Introduction

Citrus is one of the most important fruit trees worldwide. The application of genetic transformation to improve citrus has increased over the past several decades. A number of important traits, including herbicide, insect, and pathogen resistance, improved compositions, and quality of fruit, have been successfully introduced into different citrus varieties (Atkins et al. 1995; Domínguez et al. 2000, 2002a; Ghorbel et al. 2000; Peña et al. 2001; Wong et al. 2001; Guo et al. 2005; Boscariol et al. 2006; Fagoaga et al. 2006; Febres et al. 2008, 2011; López et al. 2010; Gambino and Gribaudo 2012). In citrus genetic transformations, selectable marker genes conferring antibiotic or herbicide resistance are generally required to efficiently recover transgenic plants from transformed cells. Once the transgenes have been transferred into the plant genome and the transformed shoots successfully generated from the explants, the selectable marker gene becomes unnecessary and undesirable. Moreover, the presence of marker genes raises public concerns (such as food safety and the ecological risks of the nptII or bar genes and their production) regarding the field release of transgenic plants (Ramessar et al. 2007).

Several methods have been developed to overcome these problems (Tuteja et al. 2012). For example, co-transformation or transposon-mediated transgene repositioning were used for the removal of marker genes via segregation in crops (Daniell 2002), but this is not feasible in vegetatively propagated plants that possess a very long juvenile period. Chloroplast transformation, which has been applied to crop improvement (Daniell et al. 2005), is one of the more promising containment technologies, especially cytoplasmic sterility engineered via the chloroplast genome (Ruiz and Daniell 2005), which could provide a new tool for transgene containment in plants. Molecular containment strategies based on site-recombination offer methods to directly generate marker-free plants (Daniell 2002; Hare and Chua 2002). An advanced containment method is recombinase-mediated auto-excision, which was widely applied to avoid pollen-mediated transmission of transgenes and produce marker-free genetically modified (GM) plants (Mlynárová et al. 2006; Verweire et al. 2007).

In citrus, the neomycin phosphotransferase gene (nptII) is widely used in producing transgenic plants. Elimination of this marker gene after transformation is an important step in gaining public acceptance of transgenic fruit. Some alternative methods have been reported to produce marker gene-free or marker gene-safe transgenic woody plants (Malnoy et al. 2010; Petri et al. 2011; Vanblaere et al. 2011). Using non-selection approaches, marker-free transgenic Mexican lime and sweet orange were produced. Around 1.7–8 % of regenerated shoots were transgenic when regenerated shoots were analyzed by polymerase chain reaction (PCR) (Domínguez et al. 2002b; Ballester et al. 2010; He et al. 2011). However, transgene silencing and chimerism were found in many transformed plants. Phosphomannose isomerase (PMI), encoded by the manA gene, has no adverse effects on mammalian and plant cells (Privalle et al. 2000), and has been shown to be an effective selectable marker in many different plant species (Stoykova and Stoeva-Popova 2011). However, this gene has not been widely adopted for transgenesis in citrus, although it demonstrated good efficacy of transformation in sweet orange (Boscariol et al. 2003). The ipt gene encoding the enzyme isopentenyltransferase from Agrobacterium tumefaciens is effective as a positive selectable marker gene for plant transformation (Ebinuma et al. 1997). Using ipt as a selectable marker gene, multi-auto-transformation (MAT) systems have been developed for the production of marker-free transgenic woody plants (Ballester et al. 2007; Zelasco et al. 2007; López-Noguera et al. 2009). In citrus transformation, workable transformation efficiencies of 7.2 % for citrange and 6.7 % for sweet orange were achieved by this system. The marker gene was successfully removed from 65 % of the sweet orange transformants (Ballester et al. 2007). However, several shortcomings, such as the recovery of chimeras and the incorrect excision of the RS fragment associated with the R/RS recombination system, have been reported in citrus and other plants (Ebinuma et al. 1997; Sugita et al. 1999; Ballester et al. 2008; López-Noguera et al. 2009).

In the Cre/loxP-mediated site-specific DNA recombination system, Cre recombinase specifically recognizes a repeated asymmetric 34-bp loxP recognition site and performs precise DNA excision between the two direct loxP sites (Dale and Ow 1991). The functionality of the Cre recombinase gene can be activated by a plant endogenous trigger (Mlynárová et al. 2006), a chemical inducer (Zuo et al. 2001; Petri et al. 2012), or heat shock (Luo et al. 2008; Fladung and Becker 2011) when the selectable marker gene is no longer required. The activated Cre recombinase could “auto-excise” the DNA segment between the two loxP sites, comprising Cre itself and the selectable marker gene, with 100 % efficiency (Luo et al. 2007). Given its efficient and precise excision, the Cre/loxP system has been widely used to generate marker-free transgenic plants. In this study, a novel marker-free approach mediated by the Cre/loxP system was developed to completely remove the marker genes from transgenic citrus via a single-step transformation.

