Intraspecific variation in the Populus balsamifera drought transcriptome



    1. Faculty of Forestry, University of Toronto, 33 Willcocks St., Toronto, ON M5S 3B3, Canada
    2. Centre for the Analysis of Genome Evolution and Function, University of Toronto, 25 Willcocks St., Toronto, ON M5S 3B2, Canada
    Search for more papers by this author

    1. Centre for the Analysis of Genome Evolution and Function, University of Toronto, 25 Willcocks St., Toronto, ON M5S 3B2, Canada
    2. Department of Cell & Systems Biology, University of Toronto, 25 Willcocks St., Toronto, ON M5S 3B2, Canada
    Search for more papers by this author

    1. Centre for the Analysis of Genome Evolution and Function, University of Toronto, 25 Willcocks St., Toronto, ON M5S 3B2, Canada
    2. Department of Cell & Systems Biology, University of Toronto, 25 Willcocks St., Toronto, ON M5S 3B2, Canada
    Search for more papers by this author

    1. Alberta-Pacific Forest Industries Inc. P.O. Box 8000 Boyle, Alberta, T0A 0M0, Canada
    2. Department of Renewable Resources, University of Alberta, 731 General Services Building, Edmonton, AB T6G 2H1, Canada
    Search for more papers by this author

    1. Department of Wood Science, University of British Columbia, 4030-2424 Main Mall, Vancouver, BC V6T 1Z4, Canada
    Search for more papers by this author

    1. Department of Biological Sciences, Simon Fraser University, 8888 University Drive, Burnaby, BC V5A 1S6, Canada
    Search for more papers by this author

    Corresponding author
    1. Centre for the Analysis of Genome Evolution and Function, University of Toronto, 25 Willcocks St., Toronto, ON M5S 3B2, Canada
    2. Department of Cell & Systems Biology, University of Toronto, 25 Willcocks St., Toronto, ON M5S 3B2, Canada
    Search for more papers by this author

  • Contributions:
    ETH, SDM, ALP and MMC designed research; ETH, SR, BT, SDM, ALP and MMC organised experimental logistics including transfer and establishment of biological materials; ETH, SR, and OW performed research; ETH, OW, and MMC analysed data; ETH and MMC wrote manuscript with editorial assistance from SR, OW, BT, SDM, ALP and MMC.

M. M. Campbell. Fax: +1 416 978 5878; e-mail:


Drought is a major limitation to the growth and productivity of trees in the ecologically and economically important genus Populus. The ability of Populus trees to contend with drought is a function of genome responsiveness to this environmental insult, involving reconfiguration of the transcriptome to appropriately remodel growth, development and metabolism. Here we test hypotheses aimed at examining the extent of intraspecific variation in the drought transcriptome using six different Populus balsamifera L. genotypes and Affymetrix GeneChip technology. Within a given genotype there was a positive correlation between the magnitude of water-deficit induced changes in transcript abundance across the transcriptome, and the capacity of that genotype to maintain growth following water deficit. Genotypes that had more similar drought-responsive transcriptomes also had fewer genotypic differences, as determined by microarray-derived single feature polymorphism (SFP) analysis, suggesting that responses may be conserved across individuals that share a greater degree of genotypic similarity. This work highlights the fact that a core species-level response can be defined; however, the underpinning genotype-derived complexities of the drought response in Populus must be taken into consideration when defining both species- and genus-level responses.


Trees of the genus Populus, which include poplars, aspen and cottonwoods (herein collectively referred to as poplars), are found primarily throughout the northern hemisphere (Dickmann 2001), and have many favourable attributes which have lead to their widespread use in both forestry and agriculture (Brunner, Busov & Strauss 2004). Occupying both a large geographical region and a diverse array of habitats, poplar trees must contend with a variety of environmental conditions in order to survive. Along with other environmental stresses, such as insect defoliation and rust cankers, drought is a major factor impinging on poplar growth, productivity and survival throughout its range (Hogg, Brandt & Kochtubajda 2002; van Mantgem et al. 2009). The drought sensitivity of poplar trees, which are a prominent species in many temperate forests, has posed increased concern for the future amidst predictions of changing climate and water shortages, as well as during periods of episodic drought (Schindler & Donahue 2006).

In response to water deficit, poplar trees generally display alterations in plant water status resulting in suppression of stomatal conductance and declines in productivity; however, considerable variation in drought response and tolerance has been observed within the genus Populus (Gebre & Kuhns 1991; Tschaplinski, Tuskan & Gunderson 1994; Chen et al. 1997; Monclus, Dreyer & Villar 2006; Bonhomme, Monclus & Vincent 2009). The ability of trees to adapt and survive diverse environmental variables, such as drought, is a consequence of a variety of biochemical and physiological processes, many of which are the result of stress signal perception leading to alterations in the transcriptome, resulting in an adaptive response (Kozlowski & Pallardy 2002; Lei, Yin & Li 2006). The variability in response to drought observed at the morphological and physiological level suggests that poplar trees are an excellent organism to study the molecular underpinnings of the drought response and the variation in this adaptive response to such an environmental insult.

