One of the global challenges for the next decades is the reproducible and sustainable production of wood to meet the increasing demand for energy and solid raw material. The majority of the terrestrial biomass is produced by forest trees, which are grown either in natural (primeval and secondary) forests or, with increasing significance, in tree plantations. Plantation forestry is predicted to become even more important in the future to reduce the pressure on primeval forests in an effort to support ecologically sustainable and economically profitable wood production. One substantial opportunity for plantation forestry lies in the ability to use improved domesticated tree varieties or even genetically modified (GM) trees, specifically designed for a respective end-use, e.g. low-lignin trees for pulp and paper or saccharification (bioethanol production), or high-lignin trees for solid wood combustion.

Improving trees by conventional breeding is time-consuming and often not cost-effective due to the long vegetative periods and long reproduction cycles [1]. The availability of whole genome sequences of forest trees offers the opportunity to detect novel genes responsible for important developmental processes like tree growth or wood production. In combination with the publicly accessible whole genome sequences for Populus trichocarpa [2] and Eucalyptus grandis (http://eucalyptusdb.bi.up.ac.za/), the development of new genomic tools like "Target Induced Local Lesions IN Genomes" (TILLING, [3]) or the production of genotypes carrying novel (desired) gene combinations offer the opportunity to fasten tree domestication.

The P. trichocarpa genome is approximately 403 Mb in size, arranged in 19 chromosomes and assembled into 2,518 scaffolds. The number of loci containing protein-coding transcripts is 40,668, but 45,033 protein-coding transcripts have been detected (annotation v2.2 of assembly v2.0; Phytozome v7.0; http://www.phytozome.net/poplar). However, only for a minority of these loci the functions of the protein-coding transcripts are positively known. For tree species including poplar, only very few mutants have been described that could be used to analyse specific gene function behind the mutation [4]. Induced mutagenesis combined with phenotyping tools offer significant opportunities for linking gene models with putative functions. Over the past decade, genomics reagents have become available to produce a wealth of tagged mutant plants in particular for annual model species. Mutant induction in such annual plants by T-DNA insertion or using the mobility of transposable elements (e.g. the maize Ac element or its inactive derivate Ds) in most cases was achieved using knock-out tagging, disrupting a functional pathway by element insertion in functional genes and subsequent selfing of mutagenized plants. In Arabidopsis, it is now possible to acquire a mutant of nearly every gene model by using publicly available populations of T-DNA [5, 6] or transposon [7, 8] insertional mutagenesis lines. Similarly, large scale gene tagging resources have been developed for rice [9, 10]) and barley [11, 12].

The use of loss-of-function mutations described above is not well suited for application in long-living trees. In contrast, gain-of-function strategies have significant advantages because affected genes are not disrupted but activated [13–15]. One gain-of-function approach is "Activation tagging" which means the up-regulation of an endogenous gene through presence of a tag containing strong enhancers [16] or promoters facing outwards [17, 18]. The concept behind transformation-based activation tagging is that the enhancers or the promoter are located on the T-DNA (or the transposon), and following insertion of the T-DNA close to a gene, its transcription will be activated. For Arabidopsis, large sets of "activation tagging populations" have been generated containing several T-DNA-based activation tagging vectors which are readily available from insertion collections and stock centers [19, 20].

Despite the publication of some promising reports that describe the creation of T-DNA-based activation-tagged populations in poplar [15, 21] and the identification of GA2-OXIDASE, a dominant gibberellin catabolism gene, as the first gene to be isolated from such a population [22], efficient gene tagging system for long lived forest tree species are still wanting. In order to fill this gap, Fladung et al [13] and Kumar and Fladung [23] proposed the use of a transposon-based activation tagging system for poplar. This proposal was based on the fact that the maize transposable element Ac is functional in the Populus genome [24], and re-integrations occur at high frequencies in or near coding regions [23]. Further, the majority of re-integrations were found scattered over many unlinked sites on other scaffolds than the one carrying the original integration locus, confirming that Ac does in fact cross chromosome boundaries in poplar [25].

In this paper, we describe for the first time the development of an efficient activation tagging system for aspen-Populus based on a non-autonomous "Activation Tagging Ds" (ATDs) system as described by Suzuki et al [26], in combination with a heat-inducible Ac-transposase. Four activation-tagged populations comprising in total 12,083 individuals have been produced and phenotyped. Many of the phenotypes have not been described before. Molecular analyses of individuals of the mutant population confirm the excision of the ATDs element from the original insertion locus and re-integration into or close to a gene locus, with unknown function in many cases. In a second, "blind" approach (without any phenotypic selection), 300 randomly selected individuals were PCR-screened for ATDs excision. In approximately one third of the investigated individuals, ATDs transposition was confirmed and analyses of the new genomic positions of ATDs reveal a very high percentage of tagged genes.

