Post mortem function of AtMC9 in xylem vessel elements

Summary

• Cell death of xylem elements is manifested by rupture of the tonoplast and subsequent autolysis of the cellular contents. Metacaspases have been implicated in various forms of plant cell death but regulation and execution of xylem cell death by metacaspases remains unknown.
• Analysis of the type II metacaspase gene family in Arabidopsis thaliana supported the function of METACASPASE 9 (AtMC9) in xylem cell death. Progression of xylem cell death was analysed in protoxylem vessel elements of 3-d-old atmc9 mutant roots using reporter gene analysis and electron microscopy.
• Protoxylem cell death was normally initiated in atmc9 mutant lines, but detailed electron microscopic analyses revealed a role for AtMC9 in clearance of the cell contents post mortem, that is after tonoplast rupture. Subcellular localization of fluorescent AtMC9 reporter fusions supported a post mortem role for AtMC9. Further, probe-based activity profiling suggested a function of AtMC9 on activities of papain-like cysteine proteases.
• Our data demonstrate that the function of AtMC9 in xylem cell death is to degrade vessel cell contents after vacuolar rupture. We further provide evidence on a proteolytic cascade in post mortem autolysis of xylem vessel elements and suggest that AtMC9 is part of this cascade.

Introduction

Vascular plants transport water and solutes in xylem trachea that are formed by end-to-end fusion of longitudinal tracheary elements (TEs). TEs are reinforced by deposition of secondary walls that consist of cellulose, hemicelluloses and lignin. After secondary wall deposition, the TE undergoes programmed cell death (PCD), leading to removal of the entire protoplast and formation of a hollow conduit for water transport. TE cell death is a type of plant vacuolar cell death, characterized by rupture of the tonoplast and subsequent release of vacuolar hydrolases that execute autolysis of the cellular contents (van Doorn et al., 2011a). TE autolysis is normally fully completed within a few hours after vacuolar rupture (Groover et al., 1997). Disruption of the tonoplast is followed by swelling and degradation of organelles with single membranes, such as Golgi and ER. Later, organelles with double membranes such as mitochondria are degraded (Fukuda, 1996). Although DNA is degraded rapidly within 10–20 min after tonoplast rupture, the nucleus itself seems to be the last organelle to be degraded during autolysis (Fukuda, 1996; Groover et al., 1997; Obara et al., 2001; Bollhöner et al., 2012). Although the order of events during TE autolysis is now well established, very little is known about the autolytic machinery and its regulation.

A variety of hydrolytic enzymes, including different types of nucleases, lipases and proteases, have been identified in connection to TE cell death (Thelen & Northcote, 1989; Aoyagi et al., 1998; Beers et al., 2000; Ohashi-Ito et al., 2010; Bollhöner et al., 2012; Petzold et al., 2012). ZINNIA ENDONUCLEASE1 (ZEN1), an S1-type nuclease, was shown to be required for post mortem DNA degradation in Zinnia elegans TEs (Ito & Fukuda, 2002). Serine and cysteine protease activities have been shown to increase during TE differentiation (Minami & Fukuda, 1995; Ye & Varner, 1996; Beers & Freeman, 1997; Groover & Jones, 1999) and, consequently, pharmacological inhibition of cysteine proteases in xylogenic cell cultures blocked or delayed TE differentiation, but only if applied before formation of secondary wall thickenings (Fukuda, 1996; Woffenden et al., 1998; Twumasi et al., 2010). While this suggests a role for cysteine proteases in early signalling of TE differentiation, it does not exclude roles during autolysis. Normal progression of xylem vessel autolysis was indeed shown to require the papain-like cysteine proteases (PLCPs) XYLEM CYSTEINE PEPTIDASE 1 (XCP1) and XCP2 (Avci et al., 2008). XCP1 and XCP2 are specifically expressed in xylem vessel elements (Funk et al., 2002) and localize to the vacuole, where they function in micro-autolysis. After vacuolar collapse, they participate in mega-autolysis of the cell contents (Avci et al., 2008). However, even in the absence of XCP1 and XCP2, xylem vessels are ultimately cleared and are obviously functional, suggesting involvement of additional proteases in TE autolysis (Avci et al., 2008).

In metazoans, apoptotic cell death is mediated by caspases which are cysteine proteases with specificity towards aspartate residues in their target proteins. Although caspase-like activities have frequently been demonstrated in plants in connection to various cell death processes (Bonneau et al., 2008), no orthologous genes corresponding to metazoan caspases have been found in plant genomes. Instead, plants have metacaspases, which are cysteine proteases with a structure largely similar to caspases (Uren et al., 2000; McLuskey et al., 2012; Wong et al., 2012). Like caspases, metacaspases have a catalytic Cys-His diad, a large (p20) and a small (p10) subunit. The presence of an N-terminal prodomain distinguishes type I metacaspases from type II metacaspases, which lack this prodomain but contain a linker region between their subunits (Tsiatsiani et al., 2011). Despite this structural similarity, metacaspases lack aspartate specificity and instead cleave after arginine and lysine (Vercammen et al., 2004; Watanabe & Lam, 2005; He et al., 2008) and cannot, therefore, account for the caspase-like activities in plants (Vercammen et al., 2007; Bonneau et al., 2008).