Materials and methods

Plant materials

In vitro-germinated seedlings of Jincheng orange (Citrus sinensis Osbeck) were used for transformation experiments. Seeds of Jincheng orange were collected from the National Citrus Germplasm Repository, Chongqing, China. Sterilization of citrus seeds was performed as described by Zou et al. (2008). The germination culture was maintained in darkness at 28 °C and 60 % relative humidity (RH) for 2 weeks and then changed to 28 °C, 16 h photoperiod with 45 μmol m−2 s−1 illumination and 60 % RH for an additional 3 days. Plant materials were cultured on a basal Murashige and Skoog (MS) medium (Murashige and Skoog 1962) supplemented with 30 g l−1 sucrose and solidified with 2 g l−1 Gelrite (Promega, Fitchburg, WI, USA). The pH was adjusted to 5.8 before autoclaving.

Vector construction

Molecular manipulation methods were performed as described by Sambrook et al. (2001). The ipt gene from A. tumefaciens was used as the selectable marker gene in our study. A modified version of the erGFP5 INT (gfp) reporter gene (Luke Mankin and Thompson 2001), which was interrupted by a plant intron to prevent the expression of gfp in A. tumefaciens, was used to monitor the regeneration of transgenic events.

To construct the pGLI vector (Fig. 1a), a SpeI/HindIII fragment containing the gfp expression cassette driven by the CaMV 35S promoter was inserted into a NheI/HindIII-digested pBIN 19 plasmid (GenBank No. U09365.1) to generate an intermediate vector, in which the nptII gene was replaced by gfp. In addition, a ‘CaMV 35S-MCS-nos’ fragment (two loxP sites in direct orientation were fused outside the fragment) amplified from the BinAR plasmid (Höfgen and Willmitzer 1990) by loxP-35S/nos-loxP primers was digested with BglII/NheI and inserted into the BamHI/NheI-digested intermediate vector to generate pGL-multiple cloning site (MCS). In this vector, the MCS was used for loading foreign genes. Finally, the ipt ORF fragment was amplified from A. tumefaciens by ipt-5/ipt-3 primers and digested by BglII/XhoI. The T3A transcription terminator was amplified from pER8 (GenBank No. AF309825.2) by T3A-5/T3A-3 primers and digested by SalI/XbaI. The two fragments were co-inserted downstream of the CaMV 35S promoter in the BamHI/XbaI-digested pGL-MCS to yield pGLI.

Fig. 1
figure 1

Schematic configurations of T-DNAs used for Jincheng orange (Citrus sinensis Osbeck) transformation. a T-DNA region of the pGLI construct containing the isopentenyl transferase (ipt) gene and reporter gene gfp under the control of the CaMV 35S promoter. These transgenes were inserted between two direct loxP sites. Open arrowheads indicate the presence and orientation of the loxP sites. b T-DNA region of the pGLINC and pGLI35SC constructs. The mCre recombinase gene is driven by the CaMV 35S and NosP promoters. c T-DNA region of the construct pNG as a control containing the antibiotic selectable gene nptII. d Schematic illustration of the removal of the marker gene by the Cre/loxP recombinase system. When the T-DNA was integrated into the plant genome, the expression of mCre leads to the removal of all sequences (mCre and ipt genes) between the two loxP sites from transformed cells and the production of marker-free transformants. Small arrows represent PCR primers for amplification of gfp, ipt, mCre, and post-excision fragments. RB right border, LB left border, 35S the promoter of the CaMV gene, gfp green fluorescent protein gene, nos the nos terminator sequence, L loxP recognition site, ipt isopentenyl transferase gene, T3A the polyA of the small subunit rbcS-3A gene of Rubisc (pea ribulose-1,5-bisphosphate carboxylase), nosP the promoter of the nopaline synthase gene, mCre the modified coding sequence of the Cre recombinase containing a plant intron, nptII neomycin phosphotransferase gene. H HindIII, X XbaI, Sl SalI, Sc SacI

NosP and CaMV 35S promoters were used to drive the expression of the Cre recombinase gene. To prevent the bacterial expression of cre driven by these promoters, a modified Cre gene (mCre, unpublished work), in which the coding sequence was interrupted by the intron from the PtBCP1 gene (GenBank No. JX110636), was used to investigate the deletion of transgenes. This mCre gene was T-cloned into the pUCm-T vector (Sangon Biotech, Shanghai, China) and this vector was named pUCC (the mCre gene was followed by a SpeI site). To construct the NosP::mCre fusion gene, the NosP promoter was amplified from the pBI121 vector (GenBank No. AF485783.1) by NosP-5/NosP-3 primers, then digested by HindIII/XbaI and placed upstream of mCre in the HindIII/XbaI-digested pUCC vector. To construct the 35S::mCre fusion gene, the 35S promoter was removed from the pBI121 vector by HindIII/XbaI digestion and inserted upstream of mCre in the HindIII/XbaI-digested pUCC vector. Finally, the NosP::mCre and 35S::mCre fusion genes were removed by HindIII/SpeI digestion from this vector and inserted into the HindIII/XbaI-digested pGLI to generate the pGLINC and pGLI35SC expression plasmids, respectively (Fig. 1b). Primers for the above-mentioned PCRs are:

