A unique program for cell death in xylem fibers of Populus stem


(fax +46 907866676; e-mail hannele.tuominen@plantphys.umu.se).


Maturation of the xylem elements involves extensive deposition of secondary cell-wall material and autolytic processes resulting in cell death. We describe here a unique type of cell-death program in xylem fibers of hybrid aspen (Populus tremula x P. tremuloides) stems, including gradual degradative processes in both the nucleus and cytoplasm concurrently with the phase of active cell-wall deposition. Nuclear DNA integrity, as determined by TUNEL (terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling) and Comet (single-cell gel electrophoresis) assays, was compromised early during fiber maturation. In addition, degradation of the cytoplasmic contents, as detected by electron microscopy of samples fixed by high-pressure freezing/freeze substitution (HPF-FS), was gradual and resulted in complete loss of the cytoplasmic contents well before the loss of vacuolar integrity, which is considered to be the moment of death. This type of cell death differs significantly from that seen in xylem vessels. The loss of vacuolar integrity, which is thought to initiate cell degradative processes in the xylem vessels, is one of the last processes to occur before the final autolysis of the remaining cell contents in xylem fibers. High-resolution microarray analysis in the vascular tissues of Populus stem, combined with in silico analysis of publicly available data repositories, suggests the involvement of several previously uncharacterized transcription factors, ethylene, sphingolipids and light signaling as well as autophagy in the control of fiber cell death.


Programmed cell death (PCD) is an inbuilt suicide program that is induced to ensure that organisms show appropriate developmental patterns and growth responses to external challenges. Developmental PCD in plants occurs, for instance, during embryogenesis (Bozhkov et al., 2005), leaf morphogenesis (Gunawardena, 2007), floral development (Rogers, 2006) and organ senescence (Rogers, 2005). In addition, diverse abiotic and biotic elicitors induce PCD processes, such as the pathogen-induced hypersensitive response (Hofius et al., 2007). Diagnostic hallmarks of plant PCD are similar to those found in animal apoptosis, and include chromatin condensation, oligonucleosomal fragmentation of DNA and the activation of various proteases, but the occurrence, chronological order and duration of these processes are largely dependent on the particular type of PCD that occurs. Degradation of the cytoplasmic contents by autophagy appears to be typical for plant PCD, involving formation of autophagosomes that are targeted to the vacuole (Bassham, 2007), even though autophagy can also participate in protection from cell death (Liu et al., 2005).

PCD also occurs during xylem development, for which complete autolysis of the cell contents is required to form hollow, water-transporting tracheids or vessels, which are generically known as tracheary elements (TEs) (Fukuda, 2000). Vacuolar integrity typically plays a central role in controlling the initiation and progress of PCD in TEs, as rupture of the tonoplast results in the release of hydrolytic enzymes stored in the vacuole, leading to rapid degradation of the various organelles and nuclear DNA (Kuriyama, 1999; Obara et al., 2001). A bi-functional nuclease has been shown to be responsible for the degradation of nuclear DNA in differentiating Zinnia elegans TEs in vitro (Ito and Fukuda, 2002). In addition, various serine and cysteine proteases, as well as ribonucleases, have also been implicated in the control of xylem PCD on the basis of their expression patterns (Fukuda, 2000). Large-scale gene expression analyses by microarrays have revealed putative regulators of xylem development (Demura et al., 2002; Birnbaum et al., 2003; Ko and Han, 2004; Ko et al., 2004, 2006; Ehlting et al., 2005; Kubo et al., 2005; Zhao et al., 2005), but, as the samples contained cells from several stages of xylem development, they do not allow definite identification of novel regulators of xylem cell death. Application of the laser-capture microdissection technique did not allow separation of specific stages of xylem development either (Nakazono et al., 2003). This problem could be at least partially solved by analysis of trees, in which the large size of the vascular tissues allows spatial separation of the various developmental stages of xylem.

Populus has emerged as the main angiosperm tree model system (Taylor, 2002; Jansson and Douglas, 2007), and the complete genome sequence of Populus trichocarpa has provided the information and tools required for comparative genomic approaches (Tuskan et al., 2006). Populus wood shows characteristic features of angiosperm secondary xylem, in terms of both composition and organization, with longitudinally oriented water-transporting vessel elements embedded in the bulk of xylem fibers and radially oriented parenchymatic rays. We have previously reported major transcriptional changes that occur during xylem development in the vascular tissues of hybrid aspen (P. tremula × P. tremuloides) (Moreau et al., 2005). Here we focus on the PCD of xylem fibers in Populus. Analysis of fiber development is hampered by technical complications, as the fibers located in the middle of the stem are difficult to access and their amenability to traditional microscopic techniques is poor. Hence, we have developed several novel tools for examining the major morphological and molecular changes that occur during xylem fiber cell death, involving a combination of microscopic, cytological and high-throughput gene monitoring techniques.