Variation in gene expression observed among populations is heritable (Oleksiak, Churchill & Crawford 2002; Whitehead & Crawford 2006) and, therefore, examination of the variability in the transcriptome response to drought may provide insight into the diversity and adaptation to such a response. Previous investigations dissecting the molecular underpinnings of the drought response in the genus Populus have focused on transcriptional differences among various poplar species or hybrids (Street et al. 2006; Wilkins et al. 2009). This indicates that poplar trees, regardless of species or hybrid, likely have a variety of mechanisms governing the drought response, and that this response is highly dependent on genotype. Although there is evidence that intra-specific variation in drought response can be observed among growth rates and physiological traits within a species of Populus (Lu et al. 2009), drought-induced variation in gene expression within a given species of Populus has not yet been investigated. Here, we explore the variation in transcriptome responses to drought within the species Populus balsamifera L. spp. balsamifera (balsam poplar). P. balsamifera is a dominant tree species within North America's boreal ecosystems whose range is transcontinental, and can be found growing on upland riparian sites (USDA-NRCS 2009). On account of its similarity to black cottonwood (Populus trichocarpa Torr. & Gray) and the wealth of available genomic tools (Tuskan DiFazio & Jansson 2006), P. balsamifera represents an ideal species for studying intra-specific variation in the drought response.

In this study, Affymetrix GeneChip technology was used to study the drought responsive component of the Populus transcriptome. Six P. balsamifera genotypes from various geographical regions (Fig. 1) were grown in a common growth chamber environment and the variation in gene expression in trees responding to drought was examined. These experiments aimed to test the hypothesis that there are significant differences in the trancriptomes of P. balsamifera genotypes in response to water-deficit conditions; however a common species-level response could also be assessed. We expect that the knowledge about transcriptome variation within a given species of poplar will contribute to our understanding of the adaptive responses to drought and the molecular underpinnings of these responses.

Figure 1.

Source of origin of the six P. balsamifera genotypes examined in this study.


Plant material

Dormant, 25 cm, un-rooted hardwood cuttings of six P. balsamifera genotypes (AP-947, AP-1005, AP-1006, AP-2278, AP-2298, AP-2300) were obtained from Alberta-Pacific [Forest Industries Inc. (Al-Pac), Boyle, AB, Canada]. Cuttings were imbibed for 48 h prior to planting (Desrochers & Thomas 2003) into Sunshine Mix-1 (Sun Gro Horticulture Inc, Bellevue, WA, USA; in 1 m opaque pots (10.5 cm diameter). The plants were grown in a climate-controlled growth chamber under long day conditions (16 h photoperiod, minimum light intensity: 200 µmol m−2 s−1), with a maximum day temperature of 22 °C and a minimum night temperature of 17 °C throughout the experiment. All plants were watered every 2 to 3 d to field capacity and fertilized (20:20:20, N-P-K, 1.5 g L−1, 600 mL plant−1) every 3 weeks. All plants were grown for 9 weeks prior to the onset of the water-withholding experiment, at which point they were divided into two groups, well watered (WW; n = 27–35 per genotype) and water deficit (WD; n = 27–35 per genotype). Water deficit conditions were imposed on dry plants by withholding water; wet plants were regularly watered every 2 to 3 d to maintain water status. Fifteen days following the water-withholding, the first fully expanded leaf was harvested from three individual trees from each genotype for both well-watered and water-deficit treatments at two time points: midday (MD; middle of the light period) and pre-dawn (PD; 1 h before the lights were turned on). Leaves were pooled and immediately flash frozen in liquid nitrogen for subsequent analysis. This was repeated three times in order to achieve triplicate replicates for each genotype-treatment combination at each time point.

Physiological and growth traits

A portable infrared gas analyser (IRGA; LI-6400, LI-COR Biosciences Inc., Lincoln, NE, USA) was used for measuring photosynthesis, stomatal conductance (gs) and transpiration. Beginning at the start of the water-withdrawal experiment, 9 weeks after planting, measurements were taken daily throughout the experimental period (n = 4–7 individuals per genotype/treatment). Measurements were on mature, fully expanded leaves at the midday time-point. Productivity and relative water content (RWC) measurements were made 15 d after the onset of the water-withdrawal experiment on both well-watered and water-deficit-treated plants. Productivity was measured by determination of tree height, stem circumference and total aboveground dry-weight biomass. Data analysis was performed using R (R Development Core Team 2009). Means were calculated with their standard error (SE), and compared using a two-way ANOVA. Genotype and treatment were considered as the main factors; differences between treatments and among genotypes were determined using a TukeyHSD test. Leaf RWC was calculated on a mature fully expanded leaf (n = 5 individuals per genotype/treatment). Fresh weight (FW) was recorded, and the leaf was allowed to rehydrate in distilled H2O for 24 h in the dark in order to obtain turgor weight (TW). Leaf dry weight (DW) was obtained after the leaf was dried at 70 °C for 48 h. RWC was calculated according to Barrs & Weatherley (1962) as: RWC(%) = (FW − DW) × 100%/(TW − DW).

