in a few studies (Zhong et al. 2007; Fujisawa et al. 2009), the assembly of such constructs remains challenging. With the conventional cloning methods it’s nearly impossible (Dafny-Yelin and Tzfira 2007; Naqvi et al. 2010), because of the lack of unique restriction sites when multiple genes need to be linked.

The Gateway technology is based on phage lambda site-specific recombination (Hartley et al. 2000). It relies on the action of recombinases that are capable of sequence-specific recombining compatible DNA sequences. For example the LR clonase can recombine attL with attR sites in an irreversible way and is usually used for the delivery of DNA fragments from an entry clone into a destination vector (Walhout et al. 2000). Several Gateway-compatible plant transformation vectors were designed for diverse purposes like the functional analysis of genes in planta or protein localization studies (Curtis and Grossniklaus 2003; Earley et al. 2006). Originally the Gateway technology was not laid out for stacking multiple genes in a single vector. However, the introduction of Multisite Gateway vectors opened an alternative way for recombining several DNA fragments in a single recombination step (Cheo et al. 2004; Sasaki et al. 2004; Karimi et al. 2005; Magnani et al. 2006). However, this technique is limited by the availability of more attachment sites. Chen et al. (2006) presented another method, which is called MultiRound Gateway technology, to stack multiple DNA fragments into a single vector. The system is based on two different entry vectors which can be alternately used to deliver sequentially multiple DNA fragments into a Gateway-compatible destination vector.

Here we describe an improved version of the MultiRound Gateway technology. By the use of a transformation-competent artificial chromosome (TAC)-based destination vector and a recombination deficient strain of Agrobacterium, we could increase the stability of the multigene construct as well as the efficiency of the transformation process. The constructed Gateway-compatible TAC vector possesses a cloning capacity greater than 100 kb (Liu et al. 1999; Lin et al. 2003). Therefore the vector can accept numerous DNA fragments from the entry vectors. With our system we could introduce up to eight genes located on a single destination vector including a selection marker, five enzymes, two translocators, two Scaffold attachment regions (SAR) and three putative transcription blockers (Padidam and Cao 2001) into tobacco plants. By real-time PCR we could confirm the stable expression of all transgenes for at least two generations. The expression of four of the five enzymes was also verified by enzyme activity assays.

Materials and methods

Entry vector construction

The recombination sites (attL and attR), homing endonuclease sites (I-SceI and PI-SceI), PmeI and NotI restriction sites and Multiple cloning sites (MCSI and MCSII) were synthesized in four pBS SK(+) plasmids: pBS SK(+)_G325A (PmeI-I-SceI-attL1-MCSI-attR3-PI-SceI-NotI), pBS SK(+)_G325B (PmeI-I-SceI-attL2-MCSII-attR4-PI-SceI-NotI), pBS SK(+)_G324A (PmeI-I-SceI-attL3-MCSI-attR1-PI-SceI-NotI), pBS SK(+)_G324B (PmeI-I-SceI-attL4-MCSII-attR2-PI-SceI-NotI) (Euro-fins Medigenomix, Ebersberg). The entry vector pEntry1_SpecR (Fig. 1a) was obtained by the following cloning steps. First, the spectinomycine resistance gene aadA was amplified using pWBVec8 (Wang et al. 1998) as template and primers 5′-CC tct aga CCA AGA GCT TGT CGG GAA GAT TGA AC-3′ and 5′-GA gga tcc ATG CCA TCG CAA GTA CGA GGC TTA GAA C-3′, digested with XbaI and BamHI and cloned into vector pBS SK(+)_G325A to make pBS SK(+)_G325A_SpecR. Second, the fragment PmeI-I-SceI-attL2-MCSII-attR4-PI-SceI-NotI from vector pBS SK(+)_G325B was cloned into vector pBS SK(+)_G325A_SpecR by using BamHI and XhoI to make pBS_G325_SpecR. Thirdly, the pUC origin and the bla gene was amplified using an empty pBS SK(+) vector as template and primers 5′-G TAC caa ttg CGA GCG GTA TCA GCT CAC TCA AAG G-3′ and 5′-C TGA ctg cag CGG GGA AAT GTG CGC GGA AC-3′ and cloned into vector pBS_G325_SpecR by using MunI and PstI to remove undesired restriction sites in the lacZ-fragment of the pBS SK(+) vector. The entry vector pEntry2_CmR (Fig. 1b) was obtained by the following cloning steps. First, the chloramphenicol resistance gene cat was amplified using pEG301 (Earley et al. 2006) as template and primers 5′-CC tct aga TTA GGC ACC CCA GGC TTT ACA C-3′ and 5′-GG tct aga TCA ATA AAC CGG GCG ACC TCA G-3′, digested with XbaI and cloned into vector pBS SK(+)_G324A to make pBS SK(+)_G324A_CmR. Second, the fragment PmeI-I-SceI-attL4-MCSII-attR2-PI-SceI-NotI was released from vector pBS SK(+)_G324B using BamHI and XhoI and cloned after blunting with T4 DNA polymerase into the EcoRV restriction site of vector pBS SK(+)_G324A_CmR to make pBS SK(+)_G324_CmR. Thirdly, the backbone of the plasmid was changed accordingly to pEntry1_SpecR. The entry vectors pEntry1_SpecR_iMCS and pEntry2_CmR_iMCS were obtained by exchanging the MCSI site for an inverted version.

