A genome-wide screen for ethylene-induced Ethylene Response Factors (ERFs) in hybrid aspen stem identifies ERF genes that modify stem growth and wood properties

Authors


Summary

  • Ethylene Response Factors (ERFs) are a large family of transcription factors that mediate responses to ethylene. Ethylene affects many aspects of wood development and is involved in tension wood formation. Thus ERFs could be key players connecting ethylene action to wood development.
  • We identified 170 gene models encoding ERFs in the Populus trichocarpa genome. The transcriptional responses of ERF genes to ethylene treatments were determined in stem tissues of hybrid aspen (Populus tremula × tremuloides) by qPCR. Selected ethylene-responsive ERFs were overexpressed in wood-forming tissues and characterized for growth and wood chemotypes by FT-IR.
  • Fifty ERFs in Populus showed more than five-fold increased transcript accumulation in response to ethylene treatments. Twenty-six ERFs were selected for further analyses. A majority of these were induced during tension wood formation. Overexpression of ERFs 18, 21, 30, 85 and 139 in wood-forming tissues of hybrid aspen modified the wood chemotype. Moreover, overexpression of ERF139 caused a dwarf-phenotype with altered wood development, and overexpression of ERF18, 34 and 35 slightly increased stem diameter.
  • We identified ethylene-induced ERFs that respond to tension wood formation, and modify wood formation when overexpressed. This provides support for their role in ethylene-mediated regulation of wood development.

Introduction

It is well established that treatment with ethylene (or ethylene precursors) can affect different aspects of wood formation. These include cambial growth, morphology of xylem cells, and ontogenesis of vessels, fibres and rays (Little & Savidge, 1987). Recent investigations using ethylene-insensitive hybrid aspen (Populus tremula × tremuloides) demonstrated that these responses are mediated through ethylene signalling (Love et al., 2009). In angiosperm trees, ethylene is produced during tension wood (TW) formation (Andersson-Gunnerås et al., 2003; Du & Yamamoto, 2003). TW is formed in response to gravitational and/or mechanical stimuli in leaning stems and is characterized by increased cambial growth, smaller and less frequent vessel elements, and in many species a cellulose-enriched gelatinous cell wall layer (G-layer) is formed inside the S layers in fibres (Scurfield, 1973; Hellgren et al., 2004; Felten & Sundberg, 2013). Because ethylene treatment mimics many TW characteristics it seems plausible that endogenous ethylene is a key mediator of the TW response. Indeed, the role of ethylene as an endogenous stimulator of cambial growth during TW formation was confirmed using ethylene-insensitive hybrid aspen (Love et al., 2009).

From work in Arabidopsis it is known that the ethylene signal is perceived in the cell by endoplasmic reticulum-localized ethylene receptors. Ethylene binding inactivates the receptor, which in turn de-represses the ethylene signalling pathway (Bleecker et al., 1998; Cancel & Larsen, 2002). Ethylene signal transduction downstream of the receptors involves phosphorylation, proteasomal degradation and nuclear translocation of several regulatory modules (Stepanova & Alonso, 2009; Qiao et al., 2012). This cascade leads to the stabilization of transcription factors EIN3 and EIN3-like1 (EIL1) (Chao et al., 1997; Solano et al., 1998; Guo & Ecker, 2003; Potuschak et al., 2003; An et al., 2010). EIN3 and EIL1 activate or repress expression of target genes by binding their promoters (Solano et al., 1998; Chen et al., 2009; Zhong et al., 2009; Boutrot et al., 2010; Zhang et al., 2011). The Ethylene Response Factor (ERF) family is a large gene family (Alonso & Stepanova, 2004) believed to be part of the EIN3/EIL1 targets, and the EIN3/EIL1-protein–ERF-promoter interaction has experimentally been shown for a few ERF members (Solano et al., 1998; Zhang et al., 2011; Shi et al., 2012).

Initially four ERF proteins (originally named ethylene-responsive element (ERE) binding proteins (EREBPs)) in tobacco and one in Arabidopsis were identified in a screen for nuclear proteins that bind to the 11-bp cis-element GCC box in the promoters of ethylene responsive genes (Ohme-Takagi & Shinshi, 1995; Büttner & Singh, 1997). The domain in ERFs that interacts with the GCC box typically consists of 58–59 amino acids (Ohme-Takagi & Shinshi, 1995), and is called the ERF domain. ERFs belong to the APETELA2 (AP2)/ERF superfamily, which share common structures within the conserved AP2 and ERF domains. The AP2/ERF superfamily is divided into AP2, RAV and ERF families depending on the number and specific sequence of their DNA binding domain(s): AP2 family proteins contain two AP2 domains (Riechmann & Meyerowitz, 1998), RAV family proteins contain one AP2 and B3 domain (Swaminathan et al., 2008), and ERF family proteins contain one AP2 and ERF domain (Fujimoto et al., 2000; Sakuma et al., 2002). The amino acid residues of the ERF domain that serve as the DNA binding site for the GCC box are not present in AP2 and RAV proteins (Allen et al., 1998; Fujimoto et al., 2000; Sakuma et al., 2002). The ERF family is further divided into the ERF and the Dehydration Response Element Binding (DREB) subfamilies. DREB proteins have a high affinity to dehydration-responsive elements (DRE) or C-repeat elements (CRT) found in genes involved in response to drought, temperature or salt (Stockinger et al., 1997; Liu et al., 1998; Nakano et al., 2006), whereas ERFs rather bind GCC box motifs, and variants of those such as the Jasmonate and Elicitor Response Element (JERE) (Lorenzo et al., 2003).