RNA extraction, microarray hybridisation and analysis

Total RNA was isolated from fully expanded leaves of P. balsamifera and hybridized to an Affymetrix Poplar GeneChip (Affymetrix, Santa Clara, CA, USA) as described by Wilkins et al. (2009). GeneChip expression analysis was performed using the Bioconductor (Gentleman, Carey & Bates 2004) software package AFFY (Gautier et al. 2004) in R (R Development Core Team 2009) as described in Wilkins et al. (2009). All 72 arrays were pre-processed together using GC-robust multi-array analysis [gcrma; (Wu et al. 2004)]. Expression data was filtered to eliminate probe sets with low levels of variation across samples and low levels of expression according to Wilkins et al. (2009). The pre-processed data was analysed as a multi-factorial ANOVA design (six genotypes, two treatments and two time points) using the LIMMA (Smyth 2004) package in R (R Development Core Team 2009). Treatment, genotype and time point were considered the main factors. Differential expression in response to water deficit was determined using an empirical-Bayes moderated t-statistic with a Benjamini and Hochberg adjustment to control the false discovery rate (adjusted P value cut-off of 0.05; (Smyth 2004). In order to take into consideration the magnitude of differential expression for genes that are significantly differentially expressed for treatment main effect only, probe sets were filtered according to a t-test threshold, which corresponds to a minimum fold-change of 2.0 [TREAT; (McCarthy & Smyth 2009)]. Genes were annotated using the Annotation for Probe Sets in PLEXdb (Wise et al. 2007) and Annotation Batch Function in PopGenie (Sjodin et al. 2009). All samples were uploaded to Gene Expression Omnibus (http://www.ncbi.nlm.nih.ov/geo/); accession number GSE21171.

Single-feature polymorphism (SFP) analysis

SFPs were identified using pair-wise comparisons between genotypes using Affymetrix Poplar GeneChip arrays for well-watered, midday poplar samples according to Fujisawa et al. (2009). The number of SFPs were identified for all probe sets that passed through the filtering criteria, as well as for probe sets that were either significantly differentially expressed for genotype main effect, or not.

DNA extraction and simple-sequence repeat (SSR) analysis

Total DNA was extracted according to Doyle & Doyle (1990). Seven SSR microsatellite loci were used to fingerprint the six P. balsamifera genotypes. Five of the seven loci mapped to distinct linkage group in the Populus genome (Tuskan et al. 2004); however, the remaining three have no mapping information. Electrophoresis-based SSR genotyping was performed by The Centre for Applied Genomics, The Hospital for Sick Children, Toronto, Canada.


There is intraspecific variation in productivity and physiological responses in Populus balsamifera following water deficit

To investigate the intraspecific variation of P. balsamifera in response to water deprivation, a multi-factorial experiment was conducted using six P. balsamifera genotypes. The six genotypes examined originated from five distinct geographic regions in Western Canada (Fig. 1), with varying climatic histories (Table 1). Each P. balsamifera genotype was genetically unique based on SSR microsatellite fingerprinting (Supporting Information Table S1).

Table 1.  Location and climate variables
CloneLatitude (N)Longitude (W)Elevation (m)Mean annual temperature (°C)Mean annual precipitation (mm)Degree days (>5 °C)
  • Location details and historic climatic variables adjusted for specific location and elevation using the Climate BC model described by Wang et al. (2006).

  • a

    no elevation data available.

AP-94755° 38′ 2.13″113° 23′ 53.68″7861.85651193
AP-100555° 24′ 25.95″114° 36′ 19.67″6301.85381258
AP-100655° 24′ 25.95″114° 36′ 19.67″6301.85381258
AP-227858° 46′ 13″123° 4′ 21″NAa−0.25331294
AP-229859° 11′ 19″122° 46′ 35″NAa−0.94681255
AP-230058° 51′ 14″122° 31′ 28″NAa−1.14391241

After 15 d of withholding water, decreased productivity (Fig. 2) and stomatal conductance (Fig. 3) were observed between well-watered plants and those deprived of water. Multi-factor ANOVA analysis for aboveground biomass, plant height and stem circumference revealed a significant genotype effect for all three variables; whereas, only significant treatment effect for aboveground biomass (P = 0.05, data not shown). In response to water-deficit, all six genotypes exhibited significant differences in midday stomatal conductance (gs); however, genotype AP-1006 had significant differences as early as five days after the onset of the water-deficit treatment (P = 0.1), whereas genotype AP-2300 did not exhibit significant differences until 11 d after the onset of water-deficit conditions (Fig. 3). Genotypes AP-1006 and AP-2278 showed striking differences in gs, between well watered plants and plants grown under water-deficit conditions at day 15. The differences in gs between well-watered and water-limited plants observed in other genotypes, such as AP-947 and AP-1005, was less marked. By fifteen days after the onset of the water-withholding experiment, relative water content (RWC) was significantly lower in genotypes AP-1005, AP-1006, AP-2298 and AP-2300 (P = 0.05, Supplementary Table 2). Phenotypic responses, including both physiological and morphological, to water-deficit treatment in P. balsamifera showed no significant correlation with historic climatic conditions and geographic origin (Supporting Information Fig. S1). This may reflect the high level of variation among these genotypes in their ability to respond to environmental stimuli regardless of their origins. Although their responses do not reflect historic origins, the variability observed may an important trait for survival in fluctuating environments (Clark 2010). Moreover, the variation in such adaptive traits may be particularly important in P. balsamifera as population genotypic variation and effective population size is considered low (Olson et al. 2010).