Fig. 1
figure 1

Physical maps of the two entry vectors pEntry1_SpecR (a) and pEntry2_CmR (b) and the two destination vectors pTRA_R12_CmR_ccdB (c) and pYLTAC7_R12_CmR_ccdB (d), respectively. Important regions and restriction enzymes used in previous construction steps are indicated

Full size image

The entry vectors with target genes pEntry1_PPT, pEntry2_StPEPC_TB1, pEntry1_Oac1, pEntry2_PPDK_TB2, pEntry1_SbMDH, pEntry2_HvME_TB3, pEntry1_FbCA, pEntry2_PEPS_TB2, pEntry1_EcMDH, pEntry2_EcME_TB3, pEntry2_PCK_TB2 were obtained by cloning the respective expression cassettes into the MCSI or iMCSI of pEntry1_SpecR, pEntry2_CmR, pEntry1_SpecR_iMCS and pEntry2_CmR_iMCS. For details of the cloning processes and maps of the final constructs see supplementary material (Table 3, Fig. 5). A table describing the genes to be transformed and their expected functions can also be found in supplementary material (Table 4).

Destination vector construction

For destination vector pTRA_R12_CmR_ccdB (Fig. 2a) the gateway cassette attR1-CmR-ccdB-attR2 was amplified using pEG301 (Earley et al. 2006) as template and primers 5′-CA ggc gcg ccA CGT CTT GCG CAC TGA TTT G-3′ and 5′-GA ggc cgg ccA TCG TAT GGG TAC ACC ACT TTG-3′, digested with AscI and FseI and cloned into vector pTRA, a derivative of pPAM (gi13508478). For destination vector pYLTAC7_R12_CmR_ccdB (Fig. 2b) the gateway cassette attR1-CmR-ccdB-attR2 including the SARs was removed from pTRA_R12_CmR_ccdB using SapI and PmeI and cloned after blunting of the SapI end with T4 DNA polymerase into the PmeI site of vector pYLTAC7 (Liu et al. 1999).

Fig. 2
figure 2

Schematic diagram of the first two rounds of LR recombination. The first round of LR recombination between attL1 and attL2 of the entry vector pEntry1_PPT and attR1 and attR2 of the destination vector pYLTAC7_R12_CmR_ccdB exchange the chloramphenicol resistance (CmR) for the spectinomycin resistance (SpecR) and integrate the PPT expression cassette and R3 and R4 in the destination vector. The second round of LR recombination between attL3 and attL4 of the entry vector pEntry2_StPEPC_TB1 and the new destination vector pYLTAC7_PPT reexchange SpecR for CmR and integrate the StPEPC expression cassette and attR1 and attR2 in pYLTAC7_PPT to make pYLTAC7_PPT_StPEPC. By alternating use of the two entry vectors further genes can be integrated. PPT phosphoenolpyruvate/phosphate translocator, StPEPC modified phosphoenolpyruvate carboxylase from Solanum tuberosum