LoxP-35S: 5′-AGCTAGCATAACTTCGTATAATGTATGCTATACGAAGTTATCCA TGGAGTCAAAGATTC-3′

Nos-loxP: 5′-CAGATCTATAACTTCGTATAGCATACATTATACGAAGTTATGGA CAATCAGTAAATTGA-3′

ipt-5: 5′-CAGATCTATGGATCTGCGTCTAATTTTCGGTC-3′

ipt-3: 5′-GCTCGAGCTAATACATTCCGAATGGATGACCT-3′

T3A-5: 5′-AGTCGACCAGGCCTCCCAGCTTTCGTCCGTAT-3′

T3A-3: 5′-GTCTAGAAAGCTTCCAGTCTAAGCTTCAAAAAGCCTAT-3′

NosP-5: 5′-CAAGCTTGATCATGAGCGGAGAATTAAGGGAG-3′

NosP-3: 5′-GTCTAGAGCGAAACGATCCAGATCCGGTGCAG-3′

The sequences of the restriction enzyme sites are in italics.

Additionally, to compare transformation efficiencies between the ipt and nptII selectable marker genes, the SpeI/HindIII fragment of the gfp expression cassette mentioned above was inserted into XbaI/HindIII-digested pBIN 19 to construct the pNG vector containing the nptII gene for recovering transformed shoots (Fig. 1c).

The plasmids were introduced into A. tumefaciens EHA 105. Transformants were selected on a YEB (Vervliet et al. 1975) solid medium supplemented with 50 mg l−1 kanamycin and further confirmed by restriction enzyme analysis and PCR.

Plant transformation

Transformation experiments were performed according to the protocol of Cervera et al. (1998b) with the following modifications: epicotyl segments were inoculated with the bacterial suspension for 30 min, and transferred to a co-culture medium with 2 mg l−1 benzylaminopurine (BA), 0.25 mg l−1 indole-3-acetic acid (IAA), 1 mg l−1 2,4-dichlorophenoxyacetic acid (2,4-D), and 100 μM acetosyringone. The culture was maintained in the dark for 2 days at 60 % RH and 26 °C. Then, the explants were transferred to either a hormone-free medium (ipt selection) or a selection medium with 2 mg l−1 BA, 0.25 mg l−1 IAA, and 50 mg l−1 kanamycin (nptII selection). Both media contained 500 mg l−1 carbenicillin to prevent bacterial contamination. The plates were maintained in the dark for 1 week with 60 % RH at 28 °C, and then changed to a 16-h photoperiod of 45 μmol m−2 s−1 illumination with 60 % RH at 28 °C. The explants were sub-cultured in a fresh medium at 2-week intervals. Shoots developed from the cut ends of the explants were shoot tip grafted on citrange seedling in vitro and cultured on a filter-paper bridge in a MS liquid medium (He et al. 2011) under 16 h photoperiod of 45 μmol m−2 s−1 illumination with 60 % RH at 28 °C. After 4 weeks, the generated plantlets were further grafted onto 6-month-old seedlings of Troyer citrange in the greenhouse. Transformation efficiency was defined as described by Ballester et al. (2007). Three experiments were performed with 200–300 explants per experiment.

GFP expression analysis

GFP activity in transformed cells and shoots was examined under a Leica M165 FC fluorescent stereomicroscope equipped with a 480/40-nm exciter filter, 505 nm LP dichromatic beam splitter and 510 nm LP barrier filter (Leica Microsystems, Wetzlar, Germany). The red autofluorescence from chlorophyll was not blocked by the use of any interference filter. The photographs were taken with a Leica DFC420 C digital camera and the images were assembled using Adobe Photoshop (Adobe Systems, Mountain View, CA, USA).

Transformation events, transgenic shoots, and escapes were defined as described by Domínguez et al. (2004). Each green fluorescent protuberance at the ends of the explants was defined as a transformation event. Regenerated shoots showing green fluorescence were considered transgenic while shoots showing red autofluorescence were considered escapes.

Molecular analysis

Total genomic DNA was isolated from citrus leaves using the Plant DNeasy Prep Kit (Qiagen, Beijing, China). Citrus genomic DNA samples without Ti-plasmid contamination were prepared as described by Zou et al. (2008). The presence of transgenes in the plant’s genome was further determined by PCR analysis. The primers used to amplify the gfp fragment were g1/g2 (5′-TTAGAGTTCGTCGTGTTTGTATAGT-3′ and 5′-ATGAAGACTAATCTTTTTCTCTTTC-3′). The predicted size of the amplified gfp segment is approximately 750 bp. For the detection of transgene deletion events from transgenic plants, two pairs of primers, i1/i2 (5′-CTAATACATTCCGAATGGATGACCT-3′ and 5′-AAGCTCATCATAGGCTGATCGAGGA-3′) and c1/c2 (5′-CTAATCGCCATCTTCCAGCAGGCGC-3′ and 5′-ATGTCCAATTTACTGACCGTACACC-3′), were used to assay ipt and mCre genes, respectively. The predicted 500-bp ipt and 1.2-kb mCre segments should be detected in transgenic plants if no deletion occurred. The PCR reactions were performed as follows: initial denaturation at 94 °C for 3 min, followed by 30 cycles at 94 °C for 0.5 min, 58 °C for 0.5 min, and 72 °C for 1 min, and a final cycle at 72 °C for 10 min.