The presence or absence of an activation (EDLL-motif) or repression (EAR-motif) domain in individual ERF proteins influences whether they act as transcriptional activators or repressors (Fujimoto et al., 2000; Ohta et al., 2001; Tiwari et al., 2012). Another level of complexity is generated by the fact that different ERFs show distinct temporal ethylene responsiveness (Oñate-Sánchez & Singh, 2002), and that the regulation of certain ERFs may depend on the expression of other ERF members (Oñate-Sánchez et al., 2007). This evokes the possibility of ERF signalling cascades with primary and secondary acting ERFs. Despite the fact that many ERFs are ethylene responsive, it should be noted that they can also integrate other hormonal signals such as jasmonic acid, auxin, abscisic acid and cytokinin (Finkelstein et al., 1998; Menke et al., 1999; Fujimoto et al., 2000; Lorenzo et al., 2003; Rashotte et al., 2006; Hirota et al., 2007; Zhang et al., 2008; Rashotte & Goertzen, 2010).

A recent study showed that Arabidopsis ERF1, ERF018 and ERF109 were required for the increase in cambial cell division observed in ethylene overproducing eto mutants providing evidence that ERFs are important for ethylene signalling in vascular tissues (Etchells et al., 2012). However, ERF family members are also involved in many other aspects of growth and development, and biotic and abiotic stress responses (Solano et al., 1998; Kannangara et al., 2007; Pré et al., 2008; Licausi et al., 2010). Here we have utilized the genomic sequence of the black cottonwood (Ptrichocarpa) (Tuskan et al., 2006) to comprehensively identify the whole Populus ERF gene family and update the information on the ERF phylogeny described earlier (Zhuang et al., 2008). We report expression profiles of this family in response to long- and short-term 1-aminocyclopropane-1-carboxylic acid (ACC) and ethylene treatments in stem internodes of hybrid aspen, and during TW formation in aspen (Ptremula). Additionally, we have overexpressed selected ethylene responsive ERFs and demonstrate their potential involvement in the regulation of stem diameter growth and wood chemistry.

Materials and Methods

Populus genomics

The genome mining of the Populus trichocarpa ERF family was conducted by BLAST search using Arabidopsis, rice and tomato ERF domain variants as an electronic probe for the v1.1 genome assembly of the U.S. Department of Energy Joint Genome Institute (DOE-JGI). Thereafter, the updated (v2.0) genome assembly at Phytozome was screened for Populus ERFs by matching a regular ‘R.R.{6,7}E.[RK].{6,14}R.W’ expression to translated proteins using R programming language (R Development Core Team, 2011) and with Pfam search against automatic annotations (Pfam id PF00847 for AP2 domain). In addition, the most recent genome assembly (v2.2 and 3.0) (v3.0; DOE-JGI, http://www.phytozome.net/poplar) were re-scanned with Pfam search. The phylogenetic grouping was performed by narrowing down the Populus ERFs by classifying the protein coding sequences into observed phylogenetic subgroups along with Arabidopsis and rice ERFs, and further confirming the presence of the specific motifs determining the subgroup identity as described by Nakano et al. (2006). Multiple sequence alignments of the whole Arabidopsis, rice and Ptrichocarpa ERF proteins (according to Phytozome; Ptrichocarpa v3.0 was used as a default unless otherwise indicated in Table S3) for each subgroup were conducted with CLUSTALW v2.1 (Kyoto University Bioinformatics Center; http://clustalw.genome.jp/). Phylogenetic and molecular evolutionary analyses (Neighbor-Joining, Poisson distribution model, 1000 bootstrap replicates) were conducted using MEGA v4.0 (Tamura et al., 2007).

Plant materials

Hybrid aspen (Populus tremula L. × P. tremuloides Michx.; clone T89) was used in all experiments with the exception of TW studies that were carried out with 10-yr-old forest-grown P. tremula L. In vitro experiments were carried out exactly as described in Love et al. (2009). Briefly, ACC (Sigma-Aldrich) treatment was performed with in vitro cultured plants and initiated at a standard plant height of 80 mm. ACC was delivered to the Murashige and Skoog (MS) medium to achieve a final concentration of 100 μM in the medium. This concentration has previously been found to be optimal for ACC-induced growth (Love et al., 2009). A 20-mm piece of the stem below the fourth internode from the apical meristem was sampled following a 10-h treatment. Ethylene treatments were performed with c. 2-m tall trees as described by Love et al. (2009). Stem portions (63 mm in length, internode 42 from the apex) of experimental trees were simultaneously exposed with a flow-through cuvette system to 2 ppm of ethylene and synthetic air with 350 ppm carbon dioxide with appropriate controls. Transgenic trees (pLMX5::ERF-lines overexpressing selected ERFs) for glasshouse experiments were grown in a mixture of commercial soil-sand-fertilizer mixture (Yrkes Plantjord Kronmull; Weibulls Horto, Hammenhög, Sweden) under 18-h photoperiod at 20°C : 15°C (light : dark). Trees were fertilized with c. 150 ml of 1% Rika-S (N/P/K 7:1:5; Weibulls Horto) once a week. The transgenic trees were grown to c. 1.7 m with weekly rotation to minimize positional effects. Samples for analyses were collected from the lowermost 10–30 cm of the stem (measured from soil) at harvesting. A 2-cm piece was removed from the harvested material, peeled and cleaved to remove bark and pith, freeze-dried for 48 h and used for Fourier transform infrared (FT-IR) analysis. For TW analysis forest-grown trees were selected at a natural stand near Umeå, Sweden (63°50′N, 20°20′E) and TW formation was induced by bending the trees to c. 45° for 3 wk during the most active period of cambial growth. Stem pieces from mid tree height were harvested and frozen in liquid N2. For normal wood (NW) controls, stem pieces from the corresponding position of upright trees were harvested.