Full size image

LR recombinations

The LR recombinations were performed according to the manufacturer’s instructions (Invitrogen, Carlsbad, USA) with equimolar amounts of entry and destination vectors. As negative control the LR recombinase was omitted. Aliquots of 2 μl of each reaction mixture were transformed into Escherichia coli TOP10 using electroporation. Using pTRA_R12_CmR_ccdB as destination vector, recombinants were selected on LB agar plates supplemented with 50 μg ml−1 ampicillin and 50 μg ml−1 spectinomycin after recombination with pEntry1_SpecR entry vectors and 34 μg ml−1 chloramphenicol after recombination with pEntry2_CmR entry vectors. Using pYLTAC7_R12_ CmR_ccdB as destination vector, recombinants were selected on LB agar plates supplemented with 50 μg ml−1 kanamycin and 15 μg ml−1 spectinomycin and 20 μg ml−1 chloramphenicol, respectively. Plasmids were prepared from 5 ml LB overnight cultures, supplemented with the appropriate antibiotics (same concentrations as before) according to Birnboim and Doly (1979). In case of using pYLTAC7_R12_CmR_ccdB as destination vector also 0.16 mM IPTG was added.

Plant transformation

Nicotiana tabacum cv. Petit Havana plants were grown in pots at 25 °C in a growth chamber under 16 h light:8 h dark cycle. All multigene constructs were transformed into Agrobacterium strain AGL1 (Lazo et al. 1991; kindly provided by Paul Hooykaas, Molecular and Developmental Genetics Department, Institute of Biology Leiden, Leiden University Leiden, Netherlands) and screened on LB plates containing 15 μg ml−1 kanamycin and 50 μg ml−1 carbenicillin. For verification by PCR the following primers were used (Table 1): PPT: 3472 and 865; StPEPC: 4255 and 4225; Oac1: 3971 and 4100; PPDK: 4048 and 4052; SbMDH: 4281 and 4282; HvME: 3545 and 3546; FbCA: 4309 and 4310; PEPS: 3874 and 4284; EcMDH: 3978 and 3979; EcME: 4256 and 4223; PCK: 4233 and 4234.

Table 1 Primers used for PCR verification of transformed Agrobacteria and RT-PCR analysis of transgenic plants
Full size table

Stable transformation of tobacco plants was done according to De Block (1988) and Dietze et al. (1995). Putative T0 transgenic plants were regenerated from the callus in the presence of 50 μg ml−1 hygromycin.

Real-time-PCR

Total RNA was prepared from 2 weeks old soil-grown plants by using a combined RNA-DNA extraction assay. Approximately 0.02 g leaf material were ground in liquid nitrogen and extracted with extraction buffer (0.05 M Tris–HCl, pH 7.6, 0.5 % SDS) and water saturated phenol. Reverse transcription was performed by using moloney murine leukemia virus reverse transcriptase (MMLV-RT, Promega, Madison, USA) with a random nonamer primer. Gene specific primers for quantitative RT-PCR were: Actin2: 584 and 585; PPT: 4506 and 4507; StPEPC: 4546 and 4547; Oac1: 4510 and 4511; PPDK: 4512 and 4513; SbMDH: 4514 and 4515; HvME: 4516 and 4517; PEPS: 4520 and 4521; EcMDH: 4522 and 4523; EcME: 4524 and 4525; PCK: 4526 and 4527; FbCA: 4571 and 4572 (sequences see Table 1). Actin2 was used as an internal standard (Igarashi et al. 2003). PCR amplification was done using the Platinum®SYBR®Green qPCR SuperMix‐UDG with Rox (Invitrogen, Carlsbad, USA) according to the manufacturer’s instructions with 40 cycles of 94 °C for 15 s and 60 °C for 60 s.

Extraction of leaf soluble protein and enzyme assays

Leaf discs of about 60 mg were harvested from the mid-section of the third leaf at noontime and immediately stored in ice cold extraction buffer. For determination of phosphoenolcarboxylase (PEPC), malic enzyme (ME) and malate dehydrogenase (MDH) activity, extraction buffer I (50 mM HEPES–NaOH, pH 8.0, 10 mM MgCl2, 5 mM DTT, 1 mM EDTA, 20 % glycerin, 0.01 mg ml−1 chymostatin, 0,5 mM PefablocSC was used. Samples were ground in 250 μl extraction buffer I. After centrifugation (14,000g, 5 min, 4 °C) the supernatant was transferred into a new tube and stored on ice until used. For determination of pyruvate orthophosphate dikinase and phosphoenolpyruvate synthetase activity, extraction buffer IIA (50 mM Hepes–KOH, pH 8.0, 10 mM MgSO4, 1 mM EDTA, 2.5 mM K2HPO4, 20 % Glycerin) and IIB (Buffer IIA including 5 mM DTT, 5 mM pyruvate, 1 mM PefablocSC, 2 % PolyclarAT, 0.005 mg ml−1 chymostatin) was used. Samples were ground in 250 μl extraction buffer IIB and mixed with 950 μl extraction buffer IIA. After centrifugation (14,000g, 15 min, 4 °C) 800 μl of supernatant were transferred into a new tube. Protein was precipitated with 0.312 g ammonium sulfate and centrifugation (14,000g, 10 min, 4 °C). The pellet was resuspended in 100 μl extraction buffer IIA and stored on ice until use.

