Genotype and time of day shape the Populus drought response

Authors

  • Olivia Wilkins,

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

    1. Division of Signaling Biology, Ontario Cancer Institute, 101 College Street, Toronto, Ontario M5G 1L7, Canada
    Search for more papers by this author
  • Hardeep Nahal,

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

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

    Corresponding author
    1. Department of Cell & Systems Biology, University of Toronto, 25 Willcocks Street, Toronto, ON M5S 3B2, Canada
    2. Centre for the Analysis of Genome Evolution and Function, University of Toronto, 25 Willcocks Street, Toronto, ON M5S 3B2, Canada
      *For correspondence (fax +1 416 978 5878; e-mail malcolm.campbell@utoronto.ca).
    Search for more papers by this author

*For correspondence (fax +1 416 978 5878; e-mail malcolm.campbell@utoronto.ca).

Summary

As exposure to episodic drought can impinge significantly on forest health and the establishment of productive tree plantations, there is great interest in understanding the mechanisms of drought response in trees. The ecologically dominant and economically important genus Populus, with its sequenced genome, provides an ideal opportunity to examine transcriptome level changes in trees in response to a drought stimulus. The transcriptome level drought response of two commercially important Populus clones (P. deltoides × P. nigra, DN34, and P. nigra × P. maximowiczii, NM6) was characterized over a diurnal period using a 4 × 2 × 2 complete randomized factorial anova experimental design (four time points, two genotypes and two treatment conditions), using Affymetrix Poplar GeneChip microarrays. Notably, the specific genes that exhibited changes in transcript abundance in response to drought differed between the genotypes and/or the time of day that they exhibited their greatest differences. This study emphasizes the fact that it is not possible to draw simple, generalized conclusions about the drought response of the genus Populus on the basis of one species, nor on the basis of results collected at a single time point. The data derived from our studies provide insights into the variety of genetic mechanisms underpinning the Populus drought response, and provide candidates for future experiments aimed at understanding this response across this economically and ecologically important genus.

Introduction

Because of their long lives, trees are compelled to contend with fluctuating environmental conditions over their lifetimes. Water availability is amongst the most important of these environmental fluctuations (Shafroth et al., 2000). In fact, water availability is a major determinant of the geographical distribution of tree species, as it impinges on tree establishment, productivity and, ultimately, survival (Hogg et al., 2005; van Mantgem et al., 2009). Exposure to episodic drought can also result in long-term growth reduction in established forests, and in increased mortality among young trees grown in plantations (Hogg et al., 2005).

To contend with water limitation, trees must make appropriate physiological and developmental adjustments. As is the case with their herbaceous counterparts, such changes in physiology and development are, in part, underpinned by reconfiguration of the transcriptome. Transcriptome-level changes are a prominent feature of the drought response in herbaceous plant species, including Arabidopsis thaliana (Kreps et al., 2002; Seki et al., 2002; Kawaguchi et al., 2004), barley (Ozturk et al., 2002; Talame et al., 2007), sunflower (Roche et al., 2007), wheat (Mochida et al., 2006; Xue et al., 2008) and rice (Rabbani et al., 2003; Degenkolbe et al., 2008). Examination of transcriptome remodeling in these species has revealed commonalities, including drought-enhanced transcription of genes encoding late embryogenesis abundant (lea) proteins in rice, barley and A. thaliana (Ozturk et al., 2002; Bray, 2004; Degenkolbe et al., 2008), as well as species-level differences, such as pyruvate dehydrogenase kinase 1, which is drought-induced in rice, but not in A. thaliana (Rabbani et al., 2003). Such changes in transcript abundance underpin modifications in cellular and whole plant-level responses that enable plants to better contend with water deficit (Bray, 2004; Gutterson and Zhang, 2004). Given the importance of analogous drought response mechanisms in determining the health of forest ecosystems, and in shaping the productivity of plantations for industrial end uses, there is considerable incentive to better understand how transcriptome-level changes shape drought responsiveness in trees.

Populus species, which include poplars, cottonwoods and aspens, are ecologically important temperate trees that are also grown in plantations for the purposes of pulp, paper, wood and fuel production (Tuskan et al., 2006). Water availability plays a key role in determining the distribution of given Populus species within their native ranges (Graham et al., 2002). Populus species provide an excellent opportunity to examine transcriptome-level responses to drought stimuli in tree species, as there is considerable variation in the drought response within the genus Populus, in terms of both survival and biomass accumulation (Yin et al., 2004; Marron et al., 2006; Monclus et al., 2006). The molecular mechanisms underpinning these different responses can be dissected by making use of inherent biological features of the genus, coupled with post-genome sequence molecular tools that are available for Populus. For example, the simplicity of clonal propagation of Populus genotypes allows direct comparison of watered and drought-treated plant material within one genotype, and the completely sequenced Populus genome has enabled the development of molecular tools to examine the complete Populus transcriptome (Brunner et al., 2004; Tuskan et al., 2006).