RNA extraction and cDNA synthesis

Total RNA from the in vitro-grown samples was isolated as described in Love et al. (2009), and the mRNA was further amplified according to manufacturer's protocols (MessageAmpl aRNA amplification kit; Ambion). Transgenic ERF overexpressor grown in vitro were sampled, frozen in liquid N2, and 2-cm internode samples from the base were homogenized with automated Precellys 24 (Bertin Technologies, Montigny-le-Bretonneux, France). Whole internodes from ethylene treatment (flow-through cuvettes) were harvested, frozen in liquid N2 and ground with mortar and pestle. Total RNA was extracted according to Carpenter & Simon (1998), and after the lithium chloride precipitation, RNA isolation was continued as described by Chang et al. (1993). The extracted total RNA was DNase treated (Turbo DNA-free™ Kit; Ambion), and quantified with a ND-1000 NanoDrop spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA). cDNA synthesis was performed on 2 μg of total RNA with SuperScript III (Invitrogen) reverse transcriptase according to manufacturer's instructions and cDNA was purified with a QIAquick PCR purification kit (Qiagen).

A 10–30-cm stem piece from the pLMX5::ERF overexpressors grown in the glasshouse was harvested and frozen in liquid N2. The bark was peeled off and the exposed developing wood tissue was scraped around the stem with a scalpel directly into liquid N2 (Gray-Mitsumune et al., 2004). Bark was peeled off from the TW side of the stem of trees forming TW and developing wood and phloem/cambium (inner surface of the bark) were scraped with a scalpel into N2. For NW, phloem/cambium and developing wood were scraped around the stem after peeling off the bark. Tissues were ground to fine powders using a mortar and pestle. RNA was extracted using Aurum Total RNA mini kit (Bio-Rad) and DNA was removed with the DNA-free™ kit (Ambion). The concentration was quantified with a NanoDrop (NanoDrop Technologies) and quality was assessed by gel-electrophoresis or with an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). cDNA was synthesized from 1 μg (for TW and NW material) or from 250 ng (for glasshouse material) of total RNA in a 20 μl reaction using iScript cDNA Synthesis kit (Bio-Rad).

Transcript profiling

Real-time quantitative PCR (qPCR) analysis was performed with SYBR Green and gene-specific primers for the whole Populus ERF gene family (Supporting Information Table S1). Three biological replicates from ACC and ethylene treatments were pooled into one sample, and the cDNA pool combined from several cDNA synthesis reactions was used due to multiple qPCR runs/plates for each experiment. Appropriate reference genes (Table S1) were selected using GeNorm (Vandesompele et al., 2002) (TW experiment) or by verifying stable Cq numbers with equal amounts of cDNA in qPCR reactions (ACC and ethylene experiments). Both methods estimated ACTIN 1 as the most stable reference gene. For ACC and ethylene experiments, 20–50 ng of cDNA template/reaction with three technical repeats for each ERF gene was used. Mastermix Plus for SYBR green I (Eurogentec, Seraing, Belgium) mix was used with the ABI 7000 (Applied Biosystems, Darmstadt, Germany) qPCR machine with 5 pmol of each primer. The PCR cycles were: initial denaturation at 95°C 10 min; denaturation at 95°C for 15 s, primer annealing and extension at 60°C for 1 min for 40 cycles. LightCycler® 480 SYBR Green I Master (Roche Diagnostics GmbH, Mannheim, Germany) mix was used with the LC4800 (Roche Diagnostics) qPCR machine with 5 pmol of each primer. The PCR cycles were: initial denaturation at 95°C 10 min; denaturation at 95°C for 15 s, primer annealing at 60°C for 30 s, and extension at 72°C for 1 min for 40 cycles. The average Ct value of the ACTIN 1 gene was subtracted from the corresponding Ct value of each ERF gene to obtain a normalized ΔCt value. The relative expression levels compared to control were calculated with the formula 2−ΔCt (User Bulletin 2, Applied Biosystems). The ACTIN 1 reference gene was not affected by any of the ethylene/ACC treatments, verified by the equal concentrations of control and ACC/ethylene-treated cDNA templates in qPCR reactions. The qPCR conditions for in vitro-grown pLMX5::ERF-lines were as described above for the LC4800 qPCR machine. For the TW experiment and glasshouse-grown pLMX5::ERF-lines, qPCR was carried out on 10 (TW and NW trees) or five (ERF-lines) times dilutions of cDNA template using a Bio-Rad CFX96 Real Time System with 5 pmol of each primer. The PCR cycles were: initial denaturation at 95°C for 3 min, denaturation at 95°C for 10 s, primer annealing at 60°C, and extension at 72°C for 30 s. For the TW experiment, qPCR templates comprised three biological replicates (each a pool of scrapings from two trees, i.e. in total 6 trees) for NW, and two biological replicates (each a pool of scraping from two trees, i.e. in total 4 trees) for TW. For the ERF-lines, three plants per line and 4–5 wild-type plants were analysed with qPCR. ∆Cqs were calculated for all samples. For every sample each ∆Cq was divided by the average of ∆Cq obtained in wild-type/controls for the corresponding ERF to set gene expression to a value of 1 in wild-type/controls. Detailed information about sample preparation, PCR conditions and calculation procedures are resumed according to MIQE précis guidelines (Bustin et al., 2010) in Table S2.

Generation of ERF overexpressor lines

Genomic DNA of Ptrichocarpa was extracted (Lodhi et al., 1994), and the protein coding sequence (with introns if present) of each ERF gene was amplified (Phusion DNA Polymerase, Finnzymes, Espoo, Finland) by PCR using gene-specific primers (Table S1). Each primer pair contained the attB1 and attB2 sites for the Gateway (Invitrogen) recombination reaction. The amplified attB-product was cloned into donor vector pDONR221 to create an entry clone, and further transferred into the pLMX5 destination vector (Love et al., 2009) to create an overexpression vector with pLMX5::ERF fusion. Intact inserts were confirmed by sequencing. The expression vectors were transformed into Agrobacterium strain GV3101 pMP90RK and hybrid aspen was transformed and regenerated as described by Nilsson et al. (1992).

