Sucrose synthase determines carbon allocation in developing wood and alters carbon ﬂow at the whole tree level in aspen

We studied sucrose synthase (SUS) to clarify its role in biomass formation and C metabolism at the whole tree level in hybrid aspen ( Populus tremula tremuloides ). To this we analysed source leaves, phloem, developing wood, and roots of SUSRNAi trees using a combina-tion of metabolite proﬁling, 13 CO 2 pulse labelling experiments, and long-term ﬁeld experiments. (cid:1) The glasshouse grown SUSRNAi trees exhibited a mild stem phenotype together with a reduction in wood total C. The 13 CO 2 pulse labelling experiments showed an alteration in the C ﬂow in all the analysed tissues, indicating that SUS affects C metabolism at the whole tree level. This was conﬁrmed when the SUSRNAi trees were grown in the ﬁeld over a 5-yr period; their stem height, diameter and biomass were substantially reduced. (cid:1) These results establish that SUS inﬂuences C allocation to developing wood, and that it affects C metabolism at the whole tree level.


Introduction
Wood biomass is a valuable raw material for wood-based industries including pulp, paper, sawn timber, and biofuel production (Plomion et al., 2001;Novaes et al., 2009). Wood formation depends on carbon (C) allocation to developing wood. In several tree species, C is distributed predominantly as sucrose (Rennie & Turgeon, 2009). Phloem loading in source leaves of the model tree aspen (Populus tremula L.) occurs via a symplasmic route . Phloem unloading mechanism into developing wood is less well understood but is thought to include both symplasmic and apoplasmic steps (Van Bel, 1990;Chaffey & Barlow, 2001;Mahboubi et al., 2013). Despite our increasing understanding of C transport routes in trees, the mechanism of C allocation and biomass accumulation in the different tissues remain poorly understood.
Biomass formation depends partly on the capacity of the wood tissue and other C competitor sink tissues such as shoot and root meristems to metabolize and incorporate C (Yu et al., 2015). Primary metabolic reactions are especially important in this process because they are responsible for sucrose catabolism, energy production, and the synthesis of the precursors for cell wall polymers. Studies on biomass accumulation in trees commonly focus on hormonal factors and C allocation to cell wall polymers (Dubouzet et al., 2013;Busov, 2018), while primary metabolism has received less attention (Mahboubi & Niittyl€ a, 2018). The C availability for biomass production by sink tissues depends on the photosynthetic and sucrose export capacity of source leaves (Yu et al., 2015); in recognition of this, efforts have been made to increase wood biomass by improving tree photosynthesis (Dubouzet et al., 2013;Busov, 2018). However, C acquisition, allocation, and metabolism in the different tissues of a tree depend on the relationship between sink and source tissues, which should therefore be considered jointly (Sonnewald & Fernie, 2018).
The concept of sink strength refers to a sink tissue capacity to compete for photoassimilates. This is determined by the capacity of the tissue to import C from the leaves and to synthesize macromolecules (Yu et al., 2015). The capacity for C import depends in part on enzymes that degrade sucrose. Two types of enzymes catalyse sucrose degradation in sink tissues: sucrose synthase (SUS) and invertase. SUS has been associated with C allocation, increased biomass, and sink strength (Stein & Granot, 2019). It is known to be the main sink strength determinant in potato tubers (Zrenner et al., 1995), and to control C import in young tomato fruits (D'Aoust et al., 1999). In tobacco plants, SUS overexpression increased height and biomass, indicating that it can also control C allocation in this species (Coleman et al., 2006). Conversely, evidence from rice and corn suggests that acid invertases are the main grain sink strength determinants in these species (Cheng et al., 1996;Li et al., 2013;Morey et al., 2018). The relative impacts of SUS and invertase on C allocation thus appear to depend on the tissue, species, developmental stage, and season. Consequently, it is impossible to predict the contributions of these two enzymes to C metabolism a priori.
We previously observed decreased wood density in aspen trees with reduced SUS levels, which was accompanied by a decrease in cellulose, hemicellulose, and lignin content per unit volume of wood (Gerber et al., 2014). The decrease in total polymer biomass per unit volume suggested reduced C incorporation into wood cell walls. Moreover, expression of a cotton SUS in aspen trees increased cell wall thickness and wood density (Coleman et al., 2009), while aspen SUS expression in tobacco plants increased xylem cell wall thickness (Wei et al., 2015). These observations support a central role for SUS in C allocation to wood, but they do not reveal the role of SUS in C allocation and metabolism, and biomass accumulation at the whole tree level. This prompted us to study the role of SUS at the whole tree level under glasshouse and field conditions using transgenic SUSRNAi lines. The assessment of the C metabolism was performed by combining total metabolite pool measurements and the tracking of a carbon-13 ( 13 C) flow among the central metabolites in leaves, phloem, developing wood and roots after 13 CO 2 supply.

