Woody species, including Populus, are generally recalcitrant to cell suspension culture. We have previously identified an aspen (Populus tremuloides Michx.) clone (L4) that is amenable to suspension culture from a screen of over 200 seeds (Tsai, unpublished data). This "suspension culture" trait appears to be genetically rather than developmentally controlled, as friable calli can also be routinely induced from young leaves of vegetatively propagated L4 plants maintained in a greenhouse. A typical growth curve of the L4 cell line is shown in Fig 1. With an 11-day subculture interval, the cell volume usually increased by ~400% at the end of the culture cycle. This pattern of cell growth was stable for more than 40 cycles during the experimental period.

The cells were harvested at regular intervals to determine the basal CT and PG concentrations during the culture cycle. At the start of each culture period (i.e., after subculture), CT concentrations averaged ~14.5% dry wt, but decreased gradually through the lag and early log phases to a low of ~6.5%, and then increased rapidly thereafter, reaching ~15% at the end of the culture cycle (Fig. 1). The initially high CT content corresponded with the CT level at the end of the culture cycle. The gradual decrease in CT content during early growth may be suggestive of CT breakdown and a lack of or little CT synthesis. Under normal culture conditions, the L4 cell line did not accumulate salicin or any other higher-order PGs.

Because PGs (salicin, salicortin and tremulacin) are detected in the leaves of the aspen L4 line, the inability of cultured cells to produce PGs may be attributed to the absence of PG precursors and/or spatiotemporal regulation of metabolic activities in suspension cells versus intact tissues. Feeding experiments were undertaken to test whether PGs would accumulate if putative phenolic precursors were supplied. Phenolic acids (benzoic acid, cinnamic acid, O-coumaric acid and salicylic acid) were fed at 0.2 mM, since higher feeding levels led to cell browning in preliminary trials, suggestive of toxicity. Phenolic alcohols (salicyl alcohol and benzyl alcohol), aldehydes (benzylaldehyde and salicylaldehyde) and glucosides (salicin and helicin) were fed at 1 mM. Feeding was conducted 5 days after subculture (DAS), at early log phase, and cells were harvested after 24, 48 and 96 h. The glucoside salicin was readily taken up and detected primarily as unaltered salicin in the cell extracts, although a slow conversion to its unnatural isomer, isosalicin, continued throughout the 96 hr feeding period (Fig. 2, Additional file 1). Feeding salicyl alcohol (the aglycone of salicin), salicylaldehyde and helicin (salicylaldehyde glucoside) also resulted in the accumulation of both salicin and isosalicin (hereafter referred to as salicins). In all cases, accumulation of salicins reached a plateau by 48 hr (Fig. 2), and was higher with salicin and salicyl alcohol feeding, than with salicylaldehyde and helicin feeding. Higher-order PGs were not detected during any of the feeding experiments. Feeding with all other compounds led to accumulation of their respective glucosides (see Additional file 1), but not PGs, and these compounds were not studied further. The feeding results showed that the cells are capable of accumulating both CTs (under normal condition) and simple PGs (upon precursor feeding), suggesting that the cell culture system can be exploited for the investigation of metabolic competition. Salicyl alcohol was chosen for subsequent feeding experiments because it led to a higher accumulation of salicins

When cells were fed varying levels of salicyl alcohol (0, 1, 5 and 10 mM), a dose-dependent accumulation of salicins was observed (Fig. 3A). Accumulation of salicins plateaued 24 h and 48 h after feeding with 1 mM and 5 mM salicyl alcohol, respectively, but continued to increase with 10 mM feeding over the 4-day monitoring period. The total salicins detected in the cells at the end of the experiment were 3, 17 and 29% dry weight with 1, 5, and 10 mM salicyl alcohol, respectively. At 5 mM feeding, the efficiency of salicyl alcohol uptake and conversion at the end of the 4-day period was estimated to be ~60%, as 175 μmoles of salicyl alcohol were converted to 100 μmoles of salicins.

CT levels showed a negative dose-dependent response to salicyl alcohol feeding. With 1 mM feeding, CT levels were similar to those of unfed cultures, in which CT levels decreased after 24 h (corresponding to 6 DAS), followed by a steady increase through the end of the 96 h period (Fig. 3B). This is consistent with the CT accrual kinetics typical of control cells (Fig. 1, also shown as the gray line in Fig. 3B), where CT abundance declines to a minimum at 6 DAS before exhibiting a recovery. Feeding at 5 mM caused the CT levels to decrease through 48 h (or 7 DAS), and delayed the normal pace of CT accrual by ~2 days. At 10 mM, cells showed a downward trend of CT levels and did not resume CT synthesis throughout the 96 h feeding period. At this feeding level, cell growth was also reduced by up to 28% (Fig. 3C). Together, these results are consistent with a negative effect of salicyl alcohol glycosylation on the accumulation of CTs.