Data processing

Heatmaps for ERF expression and Boxplots for pLMX5::ERF growth data were generated using the ‘heatmap’ or ‘boxplot’ function, respectively, in R (R Development Core Team, 2011). For calculating statistical differences in the growth phenotypes of the ERF overexpressors we used a nonparametric Mann–Whitney U-test in R using the ‘wilcox.exakt’ package with tied ranks for P-value estimation. The choice of a Mann–Whitney U-test was motivated by the fact that all lines per overexpressed ERF were pooled into one class (heterogeneous population) and compared to the wild-type plants of the respective growth batch.

FT-IR spectroscopy analysis

A 2-cm, freeze-dried wood piece was ground to powder in 10 ml stainless steel jars at 30 Hz for 70 s, using a bead mill (MM400; Retsch, Haan, Germany). Ten mg wood powder was mixed with 390 mg potassium bromide (KBr, infrared spectroscopy quality; Merck, Darmstadt, Germany) as an IR-transparent diluter and ground with mortar and pestle before measurements. FT-IR spectra were recorded under vacuum (4 mbar) conditions, using a Bruker IFS 66v/S spectrometer (Bruker Optik GmbH, Ettlingen, Germany) equipped with a diffuse reflectance 16-sample holder carousel accessory (Harrick Scientific Products, Pleasantville, NY, USA). In each carousel run, a background (pure ground KBr) was recorded. Spectra were collected with 128 scans at a resolution of 4 cm−1 in the region 400–5200 cm−1. Spectra were converted to data point tables using OPUS (v7.0.122; Bruker Optik). Spectra were baseline corrected using a 3-point (480, 784 and 1880 cm−1) linear baseline and total sum (area) normalized over the 480–1880 cm−1 region, using custom-built scripts (MATLAB v7.0; MathWorks, Natick, MA, USA) (Stenlund et al., 2008). The 480–1880 cm−1 region of the baseline corrected and area normalized spectra from wild type and transgenic trees were UV-scaled and subjected to multivariate analysis (OPLS-DA) (Trygg & Wold, 2002), using SIMCA-P+, v12.0 (Umetrics AB, Umeå, Sweden). OPLS-DA separates the systematic variation of pre-defined classes (here genotypes) in the FT-IR spectroscopic data into two parts: predictive and orthogonal components, which can be visualized in a Scores plot. The predictive ability of the classes in the Scores plot is given by the Q2-value. For a set of components and observations (i.e. samples) applies: the higher the Q2 value the better defined and the more distinct are the chemotypes in the classes. Predictive components are correlated to the predefined classes and can originate, for example, from cell wall chemotypes. Orthogonal components are not correlated to the predefined classes and are caused by experimental (biological or technical) variation. The contribution of different FT-IR bands to the separation of the classes can be identified from a correlation-related Loadings plot of the predictive component.

Results

Genome mining identified 170 ERF gene models in the Populus trichocarpa genome

In order to study ethylene-induced expression of ERF genes in the Populus genome, we first identified and curated the Ptrichocarpa ERF gene models for phylogenetic analysis and design of gene-specific qPCR primers. Mining and annotation of the Populus trichocarpa v1.1 genome assembly had identified 175 ERFs, of which 172 were full length and predicted to be true ERFs (Tuskan et al., 2006). Based on these annotations, further analysis of the Populus ERF gene family in v1.1 was presented by Zhuang et al. (2008). Re-mining and curation of the assembly v2.0 and v2.2 of the Populus genome in the Phytozome database identified 173 ERF gene models. Compared to v1.1, three new true ERFs were identified and seven ERFs were removed. Four of the 173 gene models were excluded because they were present in very short scaffolds not incorporated into the whole genome assembly, and they were virtually identical to four Populus ERFs already included in the remaining group of 169 (for details see Table S3). Mining of the assembly v3.0 identified still one new ERF, which is almost identical (99%) to ERF167. Thirty-nine percent of the v2.2 gene models differed from the v1.1 models and 33% of the v3.0 gene models differed from the v2.2 models, mainly due to allelic variation. All Ptrichocarpa locus names/gene models used are presented in Table S3, and the gene models that vary between consecutive assembly versions are indicated. To avoid gaps in the ERF nomenclature, a new generic number was assigned to each of the 170 identified Populus ERFs. As a result of automated annotation, some ERF gene models in the Ptrichocarpa v2.2 and v3.0 genome assemblies in Phytozome are truncated, and the predicted gene models are not optimal and/or contain sub-optimal EST support. Therefore, we created a manually curated version of ERF gene models publicly available at the PopGenIE database (http://www.popgenie.org/).

Alignment of the 170 Populus ERF proteins showed that within the ERF domain the residues Gly4, Arg6, Arg8, Glu16, Arg/Lys18, Arg26, Trp28, Leu29, Gly30 and Ala38 were conserved when compared to the GCC box binding domain in Arabidopsis ERF1 (At4 g17500; Allen et al., 1998) (Fig. S1). The exceptions were PtiERF10 where Arg8 was substituted with Lys, PtiERF92 where Arg6 was lacking, and PtiERF154 where Leu29 was substituted with Val. Despite the variation, these three Populus ERFs were considered true ERFs due to their very high homology with the other Populus ERF proteins.

The previous classification of Ptrichocarpa v1.1 ERF proteins by Zhuang et al. (2008) was conducted using the ERF classification by Sakuma et al. (2002) based on the ERF domain (groups A-1 to A-6 for DREB subfamily and B-1 to B-6 for ERF subfamily). However, the ERF classification by Nakano et al. (2006) considers all domains present in the whole ERF protein, and therefore results in more accurate phylogenetical classification with additional subgroups of ERFs. Thus, we classified the Populus ERFs, together with Arabidopsis and rice ERFs, according to protein coding sequences into phylogenetic subgroups as defined by Nakano et al. (2006) (Fig. S2). For clarity, the preceding ERF classification according to Sakuma et al. (2002) is also indicated in our phylogenetic trees that are consistent with the phylogenetic analysis of v1.1 ERF proteins (Zhuang et al., 2008). Of the 170 ERFs, 93 fall into the ERF-subfamily (divided into seven groups) and 77 fall into the DREB-subfamily ERFs (divided into five groups). However, our phylogenetic analysis suggests a re-evaluation of group V, which was split into two independent sub-clusters with DREB (subgroup Va) and ERF (subgroup Vb) subfamily proteins (e in Fig. S2). Subgroup Vb forms a cluster with 11 Populus ERFs and only one Arabidopsis orthologue and three rice orthologues. It can be noted that none of the Populus ERFs clustered into the groups Xb-L, XI, XII, XIII or XIV.

