For analysis of ABA3 transcriptional regulation 1.0 kb upstream of the PtABA3 coding sequence was amplified from Populus trichocarpa genomic DNA (Primer: Table S1) and the resulting putative ptaba3 promoter sequence was cloned into pRT99GUS [37] using restriction enzymes SmaI and EcoRI. Subsequently, the cassette consisting of the aba3 promotor, the β-glucuronidase (GUS) coding sequence was transferred into the binary vector pBIN19 [38] using HindIII restriction sites at the 5’ end of the ptaba3 promotor and the 3’ end of the terminator. AO transcriptional changes were analysed accordingly using a 1.5 kb sequence upstream of ptao2 that was defined as putative promoter sequence and fused to gusA to generate ptao2::gusA transformants. The ptao2 promoter sequence was amplified via nested PCR using P. trichocarpa genomic DNA (Primer: Table S1). The resulting fragment was cloned into the binary vector pKGWSF7 via GATEWAY technology (Life technologies, Darmstadt, Germany). Correctness of cloning was confirmed by DNA sequence analysis (Eurofins MWG Operon, Ebersberg, Germany).

Generation of constructs for PtAO2 and PtABA3 knock down transformants was accomplished according to [36] . For the ABA3-RNAi constructs two exons of ptaba3 (LG_VII 7136650-7136823 (174 bp) and LG_VII 7137117-7137323 (207 bp)) were combined in a fusion PCR using P. trichocarpa DNA as template. For AO-RNAi constructs a 410 bp fragment of the ptao2 sequence was amplified from P. trichocarpa genomic DNA (Primer: Table S1). In each case the resulting fragments were used as sense and antisense fragments and exchanged the gusA fragments in the existing RNAi construct from [36] for creating a ptaba3 or ptao2 RNAi cassette, respectively. After generation of the chimeric gene cassettes cloning into the binary vector pBIN19 was achieved via HindIII restriction sites [38] .

A 35S-promoter::gfp::35S-terminator construct was generated by fusion of the enhanced version of the cauliflower 35S-promoter and the coding sequence of the enhanced green fluorescence protein (gfp). NcoI and BamHI sites were inserted into the 5’- and 3’-regions of the full-length Arabidopsis aba3 cDNA by PCR (Primer: Table S1) to allow ligation of aba3 into the 35S-promoter::gfp::35S-terminator construct after partial digestion with NcoI and complete digestion with BamHI. Subsequently, the 35S-promotor::gfp-aba3::35S-terminator construct was partially digested with BamHI and completely digested with HindIII, which enabled ligation into in the pBIN19 binary vector [38] .

Stable transformations were performed with the P. tremula × P. alba hybrid P. × canescens (no. 7171-B4, French National Institute for Agricultural research (INRA), Paris, France). Rhizobium tumefaciens (former known as Agrobacterium tumefaciens, strain C58C1/pMP90) was used for transformation of at least 150 internodes. Regeneration from independent callus lines followed the protocol of [39] . Correct transformation of selected plant lines was validated by PCR using specific primer (Table S2). The entire regeneration and selection procedure of transgenic poplar plants lasted about a year, in which in vitro plants were cultivated on modified Murashige and Skoog Medium [40] with half strength Macro-nutrients and 20 g∙L⁻¹ sucrose (M1/2-medium). Hydroponic culture was grown as described earlier in [41] . For ABA3-RNAi plants reduced or absent AO and XDH activities were chosen as selection criteria, whereas reduced or absent AO activity with simultaneous maintenance of XDH activity were chosen as criterion for AO-RNAi plants. ABA3 overexpressing transformants (GFP-ABA3) were identified by distinct cytoplasmic fluorescence and immune-detection of GFP-ABA3 proteins. PtAO2::GUS and ABA3::GUS plants were selected based on specific GUS activities. For each experimental set three independent transgenic lines of the AO-RNAi, ABA3-RNAi, PtAO2::GUS, ABA3::GUS and the GFP-ABA3 transformants were used.

Salt stress experiments were executed using in vitro plants of PtAO2::GUS, PtABA3::GUS, AO-RNAi, ABA3-RNAi, GFP-ABA3 and WT plants. Five plants per pot were grown under sterile conditions on 72 mL M1/2 medium in the absence of NaCl. After six weeks 3 mL sterile NaCl solution were added to the medium to final concentrations of 50, 75, 100 and 150 mM. Instead of NaCl water was added to the control plants. After different time points (1, 5, 16, 24, 40 and 120 h) whole plants were harvested and roots were removed. Shoots were frozen in liquid nitrogen for further analyses. Total protein extracts were prepared from one individual of each plant line and used exemplarily for AO and XDH activity staining. Two to five plants were used for ABA measurements in replicates of three to nine.