Several recent studies have examined the relationship between genotype, transcriptome and drought response in trees of the genus Populus (Street et al., 2006; Bogeat-Triboulot et al., 2007; Caruso et al., 2008; Wilkins et al., 2009). In each of these studies, different Populus species and clones at different developmental stages were examined, different methods were used to expose the plants to ‘drought’, different tissues were collected at different times of day, different platforms were employed to investigate the drought response, and different methods were used to identify changing genes. In most of these studies, the drought response of a single species was investigated, and, in one, transcriptome modifications in response to drought were correlated with specific genotypes in a single Populus pedigree (Street et al., 2006). Although such studies begin to tease apart the within-family variation in drought response at a particular time point in this important genus, they only hint at the variation in transcriptome activity that is invoked in response to drought across the genus.

Here, we test the hypotheses that transcriptome-level responses to a drought stimulus vary between two commercially important hybrid Populus clones, NM6 and DN34, and that these responses are contingent on the time of day at which the transcriptome is examined. The drought response of hybrid poplar clones is of interest, as Populus hybrids are widespread in natural forests as well as in plantation culture (Brunner et al., 2004). The hybrid poplar clones NM6 and DN34 are used extensively in plantation culture in the Great Lakes region of North America, and are recognized for their relatively rapid and comparable growth rates, and sapling survival characteristics (Challen et al., 2001; Eckenwalder, 2001). DN34 (Populus × canadensis Moench ‘Eugenei’, Carolina poplar –P. deltoides × P. nigra) is the most prevalent Euramerican clone in cultivation in North America. NM6 (P. nigra × P. maximowiczii) is used extensively in plantation culture and in windbreaks around fields (Eckenwalder, 2001). Both clones are noted for being moderately drought tolerant (Hansen et al., 1994; Volney, 2004), but this superficial description of the response to water limitation may not reflect differences in the underlying molecular mechanisms. Here, we examine the transcriptome-level changes in response to drought over a diurnal time period, with the aim of better delineating the commonalities and differences in the drought response across the genus Populus.

Results and discussion

DN34 and NM6 have divergent physiological responses to drought

In response to water deficit, DN34 took 10 days (Figure 1a) to exhibit significant decreases in midday leaf water potential; whereas, NM6 took 12 days (Figure 1b). Although the two clones took nearly equivalent periods of time to show differences in midday leaf water potential in response to water limitation, stomatal closure in response to water limitation was notably different between the two clones (n = 9, Welch’s two-sample t-test, P < 0.05; Figure 2). Plants adjust the aperture of their stomatal pores to optimize gas exchange under changing environmental conditions. Cues from diurnal events (e.g. light availability) and episodic environmental fluctuations (e.g. decreased soil water) influence stomatal aperture. DN34 plants grown under water-limited conditions had decreased stomatal conductance during daytime time points, but were not significantly different from their well-watered counterparts at the night-time time points (n = 9 per time point, Welch’s two-sample t-test, P < 0.01; Figure 2a), indicating a reduced stomatal aperture. In contrast, NM6 trees grown under water-limited conditions were characterized by significantly lower stomatal conductance than their well-watered counterparts at all times measured (n = 9 per time point, Welch’s two-sample t-test, P < 0.01; Figure 2b).

Figure 1.

 Differences in midday leaf water potential for hybrid poplar clones (a) DN34 and (b) NM6 subjected to drought stress through water withholding. The circles in black represent data obtained for plants grown without water limitation; the circles in gray represent data for plants grown without water input for 12 days in the case of DN34 and for 14 days in the case of NM5. Each condition included measurements for nine plants, with one measurement per plant. Standard errors of the mean are presented.
Asterisks indicate a significant difference between well-watered and water-limited leaves: n = 9, Welch’s two-sample t-test, P < 0.05.

Figure 2.

 Differences in leaf stomatal conductance for hybrid poplar clones (a) DN34 and (b) NM6 subjected to drought stress through water withholding. Shaded boxes represent data obtained for plants grown without water limitation, and white boxes represent data for plants grown without water input for 12 days in the case of DN34 and for 14 days in the case of NM6. Each condition included measurements for nine plants, with one measurement per plant.
Asterisks indicate a significant difference between well-watered and water-limited leaves: n = 9 leaves, Welch’s two-sample t-test, P < 0.01.

On the last day of the water withholding experiment, the relative water content (RWC) of leaves experiencing both moisture regimes were determined at each of the four diurnal time points: in the middle of the night, before dawn, in the middle of day and late in the day. The RWC values were not significantly different between the well-watered and water-limited leaves at any time point in either genotype (n = 8 per treatment, Welch’s two-sample t-test, P < 0.05; data not shown). Studies in several other plant species have demonstrated that in response to drought, stomatal conductance and photosynthetic capacity are reduced well before alterations to whole-leaf water status are observed (Kopka et al., 1997; Miyashita et al., 2005). This may result in part because of the fact that stomatal closure is the product not only of reduced leaf water, but is also mediated by abscisic acid generated in the roots (Taiz and Zeiger, 1998). For this reason, stomatal closure and leaf water potential may be more closely correlated with plant perception of water availability. The observation that DN34 had reduced stomatal conductance only during the day, and, as such, during the highest water stress, and that NM6 had consistently reduced stomatal aperture, suggests that the two clones use different drought tolerance strategies. To test this hypothesis, remodeling of the leaf transcriptome in response to water deficit was examined for both clones.