Transcript profiling identified ERFs that respond to ACC and/or ethylene in hybrid aspen stem tissues

In order to identify candidate ERF genes involved in ethylene signalling in stem tissues of hybrid aspen, the expression of ERFs was assayed after ACC or ethylene treatments. For this purpose we designed gene-specific primers and used qPCR in order to detect ERF abundance (Table S1). ERF gene expression was determined in three independent experiments using one of two different types of plant materials produced as described in Love et al. (2009); (1) a 10-h ACC treatment (100 μM) supplied to the medium of in vitro-grown 8-cm-high hybrid aspen plantlets, and (2) a 24-h and (3) a 14-d ethylene exposure using flow-through-cuvettes covering a 63-mm stem segment of 2-m-tall glasshouse-grown hybrid aspen trees. Despite differences of the substances applied, growth conditions, and tree size between the in vitro and glasshouse experiments, both conditions resulted in typical, well-documented ethylene responses in wood formation, including stimulated cambial growth and altered xylem anatomy (Love et al., 2009). This experiment was designed to provide a general picture of genome-wide ERF expression in stem tissues under ACC and ethylene-induced and noninduced conditions differing in time and type of plant material. As a trade-off to the high sample number only one biological replicate was used (consisting of material from three pooled trees), and therefore conclusion about ethylene or ACC effects on individual ERF transcript abundance based on statistical evaluation is not possible.

Nearly all ERFs were detected by qPCR using 40 cycles as the detection limit (Table S4). Visualization of the transcript abundance in untreated stems showed similar relative expression strength of ERFs in the three experiments (Fig. 1). A significant proportion of the ERFs (72, c. 43%) responded with more than five-fold change at least in one of the experiments (Figs 1, S3). In some cases, however, this reflects changes at very low transcript abundances. A number of ERFs showed increased expression in all three experiments, indicating an early induction and long-term ethylene response. Some were present in very low abundance in noninduced tissues and were induced to a significant level by the ACC/ethylene treatment, while others showed transient induction, which diminished to the noninduced level after 14 d of ethylene treatment (Fig. 1). Only few ERF family members showed more than five-fold decrease in expression after ACC and 14-d ethylene treatment, whereas in the 24-h ethylene treatment ERFs with decreased expression were more common (Fig. S3). The ERFs that responded more than five-fold were present in almost all phylogenetic groups, with group IX ERFs showing a high proportion of ACC/ethylene response (Figs S2, S4). Taken together, the global expression analysis of ERFs in Populus stems demonstrated that transcripts of most ERFs were present in the tissues, many of the genes responded to ACC/ethylene treatment, and increased transcript abundance was more commonly observed than decreased abundance. Twenty-six ERFs that were either consistently induced in all three experiments, or transiently induced, were selected for further analyses of their function in wood formation (Fig. 1).

Figure 1.

Expression of ERF genes in hybrid aspen (Populus tremula × tremuloides) stem tissues in response to 1-aminocyclopropane-1-carboxylic acid (ACC) and ethylene. Genome-wide qPCR expression analysis of ERF genes in response to treatment with ACC (in vitro plants) and ethylene (ET; glasshouse-grown trees). To generate the heatmap, normalized relative expression data from qPCR analysis of control plants was ranked into seven classes for each experiment according to expression level (three left-most columns). Each class is attributed a colour from low (blue) to high (red) transcript abundance. Similarly, the fold increase/decrease of ERF transcript abundance in response to ACC or ethylene treatments is shown as seven classes (three right-most columns); > 10-fold decrease (dark blue), 5–10-fold decrease (middle blue), 2–5-fold decrease (light blue), 2–5-fold increase (light red), 5–10-fold increase, > 10-fold increase (dark red), < two-fold increase or decrease (light grey). All columns were clustered to create the heat map. Genes selected for further analysis are indicated in large font (all gene names are indicated in small font, similar colours correspond to similar phylogenetic subgroups). The heatmap is based on primary data in Supporting Information, Table S4. Each sample is a pool of three independent plants, and was analysed in technical triplicates. ERFs not detected in untreated samples after 40 cycles, but detected at < 35 cycles in treated samples (or vice versa) were interpreted as considerably induced and fold ratios were superior to 10-fold. ERFs detected in both untreated and treated samples between 35 and 40 cycles are assigned as noninduced because of their low transcript abundance.

ERFs induced by ethylene treatment were also induced during TW formation

Tension wood develops under endogenous ethylene formation (Andersson-Gunnerås et al., 2003; Du & Yamamoto, 2003), and many TW characteristics are induced by exogenous ACC/ethylene (Love et al., 2009). To explore whether the selected ERFs may be mediators of endogenous ethylene signalling during TW formation, their expression was assayed by qPCR in forest-grown aspen induced to form TW by bending the stem for 3 wk. The 26 selected ERFs were all detected in the wood-forming tissues. However, three ERFs (ERF14, ERF16, ERF26) were close to the detection limit in both NW- and TW-forming tissues and were therefore not interpreted as considerably expressed ERFs in this analysis (Table S4). Seventeen out of the 23 ERFs (74%) were induced more than two-fold in TW, and transcripts of eight ERFs were on average > 10-fold more abundant in TW than in NW (Fig. 2). These eight ERFs showed either very low or undetectable expression in NW (Table S4), suggesting the possibility of de novo induction of ERFs during TW formation. It is also noted that 14 out of the 17 TW-induced ERFs were induced during long-term (14 d) ethylene treatment (ERFs 3, 9, 18, 19, 21, 25, 31, 32, 33, 93, 102, 105, 132 and 139) (Fig. 1, Table S4), which is the most comparable treatment to the TW tissues sampled after 3 wk of bending. The general picture that emerges from this experiment is that a majority of the selected ERFs were also induced when TW was formed, consistent with the idea that many ERFs are induced by endogenous ethylene.