Figure 2.

Above ground biomass (a), plant height (b) and stem circumference (c) of six genotypes of P. balsamifera were calculated 15 d after the onset of the water-withdrawal experiment for both well watered (shaded bars) and water deficit treated (white bars) plants. Significant differences between genotypes and treatments (P ≤ 0.05) are denoted by small letters for all variables. Mean values and SE bars are represented.

Figure 3.

Box plot of the variation in midday leaf stomatal conductance for six P. balsamifera genotypes: (a) AP-947 (b) AP-1005 (c) AP-1006 (d) AP-2278 (e) AP-2298, and (f) AP-2300. Midday stomatal conductance for well watered plants (shaded boxes) and plants grown under water-deficit conditions (white boxes) are represented. Asterisks indicate significant difference between well-watered and water-deficit-treated plants: *P ≤ 0.1; **P ≤ 0.05; ***P ≤ 0.001. WD, Water-deficit treatment.

Table 2.  Number of SFPs identified
Genotype 1Genotype 2Number of Probe Sets with 1 SFP
All probe sets (n = 61 313)Filtered probe sets not significant for Genotype main effect (n = 10 794)Filtered probe sets significant for Genotype main effect (n = 5123)
  1. Total number of single-feature polymorphisms (SFPs) were identified in all probe sets on the Affymetrix Poplar GeneChip using SNEP [P < 0.05; Fujisawa et al. (2009)]. Genes that were identified as significantly differentially expressed (FDR = 0.05; log2(FC) cutoff = 2.0) and genes whose expression is not significantly different among genotypes were surveyed for SFPs and the proportion was calculated based on the total number of probe sets examined, respectively.

AP-947AP-100511 66340.4040.52
AP-947AP-100612 11045.5945.81
AP-947AP-227813 65746.9047.18
AP-947AP-229810 94343.1662.01
AP-947AP-230010 60239.2239.72
AP-1005AP-10067 18537.8138.16
AP-1005AP-22785 05223.5524.54
AP-1005AP-22986 93436.7137.81
AP-1005AP-23006 75934.3234.86
AP-1006AP-22783 15812.2312.65
AP-1006AP-22983 78519.1820.11
AP-1006AP-23003 81518.8319.13
AP-2278AP-22985 06628.6729.22
AP-2278AP-23005 78629.0129.77
AP-2298AP-23004 65020.1120.03

Similar reductions in gs and RWC, as well as photosynthetic capacity, have been demonstrated in poplar and other plant species under drought stress conditions (Duan et al. 2005; Giovannelli et al. 2007); however, in poplar, suppression of gs and photosynthesis occur well before changes in whole leaf water status are observed (Tardieu & Simonneau 1998; Giovannelli et al. 2007). The regulation of stomatal conductance and water status in poplar may represent an important trait for survival under fluctuating environments, particularly the extreme stresses induced by episodic drought. The marked differences in gs at day 15 suggest variability in the regulation of stomatal control and physiological response to drought. Intraspecific variation in acclimation strategies in response to drought has been observed for many other tree species (Beikircher & Mayr 2009). Regulation of morphological and physiological parameters in response to water-deficit may provide insight into the hydraulic strategies of the various P. balsamifera genotypes. The observed variation in stomatal conductance, and other physiological and morphological variables in response to water-deficit, suggested that these genotypes deployed various drought tolerance and acclimation strategies. To test this hypothesis, the variation in the transcriptome response to water-deficit conditions among the six P. balsamifera genotypes was examined.

Water deficit conditions elicit significant responses within the Populus balsamifera transcriptome

Transcriptome analysis can provide insights into similarities and differences in the mechanisms underpinning the response to water-deficit between groups of individuals. Variation in the molecular mechanism regulating the drought response in Populus suggests that genotype plays an important role in shaping the drought transcriptome (Street et al. 2006; Wilkins et al. 2009). Comparison of the drought transcriptome of two Populus hybrid genotypes indicate that there is indeed a level of conservation in the transcriptome response; however, the variable response of a given genotype cannot be overlooked (Wilkins et al. 2009). In this study we hypothesize that conserved transcriptome level responses to drought will be observed, and that the differences observed in the drought transcriptomes that are specific to an individual genotype may provide valuable insight into the molecular basis of ecologically important variation in the drought response.