To examine whether culture growth phase plays a role in cellular capacity to accumulate secondary metabolites, cultures at 2, 5, 8, or 11 DAS were fed 5 mM salicyl alcohol and sampled over a 96 h period. In general, cells at the lag (2 DAS) or early exponential (5 DAS) phases, when the CT levels were in decline, exhibited a higher capacity to accumulate salicins than cells at the mid-exponential (8 DAS) or stationary (11 DAS) phases when CT levels were increasing (Fig. 4A). At the end of the experimental period, total salicins reached ~15% and ~8% in cells fed at the earlier and later stages, respectively.

Salicyl alcohol feeding at the lag phase (2 DAS) had no effect on basal CT concentration (Fig. 4B). Feeding at 5 DAS delayed the onset of CT accumulation as described above (Fig. 3B), and reduced CT levels by 38% relative to unfed controls (Fig. 4B). At later stages of cell growth (8 and 11 DAS) when CT levels were high, salicyl alcohol feeding abolished further CT increase. These results provide support for a reciprocal metabolic competition between CT synthesis and salicyl alcohol glycosylation at several stages of culture growth.

The observed reciprocal regulation of salicins and CTs in cell cultures suggested that salicyl alcohol feeding may have affected both carbohydrate utilization and carbon flux through the phenylpropanoid pathway. To gain additional information about the metabolic response, gene expression analysis was carried out using a custom aspen 7K EST array. Cells were fed either 5 mM salicyl alcohol or water at 5 DAS, and harvested after 48 h for the analysis. The 48 h time point (corresponding to 7 DAS) was chosen because CT levels typically begin to increase by this time in unfed cells from a minimum at 6 DAS (Figs. 1 and 4B); CT levels did not exhibit the normal increase and remained near the minimum in the fed cells at 7 DAS (Fig. 4B); and accumulation of salicins in the fed cells reached its plateau after 48 h (Figs. 2, 3 &4). The 48 h feeding window therefore provided an ideal metabolic context with a reproducible difference between salicyl alcohol-fed and control cells for assessing the underlying transcriptome adjustments.

Of the 3,957 ESTs that passed a series of quality control measures, expression of 725 ESTs representing 659 non-redundant genes was found to be significantly altered (false discovery rate p ≤ 0.05) by salicyl alcohol feeding (see Methods). Most of these genes showed small expression changes, and only 35 ESTs exhibited ≥ 2-fold differences. This limited transcriptome change was not unexpected, given the observation that glycosylation of exogenous salicyl alcohol led to a delay in CT increase with no apparent effect on cell growth under the condition (48 h post feeding) studied. For further analysis, we applied a modest fold-change cutoff of 1.3 yielding 284 and 160 ESTs that were up- and down-regulated, respectively (Additional file 2).

Stress-related transcripts, such as glutathione S transferase, dehydrin, thaumatin-like protein, and germin-like protein, were among the most up-regulated genes in the salicyl alcohol-fed cells (Table 1). The most highly up-regulated EST (MTU6CR.P6.H02) corresponded to a peroxidase. This gene is poorly expressed in vegetative tissues, except cell cultures, and is strongly up-regulated in methyl jasmonate-treated cells as well as in insect-damaged aspen leaves 15 (Tsai, unpublished data). Many of the down-regulated genes were associated with phenylpropanoid and flavonoid biosynthesis (Table 1), including phenylalanine ammonia-lyase, 4-coumurate:CoA ligase, caffeoyl-CoA O-methyltransferase, cinnamoyl-CoA reductase, and leucoanthocyanidin reductase. These results are in line with the reduced CT accumulation in salicyl alcohol-treated cells.

Genes encoding fructose 1,6-bisphosphatase and sucrose-hydrolyzing enzymes, such as sucrose synthase (SuSy) and vacuolar invertase (VIN), were up-regulated by salicyl alcohol feeding (Table 1). Also up-regulated were genes comprising the glycolysis pathway, including glyceraldehyde 3-phosphate dehydrogenase, phosphoglycerate kinase, phosphoglycerate mutase, and enolase, and genes comprising the tricarboxylic acid (TCA, Krebs) cycle, including malate dehydrogenase, aconitase and isocitrate dehydrogenase (Table 1).

QPCR was used to enable higher resolution analysis of the gene expression changes observed in microarrays. Because of the limited coverage of the EST array, we expanded the analysis to include all known flavonoid biosynthetic pathway genes 6, as well as gene families associated with the transport and hydrolysis of sucrose, and the glycosylation and transport of simple phenolics 1617181920.