Fibers are largely autolyzed before cell death

Xylem fiber morphology and cell death were characterized by transmission electron microscopy analysis of Populus stems. Chemical fixation of samples, which is used in traditional electron microscopy techniques, does not often provide satisfactorily high-quality images of woody tissues. Therefore, samples were fixed using the more reliable HPF-FS techniques described by Samuels et al. (2002).

All cambial daughter cells looked identical during early xylem differentiation, with densely stained cytoplasm, a high density of various organelles, and a large central vacuole (Figure 1a). Hence, it was only after radial expansion of the cells that the various types of xylem elements, vessel elements and fibers, could be distinguished. During early differentiation of the fibers, the dense cytoplasm contained many well-defined organelles (Figure 1b). Maturation of the fibers, manifested by deposition of secondary cell walls, coincided with the presence of abundant Golgi bodies (Figure 1c). At this stage, the nuclei, mitochondria and ribosomes had a normal appearance (Figure 1c). As the vacuole increased in size during further maturation of the fibers, the size of the cytoplasm appeared to decrease and the nuclei were appressed against the cell wall (Figure 1d). Soon after bulk deposition of the secondary cell-wall material, the nucleus became round in shape, suggestive of decreased pressure from the vacuole (Figure 1e). Most strikingly, the cytoplasm became gradually less dense in appearance during late maturation of the fibers (Figure 1e–g), and eventually it became difficult to distinguish any organelles or ribosomes in the cytoplasm of the late-maturing fibers (fiber 2 in Figure 1h). Figure 1(g) shows a fiber at the very last stages of maturation with imminent vacuolar rupture, when typically only very few organelles or ribosomes were present, but instead small vesicles and dilatation of the endoplasmic reticulum were frequently observed. The rupture of the tonoplast, which is considered as the moment of death, was followed by final autolysis of the remaining cell contents and clearing of the cells (Figure 1i). Based on the vacuolar disruption hallmark and xylem viability staining (Figure 3i,k), the death of xylem fibers occurred synchronously throughout the circumference of the stem at a distance of 650–1000 μm from the cambium, depending on the tree.

Figure 1.

 Xylem fiber morphology in Populus wood as shown by transmission electron microscopy.
(a) Cambial daughter cells in the xylem.
(b) Xylem fiber at an early maturation stage with notably thin secondary cell walls. The illustrated section is located in the cytoplasm on the surface of the central vacuole, and traverses the tonoplast at several locations.
(c) Cytoplasm of a fiber at a stage of extensive secondary cell-wall formation.
(d) Nucleus of a fiber at a stage of extensive secondary cell-wall formation.
(e) Nucleus of a fiber that has finalized bulk secondary cell-wall deposition.
(f) Fiber at late maturation phase when organelles start disappearing from the cytoplasm.
(g) Fiber at the very last stage of maturation when the endoplasmic reticulum shows extensive dilatation. Very few organelles or ribosomes are present in the cytoplasm.
(h) Three fibers at late maturation phase.
(i) Three fibers at a stage when the tonoplast has ruptured and the cells are being cleared. The arrowheads indicate remnants of the cell contents in one of the fibers.
The micrographs illustrate either transverse sections from chemically fixed material (a, i) or transverse (f, h) and longitudinal (b, c, d, e, g) sections from high-pressure freezing/freeze substitution fixed material. CP, cytoplasm; G, Golgi; M, mitochondria; Nu, nucleus; PCW, primary cell wall; SCW, secondary cell wall; CV, central vacuole. Scale bars = 2 μm.

Figure 3.

 Pattern of DNA degradation as assayed by TUNEL staining and the viability of the xylem elements in Populus stem.
(a) Differential interference contrast image showing a general view of vascular tissues from the stem used for the assay.
(b–h) TUNEL staining in the vascular tissues of the stem (b) and in the the xylem at a distance of 500–750 μm from the cambium (c–h). Confocal microscopy images depicting staining of 100 μm thick sections in either the transverse (b–e) or longitudinal plane (f–h). Positive controls were treated with ApaI restriction enzyme before staining (d, g), and negative controls were stained in the absence of the TdT enzyme (e, h). Green fluorescent dots indicate nuclei that contain fragmented DNA, while lignified secondary cell walls show blue autofluorescence. Similar results were obtained in two replicate trees.
(i–k) Viability staining of the stem using nitroblue tetrazolium. Images show transverse sections of the living part of the wood from the cambial/xylem interface (left border) to the location of the xylem where all fibers were autolyzed (right border) (i), the xylem expansion and early xylem maturation stage (j), and the late maturation stage of the xylem (k). Positive signal is visible as blue/purple coloration. The xylem vessels are viable only shortly after completion of the xylem expansion [viable vessels are indicated by white asterisks in (j)]. Fibers are viable for longer, and lose their viability (as indicated by arrowheads) at approximately the same distance from the cambium in each cell row (i, k), demonstrating synchronicity in fiber cell death around the circumference of the stem.
Scale bars = 50 μm.