Using Affymetrix Poplar GeneChip microarrays, we investigated the transcript-level response to water-deficit among six P. balsamifera. Employing a multi-factorial ANOVA design (adjusted P ≤ 0.05), 280 probe sets reported on genes with significant differential transcript accumulation in response to water-deficit conditions, irrespective of the effect of genotype or sampling time (Fig. 4; Supporting Information Table S4). Many more probe sets were considered differentially expressed when no minimum threshold cutoff was applied; however, many of those probe sets had very low levels of differential accumulation of transcripts in response to water deficit. Filtering significant probe sets using a minimum threshold cutoff allows the identification of genes that are consistently differentially expressed and may prove to be more biologically meaningful (McCarthy & Smyth 2009). Consistent with previous findings among species and hybrids of Populus (Street et al. 2006; Wilkins et al. 2009), a larger proportion of genes (∼5000 probe sets) had significant differences in transcript abundance for the main effect of genotype relative to the effect of water-deficit treatment alone or the genotype-treatment interaction together. Genotype was known to play an integral role in shaping the drought response in Populus between species or hybrids (Street et al. 2006; Wilkins et al. 2009) The current study extends this finding, and highlights the importance of genotype in shaping the water-deficit response within a given Populus species.

Figure 4.

Heat map representing relative transcript abundance of all drought responsive probe sets in six P. balsamifera genotypes: AP-947, AP-1005, AP-1006, AP-2278, AP-2298 and AP-2300. Only probe sets that are significant for treatment main effect, irrespective of time of day or genotype, and are differentially expressed relative to a given threshold are represented (n = 280; FDR = 0.05, log2(fold-change)-cutoff = 2.0) for both time points: (a) midday, and (b) pre-dawn. Each column represents a biological sample, and all treatments are represented in triplicate replicates. Red indicates increased transcript abundance; green indicates decreased transcript abundance. Data are row normalized.

There is a common P. balsamifera drought transcriptome

Previously, identification of a common drought transcriptome in the genus Populus was challenging because of extensive variation in the transcriptome between hybrid genotypes, as well as variation in the transcriptome-level water-deficit response that is time of day dependent (Wilkins et al. 2009). By contrast, comparison of the drought transcriptome across six genotypes of P. balsamifera, at two time points, revealed a common transcriptome-level response to water-deficit treatment within this species (Fig. 4, Supporting Information Tables S3 and S4). The common response genes that were identified in this comparison had a significant change in transcript abundance in response to water deficit that was genotype independent (i.e. corresponded to main effect of treatment irrespective of genotype in ANOVA).

The functional roles of the probe sets that are significant for the treatment main effect for all genotypes (FDR = 0.05) and also show differential transcript abundance in response to water deficit according to a minimum threshold cutoff [log2 (fold-change) of 2.0] were classified using GO categories (Berardini, Mundodi & Reiser 2004) Supplementary Fig. 2). Overall, the largest proportion of probe sets with increased transcript abundance under water deficit conditions were those categorized as ‘other cellular processes’. By contrast, probe sets with decreased transcript abundance under water deficit conditions largely fell into the ‘protein metabolism’ category. Interestingly, for both probe sets, with increased and decreased transcript abundance under water deficit conditions, a large proportion were categorized as ‘response to abiotic or biotic stimulus or stress’, in keeping with their involvement in the water deficit response.

Many of the genes comprising the common water-deficit response, genotype-independent P. balsamifera transcriptome corresponded to genes previously identified as drought responsive in other plants (Kreps et al. 2002; Bray 2004; Street et al. 2006; Bogeat-Triboulot, Brosche & Renaut 2007; Wilkins et al. 2009). Of the 280 probe sets that were significant for the main effect of treatment irrespective of genotype, with a minimum log2 fold-change of 2.0, 29% corresponded to water-deficit-responsive genes identified in a similar experiment conducted with two hybrid poplar genotypes (Wilkins et al. 2009). In comparison to other high-throughput experiments examining the drought response in Populus (e.g. Brosche, Vinocur & Alatalo 2005; Street et al. 2006; Bogeat-Triboulot et al. 2007) a much more limited shared response was identified for P. balsamifera, as was observed by Wilkins et al. (2009).

Of the 98 probe sets that reported increased transcript abundance in response to water-deficit in this study, several of particular interest include those involved in the production of galactinol and stachyose, including GALACTINOL SYNTHASE and STACHYOSE SYNTHASE. The expression of genes encoding enzymes involved in the production of sugars from the raffinose family of oligosaccharides was also water-deficit responsive in hybrid poplar (Wilkins et al. 2009). Raffinose-derived oligosaccharides are believed to function as osmoprotectants during drought stress (Taji et al. 2002; Nishizawa, Yabuta & Shigeoka 2008). Increased transcript abundance of genes encoding enzymes that produce these compounds is consistent with a plant that is attempting to counteract water deficit. In keeping with a plant mounting a water-deficit response, the common P. balsamifera water-deficit transcriptome also included genes homologous to gene families in Arabidopsis with key roles in adjusting water balance in response to water-deficit including ABA RESPONSIVE ELEMENT BINDING FACTOR 4 (ABRE4) and EARLY RESPONSIVE TO DEHYDRATION 7 (Bray 2004). The phytohormone abscisic acid (ABA) has an extremely well established role in plant drought signalling (Bray 2004; Mahajan & Tuteja 2005). The increased transcript abundance of genes implicated into the ABA signalling pathway in P. balsamifera in response to water deficit emphasizes the central role of this compound in the drought response across diverse taxa.