Genomic DNA from leaves of transgenic plants growing in the greenhouse was used to investigate post-excision events. The two oligos, rb and lb (5′-GGCCAGTGAATTCGAGCTCGGTACC-3′ and 5′-GCGCGCGGTGTCATCTATGTTACTA-3′), specific to the T-DNA sequences outside of the two loxP sequences, were used as primers to investigate the molecular characteristics of the transgene deletions. PCR reactions were carried out as follows: 94 °C for 3 min followed by 40 cycles at 94 °C for 0.5 min, 58 °C for 0.5 min, 72 °C for 5 min, and a last extension at 72 °C for 10 min. The PCR products were cloned into a pGEM-T easy vector following the supplier’s instructions (Promega, Fitchburg, WI, USA) and sequenced.

To evaluate the effects of marker deletions on transgenic expression levels, RNA was extracted from leaves of transgenic plants in the greenhouse according to the EASYspin Plant RNA Extraction Kit manual (Aidlab, Beijing, China). First-strand complementary DNA (cDNA) was synthesized from 1 μg of total RNA with iScript™ cDNA Synthesis Kit (Bio-Rad, Hercules, CA, USA). The primers gfp-f (5′-CACCCTCGTCAACAGGATCG-3′) and gfp-r (5′-GTTGGCTTTGATGCCGTTCT-3′) were used to amplify the expression of the gfp gene in transgenic plants. To analyze the expression of the ipt and mCre genes after their deletion, two sets of primers, ipt-f/ipt-r (5′-GCGGCCAAGGCCAGAGTTAA-3′ and 5′-AATGGGCCTCAGCCGAGGTT-3′) and cre-f/cre-r (5′-AGCCGAAATTGCCAGGATCA-3′ and 5′-AACCAGCGTTTTCGTTCTGC-3′), respectively, were used to amplify the transcript products from the transgenic plants. Expression of the citrus actin gene (GenBank Accession No. GU911361.1) as an internal control was detected with the primers CtAct-f/CtAct-r (5′-CATCCCTCAGCACCTTCC-3′ and 5′-CCAACCTTAGCACTTCTCC-3′). cDNA was amplified in 20-μl reaction volumes using the 2× iQ™ SYBR® Green Supermix (Bio-Rad, Hercules, CA, USA). PCR reactions were carried out as: a pretreatment (94 °C, 3 min) followed by 40 amplification cycles (94 °C, 15 s; 56 °C, 15 s; 72 °C, 15 s), and a final extension at 72 °C for 5 min.

For Southern hybridization analysis, approximately 200 μg of genomic DNA from each transgenic and non-transgenic plant was digested with SacI, subjected to electrophoresis in a 0.8 % (w/v) agarose gel, transferred onto a nylon membrane (Hybond-N+, Amersham Biosciences, Buckinghamshire, UK) and cross-linked to the membrane with ultraviolet (UV) irradiation. The digoxigenin-labeled PCR product of the gfp gene was used as the probe. DNA labeling, hybridization, and immunological detection were performed according to the manufacturer’s instructions (DIG High Prime DNA Labeling and Detection Starter Kit II; Roche, Basel, Switzerland).

Results

Construction of pGLINC and pGLI35SC vectors

To investigate the efficacy of the Cre/loxP recombination system for the production of marker-free transgenic plants in citrus, an auto-excision system, using the ipt from A. tumefaciens as a positive selectable marker, was designed to remove the marker gene from the citrus genome (Fig. 1). Cre recombinase and ipt genes were flanked by two loxP sequences in the direct orientation. A modified gfp gene containing a plant intron as the reporter gene was placed outside the loxP sequences. The NosP and CaMV 35S promoters were selected to control the expression of the cre gene (Fig. 1b). To prevent any Cre activity driven by these promoters in bacterial cells, an intron from the PtBCP1 gene was introduced to the coding region of the cre gene, giving rise to the mCre gene. The restriction analysis and PCR demonstrated that no unexpected deletions from the pGLINC or pGLI35SC plasmids were detected in Escherichia coli or A. tumefaciens, suggesting that the intron abolished all enzymatic activity of the mCre gene in the microbial cells (data not shown). Theoretically, once the T-DNA was integrated into the plant genome, the expression of the ipt gene driven by CaMV 35S should allow for the regeneration of transformed cells. If the correct splicing of the introduced plant intron and the functionality of the Cre recombinase occurs in the plant, then the sequences (including mCre and ipt genes) between the two loxP sites would be deleted from transformed cells to produce marker-free transformants (Fig. 1d). As the control for nptII selection, the pNG vector without the mCre gene was constructed as shown in Fig. 1c.