Drought stress experiments were performed with detached leaves of hydroponically grown plants (PtABA3:: GUS, AO-RNAi, ABA3-RNAi, GFP-ABA3) or alternatively with detached leaves of in vitro plants (PtAO2:: GUS). Middle-aged leaves of poplar (height approx. 80 cm) were clipped off and dried in a glass Petri dish (x 15 cm) to a remaining fresh weight of 80% in approx. 8 to 10 h and to 60% in approx. 22 to 24 h. The complete number of leaves from in vitro poplar (6 weeks old) was handled in x 6 cm plastic Petri dishes. The drying process was observed hourly at the beginning and shortly before achieving the correct drought stress intensity. Leaf material was harvested and veins removed from hydroponically cultivated poplar after reaching the appropriate drying stage. Total leaves of in vitro grown poplar were harvested after having reached the desired drought stage and leaf samples were frozen in liquid nitrogen. As for salt stress experiments one sample from each transgenic line was used representatively to visualize AO and XDH activities. Two to five plants were used for ABA measurements in replicates of three to six.

150 mg plant material was ground in liquid nitrogen, sonified in 200 μL extraction buffer (100 mM potassium phosphate, pH 7.5, 5 mM DTT, 2 mM EDTA) and centrifuged at 4˚C. For GUS measurements 150 mg plant material were homogenized in 500 μL of 50 mM potassium dihydrogen orthophosphate and 1 mM DTT using a Precellys tissue homogenizer (6500 rpm, two times for 15 sec including 15 sec pause; Precellys 24, Bertin Technologies, Montigny-le-Bretonneux, France). After disruption plant material was sonicated and centrifuged two times (4˚C, 32,000 g, 20 min). For AO and XDH activity staining the supernatant was mixed with 15 mg polyvinylpyrrolidone (Sigma Aldrich, Steinheim, Germany) and cleared by centrifugation. Protein concentrations were determined according to [42] with bovine serum albumin as a standard.

AO and XDH activities were determined by in gel activity staining as described in [43] and [44] . Native polyacrylamide gel electrophoresis (7.5%) was used for separation of 250 μg protein per sample. Gel staining was applied overnight in 30 mL staining solution containing 100 mM potassium phosphate, 18 mg methylthiazolyldiphenyl-tetrazolium bromide and 0.25 μM phenazinemethosulphate. Indole-3-carboxaldehyde was used for AO activity detection and a mixture of 15 mg xanthine and 15 mg hypoxanthine for detection of XDH activity (substrates purchased by Sigma Aldrich).

GUS activities of AO::GUS and ABA3::GUS transformants were assayed using 50 μL protein extract in 450 μL GUS reaction buffer (50 mM potassium phosphate buffer, pH 7.0, 10 mM β-mercaptoethanol, 10 mM Na₂-EDTA, 0.1% N-lauroylsarcosine, 0.1% triton X-100 and 1 mM 4-methylumbelliferyl-D-glucuronide hydrate) at 37˚C. The reaction was stopped after 10, 20 and 30 min in 0.2 M sodium carbonate and the reaction product 4-methylumbelliferone (4-MU) was measured using a luminescence spectrometer LS30 (Perkin-Elmer, Massachusetts, USA) or the Tristar Multimode Plate Reader (Berthold Technologies, Wildbad, Germany) detecting excitation at 365 nm and emission at 455 nm. GUS activity was calculated using a 4-MU standard.

ABA concentrations were determined using the Phytodetek ABA enzyme immunoassay test kit (Agdia Inc., Elkhart, USA) according to the manufacturer’s protocol. Aqueous isolation allowed the simultaneous measurement of physiologically active ABA as well as the inactive form ABA-GE (abscisic acid-β-D-glucopyranosyles- ter).