A large number of probe sets with decreased transcript abundance in response to water-deficit in the common P. balsamifera water-deficit transcriptome were homologous to genes involved in cell wall modification, including pectin esterases and endoxyloglucan transferases. Decreased transcript abundance of these classes of genes is thought to decrease cell wall extensibility by promoting cell wall loosening or stiffening (Micheli 2001) and by controlling the cleavage of xyloglucan chains (Hyodo et al. 2003), respectively. Genes involved in cell wall modification have previously been identified as drought responsive in A. thaliana (Bray 2004). A large proportion of genes identified in the conserved P. balsamifera drought response have unknown function based on homology to other plant species. These genes that appear to be drought responsive in P. balsamifera may represent a specific water-deficit response for this species.

There is notable significant variation in drought transcriptomes across P. balsamifera genotypes

While a core set of probe sets representing a common species-level response to water-deficit treatment was identified, many probe sets reported differential expression between P. balsamifera genotypes. That is, when the water-deficit transcriptomes of the six P. balsamifera genotypes were compared using a Pearson correlation based on all drought-responsive probe sets (FDR = 0.05), some genotypes were more closely related than others with respect to their water-deficit transcriptome (Fig. 5). The similarity between all genotypes was still quite high, with minimum Pearson correlation coefficient values for any given pair-wise comparison >0.6. Genotypes AP-1005, AP-1006 and AP-2278 had the most similar drought transcriptomes; whereas, genotype AP-2300 was the most distinct with respect to the transcriptome response to water-deficit relative to the other genotypes.

Figure 5.

Pearson correlation coefficient (PCC) heat map representing the P. balsamifera drought transcriptome responses. Differential transcript abundance between well watered and water-deficit samples for the six genotypes for the drought responsive probe sets are represented (Treatment main effect; FDR = 0.05, log2(fold-change) cutoff = 2.0, n = 280 probe sets). Differential transcript abundance was calculated as the mean log2(fold-change) between well watered and water-deficit samples for a given probe-set. The PCC was determined for each pair-wise comparison, and is represented by the colour in the corresponding cell. All samples are represented on both the x- and y-axis, in the same order.

Various patterns of gene expression in response to drought were identified: genes with increased levels of transcript abundance for all genotypes, those with decreased transcript abundance across all genotypes and those with differential drought responsiveness between the six P. balsamifera genotypes. However, in the conserved set of probe sets that were drought responsive regardless of genotype (treatment main effect), notable differences in the mean log2 fold-change between well watered and water-deficit treated samples were observed. The magnitude of change in gene expression in response to water-deficit treatment varied considerably between genotypes (Fig. 6). For example, genotypes AP-1006 and AP-2278 had significantly larger fold-changes in transcript abundance levels relative to other genotypes for those genes with significant differences in transcript abundance in response to water deficit. This suggests that there was significant variation in not only transcript abundance of P. balsamifera genes that exhibited a significant treatment-genotype interaction, but also in the magnitude of intraspecific differences in gene expression for those genes whose change in transcript abundance was attributable to treatment main effect alone. This is to say that the variation in P. balsamifera water-deficit transcriptomes across the species is attributable to both qualitative and quantitative changes in transcript abundance.

Figure 6.

Box plot illustrating the interplay of genotype and treatment in shaping the drought transcriptome of six P. balsamifera genotypes. The average log2(fold-change) between well watered and water-deficit treated samples for all genes identified as significantly differentially expressed for treatment main effect (FDR = 0.05, log2(fold-change)-cutoff = 2.0, n = 280 probe sets) for probe sets with (a) decreased transcript abundance in response to WD and (b) increased transcript abundance in response to WD at the midday time point.

Qualitative variation was observed among the genotypes in the representation of genes in different functional categories in the drought transcriptomes. Qualitative differences (i.e. individual gene identities) in the transcriptomes of each genotype were analysed as a two-factor ANOVA, where treatment and time point were the two main factors. Genotype-specific responses to water deficit treatment emerged from these analyses (Supporting Information Table S3). Notably, across all genotypes, a large number of genes were predicted to be involved in protein metabolism, or response to biotic or abiotic stimulus in each genotype, similar to the results observed for the treatment main effect across all six genotypes (i.e. the common transcriptome). Nevertheless, across genotypes there was variation in the representation of given GO functional categories, with some GO categories more populated by drought transcriptome genes of some genotypes relative to other genotypes. This underscores the fact that there were qualitative differences in the nature of the drought transcriptomes across genotypes, with each genotype having a ‘GO fingerprint’ that was broadly similar to the other clones, but still relatively unique.