Plant material
Three 35S:SUSRNAi transgenic lines (Gerber et al., 2014) along with wild-type (WT) T89 hybrid aspens (Populus tremula 9 tremuloides) were grown in a glasshouse in commercial soil (Yrkes Plantjord; SW Horto, Hammenh€ og, Sweden) under an 18 h : 6 h, 20°C : 15°C, light : dark photoperiod. Trees were fertilized using 150 ml 1% Rika-S (nitrogen (N)-phosphorus (P)-potassium (K), 7 : 1 : 5; SW Horto) once a week, and were harvested when they were 2 months old. Wood samples were collected from 15 to 45 cm above the soil. In the field experiment, the same genotypes were grown for 5 yr in a field setting in V axtorp, Laholm community, Sweden (56.42°N, 13.07°E). Trees were planted with a 3 m spacing in 2011. Four trees per each transgenic line and 32 WT trees were randomly distributed over the field area. The field was harrowed twice a year during the first 2 yr following planting and grass was mowed twice a year during subsequent years. Height and diameter (at the stem base) were measured at the end of each growing season and before harvest in the summer 2016.

Sample preparation
All samples were frozen on liquid nitrogen immediately after collection and stored at À80°C until preparation. Developing wood from both glasshouse and field trees was obtained by scrapping the surface of frozen debarked stems with a scalpel while maintaining the low temperature with liquid nitrogen. Phloem samples were obtained by scrapping the inner side of the barks. Fully expanded leaves number 14-16 (counting from the top) were sampled. The tips of the roots were cut, cleaned and immediately frozen. All the obtained frozen tissues were ground with a mortar and pestle. The ground material was kept at À80°C until use.
Sucrose synthase activity determination in glasshouse grown trees The protocol was based on that of Gerber et al. (2014). Extracted samples (extraction buffer: 100 mM Tris-HCl pH 7.5, 2 mM EDTA, 5 mM DTT, a scoop of PVPP; Roche proteinase inhibitor, Roche, Basel, Switzerland) were incubated with 45 µl of reaction mix (100 mM Tris pH 7.5, 100 mM sucrose, 4 mM UDP (uridine diphosphate)) for 30 min at 25°C. A control for each sample was prepared simultaneously using 45 µl of reaction mix without UDP. The reaction product, UDP-glucose, was determined by incubating 25 µl (developing wood) or 100 µl (roots) of the reaction mix with 25 µl of the determination buffer (100 mM Tris pH 7.5, 2 mM NAD + , 0.02 u UGDH) at 340 nm. The absorbances were interpolated into a UDP-glucose standard curve. Total protein contents were determined by using the DC Protein Assay (Bio-Rad, Hercules, CA, USA).

Quantitative polymerase chain reaction (qPCR) in glasshouse and field grown trees
Total mRNA was extracted with Trizol ® (Gibco, Gaithersburg, MD, USA) according to the manufacturer´s specifications. The cDNA was prepared using the MessageAmp Premier RNA Amplification Kit (Ambion, Austin, TX, USA). The reference gene was Ubiquitin, which was chosen based on its transcript stability. Quantitative polymerase chain reactions (qPCRs) were performed using SYBR ® Green Master Mix (Bio-Rad) in a CFX96 Real Time System (Bio-Rad) with the following programme: 95°C for 5 min, then 50 cycles of 95°C for 30 s, 60°C for 15 s, and 72°C for 30 s. The employed primers are listed in Supporting Information Table S1. Primers were designed with PRIMER3 (Untergasser et al., 2012). Ratios were calculated using the equation proposed by Pfaffl (2001): Ratio = (E target ) DCp target(MEAN control À MEAN sample) ((E reference ) DCp reference(MEAN control À MEAN sample) ) -1 .

Dry weight and total C phenotype
Tissues from 2-month-old or 5-yr-old trees were dried for 1 wk at 80°C to measure the dry weight. Total C from tissues of glasshouse-grown trees was measured by elemental analyser-isotope ratio mass spectrometry (EA-IRMS; Thermo Fisher Scientific, Waltham, MA, USA). Furthermore, 5 mg of oven-dried sample were employed for each measurement. Measurements were performed with an elemental analyser (Flash EA 2000; Thermo Fisher Scientific) connected to a continuous flow isotope ratio mass spectrometer (DeltaV) (Thermo Fisher Scientific). Samples were analysed together with cyclohexanone, nicotinamide, and sucrose standards, which were standardized against standard reference materials. Data were corrected for drift and size.

Photosynthetic parameters in glasshouse grown trees
The leaf nitrogen balance index (NBI), chlorophyll index (CHL), flavonol index (FLV), and anthocyanin index (ANTH) were measured with a Force A Dualex Scientific+ TM device on 10 independent plants at midday. Chlorophyll a fluorescence (ChlF) measurements were performed on 10 leaves per plant over five consecutive days from 09:00 h to 16:00 h using a FluorPen FP100max portable fluorometer (Photon Systems Instruments, Brno, Czech Republic). The kinetics of chlorophyll fluorescence induction were estimated based on the polyphasic rise of the ChlF transient (OJIP) curve using the calculations proposed by Stirbet & Govindjee (2011). The second employed protocol applied with the FluoPen fluorometer was the Light Curve protocol (protocol LC3) using the calculations proposed by Maxwell & Johnson (2000).