Enzyme activities were determined spectrophotometrically at 30 °C. The PEPC activity was assayed by the method of Rademacher et al. (2002). The NADP-ME activity was determined according to Hausler et al. (1987), the NAD-ME activity according to Bologna et al. (2007). The NADP-MDH activity was assayed by the method of Ashton and Hatch (1983), the NAD-MDH activity by the method of Sutherland and McAlister-Henn (1985). The PPDK and PEPS activity were determined according to Ashton et al. (1990). For PEPS measurement K2HPO4 was omitted.

Results

Construction of entry and destination vectors

Four different entry vectors were designed: pEntry1_SpecR, pEntry2_CmR, pEntry1_SpecR_iMCS and pEntry2_CmR_iMCS. The sequences of the recombination sites attL1/2/3/4 and attR1/2/3/4, respectively, have been chosen according to Chen et al. (2006). A multiple cloning site (MCSI) between attL1 and attR3 serves as integration site for DNA fragments to stack into the destination vector, because integration by Gateway BP recombination is not possible. A second multiple cloning site (MCSII) allows integration of a further DNA fragment, which can be recombined simultaneously. Two I-SceI and PmeI sites, respectively, flanking the attL sites allow linearization of entry vectors and removal of the vector backbones. This will eliminate cotransformation of the entry vectors whereby cloning efficiency would be reduced. Two PI-SceI and NotI sites, respectively, flanking the selection marker allows linearization of destination vectors. It is assumed that this will increase recombination efficiency (Chen et al. 2006). Use of the homing endonuclease sites has the advantage, that these sites are extremely rare in natural sequences, whereas use of PmeI and NotI for linearization of the vectors will reduce costs.

As destination vectors two different plasmids were constructed. pTRA_R12_ CmR_ccdB is based on a pTRA vector, a derivative of pPAM (gi13508478), and contains a high copy origin (ColE1) for replication in E. coli and a broad host low copy origin (RK2) for replication in Agrobacterium. pYLTAC7_R12_CmR_ccdB is based on the TAC vector pYLTAC7 (Liu et al. 1999) and contains the single copy origin from the P1 phage for replication in E. coli and the single copy origin from Agrobacterium rhizogenes for replication in Agrobacterium. Additionally it contains the P1 lytic replicon under the control of the lac operon, which allows induction of higher copy numbers with IPTG. Both vectors possess the cis elements LB and RB for Agrobacterium mediated transfer into plant cells and two flanking SARs to reduce position effects (Grosveld et al. 1987; Stief et al. 1989; Bonifer et al. 1990). The gateway cassette consists of attR1 and attR2 recombination sites and a positive (chloramphenicol resistance, CmR) and a negative (ccdB) selection marker. CmR allows selection of the plasmid in E. coli and Agrobacterium, ccdB allows counter selection of non-recombined plasmids in E. coli.

Use of TAC vector allows better integration of transgenes

For MultiRound Gateway recombination the destination vector pTRA_R12_CmR_ccdB or pYLTAC7_R12_CmR_ccdB and the entry vectors with target genes were used. In experiments with the high copy vector pTRA_R12_CmR_ccdB only in the first two rounds of LR recombination, correct clones were obtained. For unknown reasons several attempts did not lead to correct recombinants in subsequent rounds (data not shown). In contrast, with pYLTAC7_R12_CmR_ccdB_SAR even in the seventh round of recombination nearly all clones were correct. Sporadically occurring false clones could be explained by cotransformation of entry and destination vectors. However, by previous linearization of the entry vectors, this number could be kept very low (<5 %), because linear DNA is poorly transformed in E. coli (Cohen et al. 1972; Conley and Saunders 1984; Conley et al. 1986).