In analogy to metazoan caspases, plant metacaspases have been assigned roles in stress- and hypersensitive response (HR) immunity-related cell death in Arabidopsis thaliana (van Baarlen et al., 2007; He et al., 2008; Coll et al., 2010; Watanabe & Lam, 2011), whose genome encodes three type I (AtMC1-3) and six type II metacaspases (AtMC4-9; Tsiatsiani et al., 2011). A type II metacaspase has also been implicated in developmental cell death of the embryo suspensor in Norway spruce (Picea abies; Suarez et al., 2004). The role of metacaspases during developmental cell death of Arabidopsis is not known, but the Arabidopsis thaliana METACASPASE 9 (AtMC9) has been implicated in TE PCD on the basis of its expression pattern during TE differentiation (Kubo et al., 2005; Turner et al., 2007; Ohashi-Ito et al., 2010). Also in hybrid aspen, two AtMC9 homologues exhibited expression specifically in xylem elements undergoing cell death (Bollhöner et al., 2012). AtMC9 is a type II metacaspase that undergoes autocatalytic processing and has – as the only metacaspase reported so far – an acidic pH optimum (Vercammen et al., 2004; Tsiatsiani et al., 2011). AtMC9 has also been shown to cleave the serpin-like suicide inhibitor AtSerpin1 (Vercammen et al., 2006) and to be regulated by S-nitrosylation (Belenghi et al., 2007). Here we report on a role of AtMC9 during post mortem clearance of xylem vessel elements. We present evidence that AtMC9 is specifically expressed in developing xylem vessel elements and that AtMC9 is required for efficient progression of autolysis during vessel cell death.

Materials and Methods

Plant material and growth conditions

Arabidopsis thaliana (L.) Heynh wild-type (Col-0), atmc9-1 (GABI_540_H06; Kleinboelting et al., 2012) and atmc9-2 (SALK_075814; Scholl et al., 2000), and xcp1 (SALK_084789) and xcp2 (SALK_010938; Avci et al., 2008) were used. Soil grown plants were grown in growth chambers in long day conditions (16 h light : 8 h darkness). In vitro plants were grown vertically on MS (Duchefa, Haarlem, the Netherlands) plates in growth rooms under long day conditions (16 h light : 8 h darkness, 21°C : 18°C).

Cloning and plant transformation

For translational fusions (GUS, GFP, mCherry), metacaspase genomic sequences were amplified from genomic DNA together with 5′ upstream regions of varying length and recombined into pDONR207 (Invitrogen). Resulting entry clones pENTR207-AtMC4_pro::AtMC4, pENTR207-AtMC7_pro::AtMC7 and pENTR207-AtMC9_pro::AtMC9 were recombined into pBGGUS (Kubo et al., 2005), pENTR207-AtMC5_pro::AtMC5, pENTR207-AtMC6_pro::AtMC6 and pENTR207-AtMC8_pro::AtMC8 into pKGWFS7 (Karimi et al., 2002). pENTR207-AtMC9_pro::AtMC9 was also recombined into pMDC107 (Curtis & Grossniklaus, 2003) for GFP fusion and into pKGWmCherry.

For transcriptional fusions, 1410 and 670 bp promoter fragments 5′ of the AtMC9 and BFN1 coding sequences (cds), respectively, were amplified by primers AtMC9promattB1/AMC9promattB2 and BFN1promattB1/BFN1promattB2, respectively. After recombination into pENTR207 (Invitrogen) both were recombined into pBGGN and AtMC9_pro also into pMDC110 (Curtis & Grossniklaus, 2003).

All expression clones were transformed by electroporation into Agrobacterium tumefaciens strain GV3101::pMP90RK (except pXMDC107-pAtMC9::AtMC9 which was transformed into strain GV3101::pMP90) and transferred into Arabidopsis Col-0 by floral dipping (Clough & Bent, 1998). Several independent lines were analysed to select the representative lines; at least five lines for each construct were analysed for GUS activity, 15 lines for nGFP reporter expression, and 4–5 lines for GFP and mCherry fusion protein constructs.

In order to generate the binary destination vector pKGWmCherry, we amplified the mCherry cds (Clontech; Takara Bio Europe, Saint-German-en-Laye, France) by primers SpeI-cherry-f and cherry-SacI-r and fused it 3′ to the GATEWAY cassette in pKGW + T35S (pKGW (Karimi et al., 2002) with 35S terminator cloned into HindIII site 3′ of attR2). All primer sequences are listed in Supporting Information Table S1.

Histochemical GUS staining

For histochemical GUS staining, in vitro grown seedlings or tissue samples from soil grown plants were incubated for 90 min to 4 h or overnight at 37°C in a solution containing 1 mM X-Gluc, 1 mM K₃Fe(CN)₆, 1 mM K₄Fe(CN)₆ and 0.1% Triton x-100 in 50 mM sodium phosphate buffer (Na₂HPO₄/NaH₂PO₄, pH 7.0). Samples were destained in 70% and 100% ethanol. For whole mounts, samples were rehydrated and mounted in 1 M chloralhydrate in 33% glycerol. For sectioning, seedlings were embedded in LR white resin with 10% PEG400. Samples were incubated once in LR White-PEG/ethanol (50 : 50), twice in 100% LR White-PEG (each for 10 min) and finally embedded in LR White-PEG400 with added accelerator and polymerised overnight at 65°C. Twenty micrometre sections were made using a microtome, heat fixed on glass slides and mounted in Entellan^® (Merck Millipore, Darmstadt, Germany). Light microscopic images were acquired using a Zeiss Axioplan II microscope equipped with an AxioCam CCD camera (Zeiss, Jena, Germany).

RT-PCR analysis of atmc9 mutants

Total RNA was extracted from 3-d-old seedlings grown on MS plates using RNeasy Plant Mini Kit (Qiagen) according to manufacturer's instructions. RNA was treated with Ambion Turbo DNA-free™ (Ambion, Life Technologies, Paisley, UK) according to manufacturer's instructions. cDNA was synthesised using qScript™ cDNA Synthesis Kit (Quanta Biosciences, Gaithersburg, MD, USA) with an initial denaturation step for 5 min at 65°C for samples containing 8 μl RNA (400 ng), 3 μl H₂O and 3 μl 5× qScript Reaction Mix. After cooling to 22°C, premixed 1 μl qScript RT, 1 μl 5× qScript Reaction Mix and 4 μl H₂O was added, and cDNA synthesis was run according to the manual. cDNA was 10-fold diluted with 10 mM Tris pH 8.0, and 1, 2 and 3 μl were used as template in the PCR with primers EF1αf+r, AtMC9q4f+4r and AtMC9q3f+2r, respectively. PCR was run with 35 cycles (95°C 10 s, 56°C 10 s, 72°C 40 s) using GoTag^® Green Master Mix (Promega, Madison, WI, USA). The reference gene was EF1α (At5g60390).