C labelling experiment in glasshouse grown trees
A Percival Scientific cabinet with a CO 2 system (model: PGC-7L2) was employed to supply 13 CO 2 (> 99 atom%, > 99.98%; BOC UK & Ireland, Bluebell, Dublin, Ireland) to the trees. When the experiment started, 12 CO 2 was removed from the air in the chamber using the chamber's in-built CO 2 scrubbing system, and 13 CO 2 was quickly injected up to 400 ppm. The 13 CO 2 was re-injected whenever its levels fell below 350 ppm. The CO 2 levels were measured with a WMA-4 CO 2 analyser (www.ppsyste ms.com). Groups of four plants were injected with 13 CO 2 on each experimental run, and were sampled after 4, 6 and 8 h of treatment. Unlabelled controls for each genotype were injected with 12 CO 2 in the same chamber, and sampled at 6, 8, 10 and 14 h after the start of the light period, covering the times of the day in which 13 C labelling was performed on the experimental plants. The generated labelled material was analysed by gas chromatography mass spectrometry (GC-MS) and/or two-dimensional nuclear magnetic resonance (2D-NMR) as described later.
Only metabolites with slow labelling rates (data obtained from Szecowka et al., 2013) were examined to minimize the influence of potential 12 C incorporation during sampling.

GC-MS measurements
All the labelled samples and controls were analysed by GC-MS. Frozen samples (10 mg) were extracted with 1 ml of a chloroform-methanol-water (20 : 60 : 20, v/v/v) solution containing two stable isotope reference compounds (7 ng ll À1 [ 13 C 3 ]-myristic acid and [ 2 H 7 ]-cholesterol). N-Methyl-N-(trimethylsilyl)trifluoroacetamide (MSTFA) was used as a reagent for silylation derivatization done according to Lind en et al. (2016). The GC-MS protocol was based on that of Gullberg et al. (2004). Thus, 1 ll of each derivatized sample was injected in splitless mode per run. The GC-MS system used electron impact (EI) ionization and consisted of a CTC PAL systems autosampler (CTC Analytics AG, Zwingen, Switzerland), an Agilent technologies 7890A GC system (Agilent Technologies, Atlanta, GA, USA), and a Pegasus HT GC high-throughput TOF-MS instrument (LECO Corp., St Joseph, MI, USA). The analysed fragments for the labelling calculations were selected based on Lind en et al. (2016) and Beshir et al. (2017). Corrections for the 13 C natural abundance and the presence of tetramethylsilane (TMS) groups, and computations of 13 C enrichment percentages for each identified metabolite were performed as described by Mahboubi et al. (2015) and Lind en et al. (2016).

2D-NMR analysis
The protocol was based on that of Hedenstr€ om et al., 2009. Developing wood samples belonging to SUSRNAi-1 and WT with no label (time 0) and labelled for 8 h (time 8) were analysed by 2D-NMR. Before analyses, soluble sugars and starch were removed from the samples as described in Smith & Zeeman (2006) to avoid signal interferences in the cell wall NMR measurements. Next, 20 mg of ground tissues were added to 600 ll of deuterated dimethyl sulfoxide (DMSO-d 6 ). The 2D 1 H-13 C heteronuclear single quantum coherence (HSQC) experiments were performed on a Bruker 600 MHz Avance III HD spectrometer equipped with a 5 mm BBO cryoprobe with z-gradients (Bruker Biospin, Rheinstetten, Germany). All measurements were performed at 25°C. In the pulse sequence, adiabatic 13 C-inversion and 13 C-refocusing pulses were used (Bruker pulse sequence hsqcetgpsisp.2). The sweep width was 10 ppm for the 1 H dimension and 165 ppm for the 13 C dimension. Processing and peak integration were performed using Topspin 3.6 (Bruker Biospin). The peak assignments were based on that of Kim et al. (2008).

Starch content measurement in glasshouse grown trees
Briefly, 10 mg of the tissues were extracted with 250 µl of 80% v/v ethanol and gelatinized according to Smith & Zeeman (2006). Next, 40 ll of the resuspension was treated with 6 U of alpha-amyloglucosidase and 0.5 U of alpha-amylase prepared in 50 mM acetate buffer. Negative controls were processed in the same way using 50 mM acetate buffer without the enzymes. The tubes were incubated at 37°C overnight. The treatment with the enzymes was done twice for phloem, developing wood and mature wood, and three times in the leaf samples to assure a complete starch degradation. Then, 50 ll of the samples or controls were then incubated with 75 µl of the determination buffer (HEPES pH 7.5 50 mM, NADP 0.4 mM, ATP 2 mM, G6P DHG 2 U ml À1 , HXK 2 U ml À1 , PGI 2 U ml À1 ). The absorbances were interpolated into a glucose standard curve.