Regeneration of transgenic citrus plants

Using the Agrobacterium-mediated transformation method, the above constructs were introduced into Jincheng orange (C. sinensis Osbeck). Epicotyl explants were infected with A. tumefaciens harboring pGLINC, pGLI35SC, or pGLI, and after 2 days of co-culture, the explants were cultured on hormone-free medium. The development of transgenic events was monitored by observing GFP activity with a fluorescence stereomicroscope. After 1–2 weeks of culturing, the visible calli that began to form at the ends of explants exhibited green fluorescence (Fig. 2a), indicating transgenes were successfully transformed into these cells. After two more weeks, some buds with green fluorescence emerged in these protuberances (Fig. 2b). About 1-cm-long transformed shoots showing strong green fluorescence were harvested continuously for 6 weeks after infection (Fig. 2c). In general, adventitious shoots could be obtained within 4–6 weeks after infection (Fig. 2d, e). In contrast, only a few buds were formed on the selective medium with 2 mg l−1 BA, 0.25 mg l−1 IAA, and 50 mg l−1 kanamycin when the pNG vector, which did not contain the ipt gene, was introduced into explants (Fig. 2f). These results showed that the expression of ipt improved reproduction and differentiation of the transformed cells at the first stage of in vitro regeneration when T-DNA was integrated into the citrus genome. Most of the transformed shoots could be recovered by in vitro grafting (Fig. 2g–i) and showed a normal phenotype after being grown in the greenhouse (Fig. 2j), suggesting that the transgene deletion occurred at the early stage of in vitro regeneration. Interestingly, in our experiment, about 18 % of the transformed shoots with pGLI35SC constantly displayed the ipt-shooty phenotype, which includes thicker leaves, adventitious branches, and the loss of apical dominance, from the early stage of regeneration (Fig. 2e), through post in vitro grafting (Fig. 2h, i), to the greenhouse growth stage (Fig. 2k). However, most of them did not live for a long time after being transplanted in the greenhouse.

Fig. 2
figure 2

Monitoring GFP activity in regenerated marker-free transformants of Jincheng orange (Citrus sinensis Osbeck) after transformation. a Initiation of a callus protuberance at the cut end of an explant after 2 weeks of culture on a hormone-free medium. The green fluorescent cell cluster (G) corresponds to a transgenic event and the red autofluorescent cell cluster (R) are non-transformed cells. b A developing shoot showing green fluorescence from the callus cluster after 4 weeks of culture on a hormone-free medium. c The transgenic shoot shows strong green fluorescence, whereas two escapes (arrowheads) clearly appear red after 6 weeks of culture on a hormone-free medium. df Explants transformed with pGLINC (d), pGLI35SC (e), and pNG (f) constructs after 4 weeks of culture on a hormone-free medium (d and e) or a selective medium (f). gi Transgenic shoots micrografted on citrange seedlings in vitro. A transgenic shoot containing pGLINC shows a normal phenotype (g), and the transgenic shoots containing pGLI35SC show the ipt-shooty phenotype (h and i). j and k Transgenic plants growing in the greenhouse. A phenotypically normal transformant growing in the greenhouse (j). The ipt-shooty transformant from i growing in the greenhouse while other lines from h died soon after being grafted onto a vigorous rootstock in the greenhouse. Strong green fluorescence was detected in leaves from the transgenic plants (j and k). In a, b, and c, photographs were taken under UV light (upper) and white light (lower). Bar 2 mm (a, b, c, j, and k), and 2.5 mm (di) (color figure online)

Transformation efficiency

To evaluate the transformation efficacy achieved by this auto-excision strategy, transformation events, transformation efficiency, and transgenic shoot regeneration were investigated in detail as described by Domínguez et al. (2004). Using visible green fluorescent calli at the cut ends of the explants as transformation events, the number of explants with callus protuberances showing GFP activity (Fig. 2a) was counted 2 weeks after infection with A. tumefaciens. As shown in Table 1, using ipt selection, GFP-positive calli were observed in 14.3–66.7 % of the explants. The frequencies of explants with transformation events obtained were much higher than that using nptII selection (Table 1). From 4 to 8 weeks after infection, the explants with GFP-positive shoots were counted. The results of our statistical analysis showed that high transformation efficiencies, 28.1 % for pGLINC and 13.6 % for pGLI35SC, could be achieved with ipt selection (Table 1). To evaluate the regeneration capacity of transgenic events, shoots at the ends of infected explants were randomly screened by observing GFP activity. Analysis of GFP activity showed that 36 out of 96 regenerated shoots transformed by pGLINC were GFP positive while 24 out of 120 shoots transformed by pGLI35SC demonstrated GFP activity (Table 1). For nptII selection, only 4 out of 85 analyzed shoots showed GFP activity (Table 1). These results indicated that ipt selection markedly improved transformation of sweet orange as compared with nptII selection. No GFP-positive shoots were detected, even though visible transformation events were observed in 66.7 % of explants, when the pGLI construct was used (Table 1), suggesting that the continuous expression of ipt hindered regeneration of transformed shoots.