RNA extraction and microarray analysis

Roots of 3-d-old seedlings were separated on plate from hypocotyls and cotyledons using a razor blade, collected, frozen in liquid nitrogen and homogenised with a polypropylene pestle for microcentrifuge tubes. Forty roots were pooled per replicate (n = 3). RNA was extracted with RNeasy Plant Mini Kit (Qiagen, Hilden, Germany) according to instructions and treated with Ambion Turbo DNA-free™ (Applied Biosystems) according to instructions. Microarray analysis and data normalisation were carried out by the Nottingham Arabidopsis Stock Centre (NASC) Affymetrix Service (Craigon et al., 2004) and made publicly available through the NASCArrays database (http://affymetrix.arabidopsis.info/narrays/experimentbrowse.pl), experiment reference number NASCARRAYS-567. Data analyses were performed using Microsoft Excel, GO annotations were analysed using web tools (http://arabidopsis.org/tools/bulk/go/index.jsp).

PLCP activity profiling

PLCP activity profiling of protein extracts from 4-d-old Arabidopsis seedlings roots was performed as described previously (van der Hoorn et al., 2004).

AtMC9 cleavage assay

Full-length cDNA of XCP2 was amplified with primers AtXCP2-F and AtXCP2-R, cloned into vector pET28a (Novagen; Merck Millipore, Darmstadt, Germany) using restriction enzymes BamHI and XhoI and used for in vitro transcription and translation of radio-labelled protein according to the manufacturer's instructions (TnT^® T7/T3 Coupled Reticulocyte Lysate System; Promega). Five microlitres of the 50 μl reaction were incubated at 30°C for 30 min with a range of rAtMC9 concentrations (0–250 nM), prepared in AtMC9 activity buffer (50 mM MES, pH 5.5, 150 mM NaCl, 10% (w/v) sucrose, 0.1% (w/v) CHAPS and 10 mM DTT). The inactive AtMC9 mutant (rAtMC9C147AC29A, indicated as rMC9-CA) was used as a negative control (Belenghi et al., 2007). Assay products were separated on SDS-PAGE, which were dried, exposed to phosphor imaging plates and scanned by a PhosphorImager 445SI (Molecular Dynamics Inc., Sunnyvale, CA, USA).

TEM analysis

Seedlings grown vertically for 3 d (72 h) on 1 × MS plates (1% Plant Agar (Duchefa), 1% sucrose) were fixed using 2.5% (v/v) glutaraldehyde in 0.1 M cacodylate buffer overnight at 4°C. After washing three times in buffer, the specimens were post-fixed with 1% (v/v) osmium tetroxide in the medium buffer for 1 h and washed twice in distilled water. Samples were dehydrated with 50%, 70%, 95%, 100% ethanol and infiltrated and embedded in Spurr's resin. Roots were sectioned from the tip upwards using a glass knife on Leica EM UC7 ultramicrotome (Leica Microsystems, Wetzlar, Germany). Roots were continuously oriented transversely to the root axis to allow recording of the distance from the root tip by counting sections. Ultrathin (80 nm) sections were made using a Diatome diamond knife on Leica EM UC7, collected onto copper grids, then treated with 5% uranyl acetate in water for 60 min followed by Sato's lead staining for 5 min. Sections were examined in a Jeol 1230 TEM (Jeol, Tokyo, Japan). Digital images were captured using a Gatan MSC 600CW (Gatan, Warrendale, PA, USA).

Measurement of protoxylem life time

For measurements of protoxylem life time, Z-stack images of 3-d-old roots (counterstained with 10 μM propidium iodide in MS medium) were acquired with a Leica TCS SP2 inverted confocal laser scanning microscope (cLSM) using the 488 nm line of an Ar/Kr laser. Images were processed with Leica confocal LCS Lite software and semi-automated merged with Adobe Photoshop CS3 (Adobe Systems Inc., San Jose, CA, USA) into images covering the entire root. Measurements of the distance from the last protoxylem GFP signal to the root tip were performed with Image J (http://rsbweb.nih.gov/ij/).

Analysis of subcellular localisation of fluorescent AtMC9 fusions

Seedlings were analysed using Leica TCS SP2 and Zeiss LSM780 inverted cLSM, using 488 and 514 nm Ar/Kr laser lines. For GFP analysis, 10 μM propidium iodide counterstaining in ½ MS medium was applied for 10 min and rinsed in ½ MS medium before observation.

Xylem anatomy

Hypocotyls and stem bases of 8-wk-old plants were fixed in FAA, hand-sectioned with a razor blade, stained with toluidine blue O and embedded in Entellan (Merck) or 50% glycerol. Light microscopic images were acquired using a Zeiss Axioplan II microscope equipped with an AxioCam CCD camera (Zeiss).

For root protoxylem vessel element length determination, roots of 10-d-old in vitro grown seedlings were macerated for 5 h at 95°C in a solution comprising equal volumes of 7.5% H₂O₂ and glacial acetic acid. After washing, cells were disintegrated by gentle shaking. Hypocotyl macerations were made as described (Muñiz et al., 2008). Cells were measured on DIC micrographs acquired with Zeiss Axioplan II microscope, using Zeiss Axiovision software or Image J for cell measurements.