Neutral invertase activity determination in glasshouse grown trees
The protocol was based on that of Rende et al. (2016). Briefly, 1 ll of each sample extract (extraction buffer: 50 mM HEPES pH 7, 10 mM MgCl 2 , 1 mM EDTA, 0.25 mM DTT, 1% Triton X-100, 20% glycerol; Roche proteinase inhibitor) was incubated with 75 µl of the determination buffer (50 mM HEPES pH 7, 0.4 mM NADP + , 2 mM ATP, 2 U ml À1 glucose-6-phosphate dehydrogenase, 2 U ml À1 hexokinase, 2 U ml À1 phosphoglucose isomerase). Then, 20 µl of 100 mM sucrose was added to each mixture to initiate the reactions. Controls for each sample were prepared in parallel replacing the 100 mM sucrose with HEPES pH 7. Absorbances measured at 340 nm were interpolated into a glucose standard curve.

Statistical analyses
All experiments were analysed statistically by analysis of variance (ANOVA), using Duncan's multiple range test (MRT) for multiple comparison. All analyses were performed with the INFOSTAT software (v.2008; Infostat Universidad Nacional de C ordoba, Argentina). A significance threshold of P < 0.05 was applied. The three-dimensional (3D) principal component analysis (PCA) scatter plot was created with BIOVINCI software (BioTuring Inc., San Diego, CA, USA).

SUSRNAi reduces SUS activity in developing wood and roots
The Populus genome contains seven SUS genes, named SUS1-SUS7 (Zhang et al., 2011). The SUSRNAi construct targets SUS1 and SUS2, which are the most strongly expressed isoforms in developing wood (Zhang et al., 2011;Gerber et al., 2014). SUS1 and SUS2 transcripts are also found in leaves and roots (Zhang et al., 2011), suggesting that the RNAi construct may also reduce the total SUS activity in these tissues. To clarify the SUSRNAi effects at the whole tree level, we measured SUS activity in developing wood, roots, and source leaves. In accordance with our previous results, SUS activity was reduced to near background levels in developing wood of the SUSRNAi lines (Fig. 1a), while in roots its activity was reduced to 40% of the WT level (Fig. 1b). The more modest reduction in roots is probably partly due to the expression of other SUS isoforms (Zhang et al., 2011). The SUS activity in the source leaf extracts of WT trees was close to background levels. Therefore, to assess the SUSRNAi effects in source leaves, we quantified the mRNA levels of all the SUS isoforms using qPCR. Only SUS1 and SUS5 transcript levels in the three transgenic lines were reduced to c. 30% and 50%, respectively, of the WT levels (Fig. 1c). The reduction in SUS5 mRNA levels is not a direct SUSRNAi effect, because the percentage of similarity between the nucleotide coding regions of SUS1 and SUS5 is c. 58% (Zhang et al., 2011) and the SUSRNAi does not target SUS5.
Total C is reduced in developing wood of SUSRNAi lines We performed detailed growth phenotyping of 2-month-old glasshouse grown SUSRNAi trees. The total stem fresh weight of the glasshouse-grown transgenic lines did not differ significantly from that of the WT (Fig. 2a). However, the stem dry weight of the transgenics was slightly reduced, being statistically significant in SUSRNAi-3 (Fig. 2b). Moreover, the total C content in developing wood as measured by IRMS was significantly reduced in all the silenced lines (Fig. 2c), clearly indicating that the C allocation to this tissue was reduced. These observations are consistent with our previous results documenting reduced stem wood density and increased fibre wall shrinkage upon drying in the SUSRNAi lines (Gerber et al., 2014). The total leaf fresh and dry weight were unchanged in the transgenic lines, except in SUSRNAi-3 (Fig. 2d,e). The total root fresh and dry weight were unchanged in all lines (Fig. 2g,h), as was the total C content in leaves and roots (Fig. 2f,i). Therefore, the growth defect in the SUSRNAi lines is primarily associated with wood formation, and not with leaf and root growth.
Photosynthetic light-dependent reactions are not affected in the SUSRNAi source leaves Alterations in sugar metabolism can cause alterations in photosynthesis (Sheen, 1990). To determine whether the photosynthetic performance of the source leaves in transgenic genotypes contributed to the reduction in the developing wood C content, we assayed leaf pigments and the photosynthesis light-dependent reactions. CHL, ANTH, FLV, and NBI (i.e. the ratio of the chlorophyll and flavonol indexes) showed no significant differences (Table S2). The photosynthesis light-dependent reactions can be evaluated by measuring the ChlF and computing different parameters like those of OJIP and LC (light curve) that provide information on the condition of the photosystem II and the electron transport in the thylakoid membrane. OJIP shows how the photochemical efficiency varies under different light conditions (Baker, 2008). The OJIP test equations were therefore used to calculate parameters including the PI abs , the F v /F m ratio, and the specific fluxes for the reaction centres (Table S3) included F t , Q y , F 0 and F m (Table S4). No consistent differences were observed between WT and SUSRNAi lines, indicating that the photosynthesis is not affected by the light intensity in the transgenics. Based on these results, it can be concluded that SUSRNAi does not affect pigment levels or the photosynthetic performance in source leaves. Its effects on wood are thus not due to impaired photosynthesis light-dependent reactions but more likely due to changes at the leaf metabolic level, leaf-to-wood C transport, and/or developing wood metabolism.
13 C tracking in SUSRNAi lines revealed altered sucrose turnover and C flow at the whole tree level To investigate C fluxes between organs and at the whole tree level, we developed a 13 CO 2 labelling system based on a previously reported design (Mahboubi et al., 2013(Mahboubi et al., , 2015 and used it to characterize the C fluxes in SUSRNAi-1 and SUSRNAi-3 during a 13 CO 2 pulse. Mahboubi et al. (2013) showed that 13 C can be detected in developing wood after a 4-h exposure to 13 CO 2 . We therefore exposed 6-wk-old SUSRNAi and WT trees to a 13 CO 2 pulse and collected samples 4, 6, and 8 h after the pulse start. The 13 C metabolic fate in leaves, roots, and developing wood was monitored using GC-MS. In parallel, we measured the total metabolite pools. While the total pool measurements provided a picture of the metabolic status of the tissues at fixed time points, the 13 C labelling experiments yielded insights into the movement of the 13 C among metabolites. In leaves, during active photosynthesis sucrose is derived from the Calvin cycle (Fig. 3a). Sucrose showed decreased 13 C labelling at time 6 in both lines and at time 4 in SUSRNAi-3. Fructose and glucose did not show significant differences. The labelling of Krebs cycle metabolites was not altered, except for fumarate, which was increased at time 8. The 13 C-aspartate, 13 C-serine, 13 C-phenylalanine, and 13 C-glutamate were decreased in both lines at time 6. The 13 C-threonine was decreased at time 4 in both lines, while 13 C-glycerate was decreased in SUSRNAi-1 at time 6. Total pools were unchanged, except for succinate and ketoglutarate in SUSRNAi-1 (Table S5). Thus, the C flow changes in SUSRNAi leaves were primarily evident in the analysed amino acids.
The total sucrose levels in the phloem of the SUSRNAi lines were unchanged from those in the WT, but the 13 C