Morphological assessment of xylem vessel death was difficult to perform for two reasons: the rapidity of the process, which meant very few vessels undergoing cell death were present at any time point, and the long, narrow geometry of the cells, which made them vulnerable to damage during dissection. However, on the basis of viability staining of the xylem, the vessel elements appeared to die within a distance of 400 μm from the cambium (Figure 3j).

Nuclear DNA shows partial degradation during early development of xylem fibers

Nuclear DNA degradation is commonly used as a marker for cell death, and single-cell gel electrophoresis of DNA from an intact nucleus is one method that may be used to detect this (Ostling and Johanson, 1984). In this assay, nuclei are lyzed and exposed to an electrophoretic field in an agarose microgel, in which degradation of nuclear DNA results in a ‘comet-like’ displacement of genetic material from the intact nuclei (‘the Comet head’) into a ‘Comet tail’ (Figure 2a). Comet nuclei are typically quantified in terms of two parameters: the tail length, which is related to the pattern of DNA damage, and the tail moment, which is related to the degree of nuclear DNA fragmentation (Collins et al., 2008). The nuclei of Populus xylem cells, mechanically isolated from 50 μm thick tangential cryosections, showed variation in both tail length and moment from the phloem/cambium interface (X0, 0 μm) to the location where all the fibers were completely autolyzed (X22, 1100 μm from the cambium). The total number of nuclei showing various degrees of DNA breakage is shown in Figure 2b. Sections obtained from close to the cambial area showed an initial increase in both tail length and moment, suggesting the occurrence of DNA breaks or degradation in the dividing cambial cells (Figure 2c,d). However, a replicate tree, from which sampling started after the centre of the cambial region, showed no comet nuclei until a distance of 150 μm from the cambium, suggesting that the initial increase in DNA fragmentation could be due to contamination from the phloem cells (data not shown). The tail length and tail moment increased again in the sections representing late xylem maturation (350–900 μm from the cambium; Figure 2c,d). The intensive increase across the distance of 300–400 μm from the cambium coincided with the location of vessel cell death (Figure 3j). After that point, the high level of Comet signals could have resulted from either the maturing fibers or ray cells. There was little variation in the length of the Comet tails during late xylem maturation (Figure 2e), indicating that the Comet tails originated from only one cell type (Collins et al., 2008).

Figure 2.

 Pattern of DNA degradation in Populus xylem analyzed by Comet assays (single-cell gel electrophoresis).
Nuclei were isolated from 50 μm thick tangential sections that were collected throughout the living xylem of the stem, from the cambial region (sample X0) through to completely autolyzed fibers (X22).
(a) Comet nuclei with three tail lengths. Scale bar = 10 μm.
(b) Total number of nuclei showing a Comet tail in each sample.
(c) Mean length of the Comet tail.
(d) Mean tail moment.
(e, f) Variation in tail length and tail moment within each sample. Similar results were obtained in a replicate tree, with the exception of a lack of Comet nuclei at a distance of 0–150 μm from the cambium.

The cell types that gave positive Comet signals during late xylem maturation were identified, in corresponding tissues, using the TUNEL (terminal deoxynucleotidyl transferase-mediated dUTP-FITC nick end labeling) assay, which specifically labels the free 5′-OH ends of DNA. A TUNEL-positive signal was detected in xylem rays at a distance of 100–400 μm from the cambium, especially in ray cells that were in contact with vessels (Figure 3b). In addition, TUNEL-positive cells were found in the xylem at a distance of 500–750 μm from the cambium (Figure 3b), which coincided with the high level of Comet signals observed during late xylem maturation (Figure 2c,d). A close-up of this region revealed, in both transverse and longitudinal sections, that the TUNEL signal was present exclusively in the fibers (Figure 3c,f). The TUNEL-positive signal in the fibers could not be related to the presence of, or contact with, any other cells undergoing cell death or DNA degradation, suggesting that, in contrast to the observed DNA degradation pattern in the ray cells (at a distance of 100–400 μm), the signal was entirely due to DNA degradation processes within the observed cell and that DNA degradation therefore occurs in the fibers in a cell-autonomous manner. Not all fibers were TUNEL-positive in the transverse sections, which can be explained by the fact that the length of the fibers exceeded the thickness of the sections, so many fibers were cut open and emptied during sectioning. Synchronicity in fiber development, including degradation of DNA, was supported by a simultaneous loss of viability in all fibers at approximately the same distance from the cambium (Figure 3i,k). These findings suggest that xylem fibers are the cells responsible for most of the observed increase in Comet signals during late xylem maturation. Overall, it appears that the nuclear DNA of xylem fibers is severely damaged or degraded relatively early during development, and this continues for a long time before degradation is complete.

Gene expression analysis reveals significant overlap in the transcriptional control of secondary cell-wall formation and cell death

Microarray analysis was performed on seven Populus stem samples, each representing a different stage of vascular development (Figure 4), to elucidate the molecular regulation of xylem maturation. We used the POP2 array (Moreau et al., 2005), which contains 24 735 cDNA clones from 19 cDNA libraries (Sterky et al., 2004), corresponding to 13 882 gene models and therefore approximately 30% of the 45 444 predicted Populus gene models. The microarray data for all genes are presented in Table S1.