The reciprocal competition between salicin and CT accumulation was observed under multiple feeding conditions, but varied depending on the physiological/metabolic status of the cultures. Cells exhibited a 2-fold higher capacity to accumulate salicins during early growth phases, when new CT biosynthesis was not apparent, than in mid-late growth phase cells containing high levels of CTs. Conversely, the trajectory of CT accumulation was at most only slightly impacted by salicin accrual at early growth phases, but was clearly compromised when salicyl alcohol was fed at mid-late growth phases. In other words, aspen cells responded to salicyl alcohol feeding by producing PGs and CTs in a compensatory rather than additive fashion, raising the possibility that there was a metabolic constraint to phenylpropanoid biosynthesis. Feeding methyl jasmonate, an elicitor known to stimulate phenylpropanoid gene expression 34, increased CT accumulation, but had no effect on salicyl alcohol glycosylation. Conversely, 5 mM salicyl alcohol feeding reduced the CT accumulations by 63-69% after 48 h, regardless of the methyl jasmonate levels co-administered, at 0, 1 μM or 25 μM (Payyavula, Tsai and Harding, unpublished). These results support a PG-CT tradeoff in salicyl alcohol-fed cells that is independent of the phenylpropanoid biosynthetic capacity.

In the present investigation, the PG-CT tradeoff appears to reveal a sensitivity of phenylpropanoid carbon partitioning to induced glycosylation, since the combined concentrations of CTs plus salicins in the fed cells differed little between culture ages at the end of the 4-day feeding period, and since cell growth was not affected. The reduction of CTs was, at least in part, the result of transcriptional down-regulation of phenylpropanoid and flavonoid pathway genes, a process known to be integrated with sugar sensing 35. In salicyl alcohol-fed cultures, multiple genes involved in sucrose transport and hydrolysis were up-regulated, including those encoding vacuolar sucrose transporter (SUT4), cytosolic sucrose synthases (SuSy1, 2 and 3) and vacuolar invertase (VIN2). This suggests that altered compartmentalization of sucrose hydrolysis along with increased glycosylation may have altered phenylpropanoid flux in the fed cells.

Genes involved in the Krebs cycle and glycolysis were also up-regulated in salicyl alcohol-fed cultures, implicating a diversion of metabolites into the Krebs cycle through glycolysis. One of the important processes supported by the Krebs cycle is respiration. Increased respiration is associated with increased carbon loss as CO₂. Whether this was linked to reduced CT synthesis seems unlikely due to the high availability of external sucrose. The Krebs cycle also utilizes acetyl-CoA, a precursor for malonyl-CoA. As malonyl-CoA contributes 40% of the carbon comprising CT, elevated Krebs cycle metabolism could reduce the amount of carbon allocated for CT synthesis. The Krebs cycle supplies energy as well as carbon for biosynthetic intermediates that are required for growth. If the gene up-regulation we observed was related to growth maintenance, it would mean that growth outcompeted, or had a higher priority over CT synthesis for cellular carbon in salicyl alcohol-fed cells. Whether salicyl alcohol feeding reduced cellular uptake of sucrose, thereby limiting cellular resources and CT accrual was not determined.

Essentially all of the GT1 members found to be well-expressed in cell cultures were up-regulated by salicyl alcohol feeding. The transcript levels of GT1-2 and GT1-246 were especially high, and nearly doubled in salicyl alcohol-fed cells. The data suggest that they may mediate the glycosylation of salicyl alcohol into salicin. Alternatively, because many GTs exhibit broad substrate specificity, sometimes functioning in xenobiotics detoxification 36, up-regulation of well-expressed GT1 genes might merely reflect stimulation due to a possible stress of salicyl alcohol feeding. Up-regulation of several defense-related genes in the fed cells lends further credence to a general stress response. Unequivocal identification of GT1s that specifically regulate PG accrual will require a system that synthesizes aglycone PG precursors. Nevertheless, the work reported here offers glycosylation as an important determinant of the partitioning of carbon among pathways that begin with hydroxycinnamate and benzoate substrates. After glycosylation, glycosides are likely transported and sequestered into the vacuole for storage 262728. The up-regulation of putative tonoplast-localized MRP transporters, MRP1 and MRP6, suggests their possible involvement in the transport of salicins from cytoplasm into vacuole. Salicin transport may also be mediated by vacuolar localized SUTs 19, which could implicate the possible involvement of Populus SUT4 as well.

Calli were induced from surface-sterilized leaves of greenhouse-grown Populus tremuloides genotype L-4 on semi-solid Woody Plant Medium 37 supplemented with 2.2 mg l^-1 of 2,4-dichlorophenoxy acetic acid and 3% sucrose. Suspension cultures were established with ~5 gm calli in 30 ml of liquid media in a 125 ml flask, and maintained in an orbital shaker at 120 rpm in the dark at 25°C. Cells were subcultured at 11-day intervals by inoculating 5 ml culture to 30 ml fresh medium. When indicated, replicate cultures were maintained in nephelo flasks for estimation of culture growth by the settled cell volume. Prior to measurement, approximately 10 ml of cell suspensions were allowed to settle for 25 min in the sidearm of the nephelo flask. The fraction of the suspension occupied by the cells was determined as the percent settled cell volume.