With pYLTAC7_R12_CmR_ccdB as destination vector three different multigene constructs were successfully assembled: pYLTAC7_C4_NADP-ME (a) pYLTAC7_C4_NAD-ME (b) and pYLTAC7_C4_PCK (c). The corresponding entry vectors were recombined in the stated order:

  1. a)

    pYLTAC7_C4_NADP-ME

  2. 1)

    pEntry1_ PPT

  3. 2)

    pEntry2_ StPEPC_TB1

  4. 3)

    pEntry1_ Oac1

  5. 4)

    pEntry2_ PPDK_TB2

  6. 5)

    pEntry1_ SbMDH

  7. 6)

    pEntry2_ HvME_TB3

  8. 7)

    pEntry1_ FbCA

  9. b)

    pYLTAC7_C4_NAD-ME

  10. 1)

    pEntry1_PPT

  11. 2)

    pEntry2_StPEPC_TB1

  12. 3)

    pEntry1_Oac1

  13. 4)

    pEntry2_PEPS_TB2

  14. 5)

    pEntry1_EcMDH

  15. 6)

    pEntry2_EcME_TB3

  16. 7)

    pEntry1_FbCA

  17. c)

    pYLTAC7_C4_PCK

  18. 1)

    pEntry1_PPT

  19. 2)

    pEntry2_StPEPC_TB1

  20. 3)

    pEntry1_Oac1

  21. 4)

    pEntry2_PCK_TB2

  22. 5)

    pEntry1_FbCA

Figure 2 shows the first two rounds of LR recombination, which were equal for all three vectors. In the first round, recombination occurred between pEntry1_PPT and pYLTAC7_R12_CmR_ccdB to yield pYLTAC7_PPT. Successful recombinants were selected by kanamycin and spectinomycin treatment. The second round of recombination was conducted between pEntry2_StPEPC_TB1 and pYLTAC7_PPT to make pYLTAC7_PPT_StPEPC. Correct transformants were selected by kanamycin and chloramphenicol treatment.

By alternating use of the two entry vectors we performed seven rounds of LR recombination for pYLTAC7_C4_NADP-ME and pYLTAC7_C4_NAD-ME, respectively, and five rounds for pYLTAC7_C4_PCK. The orientation of the gene expression cassettes in the entry vectors was chosen in such a way, that after recombination two consecutives cassettes are arranged head to head or separated by a putative transcription blocker if arranged tail to tail (Fig. 6 of supplementary material). According to Padidam and Cao (2001) this will prevent silencing caused by transcriptional interference. The transcription blockers are AT-rich sequences from the bacteriophage lambda.

The multigene constructs were confirmed after every round of recombination by digestion with appropriate restriction enzymes. Figure 3a shows the results after the three last rounds of recombination for pYLTAC7_C4_NADP-ME and pYLTAC7_C4_NAD-ME. The expected bands produced by digestion are indicated in Fig. 3b. The physical maps of pYLTAC7_C4_NADP-ME and pYLTAC7_C4_NAD-ME after the last round of LR recombination are shown in Fig. 3c.

Fig. 3
figure 3

Restriction analysis of pYLTAC7_C4_NADP-ME and pYLTAC7_C4_NAD-ME. After the fifth, sixth and seventh round of LR recombination the respective entry vectors (E), destination vectors (D) and derived constructs (1–6 and 1–4, respectively) were digested with BamHI or Acc65I and separated on 0.6 % agarose gels (a). As DNA size marker lambda DNA digested with PstI was used (λ). The enzymes used and the resulting fragment sizes for the respective entry vectors, destination vectors and derived recombined destination vectors (recD) are indicated in the table aside (b). The physical maps of pYLTAC7_C4_NADP-ME and pYLTAC7_C4_NAD-ME after the last round of LR recombination are showed below (c). The different gene expression cassettes are indicated as blue arrows. Used Abbreviations are the same as before. (Color figure online)

Full size image

Multigene constructs are more stable in recombination deficient Agrobacterium strain