Figure 4.

 Schematic illustration of the Populus stem sampling procedure for analyses of gene expression and the major developmental processes in vascular tissues.
Samples were taken from tree 1 for the microarray experiment, and included 50 μm thick tangential sections from phloem (Phl), cambium (C), expanding xylem (EX), the initiation of secondary cell-wall deposition stage (MX0) and maturing xylem (MX1–3). Samples were taken from trees 2 and 3 for the quantitative RT-PCR experiments, and included 13 and 15 tangential 50 μm thick samples, respectively. Gray-intensity bars indicate the various stages of xylem development in tree 1, as determined by the electron microscopy, TUNEL analyses and viability staining results. Crossed boxes indicate samples that were not used in the microarray analysis, or those missing from the RT-PCR analysis. Scale bar = 50 μm.

During xylem maturation (samples MX1–3), 2154 genes showed statistically significant up-regulation (Table S2). After hierarchical clustering analysis of these genes, two major patterns of gene expression were obtained (clusters 1 and 2) due to a major shift in gene expression between the MX0 and MX1 samples (Figure 5). Large clusters with specific expression in only one sample were not seen, although genes specific to the last xylem maturation stage formed a few small clusters (for example cluster 3, Figure 5). Quantitative RT-PCR analysis of 12 genes showed that the patterns of gene expression changes were similar throughout xylem development, although the timing/positioning varied between replicate trees (Figure S1, Table S3).

Figure 5.

 Hierarchical clustering of differentially expressed genes in the microarray analysis of Populus vascular tissues.
Gene expression ratios (log2 values as indicated by red/green colored squares) in each of the seven samples representing various stages of vascular development (for descriptions of the samples, see Figure 4).

The major shift in gene expression coincided with the onset of bulk secondary cell-wall formation in the xylem. This is illustrated also in the analysis of eukaryotic orthologous group (KOG) class (Tatusov et al., 2003) representations among the significantly up-regulated genes (Figure 6), indicating a dominance of transcriptional changes related to the formation and transport of cell-wall constituents in the xylem maturation samples MX1 and MX2. However, over-representation of the KOG class ‘amino acid transport and metabolism’ in the same samples suggested that reallocation of protein degradation products, and therefore preparation for cell death, occurred simultaneously with secondary cell-wall formation. Over-representation of the classes ‘energy production and conversion’ and ‘cytoskeleton’ may also be related to imminent cell death, which presumably requires energy for nutrient reallocation (Keskitalo et al., 2005) and involves changes in the cytoskeletal structure (Smertenko et al., 2003). Furthermore, KOG class representation suggested that the final processes to occur during xylem maturation are related to inorganic ion transport and protein turnover (Figure 6).

Figure 6.

 KOG classification of the significantly up-regulated genes in the microarray analysis of Populus vascular tissues.
The squares with gray coloration indicate statistically significant (< 0.05) over-representation of the up-regulated genes in the respective KOG classes. For descriptions of the samples, see Figure 4.

As expected on the basis of the KOG analyses, expression of genes related to secondary cell-wall formation and those related to cell death overlapped. For instance, almost all of the known cellulose (Djerbi et al., 2005) and lignin biosynthesis genes (Tuskan et al., 2006) on the array showed high expression in the first two samples from the xylem maturation phase (Figure S2). Also, several transcription factors that control secondary cell-wall formation, such as homologs of the Arabidopsis genes NST1 and MYB85 (Zhong and Ye, 2007), were significantly up-regulated in the MX1 and MX2 samples (Figure S3). Simultaneously, several stress and/or cell death-related genes were up-regulated, such as homologs of the Arabidopsis AP2-EREBP transcription factors TINY2 and RAP2.3 and of the Arabidopsis MYB transcription factor BOS1 (Pan et al., 2001; Mengiste et al., 2003; Wei et al., 2005), as well as several cell death-related proteases (Figures S3 and S4). However, a large number of transcription factors were up-regulated after the bulk of secondary cell-wall deposition, suggesting occurrence of a mode of transcriptional regulation that is unique to xylem cell death (Figure S3).

Gene expression in Populus xylem supports commonalities between various PCD processes in plants

Vacuolar processing enzymes (VPE) are the key executors of cell death in plants (Hatsugai et al., 2004; Rojo et al., 2004; Nakaune et al., 2005). Up-regulation of the Populus homologs of Arabidopsis β-VPE and γ-VPE during xylem maturation was detected, suggesting their involvement in xylem cell death also (Figure 7). Their late expression in the MX3 sample, and their previously suggested expression in the inter-fascicular fibers of Arabidopsis inflorescence stems (Ehlting et al., 2005), support their function in fiber maturation. Cathepsin B-like cysteine proteases, which are potential targets of VPE action (Shirahama-Noda et al., 2003), were also up-regulated during late xylem maturation (Figure S4). Another protease that was up-regulated during xylem maturation is a homolog of Arabidopsis metacaspase 9 (AtMC9) (Figure 7). Although there is no functional evidence for the involvement of AtMC9 in the control of cell death in Arabidopsis, the cell death-related functions of two other plant metacaspases (Suarez et al., 2004; He et al., 2008) suggest that this gene may be involved in xylem cell death.