Table 1 Investigation of transformation efficiency using different constructs in Jincheng orange (Citrus sinensis Osbeck)

Molecular analysis of marker-free transgenic plants

GFP-positive transgenic plants were subjected to PCR analysis. The specific primers g1 and g2 (Fig. 1d) were designed to confirm the integration of the gfp gene. Consequently, the predicted 750-bp gfp fragment was detected in all GFP-positive plants, and no PCR product was found in the untransformed plants (Fig. 3). Southern blot analysis further confirmed the integration of transgenes into the citrus genome and 1–4 copies of the foreign inserts were detected in transgenic plants (see Online Resource 1). These results indicated that the foreign genes were incorporated into the genomes of the GFP-positive plants and the modified gfp was an efficient reporter gene for the genetic transformation of citrus.

Fig. 3
figure 3

PCR analysis of marker-free transgenic Jincheng oranges (Citrus sinensis Osbeck) by amplification of the gfp, ipt, and mCre transgenes. Lane M DNA molecular size marker, lane P1 plasmid pGLINC as template, WT template from non-transgenic plants, lanes N2 N3, N4, N5, N8, and N16 samples from pGLINC transgenic plants, lane P2 plasmid pGLI35SC as template, lanes S1, S2, S3, S6, and S14 samples from pGLIN35SC transgenic plants. All plant DNAs were extracted from the shoots 2 weeks after being micrografted on seedlings in vitro

According to the deletion strategy in our experiment, all extraneous DNA sequences between the two loxP sites would be deleted by a CaMV 35S/NosP-driven recombinase when the T-DNA was integrated into the plant genome (Fig. 1d). As a result, ipt and mCre genes should not be detected. The two pairs of primers, i1/i2 and c1/c2 (Fig. 1d), were used to investigate deletions of the ipt and mCre genes, respectively. As shown in Table 2 and Fig. 3, PCR results revealed that neither the ipt nor the mCre fragment was detected in any of the 34 pGLINC transgenic plants; however, the 500-bp ipt and 1.2-kb mCre specific fragments were found in 4 out of 22 pGLI35SC transgenic plants. The results showed that the NosP-mediated Cre recombinase could delete the ipt marker gene from all pGLINC transgenic plants and achieve a very high deletion efficiency (100 % in this study), while the CaMV 35S promoter delivered a much lower deletion efficiency (81.8 %) (Table 2).

Table 2 PCR analysis of marker-free transgenic Jincheng orange (Citrus sinensis Osbeck)

In addition, the molecular characteristics of post-excision T-DNAs were profiled by PCR. Using the rb/lb primer pair (Fig. 1d), a 750-bp post-excision fragment should be detected in the transgenic plants when a transgene deletion occurred, otherwise an approximately 4.5-kb T-DNA fragment containing ipt and mCre genes would be amplified (Fig. 1d). PCR analysis demonstrated that only the 750-bp post-excision fragment was detected in all pGLINC and 18 pGLI35SC transgenic plants (Fig. 4a), indicating the complete deletion of the transgenes in these plants. However, both the 4.5-kb and 750-bp fragments were detected in 4 pGLI35SC transgenic plants (Fig. 4a), suggesting that the incomplete deletion of the transgene occurred in these lines. The 750-bp post-excision fragment was subjected to DNA sequencing analysis. The sequencing results revealed that a predicted 742-bp identical sequence consisting of a single loxP site was present in all lines (Fig. 4b), indicating the deletion occurred between the two loxP direct repeats, and that the Cre-mediated deletion of the transgene was highly precise.

Fig. 4
figure 4

Site-specific recombination and deletion analysis in representative transgenic Jincheng orange (Citrus sinensis Osbeck). a PCR analysis of transgene deletions from transgenic plants. Lane M DNA molecular size marker, WT template from non-transgenic plants, lane P1 plasmid pGLINC as template, lanes N2, N3, N4, N5, N8, and N16 samples from pGLINC transgenic plants, lane P2 plasmid pGLI35SC as template, lanes S1, S2, S3, S6, and S14 samples from pGLIN35SC transgenic plants. All plant DNAs were extracted from plants growing in the greenhouse. The rb/lb primer pair was used to investigate the molecular characteristics of the deletions made by the Cre/loxP system. b Sequence analysis of site-specific recombination by the Cre/loxP system. A predicted 742-bp identical sequence with a single loxP site was detected in all lines, indicating that the deletion junction was located within the two loxP direct repeats and that the Cre-mediated transgene deletion was precise