This system might prove particularly useful not only in poplar but also in other long-lived forest and fruit tree species where T-DNA-based activation tagging systems are not reliable due to the lack of high-efficiency transformation protocols.

Different mutagenesis approaches based on heterologous (transferred) transposon element systems have been successfully applied in many plant species. Most prominently, the two element maize Ac/Ds system has been successfully used to generate insertional mutants in Arabidopsis, rice or barley [12, 27–31]). In order to establish a similar transposon tagging system for trees, Fladung and Ahuja [24] transferred the autonomous Ac element to aspen-Populus and for the first time confirmed that Ac is functionally active in this tree species. Molecular evidence for Ac excision and re-integration into the genome was later provided by Kumar and Fladung [23]. Further, these authors showed that the majority of Ac genomic re-integration sites were found within or near coding regions. More recently, Fladung [25] analyzed in detail the genomic positions of Ac re-integrations by blasting Ac-flanking aspen sequences against the publicly available genome sequence of P. trichocarpa v2.0 (Phytozome v7.0; http://www.phytozome.net/poplar). The majority of re-integrations were found scattered over many unlinked sites on different scaffolds confirming that in poplar Ac is able to cross chromosome boundaries. These latest results confirmed the feasibility of the approach first suggested by Kumar and Fladung [23] to use the Ac/Ds transposon tagging system for functional genomics studies in forest tree species, and in particular, for an efficient induction of mutants.

In this study, we took advantage of the already available "Activation Tagging Ds" system (ATDs) developed by Suzuki et al [26] that contains outwards directed 35S promoters at both ends. For our study, this ATDs system was combined with the phenotypic selectable marker gene rolC [23, 32], which was cloned outside of the ATDs element so that it is active when ATDs is not excised. This gene construct was transformed into two already transgenic TRANSPOSASE-expressing aspen-Populus lines. A gain-of-function rather than a loss-of-function strategy was used as this approach does not disrupt gene expression, avoids issues of gene redundancy and allows screening to occur in a primary generation. In earlier work, an "Activation tagging" approach has been recommended as particularly practicable for application in long-living trees [13, 22].

To date, successful T-DNA-based activation tagging mutagenesis in trees has been reported only for poplar [14, 15] and GA2-OXIDASE, a gibberellin catabolism gene, was the first tree gene that was isolated from a poplar T-DNA insertion population comprising 627 individuals [22]. In the following years, other T-DNA activation tagging poplar populations were produced and screened for developmental abnormalities including alterations in leaf and stem structure as well as overall stature by Harrison et al. [21] The mutant frequency reported for the largest activation tagging poplar population (with 1,800 independent transgenic lines) was about 2.4%. In contrast, in our study, a total of 12,083 individuals were produced and screened, but our visible mutant frequency (containing also leaf and stem phenotypic alterations) was only 0.24%. However, in an additional "blind" approach (without any previous phenotyping), we determined a frequency of 32% of ATDs transpositions in randomly selected heat shocked plants. Thus, by considering only positive ATDs-tested (transposed) individuals, the mutant frequency could be raised to approximately 1%. At present, we are working on a further increase of the mutant frequency by using a positive reporter gene system combined with the ATDs system. This system only allows shoots to regenerate when the reporter gene is not active any more and thus ATDs is excised.

Thus, critically to our heat shock-based TRANSPOSASE-induction strategy, the heat shock regime itself seems to influence ATDs excision rate. This is consistent with observations made in a carefully performed study on flowering response following heat shock induction of the FT gene controlled also by the soybean Gmhsp17.5-E heat shock promoter, in which both the size of the treated plants as well as the temperature regime influenced success of flower induction [33]. Daily heat treatments (1-2 hours at 37°C) over a period of three weeks or heat treatments of shorter durations but with increased inductive temperature (from 37°C to 40°C) were reported to be successful for efficient flower induction in greenhouse grown plants taller than 30 cm [33]. In a previous study on the induction of a FLP/FRT recombination system, the soybean heat shock promoter was induced after incubation of in vitro grown transgenic poplar plants and regenerative calli at 42°C for 3 hours [34]. Transposase induction following heat treatment of in vitro grown individuals from double transgenic lines was also confirmed by RT-PCR (data not shown).