Figure 7.

 Up-regulation of some of the known regulators of cell death during xylem maturation in Populus vascular tissues as shown by microarray analysis.
Gene expression ratios (log2 values as indicated by red/green colored squares) in each of the seven samples representing various stages of vascular development (for descriptions of the samples, see Figure 4). Populus gene models were selected according to the known function of the most similar Arabidopsis gene in PCD, and according to the up-regulation of gene expression in any of the MX1–3 Populus microarray samples. The annotation for each Populus gene model includes the transcript number and the annotation of the closest Arabidopsis gene. Complete information on each Populus gene model can be found in Table S2, in which they are listed according to the transcript number.

In addition to proteases, nucleases also play key roles in the control of PCD. A gene encoding an endonuclease/exonuclease/phosphatase family protein was identified as the gene with the highest fold increase in expression in the MX1 sample (Figure 7 and Table S2). In addition, a gene homologous to the apoptosis-inducing factor (AIF) that participates in chromatic condensation and degradation of DNA during apoptotic cell death in animals (Susin et al., 1999; Galluzzi et al., 2008) was up-regulated during xylem maturation (Figure 7). Five genes homologous to AIF, annotated as monodehydroascorbate reductases (Hofius et al., 2007), are found in Arabidopsis, but their function in DNA modification remains to be demonstrated in plants.

The data also provide strong evidence for the involvement of autophagy in xylem maturation. Five Populus genes homologous to Arabidopsis autophagy genes ATG8C, D, F and I, were up-regulated in the three samples representing the various stages of xylem maturation (Figure 7). Other up-regulated genes that have been implicated in the control of autophagy include the homologs of the Arabidopsis VTI12 and TOR (target of rapamycin).

Several genes believed to protect against cell death were also found amongst the differentially expressed sets, notably genes encoding BAG (Bcl-2-associated athanogene) family proteins (Doukhanina et al., 2006), BAX inhibitors (Kawai-Yamada et al., 2004) and an ethylene-response protein similar to RAP2.3 (Ogawa et al., 2005) (Figure 7). Therefore, our results clearly indicate that pro-survival proteins, as well as pro-cell death proteins, are shared by the various plant cell-death processes.

Identification of cell-death regulators specific to the fibers

It was difficult to extract genes specifically related to fiber cell death from the expression profiles, as the gradual cell death in xylem fibers overlapped with both secondary cell-wall formation and vessel cell death. Therefore, an alternative approach was adopted that was based on comparison of the Populus transcriptome in the maturing xylem with the Arabidopsis microarray datasets in the response viewer tool of Genevestigator V2 and in five xylem-related analyses (see Experimental procedures and Table S2). As a result, 60 Populus genes were identified as fiber cell-death candidate genes (Figure 8). Obviously, the list did not represent the complete transcriptome related to fiber cell death as (i) it excludes all the factors common to fibers and vessels, and (ii) the regulatory mechanism of fiber cell death is not expected to always overlap with the senescence pathway that was used as a selection criterion in Arabidopsis.

Figure 8.

In silico identification of fiber PCD candidate genes.
Arabidopsis microarray data repositories were used for a comparative transcriptomic approach to identify fiber PCD candidate genes among the significantly up-regulated genes in the Populus microarray samples MX1–3. The color codes, annotations and samples are described in Figure 7.

The expression patterns of the 60 Populus fiber cell-death candidate genes are shown in Figure 8. Among these genes, three transcription factors, one basic-leucine zipper (bZIP) family protein and two EREBP/AP2 family proteins were up-regulated in the MX1 sample, and thus relatively early during fiber maturation. A few genes were identified that have previously been related to drought, salt or cold stress, such as the homologs of Arabidopsis CALCINEURIN B-LIKE 10, RESPONSIVE TO DEHYDRATION 21, a dehydrin family protein and a stress-responsive protein, which suggests commonalities in signaling between abiotic stress and fiber cell death, or alternatively indicates that these conditions include common changes in plant metabolism. No senescence-related transcripts were found, but the identification of two proteins that are known to be crucial for peroxisomal dynamics (PEX11D; AT2G45740) and function (KAT2/PED1; AT2G33150) provided evidence for the importance of fatty acid breakdown through β-oxidation during fiber cell death in a similar manner to that known to occur in senescing leaves. The presence of genes homologous to Arabidopsis sphingosine-1-phosphate lyase and EIN3-BINDING F BOX PROTEIN 1 support the involvement of sphingolipid metabolites and ethylene in signaling leading to fiber cell death (Figure 8). Somewhat surprisingly, phytoreceptor PHYA was also identified among the fiber cell-death candidate genes. In addition, several genes encoding proteins with unknown function were found among the fiber cell-death candidate genes, offering possibilities for recovery of novel gene functions with reverse genetics approaches. Examples of such genes are the Arabidopsis homologs of NDR1/HIN1-LIKE 1 and a MA3-domain containing protein, which are interesting due to their sequence or domain similarity to known regulators of cell death in the hypersensitive response and animal apoptosis, respectively.