Pathogenicity assays

Leaf infection with Pseudomonas syringae Pst DC3000 and Xanthomonas campestris pv. campestris Xcc8004 were performed as reported previously by infiltrating the leaf mesophyll with bacterial suspension (5 × 10⁵ CFU ml⁻¹) using a syringe without a needle (Katagiri et al., 2002; Meyer et al., 2005). For each experiment, four leaves per plant (10 plants per genotype) were inoculated. Plants were then placed under 9 h light period at 22°C and 70% relative humidity.

In order to induce HR cell death, 3-wk-old plants were infected with Pseudomonas syringae DC3000 bearing AvrB by leaf infiltration using a needleless syringe with bacteria suspension (5 × 10⁷ CFU ml⁻¹) in 10 mM MgCl₂. Ion leakage from leaf discs into ddH₂O was monitored as a change in conductivity (μS cm⁻²) with a Metler Five Easy conductivity meter.

Ralstonia solanacearum root infections were performed as previously described (Deslandes et al., 1998) by dipping roots of 4-wk-old plants into a bacterial suspension of GMI1000 wild-type strain (5 × 10⁷ CFU ml⁻¹). Disease index was scored in leaves: 0 = no symptoms, 1 = 25% wilted leaves, 2 = 50% wilted leaves, 3 = 75% wilted leaves, 4 = 100% wilted leaves (or dead plant). After inoculation, plants were transferred to a growth chamber (12 h photoperiod, 27°C, 80% relative humidity). For each experiment, at least 30 plants per genotype were analysed, in 6 (atmc9-1) and 3 (atmc9-2) independent experiments. Bacterial growth in planta was measured as previously described (Deslandes et al., 1998).

Results

AtMC9 expression is confined to root cells undergoing cell death

In order to compare expression of AtMC9 and all other type II metacaspase family members in xylem development, a series of translational reporter fusion lines was created. A GUS reporter was fused to a genomic fragment containing the promoter (AtMC_pro) and the coding sequence of each type II metacaspase. Transgenic seedlings were analysed by histochemical GUS staining. In 3-d-old seedlings, AtMC4 showed by far the strongest GUS staining among the type II metacaspase genes and was expressed ubiquitously in root tissues (Fig. 1a,b). AtMC5 was expressed in lateral root cap cells (Fig. 1c) and ubiquitously in root tissues above the elongation zone (Fig. 1d). We could not detect any expression of AtMC6, AtMC7 and AtMC8 in 3-d-old seedling roots (Fig. 1e–g), and therefore older seedlings were analysed for these lines. In 8-d-old seedlings still no expression was present in the young parts of the root for any of the AtMC6, AtMC7 and AtMC8 lines. However, in the older, upper part of the root, AtMC6 was expressed in cortical and endodermal cells (Fig. 1j), while AtMC7 expression was found in the stele (Fig. 1k) and AtMC8 in a few cells of the root, in a patchy pattern (Fig. 1l). In contrast to the other type II metacaspases with often rather broad expression domains, we observed expression of AtMC9 in only two cell types of 3-d-old roots; in lateral root cap cells (Fig. 1h) and in developing xylem vessels (Fig. 1i). Expression of AtMC9 was obvious in developing xylem vessels in all analysed parts of the plant (Fig. 1m,n). The specific expression pattern in root cap cells and vessel elements – both known to undergo PCD – strongly suggests a role for AtMC9 in developmental cell death in these particular types of cells.

Figure 1

Expression of Arabidopsis type II metacaspases. Histochemical GUS staining of transgenic Arabidopsis lines expressing translational METACASPASE:GUS fusions under control of the respective endogenous promoters. Expression profile of (a, b) AtMC4, (c, d) AtMC5, (e) AtMC6, (f) AtMC7, (g) AtMC8 and (h, i) AtMC9 in the root tip (a, c, h and left images in e–g) and in the root differentiation zone of 3-d-old seedlings (b, d, i and right images in e–g). (j) AtMC6, (k) AtMC7 and (l) AtMC8 were only expressed in the upper part of 8-d-old seedling roots. (m, n) AtMC9 expression was detected specifically in developing xylem vessel elements in the entire plant, (m) inflorescence stem cross section, (n) leaf vasculature. At least five independent GUS lines were analysed for each construct, and a representative pattern is shown. Arrowheads indicate xylem poles; if, interfascicular fibres; p, phloem; x, xylem. Bars, 50 μm.

AtMC9 is not required for xylem cell death initiation

In order to study the role of AtMC9 in xylem development, we isolated two atmc9 knock-out mutants from T-DNA insertion libraries (Scholl et al., 2000; Kleinboelting et al., 2012). Both atmc9-1 (GABI_540H06) and atmc9-2 (SALK_075814) contain the insert in the coding sequence and lack detectable full length AtMC9 transcript (Fig. 2a,b). These two mutant alleles did not display any observable alterations in the overall plant growth, anatomy of the vascular tissues or xylem cell morphology (Fig. S1). Due to the involvement of some metacaspases in HR-related cell death responses, atmc9 mutants were tested for four different pathogens; avirulent Pseudomonas syringae pv. Tomato (Pst) DC3000 avrB, virulent PstDC3000, and two vascular pathogens Xanthomonas campestris pv. campestris and Ralstonia solanacearum. The only observed difference was enhanced disease development in response to R. solanacearum associated with increased bacterial growth in the two atmc9 mutant lines (Fig. S2), which is suggestive of structural or functional defects in xylem vessel elements.