For transformation of the multigene constructs two different Agrobacterium tumefaciens strains were tested: GV2260 (McBride and Summerfelt 1990) and AGL1. Compared to GV2260, AGL1 possesses an insertion mutation in recA to improve recombinant plasmid stability (Lazo et al. 1991). As a stability assay, the destination vector pYLTAC7_PPT_ StPEPC_Oac1, which was obtained after the third round of LR recombination, was transformed into the two Agrobacterium strains. Single colonies were streaked out on new agar plates and were tested for presence of the three genes PPT, StPEPC and Oac1 by PCR. As positive control isolated pYLTAC7_PPT_StPEPC_Oac1 plasmid DNA was used, as negative control Agrobacterium cells which carried an empty destination vector were used. With GV2260 only 20–30 % of the clones were tested positive for all three genes. In contrast, with AGL1 90–100 % of the cells carried all three genes (data not shown). Hence, AGL1 was used in further experiments. Because of the hypervirulent Ti plasmid of AGL1 also higher transformation rates could be expected with this strain (Nadolska-Orczyk et al. 2000). After transformation of the multigene constructs pYLTAC7_C4_NADP-ME, pYLTAC7_C4_NAD-ME and pYLTAC7_C4_PCK in AGL1 single colonies were tested by PCR for the presence of all genes. Figure 4a shows the results for pYLTAC7_C4_NADP-ME and pYLTAC7_C4_NAD-ME. All PCR products showed the expected size (Fig. 4b). As a positive control, isolated plasmid DNA was used, as a negative control, Agrobacterium cells which carried an empty destination vector were used. Hence, although loss of a single gene was observed in some cases, AGL1 is suitable for propagation of the multigene constructs.

Fig. 4
figure 4

PCR check of multigene constructs pYLTAC7_C4_NADP-ME and pYLTAC7_C4_NAD-ME in Agrobacterium AGL1. Colonies were tested for presence of all seven genes: PPT, StPEPC, Oac1, PPDK, SbMDH, HvME, FbCA for pYLTAC7_C4_NADP-ME and PPT, StPEPC, Oac1, PEPS, EcMDH, EcME, FbCA for pYLTAC7_C4_NAD-ME, respectively. Isolated plasmid DNA was used as a positive control (K). Generated PCR products were separated on a 1.0 % agarose gel (a). Lambda DNA digested with PstI served as DNA size marker (λ). The expected bands produced by PCR are shown in the table below (b)

Full size image

Generation of transgenic plants with stable transgene expression up to T2

A total of 66 putative T0 tobacco transformants for pYLTAC7_C4_NADP-ME and pYLTAC7_C4_NAD-ME, 38 for pYLTAC7_C4_PCK and 115 for pYLTAC7_PPT_StPEPC_Oac1 were subjected to expression analysis by real-time PCR (Table 2). About 30 % of the plants transformed with the multigene constructs pYLTAC7_C4_NADP-ME and pYLTAC7_C4_NAD-ME expressed all seven genes located on the T-DNA (Fig. 7 of supplementary material). The rather high number of plants which expressed none of the genes (18 %) can be explained by the strong growth inhibition caused by some genes. HvME and especially EcME lead to serious stunting of plants. Because aim of the multigene insertions was to enhance plant growth by the introduction of a single cell C4-like CO2 concentration cycle (Table 4 of supplementary material), non-expressors were privileged in the selection process. For the construct pYLTAC7_C4_PCK with five different genes approximately 40 % of the plants expressed all genes (Fig. 7 of supplementary material). Even higher was the rate of plants which express all genes with the construct pYLTAC7_PPT_StPEPC_Oac1. More than 80 % of plants tested by real-time PCR showed expression of all three genes. The results indicated that they are likely to harbor at least one full T-DNA copy. An increase in the activity of PEPC, MDH, ME, PPDK, PEPS and PCK in these plants was also verified by enzyme assays (data not shown). The presence of some genes was also analyzed by PCR (Fig. 8 of supplementary material). In most cases the results were consistent with those from real-time PCR. Only few genes were detected by PCR but showed no expression (<5 %).