Our results demonstrated a unique cell death program in the xylem fibers of Populus stems, including a gradual degradation of the cell contents, which is clearly different from the pattern observed in xylem vessel elements (Figure 1). Even though it was difficult to observe the morphology of late-developing vessel elements, a major loss of cytoplasmic contents is believed to occur only when hydrolytic enzymes are released from the vacuole as a result of tonoplast rupture (Avci et al., 2008; Burgess and Linstead, 1984; Groover et al., 1997; Srivastava and Singh, 1972; Turner et al., 2007). A gradual degradation of the cytoplasmic contents before cell death has been observed in other developmental cell-death processes, such as petal senescence (van Doorn and Woltering, 2008). Such a pattern of cellular degradation is indicative of autophagy (van Doorn and Woltering, 2005), which is supported by the up-regulation of several autophagy-related genes, including four ATG8 genes, in the microarray analysis (Figure 7). However, typical autophagosomes were not detected in the xylem fibers of Populus stem, although this was not particularly unexpected considering the rapidity of the process (Moriyasu and Ohsumi, 1996; Yoshimoto et al., 2004). Activation of the autophagic machinery, at least at the transcriptional level, appears to be unique to fibers, as none of the autophagy-related genes were induced in Arabidopsis vessel elements in vivo or in differentiating tracheary elements in vitro (Kubo et al., 2005; Brady et al., 2007). However, it is possible that autophagy is activated in fibers for adaptive reasons other than degradation of cytoplasmic contents, such as nutrient recycling or breakdown of toxic products, thereby protecting the fibers from premature cell death. In any case, the evidence presented here for the occurrence of autophagy during xylem development is novel, and suggests an important metabolic distinction between the vessels and fibers, which coincides with the observed differences in organelle degradation patterns between these two cell types.

The nuclear DNA degradation in xylem fibers also appears gradual and/or slow (Figures 2 and 3). This pattern of DNA degradation is also strikingly different from patterns previously observed in vessel elements. For instance, the nuclei of Zinnia TEs reportedly do not show any DNA laddering or significant TUNEL staining before cell death, which would be diagnostic of gradual DNA degradation, but rather rapid and complete degradation within 20 min of tonoplast rupture (Groover et al., 1997; Obara et al., 2001). Positive TUNEL staining has been demonstrated in xylem vessels of pea plants (Mittler and Lam, 1995), but it was not clear whether the signal represented DNA degradation before or after cell death. Identification of the vacuolar ZEN1 endonuclease in Zinnia TEs (Aoyagi et al., 1998), and the demonstration of its function in nuclear DNA degradation (Ito and Fukuda, 2002), provides further evidence that vessel element DNA degradation occurs very rapidly and only after the release of hydrolytic enzymes through tonoplast rupture. Alternative or additional nucleases that are not targeted to the vacuole must be active in the fibers. Homologs to putative nucleases, such as the unknown protein with a putative function as an endo-/exonuclease and the homolog to apoptosis-inducing factor (AIF), that were up-regulated during xylem maturation in the microarray analysis (Figure 7) are potential candidates for such nucleases.

Degradative processes in the cytoplasm and the nucleus overlapped with the secondary cell-wall formation in the xylem fibers, making it difficult to distinguish the specific regulatory pathways that relate to each process. The comparative transcriptomic approach used in this study allowed us to separate these processes, and to identify regulatory pathways specific to fiber PCD (Figure 8). Several previously uncharacterized proteins were also identified, including transcription factors and completely unknown proteins. In addition, two signaling compounds were implicated in the control of fiber cell death. Ethylene is a known regulator of xylem formation, but the involvement of sphingolipid metabolites has not been previously reported, even though their function in other stress and/or cell-death processes is well established (Hofius et al., 2007).

What then is the trigger for cell death in fibers? Our data suggest at least three possible scenarios. In the first scenario, cell death is not an independently controlled process, but an inevitable fate of the fibers after completion of secondary cell-wall formation. According to this model, the fiber-specific gene expression pattern identified in our analyses merely modifies the progress of fiber cell death, but does not affect the overall outcome. However, genetic separation of cell-wall formation and cell death, as demonstrated previously in xylem vessels (Turner and Hall, 2000; Mitsuda et al., 2005; Zhong et al., 2006), casts doubt on this scenario. The second scenario involves triggering of fiber cell death in response to nutrient starvation, which is supported by activation of autophagy genes that are known to respond to nutrient starvation. Although the nutritional status of maturing fibers is not known, the presence of steep concentration gradients of sugars across the vascular tissues of Scots pine stem supports rapid impairment of the nutritional status in the course of xylem development (Uggla et al., 2001). A third scenario is based on identification of the PHYA photoreceptor gene as a fiber PCD candidate gene. A blue-light receptor (CRY2) was also significantly up-regulated in the MX3 sample (Table S2). It is currently not clear how light could affect fiber maturation in the stem, and it is therefore tempting to suggest that the photoreceptors might be involved in interplay with other stimuli such as gravity, in analogy to what has been observed in Arabidopsis hypocotyls (Lariguet and Fankhauser, 2004). Xylem fibers are known to respond to gravitropic stimulus, for instance during tension wood formation, during which a shift in the gravitational vector causes changes in xylem differentiation as well as cell death. Irrespective of the initial trigger, our results clearly demonstrate that fibers possess a unique set of morphological and transcriptional alterations, leading them to actively participate in their own demise.