The presence of ipt and mCre genes was detected in several pGLI35SC plants, which showed the ipt-shooty phenotype. To investigate expression of the transgenes, total RNAs from leaves of transgenic plants growing in the greenhouse were used to evaluate the expression level of the transgene by real-time quantitative PCR. As shown in Fig. 5, high transcript levels of the gfp gene were detected in all transgenic plants (Fig. 5a). In most of the transgenic plants, no expression of ipt or mCre was found, while strong transcript levels were detected in S1, S2, S3, and S14 pGLI35SC plants (Fig. 5b, c) that had the ipt-shooty phenotype. These quantitative PCR results further indicated that expression of the incompletely removed ipt marker gene in the four plants (Fig. 3) impaired plant growth.

Fig. 5
figure 5

Quantitative PCR analysis of marker gene deletions in transgenic Jincheng orange (Citrus sinensis Osbeck). Using leaf total RNA from the transgenic plants in the greenhouse as templates, strong expression levels of gfp were detected in all transgenic plants (a); and neither ipt nor mCre transcript products were detected in transgenic plants, except the S1, S2, S3, and S14 transgenic lines (b and c). WT template from non-transgenic plants, N2, N3, N4, N5, N8, and N16 samples from pGLINC transgenic plants, S1, S2, S3, S6, and S14 samples from pGLIN35SC transgenic plants. Error bars indicate SD values (n = 3)

Discussion

In this study, we demonstrated that combining ipt as a selectable marker gene with the Cre/loxP recombinase system could efficiently produce marker-free transgenic citrus via a single-step transformation. The ipt-type vector system has been evaluated in many plant species (Ebinuma et al. 1997; Zelasco et al. 2007; Luo et al. 2008; López-Noguera et al. 2009; Khan et al. 2011). In these reports, the removal of the selectable marker gene was mainly achieved by the Ac transposable element, or the R/RS or FLP recombination system. Ballester et al. (2007) first showed that the R/RS-mediated pMAT vector system could efficiently delete marker genes from transgenic citrus. However, R/RS-mediated transgenic recombination did not always work precisely in many citrus lines, and could generate mutant sequences in the plant genome around the recombination site, which resulted in the low level or non-expression of trait genes. In apricot, an increased lack of precision in the recombination system had also been described with the R/RS-mediated pMAT vector system (López-Noguera et al. 2009). This shortcoming makes R/RS-mediated transgenic recombination unsuitable for site-specific integration or exchanges of transgenes, which will become increasingly important for the precise, predictable, and stable production of transgenic crops in the post-genomic era (Ow 2002). Our sequencing results showed that the same post-deletion sequence containing a single loxP site was found in the transformed lines, demonstrating that the Cre-mediated deletion of the transgene was always precise (Fig. 4b). This is in accordance with the reports on the Cre/loxP system-mediated genetic deletions in other species (Mlynárová and Nap 2003; Luo et al. 2007). The precise deletion of the auto-excision system could offer the potential to manipulate site-specific integration and exchanges of transgenes in citrus transgenic breeding. Moreover, using the Cre/loxP system, the marker gene was successfully removed from >80 % of Jincheng orange lines (Table 2), giving rise to fully normal ipt-free transgenic plants. This was more efficient than transgene removal driven by the R/RS system (Ballester et al. 2007).

Selectable marker genes conferring antibiotic or herbicide resistance usually offer satisfactory transformation efficiencies. However, some concerns regarding the biosafety of these genes, such as human health effects and environmental risks, are hindering the field release of GM plants. Other alternative marker genes, such as ipt, iaaM/H, and PMI/mannose, which are believed to be biologically safe, have been used in transgenic citrus (Ballester et al. 2008). Our results demonstrated that high transformation efficiencies could be achieved with ipt as the selectable marker gene when compared with nptII selection (Table 1), suggesting that the ipt gene was a good alternative to nptII for citrus transformation. It has been shown that the overproduction of cytokinin by the ipt gene is needed for the proliferation of transgenic cells. The excision of the ipt gene rapidly terminated the overproduction of cytokinin in the transgenic cells (Endo et al. 2002). This “hit and run” changes the cytokinin level, which could cause the direct regeneration of shoots from transgenic cells on a hormone-free medium (Endo et al. 2002; Ballester et al. 2007; Zelasco et al. 2007). Similarly, in our marker gene deletion strategy, Cre-mediated deletion and ipt expression occurred simultaneously at the early stage of transformed cell proliferation. Once the deletion occurred the transcription of the ipt gene was terminated, which may have resulted in the temporal overproduction of cytokinin in the transformed cells. Even so, this temporal change of the endogenous hormone level can also improve transformed shoot regeneration on a hormone-free medium, showing that the Cre-mediated deletion at the early stage of transformation could provide enough time for ipt expression to induce shoot regeneration. In addition, transformation events were detected in 66.7 % of affected explants when the ipt gene was expressed continuously in transformed cells containing the pGLI construct. However, under the current conditions, no transgenic shoots were regenerated (Table 1). By contrast, when using the Cre/loxP system, the temporal expression of the ipt gene could markedly increase the regeneration frequency (from 0 up to 28.1 %) of transgenic shoots, although the frequency of visible transformation events was markedly decreased from 66.7 to 33.0 % (Table 1). These results indicated that the timing of the ipt-mediated cytokinin production at the early stages in transformed cells might be a key factor that affects transformation efficiency in citrus.