DN34 and NM6 had divergent transcriptome-level drought responses

Trees of the genus Populus have highly variable and genotype-specific responses to a variety of environmental stimuli. For example, hybrid poplar clones display a broad range of within-genus susceptibility to the pathogen poplar mosaic virus (Smith and Campbell, 2004). Similarly, exposure to elevated carbon dioxide and ozone induces significant clone-specific responses (Oksanen et al., 2001). Notably, response to drought also shows extensive variation across the genus (Tschaplinski et al., 1998; Marron et al., 2002; Monclus et al., 2006). Investigations of the molecular mechanisms orchestrating these varied responses have only recently begun (Brosche et al., 2005; Plomion et al., 2006; Street et al., 2006; Bogeat-Triboulot et al., 2007; Caruso et al., 2008).

The microarray study was designed as a 4 × 2 × 2 complete randomized factorial anova (four time points, two genotypes and two treatment conditions). This allowed for the identification of genes with significant differential expression between the two Populus genotypes, independent of water status or time of day (main genotype effect), and genes that were differentially expressed in drought-treated samples, regardless of genotype or time of day (main treatment effect). For each genotype and time of day, it allowed for the identification of genes in which the drought response varied with the time of day (treatment:time interactions) or genotype (treatment:genotype interactions), as identified by significant interaction terms between these effects in the anova. Finally, it enabled the identification of drought responsive genes in each genotype at each time of day (treatment.genotype.time). The anova model was parameterized by group means with a manually defined sum to zero contrast matrix (see Experimental procedures and Table S1 for more detail) to attach statistical significance to specific contrasts of biological interest within genotype or time of day, as well as to the overall main effects.

A clustered heat map of Pearson correlation between all samples, based on whole-transcriptome profiles, revealed that samples clustered according to genotype before clustering according to treatment or time of day at which they were measured (Figure 3). This observation is borne out in the anova results where the main effect of genotype is statistically significant for an order of magnitude more genes (8198 probe sets) than is the main effect of treatment (201 probe sets), or any of the other interaction terms tested (Tables 1 and S2). Genotype has previously been identified as a major determinant of the transcriptome-level drought response in Populus in both ‘pure’ species and in an F2 mapping population (Street et al., 2006). The current study differs in that a statistical contrast has been conducted over a larger number of genes, and with a greater variety of treatments, within each genotype.

Figure 3.

 Pearson correlation heat map of hybrid poplar drought transcriptomes. The heat map includes 12 798 probe sets that passed through the initial filter step [minimum expression log2(100) on at least two microarrays and minimum interquartile range (IQR) of 0.5]. All samples are represented in the same order on both axes (left to right on the x-axis and bottom to top on the y-axis). The Pearson correlation coefficient (PCC) was determined for each pair of samples. The color of each cell indicates the PCC for the contrasted samples.
Samples are described by genotype (NM6, dark grey; DN34, light grey), time of day of sample collection (midnight, MN; pre-dawn, PD; midday, MD; late day, LD; darkest to lightest grey), and by water status (well-watered samples, dark grey; water-limited samples, white).

Table 1.   Number of probe sets with significant main effects or interactions
 Number of probe sets
NM6 and DN34 together
 genotype8198
 treatment201
 treatment.MN37
 treatment.PD57
 treatment.MD146
 treatment.LD14
 treatment:genotype interaction MN1
 treatment:genotype interaction PD33
 treatment:genotype interaction MD144
 treatment:genotype interaction LD359
DN34 alone
 treatment.DN342
 treatment.DN34:time (MN.PD) interaction2
 treatment.DN34:time (MN.MD)interaction131
 treatment.DN34:time (MN.LD) interaction0
 treatment.DN34:time (MD.PD) interaction212
 treatment.DN34:time (LD.PD) interaction2
 treatment.DN34:time (MD.LD) interaction76
 treatment.DN34.MN10
 treatment.DN34.PD4
 treatment.DN34.MD490
 treatment.DN34.LD0
NM6 alone
 treatment.NM6199
 treatment.NM6:time (MN.PD) interaction142
 treatment.NM6:time (MN.MD) interaction66
 treatment.NM6:time (MN.LD) interaction0
 treatment.NM6:time (MD.PD) interaction30
 treatment.NM6:time (LD.PD) interaction564
 treatment.NM6:time (MD.LD) interaction541
 treatment.NM6.MN44
 treatment.NM6.PD318
 treatment.NM6.MD38
 treatment.NM6.LD711

The drought transcriptomes of DN34 and NM6 are influenced by the time of day

The drought transcriptomes of DN34 and NM6 were strongly influenced by the time of day at which the samples were collected. The treatment main effect within each genotype was significant for a small number of genes (two in DN34 and 199 in NM6). However, the treatment contrast within each time point for each genotype was significant for a much larger number of genes (Table 1), indicating that even within a genotype, the genes differentially regulated by drought are not conserved across time points, but rather constitute time of day-dependent responses.