Phenolic compounds with concentrations ranging from 0.2 to 5 mM were administered to the cultures at 5 DAS unless otherwise specified. Phenolic compounds were dissolved in DMSO for use in initial feeding trials, and feeding volumes ranged from 12 to 32 μl per flask. Control flasks received blank DMSO, and DMSO alone (with variations in feeding volumes) did not lead to accumulation of the various glucosides of interest. In subsequent feeding trials, salicyl alcohol was dissolved in water and feeding volumes ranged from 60 to 600 μl per flask for the dose-dependent experiments (control flasks received 300 μl of water). For the time-course (culture age) study and the microarray analysis, salicyl alcohol (5 mM final concentration) or water was administered at a fixed volume of 300 μl. Cells were harvested at regular intervals by low vacuum filtration, snap frozen in liquid nitrogen and stored at -80°C until use.

Freeze-dried samples (5 mg each) were extracted in 800 μl of cold methanol for 20 min in a cold ultrasonic bath and centrifuged at 15,000 g for 5 min. The methanol extracts (5 μl) were injected into an Eclipse XBD-C18 column (5 μm, 2.1 × 150 mm) and analyzed by HPLC-UV/MS (Hewlett-Packard 1100 Series, Agilent Technologies) at a flow rate of 0.2 ml/min using solvents A (10 mM formic acid, pH 3.4) and B (100% acetonitrile) according to the following gradient: 0 to 15 min, 0% to 70% B, 15 to 17 min, 70% to 100% B, 17 to 19 min, 100% to 0% B, and 19 to 30 min, 0% B. Glucosides (salicin, isosalicin, cinnamoyl-glucoside, O-coumaroyl-glucoside, salicyloyl-glucoside, benzyl alcohol-glucoside, benzoyl-glucoside, and helicin) were identified by UV absorbance and mass spectral data (Additional file 1). Concentrations of both salicin and isosalicin were estimated by a calibration curve developed using authentic salicin (Sigma). Total CTs were estimated as described 32 according to Porter et al., 38, using a standard curve developed based on purified aspen leaf CTs.

RNA was extracted from frozen cells according to Chang et al. 39 and treated with Turbo DNase according to the manufacturer's instructions (Ambion). Aminoallyl-labeled cDNA synthesis, Cy-dye coupling, purification, EST microarray processing were carried out as described 32, except for hybridization which was performed in a humid hybridization oven (Boekel Scientific). An equal amount (50 pmole) of Cy3- and Cy5-labeled cDNA from control and salicyl alcohol-fed cells was mixed and used for each hybridization. The experiment consisted of three biological replicates (three controls randomly paired with three treatment groups) with a technical dye swap for each pair of samples, giving rise to a total of 6 hybridizations.

Hybridized slides were scanned with a Genepix 4000B scanner (Axon Instruments) and the florescence signal intensity was quantified using the GenePix Pro 5.1 software (Axon Instruments). Probes with signal intensities in both channels greater than two standard deviations from the background signal were flagged as present, and other irregular or low-quality spots were manually flagged. Data were normalized by theLOWLESS (locally weighted linear regression) algorithm implemented in the GeneSpring 7.3.1 software (Agilent). Probes (excluding spike controls) that passed the following quality control measures were subjected to statistical analysis: present in at least four replicates (4,132 spots) and with a coefficient of variation among replicates less than 35% (3,957 spots). A total of 1,157 ESTs were found to be differentially expressed based on t-test, with a Benjamini and Hochberg false discovery rate for multiple testing correction at P ≤ 0.05. The list was further filtered to retain 444 probes with (1) a raw hybridization signal ≥ 100 in at least four of the six replicates (725), and (2) a fold-change cutoff of 1.3. The microarray data has been submitted to GenBank Gene Expression Omnibus repository under accession number GSE18360.

cDNA was synthesized using DNA-free total RNA, anchored oligo(dT)20 primers and SuperScript II reverse transcriptase (Invitrogen). Relative transcript abundance was analyzed by Q-PCR using cDNA derived from 2.5 ng of total RNA, gene-specific primers and ABsolute QPCR SYBR Green Mix (ABgene) with ROX as an internal reference. Relative expression of the genes of interest was calculated by the ΔC_T method as described 6 using ubiquitin-conjugating enzyme E2 and elongation factor 1-β as the housekeeping genes. Primer information is provided in Additional file 3, except phenylpropanoid gene primers as reported 6.