Figure 2.

Expression of Ethylene Response Factors (ERFs) in response to tension wood (TW) induction. Ratio of ERF expression of 23 selected ERFs during TW and normal wood (NW) formation in forest-grown aspen (Populus tremula). TW samples consisted of two biological replicates (each replicate was a pooled sample from two trees), NW samples consisted of three biological replicates (each replicate was a pooled sample from two trees). Ratios were calculated from each of the two TW samples to the average of all three NW samples. The upper and lower limits of error bars indicate the range of the two ratios obtained. The horizontal line indicates two-fold change. The graph is based on primary data in Supporting Information, Table S4.

ERFs can modify cambial growth and wood chemistry when overexpressed in wood-forming tissues

Twenty-four of the selected ERFs were overexpressed in hybrid aspen to investigate their potential to modify cambial growth or wood chemistry (ERF26 and ERF93 were not overexpressed as they only were discovered after the initiation of this experiment in the updated version of the Ptrichocarpa genome). To direct the overexpression to wood-forming tissues we used the LMX5 promoter that is preferentially active in developing wood (Love et al., 2009). Transgenic overexpression lines were produced for 20 of the selected ERFs – viable transformed plants did not regenerate with constructs for four ERFs (ERF3, 9, 14 and 103). For each construct, multiple in vitro lines were regenerated and screened for ERF transcript abundance by qPCR (Table S5). Overexpressing lines (mostly five per construct) were grown in triplicates in glasshouse to a height of c. 1.7 m during spring/early summer. Because of the large number of trees, the growth experiment was performed in two batches, each containing the full set of lines for the respective construct (Fig. 3).

Figure 3.

Relative difference of diameter and height growth of pLMX5::ERF overexpressing hybrid aspen (Populus tremula × tremuloides) trees compared to wild-type in the respective growth batches. Each box comprises the pooled data from four to five lines (two lines for ERF30) with three replicate trees per line relative to the average data from 16 wild-type trees. The horizontal line inside the box designates the median, the box around it indicates the interquartile range and contains 50% of the measured values. Whiskers refer to minimum and maximum values, whereas circles above and below the upper and lower whiskers show outlier values. All trees for all lines per gene were pooled to one class and compared to the 16 wild-type plants of the respective growth batch using a Mann–Whitney U-test. Significant changes are indicated by asterisks: *, < 0.05; **, < 0.01.

The most striking phenotype resulted from the overexpression of ERF139; all the transgenic lines were much smaller (up to 80%) than wild-type trees (Figs 3, 4); pLMX5::ERF139 lines were also the only case where a clear difference in wood anatomy could be observed under light microscopy. In particular the vessel elements were small in pLMX5::ERF139 lines (Fig. 4). The most affected line in terms of height and wood cell size had the highest ERF139 transcript levels (Figs 4, S5). In addition to ERF139, overexpression of four ERFs (ERF18, 34, 35 and 105) significantly increased, and two ERFs (ERF71 and 85) suppressed the diameter growth (Fig. 3). The increase in diameter growth was confirmed for ERF18, 34 and 35 overexpressors in another experiment conducted during late autumn/winter. In this experiment also overexpression of ERF71 and 85 showed an increase in diameter growth, whereas no difference was observed for ERF105 (Fig. S6). This result suggests that environmental conditions (such as seasonal effects) are important for the effect observed from overexpression of some ERFs on diameter growth. This was particularly striking for ERF85.

Figure 4.

Plant height and wood anatomy of pLMX5::ERF139 overexpressing hybrid aspen (Populus tremula × tremuloides) trees. Nine-week-old wild-type (a) and pLMX5::ERF139 (b) trees. Height data for each individual biological replicate of ERF139 overexpressing line and average height data for wild-type (= 16), error bar indicates ± SD (c). Micrographs of representative stem sections collected 20 cm above soil stained with safranin/alcian blue from wild-type (d), pLMX5::ERF139 line 5 (e) and pLMX5::ERF139 line 6 (f). In particular vessel elements were seemingly smaller. In the most extreme line (6) aberrant ray cells were observed, i.e. small and round rather than long elongated cells as well as disorganized fibres – (arrow in f and magnified in g). V, Vessel; F, fiber; R, ray.

In order to identify modifications in wood chemistry and/or -structure in the ERF overexpressing lines, we used diffuse reflectance FT-IR spectroscopy to screen all trees. Spectra were obtained from wood powder and the data were evaluated by OPLS-DA analysis (Trygg & Wold, 2002). Differences between transgenic lines and wild-type were visualized by Scores plots, and the major spectral bands that contributed to the separation were evaluated by Loadings plots. Spectra from trees overexpressing ERFs 18, 21, 30, 85, and 139 gave reliable models that separated the transgenic lines from wild-type trees (Table 1). This difference in chemotypes was confirmed for these overexpressed ERFs in a second experiment (Table S6).

Table 1. Ethylene Response Factors (ERFs) that modified wood chemistry when overexpressed in hybrid aspen (Populus tremula × tremuloides)
Overexpressed geneOPLS-DA modelaChemical alterationsb
  1. The FT-IR spectra of wood powder from these transgenic lines (data are from 4 to 5 lines per overexpressed ERF (two lines for pLMX5::ERF30) with three biological replicates per line) separated from that of wild-type trees (= 16). The chemical alterations are deduced from Loadings plots of the OPLS-models. For details see Supporting Information, Fig. S7.