For immunodetection 50 μg protein were separated by 7.5% SDS polyacrylamide gel electrophoresis (SDS- PAGE) prior to transfer onto nitrocellulose membrane (Hybond-P, GE Healthcare, Fairfield, USA) by electro-blotting. As protein marker PageRuler™ Prestained Protein Ladder (Fermentas, St. Leon-Rot, Germany) was used. The primary anti GFP-antibody (Anti-GFP rabbit polyclonal serum, Invitrogen, Darmstadt, Germany) was diluted 1:750 prior to use. The detection was finalized with an alkaline phosphatase-conjugated anti-rabbit antibody (Sigma Aldrich), diluted 1:7500 prior to use, and BCIP/NBT staining reagents (Promega, Fitchburg, USA).

Fluorescence of the lower leaf epidermis was visualized with the confocal laser scanning microscope (cLSM) 510 META (Release Version 4.2 SP1) connected to an Axiovert 200M (Carl Zeiss, Oberkochen, Germany). The specimens were examined either using the C-Apochromat 40x/1.2 water-immersion objective for detailed pictures with excitation using the argon laser (488 nm line for GFP and chlorophyll auto-fluorescence). The emitted light passed the primary beam splitting mirrors UV/488/543/633 and was separated by a secondary beam splitter at 545 nm. Fluorescence was detected with filter sets as follows: BP 505-530 for GFP and LP 650 for red auto-fluorescence of chlorophyll. When appropriate, the bright field images of samples were imaged with the transmitted light photomultiplier, and the Lambda-mode was used to examine the spectral signature of fluorophores. All images were processed with LSM Image Browser Release 4.2 (Carl Zeiss).

Arabidopsis thaliana AO genes comprise a multigene family of four isoforms: aao1 (AT5G20960), aao2 (AT3G43600), aao3 (AT2G27150) and aao4 (AT1G04580) [45] . P. trichocarpa shotgun sequence assemblies were annotated, preferentially based on their homologies to A. thaliana genes [27] , which established the basis for identification of P. trichocarpa genes. NCBI (National Center for Biotechnology Information, www.ncbi.nlm.nih.gov) recorded four Poplar AO isoforms: ptao1 (GI: 224104955), ptao2 (GI: 224131694), ptao3 (GI: 224145733) and ptao4 (GI: 224145729). ptao1 presents most notable similarities to aao2 and aao4, and ptao2 presents similarities to aao1 and aao3. ptao3 and ptao4 represent truncated sequences (Supporting Information Figure S1) and are thus likely to be derived from incomplete gene duplication events. Therefore, two AO isoforms were expected for P. trichocarpa [18] with PtAO2 being the most likely candidate for representing the ABA-synthesizing protein isoform due its similarity to AAO3.

In Arabidopsis, the Moco sulphurating enzyme ABA3 (At1g16540) regulates the activation of AO and XDH by inserting a sulphur atom into the Moco to generate its sulphurized form, which is required by AO and XDH to accomplish their tasks in drought and salt stress reactions [13] [30] . ptaba3 is described in the P. trichocarpa database (www.phytozome.net) as POPTR_0007s08330/PtABA3 (“similar to ABA DEFICIENT 3”) representing an ‘ortholog of Arabidopsis At1g16540’. The comparison of the aba3 and the ptaba3 domain structures clearly demonstrates this orthology (Supporting Information Figure S2).

Generation of the different transgenic poplar lines was accomplished using a minimum of 150 internodes for each construct, from which 7 to 44 independent callus lines were established. According to the used constructs, 33 to 128 plants were regenerated (Supporting Information Figure S3). Three plant lines from independent original calli were chosen for further analyses. Plant lines for promoter-GUS transformants (PtABA3::GUS and PtAO2::GUS) were selected with regard to similar expression levels of gusA: no. 8, 27 and 50 for PtABA3::GUS and no. 12, 19 and 20 for PtAO2::GUS (Supporting Information Figure S4).

AO- and ABA3-RNAi lines were chosen after PCR verification of the insertion of the RNAi cassette and based on AO and XDH activity staining results. Among AO-RNAi transformants, lines with residual AO activity but unaffected XDH activity were selected. ABA3-RNAi plants presenting neither AO nor XDH activity were used (Figure 1 & Figure 2).

Expression of GFP-ABA3 fusion proteins in transgenic lines was detected by immunoblot analyses using a GFP-specific antibody. GFP-ABA3 expressing poplars no. 6, 14b and 28 presented a distinct band in the range of 117 kDa representing the calculated size of the fusion protein (92 kDa ABA3 plus 27 kDa GFP, Figure 3(a)). Confocal laser scanning microscopy of the fusion protein in poplar leaves verified the transformation. Moreover, GFP-specific fluorescence was completely precluded from organelles and gave a clear cytoplasmic localization (Figures 3(b)-(d)). The lambda mode verified the GFP-specific emission at 515 nm (Figure 3(e)).