Research
New Phytologist incorporation rate into the phloem sucrose pool was reduced in both SUSRNAi lines at time 6 ( Fig. 3b), suggesting that SUSRNAi reduced the phloem sucrose loading rates.
Carbon in roots is derived from sucrose imported via the phloem. Root 13 C-sucrose was only reduced in SUSRNAi-1 at time 8 (Fig. 3c), while 13 C-fructose and 13 C-glucose did not show changes (Supporting Information Fig. S1). The same was observed for the Krebs cycle metabolites 13 C-citrate, 13 C-malate and 13 C-succinate. However, 13 C-ketoglutarate level was reduced in both lines at time 4 and time 6, and in SUSRNAi-1 at time 8. The 13 C-glutamate, 13 C-aspartate, 13 C-phenylalanine and 13 Cserine did not show differences, but both lines showed decreased levels of 13 C-valine at time 8 and of 13 C-shikimate at time 6. Total pools for these metabolites did not differ significantly from WT ( Fig. 3c; Table S6). Therefore, SUSRNAi changed the 13 C flow through some individual metabolite pools in the roots, but there was no overall change.
Sucrose import fuels the central metabolism also in developing wood (Fig. 4a). The 13 C-sucrose enrichment in developing wood of SUSRNAi lines was decreased in comparison to WT in both lines at time 6 and in SUSRNAi-1 at time 8. The 13 C-glucose and 13 C-fructose, which can be produced from sucrose cleavage, were reduced in SUSRNAi-1 at time 8 and in both lines at time 6 and 8, respectively. Apart from 13 C-malate, the Krebs cycle metabolites showed alterations as well: 13 C-citrate was reduced in SUSRNAi-1 at time 6; 13 C-ketoglutarate was reduced in both lines at time 6 and 8; 13 C-succinate was reduced in SUSRNAi-3 at times 4, 6 and 8; and 13 C-fumarate was reduced in both lines at time 8. The amino acids 13 C-serine and 13 C-phenylalanine showed decreased enrichment in both lines at times 6 and 8. The 13 C-glycine and 13 C-glycerate, which derive from 13 C-serine, were reduced in both lines at time 6 and reduced in SUSRNAi-3 at time 8, respectively. The 13 C-aspartate was reduced in SUSRNAi-1 at time 6 and 8. The 13 C-beta-alanine was reduced in SUSRNAi-3 at time 4, while 13 C-asparagine was reduced in SUSRNAi-3 at time 4 and in both lines at time 6. The 13 C-glutamate was reduced in SUSRNAi-1 at time 6. The 13 C-GABA and 13 C-isoleucine did not show differences. Thus, there is an overall 13 C distribution decrease in the central metabolites of developing wood in the transgenic trees, which is linked to the reduction in 13 C-sucrose. Moreover, a PCA of these data clearly showed that the overall 13 C labelling profile of the metabolism in transgenic lines differs from the WT trees (Fig. 4b).
Total sucrose levels in the SUSRNAi developing wood did not differ significantly from those in WT trees between 6 and 14 h after the start of the light period (Fig. S2). Total pools of glucose and fructose in developing wood were also not significantly changed. However, all the three sugar levels exhibited increasing Fig. 2 Phenotype of 2-month-old wild-type (WT) and SUSRNAi hybrid aspen (Populus tremula 9 tremuloides) trees grown in the glasshouse. Top panels: total stem fresh weight (a), total stem dry weight (b), and C content of developing wood (c). Middle panels: total leaf fresh weight (d), dry weight (e), and C content (f). Bottom panels: root fresh weight (g), dry weight (h) and C content (i). Mass fraction of C was measured by IRMS. Bars indicate the average and error bars indicate the standard error. ANOVA, Duncan test, n = 4. Different letters indicate significant differences (P < 0.05).
Ó 2020 The Authors New Phytologist Ó 2020 New Phytologist Foundation New Phytologist (2021) 229: 186-198 www.newphytologist.com trends at some time points. This is consistent with the results of Gerber et al. (2014), who observed a similar effect in the total pools of these sugars. The differences in significance between these two works may be due to the use of different analytical techniques and statistical methods. Most of the 14 other measured metabolites including aspartate, glycine and phenylalanine did not show differences between the SUSRNAi lines and the WT trees (Fig. S2). However, at one time point SUSRNAi-3 contained increased levels of asparagine (8 h), glutamate (10 h) and serine (10 h), while for SUSRNAi-1 increases were observed in beta-alanine (14 h) and isoleucine (10 h). The latter line also showed decreased levels in citrate (8 h), fumarate (6 h) and malate (6 h). Both lines showed decreased levels in ketoglutarate (6 h). Succinate and glycerate had decreased levels in SUSRNAi-1 (6 h) and increased levels in SUSRNAi-3 (10 h). These findings show that the transgenic trees had alterations in the steady-state levels of some metabolites in developing wood, which are indicative of rearranged central metabolism in response to the reduced SUS activity.
The decreased 13 C-sucrose and the overall decreased 13 C labelling of central metabolites in developing wood strongly suggest that the total C decrease in this tissue in the transgenic lines (Fig. 2c) is due to the decreased sucrose incorporation. The decreased C flow in wood would have an impact on the Fig. 3 Central metabolism in leaves (a) and sucrose metabolism in phloem (b) and roots (c) of 6-wk-old wild-type (WT) and SUSRNAi hybrid aspen (Populus tremula 9 tremuloides) trees grown in the glasshouse. Carbon-13 ( 13 C) enrichment of metabolites ((a) and left panels in (b) and (c)) and sucrose total pools (right panels in (b) and (c)) were measured by GC-MS. To obtain the 13 C enrichment of metabolites, trees were supplied with 13 CO 2 for 4, 6 and 8 h. Total pool values in phloem and roots are relative to WT samples collected 6 h after the start of the light period. Bars indicate the average and error bars indicate the standard error. ANOVA, Duncan test, n = 4-3. Different letters indicate significant differences (P < 0.05). (2021)