  2. a

    All models have one predictive and three orthogonal components. Q2, R2X and R2Y values are all cumulative.

  3. b

    Based on area-normalized diffuse reflectance FT-IR spectra. Changes reflect alterations in proportions relative to wild-type, and not absolute amounts.

ERF18 Q2 = 0.558, R2X = 0.805, R2Y = 0.767Higher amounts of glycosidic linkage (1150 cm−1) and carbohydrates (likely cellulose, 710, 900, 1320 cm−1). Lower amounts of lignin, particularly G-type (1510 cm−1, the 1595 cm−1 band remains unaltered)
ERF21 Q2 = 0.597, R2X = 0.761, R2Y = 0.751
ERF30 Q2 = 0.627, R2X = 0.826, R2Y = 0.807
ERF85 Q2 = 0.736, R2X = 0.853, R2Y = 0.808Lower amounts of glycosidic linkage (1150 cm−1) and carbohydrates (likely cellulose, 900, 1320 cm−1). Higher amounts of lignin S to G ratio less altered (1510 cm−1, 1595 cm−1).
ERF139 Q2 = 0.782, R2X = 0.908, R2Y = 0.898Lower amounts of glycosidic linkage (1150 cm−1) and carbohydrates (unspecific, 1000-1100 cm−1 region, 900 and 1320 cm−1). Higher amounts of lignin, particularly G-type (1510 cm−1, the 1595 cm−1 band is less affected)

FT-IR spectral bands are generally not diagnostic for any particular cell wall polymer, but rather reflect chemical bonds that may be present in several wall components (Gorzsás et al., 2011). Interpretation of Loadings plots should therefore be done with care and preferentially be based on the co-variation of several bands (for band interpretation see Fig. S7). With those precautions in mind, the Loadings plots indicate increased lignin and decreased polysaccharides in pLMX5::ERF85 and pLMX5::ERF139, and decreased lignin in pLMX5::ERF18, pLMX5::ERF21 and pLMX5::ERF30, coupled with an increased proportion of polysaccharides in pLMX5::ERF18 and pLMX5::ERF30 and somewhat less so for pLMX5::ERF21 (Table 1). Together the data indicate that the above-mentioned ACC/ethylene-inducible ERFs have the potential to influence wood development and/or deposition of secondary cell wall polymers.

Discussion

Ethylene is an important signalling substance in wood development and is in particular involved in the remodelling of wood formation upon TW induction (Love et al., 2009). ERFs serve as important integrators to mediate ethylene responses (Gutterson & Reuber, 2004), and in Arabidopsis ERFs were shown to be essential for ethylene-stimulated cambial cell division (Etchells et al., 2012). In a genome-wide screen we demonstrated that a large set of Populus ERFs responded to both ACC and ethylene when applied to woody stems, and to TW formation when endogenous ethylene is produced. We further identified ERFs that modify wood chemistry and diameter growth when overexpressed in wood-forming tissues.

ERFs consist of a large gene family in Populus with many gene duplicates. Bioinformatic analysis and gene model curation identified 170 ERFs, which is much more than in Arabidopsis and rice with 122 and 139 ERFs, respectively (Nakano et al., 2006). From the phylogenetic analysis it is evident that Arabidopsis orthologues to Populus ERFs are in many cases not obvious due to the greater number of gene duplication events. These duplication events in the Populus genome have also resulted in enlargement of other gene families such as genes involved in lignocellulosic wall biosynthesis and disease resistance (Tuskan et al., 2006).

Transcripts for the majority of the annotated ERF family members were detected in stem tissues of hybrid aspen, although the abundance was in many cases very low (Fig. 1, Table S4). A large number of ERFs also responded to ACC/ethylene treatment. This is in agreement with a microarray-based screening of ERF responsiveness in different organs and tissues of Arabidopsis (e.g. Feng et al., 2005), and there are also other examples where selected ERFs have been shown to be ethylene-induced (Ohme-Takagi & Shinshi, 1995; Büttner & Singh, 1997; Fujimoto et al., 2000; Etchells et al., 2012). Ethylene induction of ERFs will depend on the developmental and tissue context, and the duration of the ethylene signal. Therefore, it is difficult to make any general conclusion about which ERFs are ethylene induced and which are not, but it is clear from our and other studies that ERFs are important components of ethylene signalling. However, the responsiveness of ERFs is not exclusive to ethylene because experiments with ethylene-insensitive Arabidopsis have shown that ethylene-induced ERFs can respond to other hormones such as jasmonic acid to integrate environmental stimuli (Lorenzo et al., 2003; Pré et al., 2008). Our genome-wide screens revealed also a significant number of ERFs that were expressed in wood-forming tissue, but did not respond to ethylene (Fig. 1, Table S4). This is not surprising because this large family of transcription factors must have additional functions beyond ethylene signalling.

The genome-wide investigation of ERFs responding to ACC/ethylene offers an access to identify mechanisms underlying ethylene-mediated processes in wood development in Populus. ERFs that were consistently, or in some cases transiently, induced in our experiments were investigated further for their expression in TW and their potential to affect wood formation when overexpressed in Populus. All of the selected ethylene-responsive ERFs were detected in the cambial region (phloem, cambium and developing xylem) in field-grown aspen trees (Fig. 2, Table S4). Many ERFs were also induced during TW formation when endogenous ethylene is produced, in agreement with the role for ERFs in ethylene signalling. However, some ERFs that showed consistent induction by ethylene treatment were not responding to long-term bending. One possible explanation is that these ERFs are more important in bark tissues (which for practical reasons were included in the genome-wide screen for ACC/ethylene responses) rather than in cambial region and involved in, for example, ethylene-mediated defence responses.