Experimental procedures

Plant material and growth conditions

Micropropagated Populus (Populus tremula L. × P. tremuloides Michx.) clone T89 trees were grown in the greenhouse under natural day length, supplemented, during the winter season, with 12 h daily illumination from metal halogen lamps. The trees were sampled during the spring or the summer when they were 6–12 months old, by collecting stem samples for each type of analysis as close to each other as possible from each of three replicate trees (trees 1–3). P. trichocarpa Torr. & A. Gray × P. deltoides L. (clone H11-11) and P. tremula × P. tremuloides (back-crossed to parent P. tremula) trees were also grown in an open field and sampled for electron microscopy analysis using HPF-FS, as described below.

Electron microscopy

Longitudinal stem sections were collected from P. trichocarpa × P. deltoides and P. tremula × P. tremuloides trees using a sharp double-edged razor blade prior to processing for HPF-FS, as described by Samuels et al. (2002) with minor modifications. Samples were kept overnight in 5% Spurr resin (Sigma, http://www.sigmaaldrich.com), then for 3 h in each of 8, 25, 50 and 75% resin, and finally 12 h in 100% resin. After two additional incubations for 4 h in 100% resin, samples were polymerized overnight at 70°C. For transmission electron microscopy (TEM), sections about 70 nm thick were taken with an ultramicrotome and treated as described by Samuels et al. (2002), except for the uranyl acetate step being shortened to 20 min. Sections were examined under a Hitachi H7600 transmission electron microscope (http://www.hitachi.com). For conventional fixation, Populus tremula × tremuloides stem sections from trees 1–3 were embedded in Spurr resin as described by Rensing (2002) before TEM examination.

TUNEL assay

Transverse and radial cryosections (100 μm thick) were taken from the vascular tissues of trees 1–3. The sections were fixed on coated slides in freshly prepared 4% paraformaldehyde in PBS pH 7.4 for 20 min at room temperature, and analyzed using a terminal deoxynucleotidyl transferase (TdT)-mediated dUTP-FITC nick end labeling (TUNEL) assay kit according to the manufacturer’s instructions (Roche, http://www.roche.com). The positive control included DNA digestion by the restriction enzyme ApaI for 1 h at 37°C prior to TUNEL staining. The TdT enzyme was omitted from the TUNEL assay for the negative control. The slides were scanned using a confocal Leica TCS SP2 AOBS scanning system mounted on a Leica DM IRE2 inverted microscope employing Leica TCS software (http://www.leica.com), with excitation at 488 nm and emission at 500–550 nm. The cell walls were stained with 0.1% calcofluor (excitation at 405 nm/emission at 410–450 nm). The obtained images were projections of 120 xyz confocal scans to tissue depths of 60–80 μm.

Viability staining

Populus stem sections were incubated for 1 h under strong light conditions in 50 mm sodium phosphate buffer (pH 7.6) containing 50 mm sodium succinate and 500 mg/L nitroblue tetrazolium. Sections were cut by hand or a vibratome, and analyzed using an Axioplan 2 microscope (Zeiss, http://www.zeiss.com).

Single-cell gel electrophoresis (Comet assay)

Tissue-specific sampling was performed by taking 50 μm thick tangential cryosections (Uggla et al., 1996) from the living xylem of trees 1–3. Nuclei were isolated as described by Gichner and Mühlfeldová (2002) with the following modifications; each section was chopped using scissors in 18 μl 0.4 m Tris/HCl pH 7.5, and then shaken on ice at 250 rpm for 1 h. The neutral Comet assay was performed as described by Olive and Banáth (1995) with slight modifications. Buffer containing the nuclei, but no visible debris, was mixed with 70 μl of 1% agarose prior to mounting between a frosted 1% agarose-coated microscopy slide and coverslip. After gelling on ice, coverslips were removed and the slides were incubated on ice in a lysis buffer (2.5 m NaCl, 100 mm EDTA, 10 mm Tris/HCl pH 7.5, 10% DMSO, 2% Triton X-100 and 2% sodium lauryl sulfate) for 1 h with shaking at 50 rpm. The slides were subsequently washed three times in 0.4 m Tris/HCl (pH 7.5), subjected to horizontal electrophoresis in 1 × Tris-Acetate EDTA buffer at 1 V cm−1 for 20 min, washed three times in distilled water, then treated for preservation as described by Woods et al. (1999) prior to staining and analysis by dehydration and subsequent rehydration. All Comet nuclei were visualized and recorded using an Axioplan 2 epifluorescent microscope (Zeiss). The Comet tail lengths and moments were analyzed using Comet Assay IV software (Perceptive Instruments, http://www.perceptive.co.uk/applications/comet.htm).