Research
New Phytologist cell wall polymer biosynthesis. Indeed, the amounts of cellulose, lignin and hemicellulose per volume of wood were reduced in all the transgenic lines in comparison to WT (Gerber et al., 2014). Thus, the alterations in the content of the polymers on a volume basis were due to an overall decrease in the C amount allocated to the cell walls. This conclusion was supported by NMR spectroscopy measurements comparing the 13 C content of SUSRNAi-1 and WT cell wall polymers at times 0 and 8. The data point to a reduction in the 13 C levels in total cell wall (Fig. S3a), total G and S lignin (Fig. S3b) as well as in methoxy groups (Fig. S3c) and in individual structures related to lignin ( Fig. S3d-g). With respect to cell wall carbohydrates, the effect was more subtle although some individual peaks seem to have reduced labelling (Fig. S3b,h).

SUSRNAi does not affect starch levels in glasshouse grown trees
Starch is an important C sink and storage reserve in some tree tissues (Dietze et al., 2014). Therefore, we measured starch levels in developing wood, mature wood, phloem, and source leaves to assess the possible role of SUS in starch metabolism in glasshouse grown aspen. No statistically significant differences between the transgenic and WT trees were found in any of the tissues www.newphytologist.com (Fig. S4). Worth noting is that the developing wood starch content of WT trees was 0.7 nmol of glucose per milligram fresh weight, which is equivalent to 0.013% (w/w) of fresh developing wood. Unda et al. (2017) found that in 5-month-old glasshousegrown hybrid poplar starch accounted for c. 2.5% of developing wood on a dry weight basis. These amounts are well below the cell wall polymer contribution to developing wood biomass, which is above 80% (w/w) (Gandla et al., 2015;Wang et al., 2018). Thus, we found no evidence of SUS involvement in starch metabolism during active growth, or that starch would constitute a significant C sink in aspen developing wood.
SUSRNAi causes strong growth reduction under field conditions Aspen trees in natural stands undergo seasonal growth and are exposed to many environmental stresses. Therefore, to fully evaluate the SUSRNAi effects on tree growth, we grew the transgenic trees in a 5-yr field experiment. The stem height and diameter of the SUSRNAi aspen trees were initially identical to those of WT trees but began to show growth defects in subsequent years. We observed consistent reduced height from year 4 onwards, while stem diameter was reduced at year 5 (Fig. 5a,b). The growth defects were more pronounced in SUSRNAi-3. In the fifth year the trees were cut, and the total aboveground biomass measured. The fresh and dry stem weights of the transgenic lines were reduced by up to 80% compared to those for WT trees (Fig. 5c,  d). To confirm that the SUSRNAi was still functional in the transgenics, SUS2 expression was assessed by qPCR and shown to be reduced (Fig. S5). These data prove that SUS is critical for aspen growth and biomass accumulation under field conditions.

Acid invertase activity is increased in glasshouse SUSRNAi lines
Despite the metabolic changes in the SUSRNAi lines, stem biomass was only slightly reduced in the glasshouse-grown transgenic lines (Fig. 2). Since sucrose in the sink tissues is cleaved either by SUS or invertases, we hypothesized that invertases may compensate for SUS defects in the transgenics. Interestingly, the cytosolic neutral invertase activity in the developing wood of SUSRNAi lines was comparable to that of WT trees, but the acid invertase activity was increased (Fig. 6). This suggests that the acid invertase activity located in the apoplast and/or vacuoles can compensate for the lost SUS activity. However, the observed reduction in total C (Fig. 2c), and 13 C labelling over time (Fig. 4) in the developing wood of glasshouse-grown trees suggest that the acid invertase compensation was only partial. Moreover, field grown lines began to show a consistent reduction in stem height from year 4 and in stem diameter at year 5 (Fig. 5), suggesting that a partial compensation mechanism also operates under field conditions. Thus, the sustained reduction in SUS activity over multiple years under field conditions led to accumulative impairment of C incorporation to wood causing biomass reduction. Therefore, it can be concluded that SUS activity is a central determinant of wood formation and tree growth.