Table 2 Expression analysis of T0, T1 and T2 plants by real-time PCR
Full size table

We allowed two transgenic lines of each construct that expressed all transgenes to mature and set seed and determined the inheritance and stability of the T-DNA in the next generations. Again the plants were subjected to expression analysis by real-time PCR (Table 2, Fig. 9 of supplementary material). T1 plants of lines L4 and L13 of pYLTAC7_C4_NADP-ME, L25 and L29 of pYLTAC7_C4_PCK and L4 and L11 of pYLTAC7_PPT_StPEPC_Oac1 exhibited a plausible single-insert segregation pattern when grown on a hygromycin-containing medium. In progenies of line L9 of pYLTAC7_C4_NAD-ME there were plants which expressed 7 genes as well as plants which expressed only 6 genes. They were likely to carry two T-DNAs, an entire and a truncated version. Nearly all progenies of line L22 of pYLTAC7_C4_NAD-ME grew on hygromycin-containing medium and showed expression of all transgenes. Presumably they carried two or even more T-DNAs. Again an increase in enzymatic activities in these plants was verified by enzyme assays (Fig. 10 of supplementary material).

We further analyzed one transgenic line of each construct in the T2 generation to demonstrate inheritance and stability of T-DNA inserts in subsequent generations. Line L13–12 of pYLTAC7_C4_NADP-ME was likely to be homozygous. All progenies expressed all genes located on the T-DNA. T1 plant of line 9–24 of pYLTAC7_C4_NAD-ME exhibited a segregation pattern similar to L9. T1 plants of line L25–39 of pYLTAC7_C4_PCK and line 11–15 of pYLTAC7_PPT_StPEPC_Oac1 showed a plausible single-inherit segregation pattern like the respective parental plants. These results supported the notion that our plants carried full T-DNA inserts and passed them down to next generations. Hence, our data clearly show that multigene expressing plants can be produced using our MultiRound Gateway system.

Discussion

In this work, we present our improved version of MultiRound Gateway technology and demonstrate that it is suitable for the generation of multigene expressing plants. Using our constructed low copy destination vector pYLTAC7_R12_CmR_ccdB we assembled multigene constructs with up to eight transgenes including a selection marker by seven rounds of LR recombination. Each transgene possessed its own promoter and termination region corresponding to more than 26 kb. In contrast, with the high copy destination vector pTRA_R12_CmR_ccdb, similar to that used by Chen et al. (2006), only two rounds of LR recombination were successful. Our proof of principle experiments don’t match the limit of the system, because the cloning capacity of the TAC-based destination vector is larger than 100 kb (Liu et al. 1999; Lin et al. 2003), whereas the number of gene-assembly rounds is not a limiting factor within a reasonable range. Dependent on cloning capacity of the destination vector the procedure maybe repeated nearly an infinite number of times. The orientation of transgenes in the destination vector can be freely designed by appropriate cloning into the multiple cloning sites of the entry vectors. We used four different MultiRound Gateway-constructed binary vectors with three, five and seven transgenes located on the T-DNA for stable genetic transformation of tobacco plants. We showed that numerous plants stably expressed the entire T-DNA molecule-encoded transgenes for at least two generations. Thus, our system represents a powerful tool for the construction of multigene plant transformation vectors overcoming the drawbacks of traditional cloning methods.

Another possibility for the assembly of multigene constructs was presented by Lin et al. (2003). The method was based on the Cre/loxP recombination system and the introduction of homing endonucleases. Lin et al. stacked ten foreign DNA fragments into a TAC vector by eight rounds of recombination. A major drawback of their system was that every recombination resulted in the integration of the whole donor vector. The undesired backbone of the plasmid and the redundant loxP site had to be removed by digestion with a homing endonuclease. Because of the asymmetric nature of the homing endonuclease sites the ends were not compatible to each other. The recyclization of the TAC vector required the aid of a compatible linker. In comparison to our method, this procedure was quite time-consuming.