Figure 2

Timing of xylem cell death and expression of cell death related genes in the atmc9 mutants. (a) T-DNA insertion sites of atmc9-1 and atmc9-2 in the AtMC9 gene. The position of the p10 and p20 protein domains is indicated below in grey. (b) Detection of AtMC9 and EF1α (control) expression by RT-PCR in Arabidopsis Col-0 and the mutants. atmc9-1 contains truncated AtMC9 transcript, detected by primer pair AtMC9q3f+2r that amplifies an AtMC9 fragment 3′ of the atmc9-1 insertion site. (c–e) The moment of protoxylem cell death in the atmc9 mutant and the wild-type. Values depict the distance from the root tip to the location where protoxylem vessel elements lost the nuclear localised GFP signal from (c, d) AtMC9_pro: nGFP and (e) BFN1_pro:nGFP reporter lines. Data are means (± SD) of three to six roots. (f, g) Expression level of (f) the metacaspase gene family and of (g) xylem-expressed cell death-related genes such as the PLCPs XCP1, XCP2, CEP1, CEP2 and XBCP3, the serine protease XSP1 and the nuclease BFN1. (Col-0, black bars; atmc9-1, light grey bars; atmc9-2, mid grey bars.) Note the truncated transcript (b) in atmc9-1, causing the AtMC9 level to appear as wild-type in the microarray analysis. Level of expression is plotted as mean signal intensity (± SD) from a microarray analysis of 3-d-old seedling roots (n = 3).

In order to assess xylem development in more detail, we analysed expression of xylem-specific reporters in the atmc9-1 and atmc9-2 mutant backgrounds. These reporter lines express GFP with a nuclear localisation signal (nGFP) under control of promoters of two xylem-specific genes; AtMC9 and BIFUNCTIONAL NUCLEASE1 (BFN1; Fig. S3). BFN1 is a homologue of the Zinnia ZEN1, which is required for post mortem DNA degradation in Zinnia TEs (Ito & Fukuda, 2002). Both AtMC9 and BFN1 are expressed late during xylem differentiation (Turner et al., 2007; Farage-Barhom et al., 2008; Figs 1, S3), and any alteration in late differentiation of xylem elements in a mutant is therefore expected to affect the spatio-temporal expression profile of these reporters. Reporter expression was analysed in the two protoxylem strands of the primary root of 3-d-old seedlings. Protoxylem vessel elements were chosen because they express AtMC9 late during differentiation (Figs 1, S3) and because they display a well-defined developmental series of differentiation and cell death that is easily detectable in young, in vitro grown seedling roots.

Both reporter lines exhibited strong GFP signal in late maturing protoxylem vessel elements in a similar manner in both the atmc9 and the wild-type backgrounds (Fig. S3). A very clear and sudden disappearance of the nGFP signal in the two protoxylem cell files was observed in both reporter lines (Fig. S3), most likely due to quenching of GFP fluorescence as a result of cellular acidification upon vacuolar rupture (Kneen et al., 1998; Llopis et al., 1998). The nGFP fluorescence disappeared at the same distance from the root tip in atmc9 background as in wild-type (Fig. 2c–e), suggesting that the lifetime of the xylem elements is not altered in atmc9 and, hence, that AtMC9 is not involved in initiation of cell death in vessel elements.

A microarray analysis was performed to test whether a xylem cell death phenotype in atmc9 was masked by upregulation of an alternative cell death pathway. Candidates that could operate in such a pathway are other metacaspase family members and hydrolases, such as the nuclease BFN1 and the PLCPs XCP1, XCP2, which are known to function in xylem cell death (Ito & Fukuda, 2002; Avci et al., 2008), as well as the xylem-expressed PLCPs XYLEM BARK CYSTEINE PEPTIDASE 3 (XBCP3), CYSTEINE ENDOPEPTIDASE 1 (CEP1) and CEP2, and the XYLEM SERINE PROTEASE 1 (XSP1). Gene expression analysis revealed no clusters of functionally related genes that were differentially expressed in 3-d-old seedling roots of the mutants in comparison to wild-type (data not shown). Expression of metacaspase genes, except for AtMC9, was not altered significantly, and neither that of the putative xylem cell death related genes (Fig. 2f,g). In conclusion, atmc9 mutants do not upregulate any other putative regulators of xylem cell death on the transcriptional level to compensate for the absence of AtMC9 in xylem development.

AtMC9 is required for fast post mortem xylem vessel autolysis

A transmission electron microscopy (TEM) analysis of the primary root vessel elements was undertaken in atmc9-2 and wild-type seedlings to verify the results on cell death initiation obtained with the reporter gene analysis, and to analyse post mortem events that were not possible to visualise with the help of the reporter lines. For the TEM analysis, roots of 3-d-old seedlings were sectioned from the tip upwards keeping track on the distance from the root tip. Preliminary TEM analyses together with the reporter gene analyses had revealed that protoxylem cell death occurs at a distance of 1500–2000 μm from the root tip, and ultrathin cross-sections were therefore taken every 100 μm within this region and analysed by TEM (Fig. 3a). The average length of mature protoxylem vessels elements was c. 180 μm in both the wild-type and the atmc9-2 mutant (Fig. S1). Each cell is therefore represented on either one or two TEM images. This painstaking sampling procedure allowed tracking the exact position and comparison of the different developmental stages between the atmc9 mutant and the wild-type.

Figure 3

Progression of cell death in protoxylem vessel elements. (a) Expression pattern of AtMC9_pro:: nGFP in 3-d-old Arabidopsis seedling root. Detailed TEM analysis of the protoxylem vessel elements was performed at a distance of 1500–2000 μm from the root tip where cell death was initiated by vacuolar rupture and executed by autolysis. (b) Detailed image of AtMC9_pro:: nGFP expression in the two protoxylem files, characterised by the spiral patterning of their secondary cell walls, stained by propidium iodide. (c–i) Transmission electron micrographs from transverse sections of several Col-0 roots, illustrating different stages of protoxylem vessel death within the selected 1500–2000 μm distance from the root tip. Micrographs are from several different roots as, due to the fast rate of autolysis, not all stages can be observed in a single protoxylem cell file. (c–e) Intact cells showing partly dilated ER and vesicle-rich cytoplasm. (f, g) Early stages of autolysis, with the tonoplast at least partly ruptured but organelles still intact. (h) Late stage of autolysis with grainy remnants and no detectable subcellular structures such as membranes or organelles. (i) A cleared vessel after final autolysis. Bars: (a) 500 μm; (b) 30 μm; (c–i) 1 μm. Asterisks indicate protoxylem vessel element. e, ER; g, Golgi; m, mitochondrium; n, nucleus; sw, secondary cell wall; v, vacuole.