In DN34, the treatment:time-of-day interaction term was significant for a large number of genes. The extent of this significance was primarily determined by midday-specific drought-induced changes in transcript abundance (Figure 4a; Table 1). Although there was a larger number of genes with conserved drought responsiveness throughout the day in NM6 than in DN34 (Table 1), there was an even larger number of time of day-specific responses to drought. Whereas most of the observed drought response occurred at the midday time point in DN34, distinct time of day-specific transcriptome changes were observed at all times of day in NM6 (Figure 4b; Table 1). Notably, the times when the greatest variation occurred were dawn and late in the day, indicating that different transcriptome programmes were invoked in a time of day-specific manner.

Figure 4.

 Heat map representing the relative transcript abundance of drought-responsive genes in hybrid poplar clones (a) DN34 and (b) NM6. The probe sets included correspond to transcripts with a significant main treatment effect (treatment.DN34 or treatment.NM6, adjusted P < 0.05), without treatment:time-of-day interaction (black sidebar), or to transcripts with significant treatment:time-of-day interaction (red sidebar). Each column represents a discreet biological sample, and all treatments are presented as biological triplicate replicates. Red indicates higher, and green indicates lower, levels of transcript abundance. Data are row normalized.
W, well-watered samples; D, water-limited samples; MN, samples collected in the middle of the night; PD, samples collected before dawn; MD, samples collected in the middle of the day; LD, samples collected late in the day.

Time of day-specific transcript accumulation patterns were previously described in A. thaliana under a variety of environmental and genetic conditions (Harmer et al., 2000; Schaffer et al., 2001; Blaesing et al., 2005; Covington and Harmer, 2007; Covington et al., 2008; Michael et al., 2008a,b). Under conditions of sufficient water and nutrients, between 30 and 50% of transcripts detectable in the A. thaliana rosette showed significant and reproducible diurnal changes (Blaesing et al., 2005). Furthermore, under conditions of environmental stress, the circadian clock has been shown to modulate an organism’s response differently depending on the time of day at which the stimulus is applied via the time-dependent gating effect (Covington and Harmer, 2007). For instance, reduced levels of sugars, which have been identified along with the circadian clock as major determinants of diurnal regulation of gene expression, led to the altered expression of hundreds of genes involved in sugar metabolism in a coordinated diurnal fashion in A. thaliana (Blaesing et al., 2005). In A. thaliana, there are a limited number of transcriptome-level studies that directly examine stress responses through time: either at 24-h intervals (Kreps et al., 2002; van Leeuwen et al., 2007) or at multiple time points in a single light period (Kilian et al., 2007). Notably, none of these published studies directly examined the consequences of diurnal changes on the stress response.

The NM6 drought transcriptome suggests that the phase of the regular diurnal cycle of transcript abundance has been shifted by drought

In response to water deficit, a group of 57 probe sets reported altered diurnal patterns of transcript abundance in NM6 (Table S3). These probe sets were identified based on their transcript abundance patterns, which reported significantly decreased transcript accumulation in response to water limitation in leaves collected before dawn, and significantly increased transcript accumulation in response to water limitation in leaves collected late in the day (Figure 5). In each case, transcript accumulation in the water-limited leaves collected before dawn presages midday transcript abundance in both well-watered and water-limited leaves, and transcript accumulation in the water-limited leaves collected late in the day presages transcript abundance in the middle of the night under both water regimes (Figure 5). No probe sets were identified in DN34 that had an equivalent phase shift in the diurnal pattern of transcript abundance in response to drought as had been observed in NM6. The drought-induced phase shift in transcript abundance observed in NM6 emphasizes the importance of time of day in shaping stress responses in plants. Moreover, these findings suggest that the relevance and value of studies interrogating stress responses would be enhanced by the precise reporting of time during the diel cycle, when the data were collected (Brazma et al., 2001).

Figure 5.

 Drought shifts the phase of the regular diurnal cycle of transcript abundance of 57 genes in NM6. Each point represents the mean of three biological replicates. Data for each probe set are mean-centered. Well-watered samples are represented as blue lines, and water-limited samples are represented as red lines.