Discussion
Our results show that field grown SUSRNAi trees accumulate less stem biomass (Fig. 5), and the glasshouse grown transgenics contain less total C in developing wood (Fig. 2). Since stem biomass formation is dependent on the C derived from sucrose metabolism in sink tissues (Lemoine et al., 2013;Ruan, 2014), we performed labelling experiments with 13 CO 2 to investigate whether the reduced biomass and total C could be ascribed to altered 13 C-sucrose transport and/or metabolism in the transgenic lines. The reduced 13 C-sucrose in the developing wood of SUSRNAi trees suggested a reduction in sucrose import (Fig. 4a). The reduced 13 C-sucrose level in developing wood (Fig. 4) is also consistent with a reduction in de novo sucrose synthesis, however this process is not believed to contribute significantly to the C flux during wood formation (Roach et al., 2017). The reduction in 13 C sucrose was reflected in the decreased 13 C labelling of many central metabolites (Fig. 4a); this trend was also observed for the 13 C accumulation in cell wall polymers (Fig. S3). Since also the cell wall polymer mass per volume unit is reduced in the transgenic lines (Gerber et al., 2014), it can be concluded that C allocation to cell walls is decreased and the cause of the biomass loss (Fig. 5).
The mild phenotype of the glasshouse grown SUSRNAi trees (Fig. 2b) suggested the existence of an alternative compensatory sucrose cleavage mechanism involving invertases. We observed no significant changes in neutral invertase activity in the SUSRNAi lines, but the acid invertase activity was increased (Fig. 6). Compensation by acid invertases when SUS activity is reduced has also been reported in other species (Zrenner et al., 1995;Baroja-Fern andez et al., 2009). The increased acid invertase activity also suggests that these invertases influence C allocation in developing wood. Cell wall acid invertases have coevolved with the vasculature and are believed to contribute to phloem unloading (Wan et al., 2018). The tendency of increased glucose and fructose levels ( Fig. S2) in SUSRNAi trees resembles that seen in potato tubers with reduced SUS, in which the hexose increase was attributed to a compensatory increase in invertase (Zrenner et al., 1995). A similar compensation mechanism may also function in the developing wood of SUSRNAi trees. However, the observed phenotypes in the transgenics including the reduction in total C (Fig. 2c) and 13 C-metabolite labelling ( Fig. 4) suggest that acid invertases only partially compensate for the SUS loss. This explanation is in line with the gradual growth reduction in the field grown SUSRNAi trees (Fig. 5). The seasonal changes and environmental effects under natural field conditions are likely to exacerbate this effect. Environmental stress conditions have been shown to exacerbate SUS defect phenotypes in other species (Ricard et al., 1998;Wang et al., 2014;Takehara et al., 2018). Moreover, in aspen seasonal changes in SUS activity were observed in the outer trunk wood (Schrader & Sauter, 2002), suggesting that SUS may be involved in the activity-dormancy and dormancy-activity transitions.
SUS catalyses the sucrose degradation into fructose and UDPglucose, the latter being the substrate for cellulose synthesis. Although it has often been hypothesized that SUS is essential for directly providing UDP-glucose to the cellulose synthase complex  (Amor et al., 1995;Haigler et al., 2001;Stein & Granot, 2019), near complete silencing of SUS in aspen developing wood did not abolish cellulose biosynthesis (Gerber et al., 2014). However, SUS overexpression in aspen can increase the cellulose content (Coleman et al., 2009). Consequently, SUS can contribute to the UDP-glucose provision for cellulose synthesis in aspen trees but is not essential for this process. Instead, cytosolic invertase silencing significantly reduced cellulose levels in developing wood of aspen, suggesting that the cellulose biosynthesis pathway involves hexokinase and UDP-glucose pyrophosphorylase (Rende et al., 2017), while SUS plays an essential role in controlling total C allocation.