TEM analysis of the series of successive cross-sections illustrated clearly the different stages of protoxylem vessel death, including initiation of cell death by tonoplast rupture, autolysis and clearance of the cells within the distance of 1500 to 2000 μm from the root tip (Fig. 3c–i). The images representing these stages were taken from several different roots as the different stages of cell death cannot be detected within one and same root due to the rapid progression of cell death. Before vacuolar rupture, the cells sometimes showed numerous vesicle-like structures in the cytoplasm and dilated ER (Fig. 3c–e). No signs of nuclear degradation were detected before tonoplast rupture. The moment of cell death, that is the distance from the root tip when vessel elements had lost tonoplast integrity, was defined as the moment when the border between cytoplasm and vacuolar lumen became blurry on the TEM images (Fig. 3f). Directly after tonoplast rupture, organelles were still intact and vacuolar contents became intermixed with the cytoplasm (Fig. 3f,g). This led to rapid degradation of ER membranes and organelles such that it was not possible to follow the order of events during cellular disintegration. Following this, the cell lumen was filled with a granular mass without recognisable substructures, presumably representing late stages of autolysis (Fig. 3h). Finally, the cell lumen was completely cleared (Fig. 3i).

Cell death of protoxylem vessel elements occurred, judged on the basis of vacuolar rupture, at approximately the same time in atmc9-2 and wild-type roots. Out of 36 analysed micrographs within the range of 1500–2000 μm (Fig. 4), intact cells were observed on 8 and 10 micrographs from wild-type and atmc9-2 mutant specimens, respectively (Fig. 4). However, there was a clear difference in the speed of post mortem autolysis after vacuolar rupture. While vessel elements of the wild-type were completely cleared in 22 instances out of 36, only seven out of 36 fulfilled this criterion in the atmc9-2 mutant (Fig. 4). From this it follows that atmc9-2 cells were in most cases (19 out of 36) in different stages of autolysis, while this was true for the wild-type in just 6 cases out of 36 (Fig. 4). One could estimate on the basis of these results and the known length of the protoxylem vessel elements that autolysis was ongoing on average in two to three cells of an atmc9-2 vessel but only in one cell in a wild-type vessel. These results suggest that AtMC9 is not required for vacuolar rupture itself but for fast and efficient post mortem autolysis of xylem vessel elements.

Figure 4

Electron microscopy analysis of protoxylem vessel cell death in the Arabidopsis atmc9-2 mutant and the wild-type. Electron micrographs of primary vascular tissues of the roots in the atmc9-2 mutant and the wild-type. Transverse sections were taken from 3-d-old seedling roots with the ultramicrotome, starting from the tip, and analysed at every 100 μm within the distance of 1500–2000 μm from the root tip using TEM. Protoxylem vessel elements are the centrally located cells on each micrograph, marked with asterisks. In three biological replicates (root 1–3) with two protoxylem strands each, up to c. 20 vessel elements were analysed per genotype. Circles are a schematic view of the morphological appearance of the protoxylem cells; filled black circles indicate intact cells, grey circles cells in a phase of autolysis and white circles cleared protoxylem vessel. Data are shown for one of two independent experiments with similar results.

AtMC9 influences activities of papain-like cysteine proteases

The phenotype of the atmc9-2 mutant resembles that reported for the xcp1 xcp2 double mutant (Avci et al., 2008), prompting us to consider a mechanistic link between AtMC9 and the PLCPs XCP1/XCP2. To study a direct effect of AtMC9 on XCP2 in vitro, we tested whether recombinant AtMC9 is able to cleave XCP2 in a TNT-protease assay. The XCP2 protein that was expressed in a cell-free lysate system (TNT; Promega) was present in two forms of c. 50 and 35 kDa, respectively (Fig. 5a), consistent with the known processing pattern of XCP2 and PLCPs in general (van der Hoorn et al., 2004; Avci et al., 2008; Richau et al., 2012). Addition of recombinant AtMC9 lead to degradation of both these forms and no additional bands of smaller size appeared, indicating rather complete degradation of XCP2 in vitro.

Figure 5

AtMC9 acts on papain-like cysteine proteases in vitro and in planta. (a) Cleavage of TNT-expressed XCP2 by increasing concentration of recombinant AtMC9 (as indicated above the radiography image). Bands at 50 and 35kDa represent XCP2 and its processed form. AtMC9-CA is an inactive variant of AtMC9. (b) Activity profiling of PLCPs in total protein extracts (15 μg from each genotype) from 4-d-old Arabidopsis seedlings roots. Active PLCPs were labelled with the biotinylated E-64 derivate DCG-04, separated on SDS protein gel, and subsequently transferred onto nitrocellulose membrane (Bio-Rad) and detected with streptavidin-HRP (Strep-HRP, 1 : 3000, Ultrasensitive, Sigma). For specificity determination, excess E-64 is added to the initial labelling procedure. Equal protein loading is shown on Coomassie stained gels below. Protein extracts from seedling roots were used and, therefore, no strong Rubisco band is visible. Both experiments were performed twice.