The interplay of genotype and time of day in shaping Populus drought transcriptomes is complex

Treatment:genotype interactions at each time of day and were significant for a total of 496 probe sets when all time points were considered (False Discovery Rate (FDR)-adjusted P < 0.05; Tables 1 and S3). The anova-derived treatment:genotype interaction terms were significant for the largest numbers of genes at the midday (144 probe sets) and late-day (359 probe sets) time points. One hundred and thirteen of the 144 probe sets that had a significant treatment:genotype interaction term in the midday time point were also identified as changing significantly in response to drought in genotype DN34 at midday (contrast treatment.DN34.MD); in contrast, only one was identified as changing significantly in response to drought in genotype NM6 at midday (contrast treatment.NM6.MD) (Table S2). None of the 359 probe sets that had a significant treatment:genotype interaction term late in the day were also identified as changing significantly in response to drought in genotype DN34 late in the day (contrast treatment.DN34.LD), whereas 324 were identified as changing significantly in response to drought in genotype NM6 late in the day (contrast treatment.NM6.LD) (Table S2). This suggests that time of day is a key factor in the determination of the drought transcriptome in both DN34 and in NM6, and, furthermore, that the time of greatest drought-induced transcriptome change occurs at different times of day in the two genotypes. The different responses were confirmed for five genes by quantitative RT-PCR. All genes showed the same trend in transcript abundance accumulation as was observed in microarray analysis (Figure S1).

Markedly, at times of day with the fewest drought-responsive genes (midnight and before dawn), a large proportion of those genes had a significant main treatment effect, but lacked the treatment:genotype interaction (Figure 6a,b; Table 1). By contrast, at times of day with the largest number of drought-responsive genes (midday and late day), the majority of those genes had a significant treatment:genotype interaction (Figure 6c,d; Table 1). The data suggest that commonalities in the poplar drought response (i.e. across the genus) are most likely to be identified at the midday time point. Nevertheless, the data presented here also suggest that significant differences in the drought transcriptome between genotypes are also likely to be manifest at the midday and late-day time points. Taken together, the data support the contention that midday and late-day time points might be fruitful points for the identification of a core Populus drought response, as well as the transcript abundance changes that contribute to genotype-specific differences in drought responsiveness.

Figure 6.

 Heat maps representing the relative transcript abundance of drought-responsive genes at four time points in hybrid poplar clones DN34 and NM6. The four time points are presented separately: (a) the midnight heat map includes 38 probe sets; (b) the pre-dawn heat map includes 80 probe sets; (c) the midday heat map includes 255 probe sets; and (d) the late-day heat map includes 366 probe sets. Included probe sets correspond to transcripts with significant main treatment effects (adjusted P < 0.05) at one time point (treatment.MN, treatment.PD, treatment.MD or treatment.LD), without a treatment:genotype interaction (black sidebar), or to transcripts with significant treatment:genotype interaction (adjusted P < 0.05) for the specified time (red sidebar). Each column represents a discreet biological sample, and all treatments are presented as biological triplicate replicates. Red indicates higher, and green indicates lower, levels of transcript abundance. Data are row normalized.
W, well-watered samples; D, water-limited samples.

Characterization of a Populus genus drought-responsive transcriptome

Defining a genus-wide drought response is challenging: not only because of the divergent responses between species, but also because of the variation within a single genotype over a diel period. Characterization of a conserved drought response across multiple genotypes must, therefore, either limit its scope to a specific time of day or consider the envelope of drought-induced transcriptomal changes at multiple time points. In the present study, probe sets significant for the treatment main effect at each time of day were identified (Table S3), and their expression levels in each treatment were examined (Figure 7). It is apparent that transcript abundance in many of these statistically identified probe sets were indeed changing in both genotypes and at multiple times of day. However, it is also apparent that attempts to identify a conserved drought response based on samples collected only at one time of day, for example midday, would grossly underestimate the level of conservation of drought responsive genes between the genotypes.

Figure 7.

 Relative transcript abundance of genes with conserved drought response in both DN34 and NM6. Probe sets with a significant main treatment effect for each time point are included (treatment.MN, treatment.PD, treatment.MD and treatment.LD). (a) Fifty-five probe sets were characterized by increased transcript abundance, and (b) 153 probe sets were characterized by decreased transcript abundance, in response to drought. Each line represents the relative transcript abundance of a single probe set: the left terminus of each line represents the relative transcript abundance in well-watered samples, and the right terminus of each line represents the relative transcript abundance of the water-limited samples. Each terminus represents the mean of three biological replicates. The relative expression for each probe set is shown for all time points. Data for each probe set are mean-centered.
W, well-watered samples; D, water-limited samples; MN, samples collected in the middle of the night; PD, samples collected before dawn; MD, samples collected in the middle of the day; LD, samples collected late in the day. Annotated lists of the probe sets included in this figure are included in Table S4.

Comparison of genes comprising the ‘conserved’ drought response in the present study against the drought transcriptomes and/or proteomes of several recently published Populus drought studies (Brosche et al., 2005; Plomion et al., 2006; Street et al., 2006; Bogeat-Triboulot et al., 2007) reveals a much more limited ‘conserved’ drought response (Table 2). The previous studies differed significantly from the present study, as well as from each other, on several important points: they used different species (P. euphratica, P. canadensis, P. trichocarpa and P. deltoides, F2-mapping population), different molecular platforms (POP1 array, P. euphratica spotted cDNA array, reverse RNA dot blot, MALDI-TOF-MS), different methods of inducing a drought response (PEG6000, withholding water, geographical differences), leaves at different developmental stages, or applied different levels of drought stress and different statistical approaches for identifying drought responsive genes.