Possible explanations for the overall relatively low frequency of ATDs transposition could be silencing effects due to double insertion of the ATDs element or chromosomal position of the original (donor) ATDs locus. Early evidence for a relationship between T-DNA copy number and repeat formation as well as promoter methylation in poplar has been provided by Kumar and Fladung [35]. However, among the 23 different double transgenic lines carrying one to four copies of ATDs, no notable correlation was found between copy number and mutant frequency.

Alternatively, in ten (N82-3, -4, -5, -7, -8, -10, -11, N92-3, N95-2, -3) out of 23 primary double transgenic, non-ATDs transposed lines, annotations of the ATDs donor locus flanking genomic sequences revealed insertion into or nearby genes. These ten lines, which themselves can be considered as T-DNA tagged variants, yielded only twelve ATDs-tagged variants. On the contrary, analysis of genomic sequences flanking ATDs donor loci in the two lines with the highest number of phenotypically tagged lines (N82-2 with 5 and N82-14 with 7) revealed no transcript annotation. A similar trend was observed in our anonymous approach. Here, randomly selected heat-shocked plants were first PCR-screened for successful ATDs excision, and, in a second step, ATDs excision-positive plants were analyzed for genomic localization of new ATDs insertion sites. Out of 128 tested plants from six of the above mentioned ten lines with annotations, 30 positive ATDs excisions (23.4%) and 7 BLAST hits (5.5%) were detected. However, three lines without any positive annotation of the ATDs donor locus flanking genomic sequences (N82-14, -15, N92-1) revealed 34 positive ATDs excisions (59.6%) and 16 BLAST hits (20.2%) in 57 tested plants.

The variations in phenotype in some of the ATDs-tagged mutants might be similar to those observed by Harrison [36] explaining partial silencing of the shriveled leaf mutant due to methylation effects. A positive correlation between 35S enhancer element methylation and low frequency of T-DNA-based activation tagging was reported by Chalfun-Junior et al. [37] Further, an early report describes the influence of endogenous and environmental factors on 35S promoter methylation [38]. Because ATDs is carrying both the four repeats of enhancer elements as well two 35S promoters, variations of mutant phenotypes are possible. Further, to confirm that the variants obtained are truly transposon-tagged and possibly also to explain observed phenotype variations, we already have initiated heat-shock treatments of some variants to restore the wildtype phenotype. Further, semi-quantitative PCR analyses are underway to confirm the activation of the transcripts in the variant lines.

Tagging approaches based on T-DNA insertion are effective only for plant species (like Arabidopsis and poplar) that can be easily transformed and for which high frequencies of tagged lines can be obtained [28]. One possible advantage of T-DNA based activation tagging could be that even T-DNA insertion sites are not randomly distributed in the genome but do show some insertion site preferences to the 5'UTR of a gene coding region [39]. For transposable elements, however, new insertion sites were found scattered throughout the genome at many unlinked sites [28, 31, 40]. But similar to results obtained for poplar [25], also for Arabidopsis a preferential transposon insertion around transposon donor sites was found by Raina et al. [41] However, in many of the heat shocked ATDs-excision positive tested plants analyzed in this study, the scaffolds revealing the new ATDs insertion loci are unlinked to those harbouring the original donor locus.

The fact that a transposon is able to jump to other chromosomes, thus passing chromosomal boundaries, leads to the convenient situation that only a few primary transposon transgenic lines are required for the establishment of large transposon tagging populations in order to tag at least theoretically every gene in a tree genome. This would be difficult to achieve through T-DNA tagging as plant transformation is time consuming and, therefore, the genome can't easily be saturated with T-DNA tags.

For both T-DNA and transposon activation tagging, the strategy followed so far was to first phenotype an existing tagging population and then to determine new genomic insertion loci of a tag. Based on our results presented here, we propose a novel strategy of activation tagging that is supported by the demonstrated power of the ATDs transposon approach and the simplicity to induce ATDs transposition in vitro at least in some lines. The ATDs-based strategy allows first the production of a very high number of independent ATDs-transposed plants that can be screened for new ATDs flanking genomic loci. The sequences obtained in this way can then be subjected to BLAST analyses, and finally based on this in silico research, variants of specific interest can be selected, transferred to and investigated in the greenhouse.