In order to test whether AtMC9 affects activity of XCP1/XCP2 in vivo, we performed probe-based activity profiling of PLCPs in protein extracts of seedling roots (van der Hoorn et al., 2004). Activity profiling is the most informative method to analyse regulation of proteases because it reveals the actual activity and not merely the amount of protease. Root protein extracts were incubated with DCG-04, a biotinylated derivative of the PLCP inhibitor E-64. DCG-04 binds covalently to the active protease, inhibits the cleavage mechanism and can be detected by streptavidine-peroxidase conjugates (van der Hoorn et al., 2004).

The double mutant xcp1 xcp2 was included to visualise the location of XCP1/XCP2 activity on the gel. Also, an atmc9-2 xcp1 xcp2 triple mutant (Fig. S1a) was created to further test the link between XCP1/XCP2 and AtMC9. In this assay, three main PLCP activities of c. 35, 40 and 42 kDa were detected (Fig. 5b). No differences in those activities were observed between atmc9 mutants and wild-type plants (Fig. 5b). xcp1 xcp2 activities did not differ from wild-type either, indicating that a loss of 35 kDa activity due to lack of functional XCP1/XCP2 is masked by other PLCP(s) in Arabidopsis seedling roots. Interestingly, a weaker 35 kDa signal was present in the atmc9-2 xcp1 xcp2 triple mutant (Fig. 5b), suggesting that activity of these other PLCPs is affected in the triple mutant. Altogether, these results could not support an effect of AtMC9 on the main activity of XCP1/XCP2, but interestingly indicate that AtMC9 influences the activity of some unknown PLCPs in xylem elements of Arabidopsis.

AtMC9 is located in aggregates during final autolysis

In order to further understand the function of AtMC9 during protoxylem maturation, subcellular localisation of AtMC9 was investigated in transgenic Arabidopsis lines carrying translational AtMC9:GFP and AtMC9:mCherry fusions under control of the endogenous AtMC9 promoter (AtMC9_pro). The expression of the fluorescent fusion proteins was highly specific to root cap and xylem cells (Fig. 6), as expected from the earlier analysis of the AtMC9_pro::GUS line (Fig. 1). The expression became detectable in protoxylem cells when secondary cell wall thickenings were already present. Initially, AtMC9:mCherry was observed evenly in the cytoplasm (Fig. 6a). Later during vessel maturation, when cells already had thick secondary walls, mCherry fluorescence was present in the entire vessel lumen, presumably due to vacuolar rupture (Fig. 6b). After vacuolar rupture and just before complete autolysis of the cell contents, AtMC9:mCherry was localised specifically in small aggregates (Fig. 6c,d). Localisation of AtMC9:GFP was equal to that of AtMC9:mCherry, demonstrating similarly a clear shift from the even distribution in the cytoplasm to the cytoplasmic aggregates concurrent with the vacuolar rupture (Fig. 6g,h). A similar shift in AtMC9 localisation was observed also in root cap cells. Both fluorescent fusion proteins were detected in the cytoplasm of intact root cap cells (Fig. 6e,f,i), and AtMC9:mCherry, presumably due to higher pH-stability of its fluorescence, also in the entire cell lumen after vacuolar rupture and in cellular remnants (Fig. 6e,f). Transgenic lines carrying a transcriptional AtMC9_pro::GFP fusion construct only showed evenly distributed, nonaggregated GFP throughout the cytoplasm (Fig. 6j–l) of root cap and xylem cells. This suggests that the formation of fluorescently labelled aggregates depends on the presence of AtMC9 in the fusion protein. Further, also in lines expressing the AtMC9:GFP fusion construct under control of the CaMV 35S promoter, fluorescence was evenly distributed throughout the cytoplasm of most cells (Fig. 6m). This suggests that high contents of AtMC9:GFP alone are not sufficient to induce aggregate formation and that autolysis may be required. However, no fluorescence could be observed in autolysing xylem cells, likely due to low activity of the 35S promoter at this developmental stage.

Figure 6

Sub-cellular localisation of AtMC9 during vessel cell death. Expression of (a–f) AtMC9_pro::AtMC9: mCherry, (g–i) AtMC9_pro::AtMC9:GFP, (j–l) AtMC9_pro::GFP and (m) 35S_pro::AtMC9:GFP in (a–d, g, h, j–m) protoxylem vessel elements and (e, f, i, l) root cap cells. (a–c) AtMC9: mCherry localisation (a) in the cytoplasm of protoxylem vessel elements containing intact vacuole, (b) in the cell lumen directly after vacuolar rupture and (c, d) in aggregates during later stages of autolysis. (d) Two consecutive vessel elements show the different subcellular localisation of AtMC9: mCherry at different stages of vessel cell death, the lower one showing early autolysis after vacuolar rupture and the upper one formation of aggregates before complete clearance of the cell. (e, f) AtMC9: mCherry localisation in (e) the cytoplasm of intact root cap cells, in the entire cell lumen after vacuolar rupture and (f) in aggregated remnants of dead cells. (g–i) AtMC9:GFP localisation in (g) cytoplasm during early vessel differentiation, (h) during late vessel differentiation with obvious aggregate formation and (i) in root cap cells. Cell walls are visualised by differential interference contrast (DIC) or propidium iodide (PI) staining. Asterisks indicate cells with ruptured vacuole. Arrowheads indicate aggregates. v, vacuoles. Bars, 10 μm.