Table 2.   Drought-responsive genes identified in DN34 and NM6 that have also been identified in other Populus drought transcriptome studies. The table lists the Affymetrix probe set identifier on the Poplar Whole Genome GeneChip, the homologous gene in the Arabidopsis thaliana genome, a short annotation of the A. thaliana gene and the Populus studies in which they have previously been published
Probe set identifierGene model nameA. thaliana gene IDAnnotation
  1. 1Caruso et al. (2008).

  2. 2Bogeat-Triboulot et al. (2007).

  3. 3Street et al. (2006).

Ptp.5681.1.S1_s_ateugene3.00150294AT1G684901Expressed protein
Ptp.1264.1.S1_s_at/ PtpAffx.44523.1.S1_a_atgw1.VIII.1883.1AT1G698802Thioredoxin H-type 8
Ptp.4965.1.S1_x_at AT3G093902,3METALLOTHIONEIN 2A
PtpAffx.249.695.A1_a_at/ PtpAffx.618.2.S1_a_at/ PtpAffx.618.2.S1_x_at/ Ptp.4965.1.S1_x_atgw1.I.1300.1/ gw1.I.1300.1/ estExt_fgenesh4_pg.C_2080003/ eugene3.01200081/AT3G093902,3METALLOTHIONEIN 2A
PtpAffx.2311.2.A1_atestExt_fgenesh4_pm.C_LG_I0800AT3G473402DARK INDUCIBLE 6
Ptp.3162.1.S1_at/ PtpAffx.1458.1.S1_atfgenesh4_pm.C_LG_II000817/eugene3.00140486AT3G618902HOMEOBOX PROTEIN 12

Of the 201 probe sets exhibiting a significant main treatment effect in the present study, 47 (32%) correspond to genes that have been identified in other high-throughput drought experiments conducted in A. thaliana or in Populus (Bray, 2002; Kreps et al., 2002; Kawaguchi et al., 2004; Street et al., 2006; Bogeat-Triboulot et al., 2007; Huang et al., 2008). Several of the genes models identified in the present study are homologous with genes in families in A. thaliana that are known to play roles in water balance, stress response and photosynthesis, including RESPONSIVE TO ABA 18 (RAB18) (Mantyla et al., 1995), GALACTINOL SYNTHASE (GolS) (Taji et al., 2002) and RESPONSIVE TO DEHYDRATION 22 (RD22). Four probe sets, corresponding to genes that are homologous with members of the PLASMA MEMBRANE INTRINSIC PROTEINS family (AtPIP2;2 and AtPIP2;7), had significant main treatment effects. PIPs represent the largest subgroup of aquaporins, membrane proteins that facilitate the movement of water across the plasma membrane, in the A. thaliana genome (Jang et al., 2004). PIPs are important for maintaining osmotic balance in cells, and some, including AtPIP2;2 and AtPIP2;7, have been shown to be downregulated in response to drought (Jang et al., 2004). Genes were annotated using the Annotation Batch Function in PopGenie (Sjordin et al., 2009).

It is striking that so few genes comprise a common ‘Populus drought response’. Nevertheless, this finding is consistent with a recent study that examined within-species variation in the transcriptome-wide response of A. thaliana to exogenously applied salicylic acid (SA) (van Leeuwen et al., 2007). This study determined that only a limited number of genes (38 of 3620 SA-responsive genes) had similar expression profiles across multiple A. thaliana ecotypes. Moreover, similar to the results with different hybrid poplar clones described here, different A. thaliana ecotypes varied significantly in the timing of transcriptome alterations at different times after the application of SA. Taken together, the findings with A. . thaliana and Populus emphasize the fact that time is an important dimension to consider when examining genotype-derived differences in transcriptome-level responses to a stimulus. For example, genotype-specific variation in the transcriptome might not arise merely from simple qualitative binary (on/off) switches in gene expression, but from more subtle shifts in the timing of the response.

Differences in experimental design between the present and past studies undoubtedly account, in part, for the small number of genes that are shared between the different analyses. However, the present study indicates that genotype and time of sample collection may equally contribute to the small number of genes identified in the intersection set of the different experiments. Future assessments of the Populus drought transcriptome will be aided by a broader view of the response, integrating genotype and time into the analyses. As such, the data presented herein provide a useful baseline for future studies aimed at the dissection of the Populus drought transcriptome in other genotypes, at other times of day, and relative to other developmental and stimulus-response processes. To better enable the use of this baseline, we have developed an expandable online tool to visualize transcriptome changes in response to drought in Populus. The tool builds on the PopGenExpress platform (Wilkins et al., 2009), and expands it to include tools to facilitate the examination of genetic networks. These tools enable the intuitive visualization of the expression of individual genes across experimental conditions (http://bar.utoronto.ca/efppop/cgi-bin/efpWeb.cgi?dataSource=PoplarTreatment; Figure 8), as well as the identification and visualization of genes that are co-expressed with a gene of interest across selected conditions (http://bar.utoronto.ca/eapop/cgi-bin/ntools_expression_angler.cgi; Figure S2), genes that are expressed according to a user-defined pattern (http://bar.utoronto.ca/eapop/cgi-bin/ntools_expression_angler.cgi; Figure S3) and patterns of transcript abundance in a user-defined group of genes (http://www.bar.utoronto.ca/ebpop/; Figure S4). These tools should better enable community-based analysis of the Populus transcriptome.