Microarray analysis

A longitudinal stem piece, including the majority of the cortical and phloem tissues as well as the majority of the woody tissues, was cut with a razor blade from the base of tree 1 and frozen on dry ice for the microarray analysis. Immediately above, a second sample was taken for electron microscopy analysis to define the location of the various stages of xylem differentiation. The frozen sample was sectioned in the tangential plane at 50 μm with a cryomicrotome as described by Uggla et al. (1996). Seven cryosections, representing various stages of vascular development, were chosen for the microarray analysis (Figure 4). The intervening sections were used to analyze the DNA degradation pattern for verification of the developmental stages of the selected samples. Isolation of mRNA from the cryosections and synthesis of cDNA were performed as described by Moreau et al. (2005).

The amplified cDNA samples were hybridized for microarray analysis in an all-versus-all design, using the POP2 array (Moreau et al., 2005) with an automated slide processor (Amersham Lucidea SlidePro; http://www.amersham.com). The slides were scanned and the resulting images were analyzed as described by Smith et al. (2004), then the results were imported to the microarray repository UPSC-BASE (Sjödin et al., 2006) for stepwise normalization (Wilson et al., 2003) and restricted linear scaling (Ryden et al., 2006). Images, raw data and normalized data are stored and are freely accessible as experiment UMA-0064 in UPSC-BASE. Redundancy was removed by median aggregation of multiple clones corresponding to the same gene models. To create an in silico reference point (Diaz et al., 2003), linear model transformation using LIMMA (Wilson et al., 2003) was implemented.

Gene models that had a corrected P value < 0.05 were regarded as significant, and annotations were extracted for them from PopulusDB (Sterky et al., 2004) and the Arabidopsis Information Resource (Huala et al., 2001). Proteases (García-Lorenzo et al., 2006), transcription factors (Riano-Pachon et al., 2007; Zhu et al., 2007) and KOG classes (Tuskan et al., 2006) were downloaded and tested for over-representation using Fisher exact tests. The gene models for the proteases and transcription factors are indicated in Table S1. Heatmaps were produced using Pearson correlation with the average agglomeration method in the statistical software package R (Ihaka and Gentleman, 1996).

In silico selection of the fiber cell-death candidate genes

Genes were selected that showed significant up-regulation in the xylem maturation zone (MX1–3) of Populus (2156 genes; Table S2), and compared to existing Arabidopsis microarray datasets to identify transcriptional changes specifically related to fiber cell death. Gene expression data for the closest Arabidopsis homologs were gathered from the response viewer tool of Genevestigator V2 (http://www.genevestigator.ethz.ch; Zimmermann et al., 2004). Of the 83 stress-related experiments included in the V2 Genevestigator tool, most of the common marker genes for xylem cell death (see text highlighted in red inTable S2) were found be highly expressed in the ‘PCD:Senescence’ experiment (senescing cell cultures of Arabidopsis; Swidzinski et al., 2002), which suggests that this experiment is indicative of transcriptional regulation during xylem cell death. Consequently, as the first selection criterion, Populus genes were selected for which the Arabidopsis homolog had a value greater than two in the ‘PCD:Senescence’ experiment. As the second selection criterion, genes with apparent expression in tracheary elements or vessels were excluded on the basis of (i) up-regulation (ratio > 1.3) in the Arabidopsis Genevestigator experiment ‘Hormone:BL/H3BO3’, which represents in vitro gene expression in differentiating tracheary elements (Kubo et al., 2005), (ii) enrichment in the maturing xylem (Brady et al., 2007) and (iii) up-regulation during stele development (Birnbaum et al., 2003). As a third selection criterion, genes were included only if they were previously identified in any of three Arabidopsis microarray datasets representing secondary xylem development (Ko and Han, 2004; Ehlting et al., 2005; Zhao et al., 2005). As Arabidopsis xylem lacks ray cells, this last selection criterion allows exclusion of transcriptional changes in Populus that possibly originated from the ray cells. All data from the response viewer tool of Genevestigator V2 and from the additional microarray datasets for the selected Populus genes are shown in Table S2.


We wish to acknowledge Kermit Ritland (Department of Forest Sciences, University of British Columbia, Canada) for providing populus material and Lenore Johansson (Umeå Plant Science Centre, Umeå University, Sweden), Garnet Martens and Derrick Horne (University of British Columbia Bioimaging Facility, Canada) for help with the electron microscopy analyses. This work was supported by the Swedish National Graduate School for Genomics and Bioinformatics, the Swedish Research Council Formas, the Carl Trygger Foundation and the Swedish Research Council.