While natural variation in the transcriptome response to various environmental stimuli has not been documented in poplar, it has previously been described in A. thaliana (Kreps et al. 2002; Hannah et al. 2006; van Leeuwen et al. 2007). Variation in the transcriptome response to cold stress between various accessions of A. thaliana highlighted the complexities of such a response. These stress-induced alterations in gene expression suggested that not only differential expression of genes, but also the variation in the magnitude of expression is likely to influence the variation in acclimation capacity of these accessions (Hannah et al. 2006). Consistent with this, analysis of the A. thaliana salicylic acid response revealed extensive transcriptome variation, where relatively few genes responded similarly across the A. thaliana accessions (van Leeuwen et al. 2007). These findings suggested that A. thaliana ecotypes differentiate to a greater extent in terms of environmental responsiveness, such that each ecotype is well matched to local environmental conditions. The higher level of commonalities in the P. balsamifera water-deficit transcriptome response is likely a consequence of the relatively low population genetic variation found in P. balsamifera (Olson et al. 2010). The higher level of commonalities in the P. balsamifera water-deficit transcriptome response may be reinforced by the fact that a broad range, long-lived species, like P. balsamifera, must retain a more generalist response to environmental stimuli across its range. Future studies could test this hypothesis by examining the intraspecific variation in P. balsamifera transcriptome responses to a wider variety of environmental stimuli.

Time of day shapes the P. balsamifera drought transcriptome

The P. balsamifera water-deficit transcriptome is not only shaped by genotype, but also by the time of day. The interaction between time point and water-limitation revealed 129 probe sets with significant differences in transcript abundance (Supporting Information Table S3); however, when the interaction between the time of day and treatment was assessed within each individual genotype, particular genotypes, such as AP-1006, had a larger cohort of probe sets with differential transcript abundance relative to others, such as AP-2298. As previously observed with hybrid poplar genotypes, the transcriptome-level response to water-deficit conditions were influenced by time of day, and time of day was an important factor when considering the conserved drought response in Populus (Wilkins et al. 2009). However, in P. balsamifera, the time of day treatment interaction was less significant than that observed previously between the hybrid poplar genotypes (Wilkins et al. 2009). The magnitude of transcriptome differences observed with the hybrid poplar genotypes relative to that observed between the P. balsamifera genotypes may be attributable to transgressive effects. Transgressive effects have been observed in interspecific hybrids in other plant genera (Lai et al. 2006), and might be expected in the hybrid poplars, but would be lacking in the pure P. balsamifera genotypes.

The extent of transcriptome-wide transcript abundance change enables P. balsamifera to sustain growth under water-deficit conditions

This study demonstrates the complexities of the drought response within a given species of Populus. Genotypes with strong physiological responses to water-deficit conditions tended to have increased magnitude change in expression of genes that were significant for the treatment main effect (Fig. 6, Supplementary Fig. 1b). Genotype AP-1006 had the most rapid decline in stomatal conductance in response to the imposition of water-deficit conditions, and also showed the largest mean log2(fold-change) between well-watered and water-deficit treated samples for all probe sets significant for treatment main effect. Correlation between magnitude of cold tolerance and amplitude of gene expression [log2(fold-change)] has been observed among Arabidopsis accessions (Hannah et al. 2006). It has been hypothesized that increased capacity for cold acclimation may be related to the observed changes in the transcriptome; however, supporting evidence revealed that reduced cold acclimation in accessions with reduced capacity for cold tolerance was not supported by metabolic activity. This suggests that the mechanisms that form the foundations of complex phenotypic traits, such as cold tolerance, or drought acclimation are likely controlled by a large number of transcriptome changes, rather than individual genes.

Tree growth, another complex phenotypic trait, is also underpinned by genetic factors that respond to environmental stimuli (Grattapaglia et al. 2009). Consistent with this, the capacity of P. balsamifera to sustain growth during drought was positively correlated (R2 = 0.776, P = 0.02) with the magnitude of change in transcript abundance across the remodelled transcriptomes (Fig. 7). This suggests that it is not merely the nature of genes that enables plant growth during drought, but also the magnitude of change in transcript abundance for genes that are drought responsive. While most studies emphasise the importance of changes in the specific ‘cohort’ of genes expressed in response to a stress stimulus, the results presented here indicate that the magnitude of change in transcript abundance for all genes across the transcriptome is every bit as important in buffering the response. It is noteworthy that sustained growth under drought conditions and the magnitude of drought-induced, transcriptome-wide changes transcript abundance were the only two factors that showed a significant correlation in this study (Supplementary Fig. 1). This underlines the often overlooked role of magnitude of transcriptome-wide changes in transcript abundance as a capacitor for growth in response to key environmental stimuli, and provides a balanced counterpoint to the focus on the role of individual genes.

Figure 7.

The relationship between the magnitude change in gene expression and the difference in plant height between well watered and water-deficit treated P. balsamifera trees. Linear regression analysis revealed a significant relationship between these two variables (P = 0.02033). The coefficient of determination (R2) is shown in the figure panel. AP-947 (inline image), AP-1005 (▵), AP-1006 (+), AP-2278 (□), AP-2298 (○) and AP-2300 (×).