Chimerism, in which both excised and non-excised T-DNA sequences coexist because of non-complete deletions in the cells of transgenic plants, is a drawback of the recombination-based marker-free systems (Sugita et al. 1999; Mlynárová and Nap 2003; Ballester et al. 2007). PCR analysis demonstrated that the expression of the Cre recombinase driven by the CaMV 35S promoter did not completely delete the ipt gene in transgenic plants and that four chimeric plants were detected from the 22 pGLI35SC transgenic plants (Figs. 3, 4; Table 2). As a result, the ipt gene was continuously expressed in some cells of these plants (Fig. 5c), and the obtained transgenic plants exhibited the ipt-shooty phenotype (Fig. 2h, I, k). Finally, three of these shoots died within 6 months of being transferred to the greenhouse. These results indicated that the imperfect deletion of the ipt gene caused these transgenic plants to exhibit the ipt-shooty phenotype. Conversely, no transformed shoots with pGLINC, in which the Cre recombinase was driven by the NosP promoter, showed the ipt-shooty phenotype after in vitro grafting (Fig. 2g, j). PCR analysis confirmed that ipt and mCre genes were deleted completely from these transgenic plants (Table 2; Fig. 3). These data demonstrated that expression of the Cre recombinase controlled by the strong CaMV 35S promoter was more prone to cause chimeric deletions when compared with the weak NosP promoter. The chimeric deletions were presumably caused by the incomplete recombination action of the continuously present Cre protein (Mlynárová and Nap 2003). Therefore, the suitable expression of the recombinase gene is possibly an important factor in reducing the frequency of incomplete deletions. In our experiment, the NosP promoter was a suitable driver for the Cre-mediated marker-free system in citrus. The control of transgenic deletions by a tissue-specific promoter directing recombinase expression could provide other good candidates that avoid the problems of chimerism and inefficient deletions (Verweire et al. 2007; Van Ex et al. 2009).

In our study, the production of marker-free transgenic plants was performed by a “single-step transformation” (Endo et al. 2002; Zelasco et al. 2007). In this kind of transformation, the ipt-shooty intermediate of transformed shoots was hardly detected (Endo et al. 2002). It showed that most of the transgenic shoots, through an ipt-shooty intermediate, were marker-free chimeras (Table 2). As a result, it is difficult to screen transgenic shoots by observing the ipt-shooty intermediate at the early stage of transformation. In addition, the gfp reporter gene is not needed in the practical manipulations of transgenic plants with target traits, although its utility is very convenient for detecting transgenic shoots. Molecular techniques are, however, essential for screening transgenic shoots. Thus, the regeneration capacity of transgenic shoots is vital for the application of the auto-excision system. Our investigation demonstrated that in a scale analysis of 96–120 shoots, 20.0–37.5 % of the regenerated shoots were transformants (Table 1). Moreover, it was confirmed that the marker gene had been removed completely from most of these transgenic shoots (Table 2). These results showed that in combination with the ipt marker gene, Cre-mediated deletions could efficiently improve the regeneration of transgenic shoots from the ends of explants, which would save both time and effort when screening marker-free transgenic plants.

Our study showed that the auto-excision system could efficiently produce marker-free transgenic plants from in vitro-germinated seedlings. However, further evaluation of this system is required using the transformation of mature explants that are vital for many citrus varieties (especially, seedless varieties), because the regeneration of transgenic plants from mature tissues of citrus could overcome the problems of the long juvenile period and high heterozygosity (Cervera et al. 1998a; Gambino and Gribaudo 2012). Ballester et al. (2007) reported that marker-free transgenic shoots were efficiently regenerated from internodal segments of greenhouse-grown pineapple sweet oranges, suggesting that the ipt-mediated increase of endogenous cytokinin could improve regeneration of transgenic shoots from mature explants. Thus, the auto-excision system mediated by Cre/loxP could also be efficient in genetic transformations with mature explants. The generation of marker-free transgenic plants by the presented auto-excision system also requires further evaluation in different citrus varieties to confirm it could effectively delete marker genes from transformed commercial cultivars. In conclusion, we developed a highly efficient Cre-mediated method for the production of marker-free transgenic citrus using ipt as the selectable marker gene. By this approach, high transformation efficiencies and 100 % marker-free transgenic plants were achieved. Our results demonstrate that this deletion system offers a good alternative transgenic manipulation that produces marker-free transgenic citrus, which may relieve public concerns regarding the safety of GM organisms.