The growing behavior of all transgenic poplar lines was studied in in vitro and hydroponic cultures and compared to the appearance of WT plants. Morphological characteristics of the plants (size of the plant, leaf size and shape, stem diameter, guard cell size and guard cell numbers per leaf area) did not differ between the different transgenic plants and the WT. At non-stressed conditions, the ABA content of ABA3-RNAi, AO-RNAi and GFP-ABA3 poplar transformants ranged around basal WT level (Figure 4 & Figure 5).

Promoter-GUS fusion constructs were used to investigate the transcriptional regulation of ptaba3 and ptao2. Therefore, changes in GUS activity of PtAO2::GUS and PtABA3::GUS transformants were analyzed under salt and drought stress conditions. In general, GUS levels of PtAO2::GUS transformants appeared to be three times higher in comparison to ABA3::GUS plants.

When subjected to drought stress conditions at which plants were allowed to lose 20%, 40% and 60% of fresh weight, GUS activities of ABA3::GUS plants no. 8, 27 and 50 remained stable. For the same plants, 50, 100 and 150 mM NaCl stress for 1, 5, 24 and 120 h likewise did not alter GUS activities and thus ptaba3 transcription levels (Supporting Information Figure S5). Similar, PtAO2::GUS transformants no. 12, 19 and 20 did not show considerable changes in GUS activity, neither when dehydrated to 80 and 60% of their respective initial fresh weight, nor when exposed to 75 or 150 mM NaCl for 16 and 40 h (Supporting Information Figure S6).

Leaves of WT and GFP-ABA3 overexpressing plants indeed had the highest AO and XDH activities, while AO and XDH or only AO activities, respectively, were significantly diminished or not detectable in leaves of ABA3-RNAi and AO-RNAi plants (non-treated material in Figure 1 & Figure 2).

High salinity for different time intervals in different concentrations generally did not result in phenotypical changes in any of the investigated plant lines. Instead, plants appeared healthy and did not exhibit clear visible

symptoms such as stress-induced senescence. Moreover, no salinity-induced change in AO or XDH enzyme activity was observed in any of the transgenic lines, independent from the salt concentration or the duration of treatment (Figure 1). However, 16 h after application of NaCl all plants presented a significant increase in ABA levels at both concentrations 75 and 150 mM. Interestingly, this salt stress-induced increase of ABA levels was found to be only transient since ABA levels dropped near to control levels when NaCl exposure times were extended to 40 h. An important observation in these experiments concerns the total ABA levels in WT and GFP-ABA3 transformants relative to the ABA3- and AO-RNAi lines: while WT and GFP-ABA3 accumulated up to 6000 ng ABA g⁻¹ FW⁻¹ after 16 h of salt stress exposure, plants of both RNAi lines revealed a markedly lower increase of only up to 2000 ng ABA g⁻¹ FW⁻¹ (Figure 4). Although total ABA levels have dropped after 40 h of salt stress in all plants, the differences between WT and GFP-ABA3 overexpressing plants (about 1000 ng ABA g⁻¹ FW⁻¹) relative to ABA3- and AO-RNAi plants (about 500 ng ABA g⁻¹ FW⁻¹) were still statistically significant, indicating a crucial role of PtAO2 and PtABA3 for salt stress-induced ABA synthesis in poplar.

When subjecting WT, GFP-ABA3, ABA3-RNAi and AO-RNAi leaves to drought stress by allowing them to dehydrate to 80% and 60% of their initial fresh weight, all treated leaves clearly appeared flaccid. However, no phenotypical differences were evident between the various transgenics and controls, and also AO and XDH activities remained constant along the drought stress procedure in all investigated poplar leaves (Figure 2). Yet, neither 20% nor 40% drought stress-induced loss of fresh weight resulted in markedly altered AO and/or XDH activities relative to untreated leaves. In contrast, ABA levels were strongly enhanced with an overall 5-fold increase after 20% loss of fresh weight and a 15- to 20-fold increase after 40% loss of fresh weight in leaves of all plant lines (Figure 5). Unlike salt stress treatment however, drought stress did not expose significantly different response efficiencies between the individual transgenic and control plants, suggesting that the ABA-mediated drought stress response is relatively independent from altered AO and/or XDH activities.