It’s particularly time-saving when transfer and homologous recombination are performed in vivo in E. coli (Muyrers et al. 2001; Warming et al. 2005; Li and Elledge 2005). However, this strategy is not suitable for constructs with repeated elements, because sequences between these elements are prone to deletion. Yet, Chen et al. (2010) developed a technique called MISSA (multiple round in vivo site specific assembly), which relies on a combination of the Cre/loxP and the lambda phage recombination system, and which enable assembly of DNA sequences in vivo. The system is based on conjugational transfer, which is driven by donor strains, and two in vivo site-specific recombination events, which are mediated by Cre recombinase and site-specific lambda phage recombination proteins in recipient strains. After introduction of entry vectors into donor strains the genes of interest can be transferred into recipient vectors by mixing bacterial strains. In their proof-of-principles experiments Chen et al. (2010) constructed a multigene vector, which was generated by fifteen rounds of in vivo recombination. The cloning efficiency of the MISSA system was 60 up to 80 % depending on the used donor vectors. A main source of false clones was the homologous recombination events. Hence the repeated use of the same elements should be avoided as much as possible. Using the low copy destination vector pYLTAC7_R12_CmR_ccdB the cloning efficiency of our MultiRound Gateway system was nearly 100 %, despite the use of several repeated elements. For example the pYLTAC7_C4_NAD-ME construct contained three direct repeated doubled 35S promoters, three direct repeated pA35S termination regions, three direct and one inverted repeated 3′g7 termination regions, two inverted repeated RbcS promoters from Chrysanthemum morifolium and two direct repeated SAR elements. Sporadically occurring false clones could be explained as cotransformation of entry and destination vectors. However, by linearization of entry vectors this could be almost completely eliminated.

A drawback of recombination based systems is that plasmids obtained with these strategies contain additional sequences at the junction sites. The Golden Gate cloning method reported by Engler et al. (2008) overcame this problem. The strategy is based on the use of type IIs restriction enzymes, which cut outside of their recognition sequence such as BsaI. With proper design of the cleavage sites, two fragments cut by such an enzyme can be ligated into a product lacking the original restriction site. Other ligation products containing the original restriction site will be re-cut. Besides the feature to leave no additional sequences in the destination vector the method allows ligation of multiple fragments. However, subsequently, no further fragments can be added. It’s also not clear, if the method is still efficient for large fragments. However, the method could be used for the construction of the expression cassettes in the entry vectors, which could easily be adapted for this purpose. A worthwhile alternative to simplify construction of entry vectors is the In-Fusion system from Clonetech (Zhu et al. 2007). By recombining entire expression cassettes instead of single elements the additional sequences are located between the cassettes, where no negative effects are expected.

The irreversible nature of the recombination-based reactions does not enable the modification of existing multigene vectors. A versatile and modular system for the assembly of such vectors was recently described by Zeevi et al. (2012). Their system is composed of a destination vector that has been engineered to carry an array of unique recognition sites for zinc finger nucleases (ZFN) and homing endonucleases. Using this method they produced binary vectors that carried up to nine genes. However, the system is limited by the availability of different ZFNs and homing endonucleases, respectively.

In contrast to E. coli, the constructed multigene vectors were to some extent unstable in Agrobacterium. Sometimes one or more genes were lost during transformation or propagation in Agrobacterium. However, the instability should not be attributed to MultiRound Gateway but to the transgenic DNA sequences themselves. DNA sequences between the several repeated elements seemed to be susceptible to deletion due to homologues recombination, although AGL1 is a recA negative strain. Maybe a conjugational transfer of the multigene vector would result in gene loss less frequently than electroporation. For the observed gene loss in some transgenic tobacco plants also other factors should be considered, for example, homologues recombination in plants or during the process of T-DNA integration.

We observed transgenic lines which expressed all the transgenes over at least two generations (Table 2). Yet, we could not identify a clear correlation between the expressions of different transgenes located on the same T-DNA or between different lines. For example, pYLTAC7_C4_ NADP-ME lines that showed high levels of PPT expression did not necessarily exhibit high PPDK or FbCA expressions. Thus, it’s not possible to determine the efficiency of a specific promoter in a given multigene array due to differences in the gene expression levels which did not correlate with their promoter types. Similar results reported Fujisawa et al. (2009) and Zeevi et al. (2012). Comparably to Chen et al. (2010) variable expression levels of individual transgenes was not only observed between different transgenic lines but also between individual transgenes driven by the same promoter in a given multigene array. Similar promoters seem to behave differently depending on their arrangement in a multigene cluster. Furthermore gene expression is dependent on the integration locus. If they become inserted into a transcriptionally active region (Koncz et al. 1989; Kertbundit et al. 1991), expression may be influenced by regulatory sequences of nearby host genes.

In summary, the improved version of the MultiRound Gateway technology described here represents a powerful, highly efficient tool for multigene plant transformation. But its utility is not restricted to plant applications. Any appropriate DNA components can be assembled in various vectors modified according to this study for different purposes, such as transfer of multiple genes of interest to yeast, insect, or mammalian cells.