Figure 8.

 A PopGenExpress electronic fluorescent pictograph (eFP) displaying drought-induced transcript accumulation patterns of the Populus homolog of XERICO (Ptp.7110.2.S1_at) in DN34 and NM6. The relative transcript accumulation in response to drought is presented. In all cases, red indicates higher levels of transcript accumulation and blue indicates a lower level of transcript accumulation, relative to transcript accumulation levels in the corresponding well-watered sample.

Conclusion

This study demonstrates that it is not possible to describe a genus-wide drought transcriptome for Populus based on an examination of the drought transcriptomes of one or a few Populus genotypes. An attempt to do so would fail to capture the diverse molecular mechanisms employed by the genus in response to this recurrent environmental stress. Specifically, comparison of the drought transcriptomes of the two hybrid Populus clones used in this study indicates that although there are a number of conserved transcriptome-level changes between the genotypes, there are many more changes that appear to be particular to the drought response of one or the other genotypes.

Critically, this study also shows that drought-induced transcriptome changes are dependent on the time of day at which they were measured. Transcriptome measurements at a single time of day not only miss components of the drought transcriptome, but may also be misleading in terms of the degree of transcriptome-level response orchestrated by the organism. Full characterization of the envelope of transcriptome adjustments that occur in response to a stimulus should factor in the time of day. Such considerations will be crucial in attempts to associate differences in gene expression with allelic variation to explain the genetic basis of environmental response traits.

That genotype and time of day should influence the way in which an organism responds to its environment is almost intuitive. Despite this fact, there are few studies that have explored the interplay between stimulus response, genotype and time of day in any organism. This is surprising, given the prominent role that the combination of genotype and the way that it responds to a stress, contingent on the time of day, is likely to play in acclimation and, ultimately, survival of the organism in question. The data presented here provide a clearer picture of the role played by genotype and time of day in shaping the ability of a prominent tree genus to contend with a key environmental insult: drought.

Experimental procedures

Plant material

Unrooted stem cuttings of two hybrid Populus clones, DN34 (P. deltoides × P. nigra) and NM6 (P. nigra × P. maximowiczii) (L.A. Quality Products Ltd., Portage la Prairie, Manitoba, Canada), were imbibed in cold water for 48 h prior to planting. The plants were established and grown in Sunshine Mix number one (Sun Gro Horticulture, http://www.sungro.com) in tall opaque pots (1-m length; 10-cm diameter) in a climate-controlled glasshouse with a 16-h photoperiod, a maximum daytime temperature of 22°C, a minimum night-time temperature of 17°C and with a minimum daytime light intensity of 150 μmol m−2 s−1. All plants were watered with tap water every 3 days to restore the soil to field capacity. The trees were fertilized (20:20:20, N-P-K, 1.5 g L−1) at a rate of 600 ml plant−1 every 7 days, and were last fertilized 3 days before the onset of the water-withholding experiment.

Saplings were grown without water limitation for 6 weeks, at which time one half of each population was grown without further input of water, and the other half was grown as before, without water limitation. The leaf water potential of well-watered and water-limited trees was measured daily using a pressure bomb (Model 1000 Pressure Chamber Instrument; PMS Instrument Co., http://pmsinstrument.com) until such time as a statistically significant difference in midday leaf water potential was observed between the water-limited trees and the well-watered trees for three consecutive days – 12 days for DN34 and 14 days for NM6 (n = 9 plants per treatment per genotype, Welch’s two-sample t-test, P < 0.05). Once this difference was observed, the first fully expanded leaf was harvested from three trees in each population, both well-watered and water-limited, at four times in a 24-h period: pre-dawn (1 h before the lights were turned on), midday (middle of the light period), late day (1 h before the lights were turned off) and midnight (middle of the dark period). The leaves from each population were pooled and flash frozen in liquid nitrogen. This was repeated three times, such that there were three replicate samples collected for each treatment at each time point for each genotype. All experiments were performed such that physiological measurements (n = 9 individuals), including photosynthetic rates and stomatal conductances, via gas exchange, were captured simultaneously with the collection of the leaf tissue used for the transcriptome analysis using the LI-6400XT Portable Photosynthesis System (LI-COR Biosciences, http://www.licor.com).