Overexpression of ERF139 in the cambial tissues of hybrid aspen resulted in major changes in wood development and wood chemistry, and produced dwarf trees (Fig. 4). It is not possible to conclude about the primary and secondary effects of ERF139 overexpression. We hypothesize that vessel cell expansion is a primary target of ERF139, and that it has a function in inducing this key characteristic of TW. Reduced vessel diameter will have a negative impact on transport efficiency of water and nutrients, and hence also on plant growth. Dwarfed trees and altered xylem anatomy will, in turn, unavoidably lead to an altered wood chemistry. Several of the other ethylene responsive ERFs modified diameter growth and/or wood chemistry when expressed in wood-forming tissues, without having any major effects on plant development and wood anatomy (Figs 3, S6, Table 1). It is clear that the effect of the overexpression of some ERFs on growth differed between experiments performed at different times of the year. This suggests an interaction between the environment and ERF signalling in growth regulation. However, overexpression of ERFs 18, 34 and 35 consistently stimulated diameter growth. The effects on wood chemistry observed from overexpression of five of the ERFs were also consistent between experiments.

Whereas the stimulating effect of ethylene on cambial cell division is well-established, the effects of ethylene on wood cell wall biosynthesis and resulting cell wall chemistry are poorly understood. Although ethylene treatment seems to stimulate post-harvest lignification in crop species (Liu & Jiang, 2006), it was demonstrated to primarily affect the carbohydrate matrix in xylem cell walls when applied to Norway spruce seedlings (Ingemarsson, 1995). It should be noted that the FT-IR analysis indicated that overexpression of ERF18 and ERF21 enhanced the carbohydrate content in wood. This was not due to any visible induction of cellulose rich G-layer in these trees that is typical for TW fibres in P. tremula. However, increased carbohydrate content is also a characteristic of S2 cell wall layers of TW fibres in species that do not form G-layers (Felten & Sundberg, 2013). The exact targets of the ethylene-induced ERFs identified here, which would affect wood chemistry, remain to be established, but they may interact with the transcriptional machinery underlying secondary cell wall biosynthesis (Zhong et al., 2008) to mediate plasticity in polymer composition of wood cell walls.

Transcriptome analyses have suggested that members of the ERF family are important in xylem development in Arabidopsis, where AtERF1 and AtERF2 were induced by artificial weight application (Ko et al., 2004). Furthermore, ERF1, together with ERF018 and ERF109, were shown to have a function in cambial cell division in Arabidopsis (Etchells et al., 2012). These ERFs were induced by ethylene and were essential for cambial growth stimulation in ethylene overproducing eto mutants. Increased ethylene signalling through these ERFs was further shown to compensate for cambial growth in pxy mutants, and ethylene and PXY signalling were shown to independently regulate cambial cell division. However, the observation that cambial cell division was in general not affected in ethylene-insensitive mutants of Arabidopsis (Etchells et al., 2012) suggests that ethylene signalling is mainly functioning in response to environmental conditions and stress. The transcripts encoding the closest Populus orthologues to AtERF018 (PtiERF136 and PtiERF137) and AtERF109 (PtiERF68, PtiERF69 and PtiERF77) were not strikingly induced by ACC/ethylene in our experiments, except for ERF69 after 14 d of ethylene treatment. However, transcripts coding for the two closest Populus orthologues to AtERF1 (PtiERF14 and PtiERF16) were highly induced by ACC/ethylene, but not during TW formation. No phenotype was observed in ERF16 overexpressors, and the attempt to overexpress ERF14 did not result in any transformants. ERF function in cell wall biosynthesis was suggested in a study where heterologous expression of Arabidopsis SHINE (SHN) in rice was shown to dramatically change the composition of cellulose, hemicellulose and lignin without affecting the overall growth (Ambavaram et al., 2011). The effect of SHN on rice secondary wall biosynthesis may, however, not reflect its endogenous function because in Arabidopsis SHN was described as being involved in wax/cutin lipid regulation and drought tolerance (Aharoni et al., 2004; Kannangara et al., 2007). Nevertheless, its interaction with cell wall biosynthesis supports the idea that some ERFs have binding motifs that directly or indirectly regulate secondary cell wall biosynthesis.

Given the role of ethylene as an important signalling molecule in plant adaptation to changing environments, we can expect that ERFs have important functions in generating plasticity in wood such as TW formation to make it sensitive and reactive to environmental stimuli. Overexpression of ERF139 resulted in small vessels, which is one of several hallmarks of TW. ERF18 and ERF21 overexpression enhanced the carbohydrate content of wood, another feature of TW-S2 walls in species that do not form G-layers (Felten & Sundberg, 2013). However, TW also exhibit many other characteristics and most likely involves not only a complex regulatory network including ERFs, but also other signalling components.

Acknowledgements

We thank: Drs Hannes Kollist and Bahtijor Rasulov for technical help in ethylene treatments of the Populus stems; Jenni Kiiskinen and Tuomas Puukko for technical help in laboratory analyses and glasshouse experiments; Pekka Lönnqvist, Leena Grönholm, Marjukka Uuskallio and Marja Tomell for nursing the plant material; Dr Jarkko Salojärvi for help in Populus genome mining; Carin Olofsson for help with growth measurements; Kjell Olofsson for sectioning and Gunilla Malmberg for Populus transformation and maintaining transgenic trees. We acknowledge Manoj Kumar for providing TW material and are grateful to Nathanial Street for integrating ERF gene models into the PopGenIE database. This work was supported by grants from FORMAS (FuncFiber/Bioimprove -centre of excellence in wood science), Vetenskapsrådet and the Swedish Energy Agency, EUProgramme Renewall (FP7/2007–2013), VINNOVA, and Bio4Energy, the Swedish Programme for renewable energy to B.S. and by grants from TEKES and Forest Cluster - Strategic Centre for Science, Technology and Innovation (FuncWood, EffTech and EffFibre) to J.K. The work conducted by the US Department of Energy Joint Genome Institute is supported by the Office of Science of the US Department of Energy under Contract No. DE-AC02-05CH11231.

Ancillary