The extent of differences in drought-responsive transcriptomes between P. balsamifera clones positively correlated with the extent of intraspecific DNA sequence variation

The differences and commonalities in water-deficit-induced transcript abundance patterns may be attributable to sequence variants from one P. balsamifera genotype to another. One advantage of Affymetrix GeneChip data of the sort described herein is that probe-level data provide a relatively simple means by which to assess sequence polymorphisms between pairs of genotypes. Genome-wide sequence polymorphisms, known as single feature polymorphisms (SFPs) can be identified using Affymetrix GeneChip data (Luo et al. 2007; Gupta, Rustgi & Mir 2008; Fujisawa et al. 2009). When pair-wise comparisons of the single-probe level hybridisation data for the six P. balsamifera genotypes were examined, across all probe sets on the Affymetrix Poplar whole-genome GeneChip, the number of SFPs found in any given pair-wise comparison between genotypes varied from approximately 3100 to 13 000 (Table 2). Genotypes that appeared to be most divergent based on increased SFP occurrence between the genotypes also showed decreased commonalities in drought-responsive genes. For example, genotype AP-947 and AP-2300 were divergent with respect to their drought transcriptomes (Fig. 5) and also had >10 000 SFPs. Conversely, genotype AP-1006 and AP-2278 were more closely related with a high degree of similarity for genes significantly expressed in response to water-deficit, and also had the least number of SFPs between them.

Notably and importantly, transcript abundance differences observed between genotypes were not attributable to the number of SFPs identified for a given pair-wise comparison. The proportion of SFPs identified for any given pair-wise comparison was the same for genes with significant differences in transcript abundance in response to water-deficit in combination with genotype (i.e. they had a significant genotype-treatment interaction) and for those where genotype played no role in water-deficit-induced changes in transcript abundance (i.e. determined by treatment main effect only) (Table 2). These data are important in that they reveal that the differences in transcriptome observed between two genotypes were not merely attributable to differences in hybridization on account of sequence polymorphism, but conversely suggest that most of the intraspecific differences in water-deficit-induced changes to the P. balsamifera transcriptome were likely attributable to non-coding cis-acting sequences.

Intriguingly, the degree of relatedness between P. balsamifera genotypes, as defined by frequency of SFPs, did not correspond to the geographical origin of the genotypes. That is, pairs of genotypes that were acquired nearby had as many pair-wise SFP differences as pairs that were acquired from two very different locations. This finding has two important implications. First, inasmuch as the six genotypes reported here were representative of P. balsamifera, SFP-inferred relatedness does not reflect geographic distance between genotypes. Second, and more importantly, SFP-inferred relatedness corresponded to transcriptome-level relatedness for the water-deficit-induced transcriptome. That is, pairs of genotypes with fewer SFPs had more closely related transcriptome profiles; whereas, pairs with greater numbers of SFPs had more distinct transcriptomes. These findings suggest that genetic relatedness is likely to be an indicator of a shared water-deficit response. Moreover, the findings suggest that while local environments must play a role in the selection of specific phenotypic responses in P. balsamifera, the responses of some genotypes appear to be relatively robust across a large geographical distance for this species. Taken together, this suggests that both local adaptation and phenotypic plasticity might be underlying factors determining the wide geographical range of P. balsamifera.


Although there was a common, shared water-deficit-induced transcriptome level response for P. balsamifera, the amplitude of gene expression for the shared water-deficit transcriptome varied among genotypes. Larger changes in the absolute magnitude of transcript abundance for probe sets that were significant for treatment main effect were observed for genotypes that had more rapid declines in their physiological status in response to drought. Phenotypic traits, such as growth, are correlated with genetic responsiveness to drought. Genotypes that had the capacity to sustain growth under water-limitation also exhibited increased magnitude change in the remodelled transcriptome. Genotypes that had greater commonalities in their drought transcriptomes in response to water-deficit also had fewer SFP differences, suggesting that responses may be conserved across individuals that share a greater degree of genotypic similarity. Moreover, the lack of correspondence between pair-wise SFP differences and geographical origin between genotypes suggests that some genotype-derived responses are locally adapted, while others are spread widely on the landscape. Together, these findings better define within-species variation in the response of an important genus to a key environmental challenge, and raise testable hypotheses regarding the mechanisms underpinning the drought response in poplars, and how these shape the distribution of poplars on the landscape.


We are most grateful to Bruce Hall and Andrew Petrie for excellent greenhouse assistance, John McCarron for experimental set up, Joan Ouellette for technical assistance, and Dave Kamelchuk (Al-Pac) for collecting all the plant materials. We are also most grateful for incredibly useful comments on the draft manuscript provided by two anonymous reviewers. Research infrastructure and technical support was generously provided by the Centre for Analysis of Genome Evolution & Function at University of Toronto. OW was generously supported by a Natural Science and Engineering Research Council of Canada (NSERC) Canadian Graduate Scholarship (CGSD). SDM is a Canada Research Chair. This work was generously supported by funding from NSERC, the Canada Foundation for Innovation (CFI), and the University of Toronto to SDM, ALP and MMC.