RNA extraction and microarray hybridization

Plant material was ground to a fine powder under liquid nitrogen, and total RNA was extracted from each sample using the procedure described by Chang et al. (1993). RNA quality was determined electrophoretically. For each sample, 10 μg of total RNA was reverse transcribed, labeled and hybridized to the Poplar Genome Array, according to the manufacturer’s protocols (Affymetrix, http://www.affymetrix.com) at the Centre for the Analysis of Genome Evolution & Function at the University of Toronto. For each treatment, RNA was extracted from three replicate biological samples, and each was hybridized to a Poplar GeneChip. The Poplar Genome Array includes 61 251 probe sets representing more than 56 055 transcripts (Affymetrix). The GeneChip probe design is based on the Joint Genome Institute’s P. trichocarpa genome project’s predicted gene set v1.1 and all publicly available expressed sequence tag (EST) and mRNA sequences for all Populus species available through UniGene Build #6 and GenBank in spring 2005. The array queries transcripts derived from 13 Populus species in addition to the fully sequenced P. trichocarpa genome.

Microarray analysis

GeneChip data analysis was performed using the BioConductor suite (Gentleman et al., 2004) in r (a language and environment for statistical computing; http://www.R-project.org) (R Development Core Team, 2009) using the affy package (Gautier et al., 2004). All 48 microarrays were preprocessed together using GC-robust multi-array analysis (gcrma) (Wu et al., 2004). Expression data were filtered to remove probe sets with low expression and low variance across all arrays (genefilter, Gentleman et al., 2009; minimum intensity of 100, on a minimum of 10% of arrays, and minimum interquartile range of 0.5 on the log2 scale).

The preprocessed data were analysed as a 4 × 2 × 2 factorial complete randomized anova design (four time points, two genotypes and two treatments) using the limma (Linear Models for Microarrays) package (Smyth, 2005) in r (R Development Core Team, 2009). The use of limma differs from the standard linear model approach only in that it moderates the denominator of the F-statistic by applying an empirical Bayes smoothing to the standard errors based on the distribution of all genes, to provide more stable inference and improved power in microarray experiments (Smyth, 2004). The linear model was parameterized by group means with a manually defined sum-to-zero contrast matrix (see Table S1) to test directly for the contrasts of interest: the main and interaction effects overall, as well as treatment effects within each genotype and within each time of day, treatment:time-of-day effects within each genotype and treatment:genotype interactions within each time of day. Although the group means parametrization is less common in other fields of statistics than the Helmert or treatment contrast parametrizations, it is equivalent to using these parametrizations in combination with a contrast matrix to extract additional effects of interest, and it is implemented in limma with a convenience function to make complex contrasts more straightforward to parameterize. A Benjamini–Hochberg false discovery rate of 0.05 was applied to the output of all tests. All samples have been uploaded to the Gene Expression Omnibus (http://www.ncbi.nlm.nih.ov/geo/); accession number GSE15242.

Quantitative RT-PCR validation of transcript abundance

Five genes with differential patterns of expression across treatments were selected for qRT-PCR validation. Total RNA was prepared as described above, and 5 μg of each sample was used for cDNA synthesis from oligo(dT)18 with SuperScript II Reverse Transcriptase (Invitrogen, http://www.invitrogen.com), following the manufacturer’s instructions. Real-time qRT-PCR was performed using the iCycler iQ real-time PCR detection system (Bio-Rad, http://www.bio-rad.com), using the iQ SYBR green supermix (Bio-Rad). The relative transcript abundance for each gene was determined using the Pfaffl method (Pfaffl, 2001). Data were normalized to the geometric mean of two control genes, ACTIN-1 and ACTIN-5. Where it was possible, primer sets were designed to span an intron to prevent amplification of genomic DNA. Melt-curve analysis was performed following amplification to confirm the specificity of the amplification reaction See Table S5 for primer sequences.

Poplar gene expression (PopGenExpress) database expansion and integration with data mining tools

The drought transcriptome data set was integrated into the PopGenExpress platform (Wilkins et al., 2009). The PopGenExpress data mining tools were expanded to include Expression Browser, Expression Angler and Project Browser, based on the Bio-Array Resource database framework (Toufighi et al., 2005). Database tables were generated using MySQL to archive meta-information for the samples used to produce the transcriptome data, as well as to store the expression data themselves, after the MIAME convention (Brazma et al., 2001), as described previously (Wilkins et al., 2009). For the PopGenExpress database, Affymetrix data were normalized using the GCOS algorithm, with a Target Intensity (TGT) value of 500. Diagrammatic representations of poplar were generated and processed as described previously to create an input for the electronic fluorescent pictograph (eFP) browser (Wilkins et al., 2009).

Contributions

OW and MMC designed the research; OW performed the research; OW and MMC analysed the data; LW guided and conducted the statistical analyses; OW, HN, NJP and MMC devised and developed new analytical tools; OW and MMC wrote the manuscript with editorial assistance from NJP, LW and HN.

Acknowledgements

We are most grateful to Bruce Hall and Andrew Petrie for excellent glasshouse assistance, Joan Ouellette, Julia Romano, Sherosha Raj and Bronwyn Rayfield, for technical assistance, to Josephine McKeever for renderings for the eFP browser and to Tam McEwan for providing the hybrid Populus plant material. We are also most grateful for excellent suggestions and insights provided during the review process. 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). This work was generously supported by funding from NSERC, the Canada Foundation for Innovation (CFI) and the University of Toronto, to MMC.

Ancillary