When leaves transition from sink to source tissues, SUS levels fall significantly (   www.newphytologist.com Zhu et al., 2018). Consequently, SUS levels in mature leaves are low (Pavlinova et al., 2002;Thirugnanasambandam et al., 2019). In line with this, SUS activity in source leaves of glasshouse grown aspens was close to the background level. Leaf SUS is thought to be involved in sucrose degradation during the day (Rolland et al., 2006). Hence, several mechanisms could be affecting the sucrose labelling in the SUS leaves (Fig. 3a), including reduced sucrose synthesis, or differential labelling kinetics of the different subcellular sucrose pools (Kruger et al., 2007;Fernie & Morgan, 2013;Beshir et al., 2017). Moreover, the alteration in 13 C labelling is not limited to sucrose in leaves, but several other central metabolites also show differences, especially amino acids (Fig. 3a). This may be considered a regulatory effect generated locally in the leaves or triggered by a sink-to-source signal from the developing wood in response to the reduced sucrose and C allocation (Ainsworth & Bush, 2011). However, the SUSRNAi source leaves do not show marked phenotypic ( Fig. 2; Tables S3, S4) or metabolite pool alterations ( Fig. S4d; Table S5), which suggests that SUS does not have an essential role in these leaves during active growth. The role of SUS in aspen roots under glasshouse conditions seems even more limited. The fact that the roots of SUSRNAi lines show no phenotypic alterations (Fig. 2g-i) and that they have a few metabolites with altered 13 C labelling ( Fig. S1) suggests that they can compensate for the reduced SUS activity. Similarly, no obvious role for SUS was observed in maize roots under aerobic conditions, but it was important under anaerobic condition such as caused by flooding (Ricard et al., 1998). However, it is also possible that the limited effect of SUS silencing in aspen roots was due to residual SUS activity in this tissue.
Because all the analysed tissues had reduced SUS activity or reduced SUS mRNA levels, it is probable that the observed effects are initiated locally. This is especially true for developing wood, which has high SUS activity in WT and only residual SUS activity in the SUSRNAi lines (Fig. 1). However, systemic effects could be occurring simultaneously, including additive effects on the 13 C-sucrose transport across tissues. This is illustrated by the decreased 13 C-sucrose enrichment in the source to sink transport direction in the SUSRNAi lines (Figs 3, 4), especially from phloem towards developing wood. Since SUS has such low levels in source leaves (Fig. 1), it is not likely that the effect seen in developing wood is merely an effect generated in the leaves and then cascaded to the sink tissues. However, a feedback regulation produced by the reduced developing wood sink activity, which would impact the C metabolism in leaves, is possible (Ainsworth & Bush, 2011). Probing such a mechanism could be done through grafting experiments testing different combinations of WT and SUSRNAi scions and stocks. Indeed, the C flow alteration at the whole tree level together with the altered sucrose turnover in leaves, phloem and developing wood raises the interesting possibility that SUS could participate in a signal generation involved in the source-sink metabolic coordination.
To summarize, the decrease in SUS activity leads to decreased sucrose C incorporation in developing wood, which is subsequently reflected in the decreased C flow in the central metabolism. This determines a lower C allocation whose output is a reduced total C and reduced biomass in wood. This provides evidence that SUS is involved in determining C allocation in aspen developing wood and, thus, in affecting the biomass accumulation into this tissue. Likewise, SUS silencing affects the C metabolism at the whole tree level as shown through 13 C labelling. The alteration in C metabolism at the whole tree level suggests that SUS could participate in the sink-source metabolic coordination.

Supporting Information
Additional Supporting Information may be found online in the Supporting Information section at the end of the article.     Table S1 Primer list used in the qPCRs.