Discussion

The Arabidopsis type II metacaspase AtMC9 was shown in this study to participate in post mortem autolysis of cell contents during xylem cell death. Several different types of proteases were earlier implicated in xylem cell death on the basis of their expression patterns or enzyme activities (Minami & Fukuda, 1995; Ye & Varner, 1996; Beers & Freeman, 1997; Yamamoto et al., 1997; Woffenden et al., 1998; Groover & Jones, 1999; Fukuda, 2000; Zhao et al., 2000; Twumasi et al., 2010; Han et al., 2012), but functional evidence existed only for the PLCPs XCP1 and XCP2 (Avci et al., 2008), which were shown to participate in xylem cell autolysis in a similar manner to AtMC9. In the xcp1 xcp2 double mutant, cellular remnants were present in 2–3 vessel elements further up in the root than in the wild-type (Avci et al., 2008). Hence the degree of delay in xcp1 xcp2 vessel autolysis corresponds well with the delay observed in atmc9-2 (Fig. 4). It is curious that both AtMC9 and XCP1/XCP2 exert only rather mild effects on the autolytic process. Indeed, vessels were finally completely hydrolysed in both atmc9 and xcp1 xcp2 mutant backgrounds. A reason for this could be redundancy within gene families, but this appears unlikely in the case of AtMC9. Although AtMC9 did not seem to be the only type II metacaspase that was expressed in xylem elements, it was the only one that was strictly confined to differentiating xylem elements (Fig. 1), and it has previously been identified as the only metacaspase induced during TE differentiation in vitro (Turner et al., 2007). Also, AtMC9 appears unique among metacaspases in its strict requirement for an acidic pH and that it does not require Ca²⁺ for activation (Vercammen et al., 2004; Watanabe & Lam, 2005; Turner et al., 2007; He et al., 2008). It is therefore plausible that AtMC9 has a specific and nonredundant function in post mortem autolysis of xylem elements, although it does not rule out the possibility that parallel pathways function in completing autolysis in the absence of AtMC9.

AtMC9 did not affect initiation of the cell death process (Figs 2, 4). This is not surprising considering its requirement for acidic pH that is reached only after vacuolar rupture. But it is in striking contrast with the function of other type II metacaspases that are implicated in initiating their respective cell death processes. AtMC8 is required for induction of PCD triggered by oxidative stress (He et al., 2008) while AtMC4 positively regulates HR cell death (Watanabe & Lam, 2011). AtMC5 can activate caspase-like proteases and induce apoptotic-like cell death in yeast (Watanabe & Lam, 2005). Similarly, mcII-Pa was shown to activate a caspase-like VEIDase, required for PCD in the spruce embryo suspensor (Bozhkov et al., 2005). With AtMC9 acting only post mortem, the trigger of cell death in xylem elements remains unknown. Only few studies have tried to find answers to this intricate question. Kuriyama (1999) demonstrated that tonoplast permeability changed before its rupture, and proposed the presence of specific triggers or signals that would control tonoplast integrity and hence cell viability. Groover & Jones (1999) proposed a role for an unknown extracellular serine protease that accumulates in TE cell walls and induces cell death after reaching a critical threshold concentration. It is also possible that vessel cell death does not require a specific trigger, but is an inevitable fate of cells that spent all resources for secondary wall synthesis. This scenario is supported by two main issues; the inability to separate secondary wall formation from cell death in xylem vessels by genetic or pharmacological means (Bollhöner et al., 2012) and the transcriptional coregulation of the secondary cell wall synthesis and cellular hydrolysis by the VASCULAR NAC DOMAIN 6 (VND6) and 7 (VND7) transcription factors (Ohashi-Ito et al., 2010; Yamaguchi et al., 2010, 2011; Zhong et al., 2010b). Consequentially, the only specific factors for cell death were actually the ones responsible for post mortem autolysis of cell contents. Simultaneous activation of the expression of both secondary wall- and cell death-related genes infers that the cell death machinery must be prevented from functioning prematurely during secondary wall biosynthesis. Requirement for an acidic pH, as is the case for AtMC9, could be one mechanism guaranteeing a function entirely post mortem after vacuolar rupture.

The effect of loss of AtMC9 function raises the question of the importance of post mortem autolysis in the functioning of xylem vessel elements. Our results, together with the reports on the ZEN1 nuclease and XCP1/XCP2, could actually be interpreted in such a way that the efficient enzymatic removal of TE cell contents may not primarily serve to clear the tubes for water transport. The absence of growth phenotypes in mutants of these hydrolytic enzymes suggests that remnants of incomplete autolysis may not be as devastating for water flux as expected (Groover & Jones, 1999). Instead, efficient autolysis may have other functions such as removal of nutrients in order to prevent pathogen growth in the dead cells, as indicated by the faster wilt response of atmc9 mutants upon infection with the xylem-colonising bacteria R. solanacearum (Fig. S2). Not all proteins are degraded either; some can be found quite intact and may even be functional in the xylem sap (Kehr et al., 2005; Ligat et al., 2011).

An interesting question remains about the targets of AtMC9. So far, only one target protein has been reported for plant metacaspases, the spruce tudor staphylococcal nuclease (TSN), targeted by the type II metacaspase mcII-Pa. Human TSN functions in RNA processing and splicing, but the function of the plant TSN has remained elusive even though a role in cell death regulation of pollen tapetum was demonstrated in Arabidopsis (Sundström et al., 2009). TSN activity has also been linked to stress tolerance and selective stabilisation of mRNAs encoding for secreted proteins, including numerous protease inhibitors (dit Frey et al., 2010). It is tempting to speculate also that AtMC9 is part of a regulatory cascade having only few specific target proteins in xylem elements. An obvious candidate for such a target is XCP1/XCP2 due to the similarity of the atmc9 and xcp1 xcp2 mutant phenotypes (Fig. 4; Avci et al., 2008). Although we found that XCP2 is degraded in vitro in presence of recombinant AtMC9, we could not address an effect on XCP2 activity in vivo, probably due to the masking activity of other PLCPs. Our results demonstrated, however, that AtMC9 influenced the activity of yet unidentified PLCP(s) in the absence of XCP1 and 2 (Fig. 5). Detailed expression studies and proteomic analyses are needed to reveal identities of these additional PLCPs.

It was interesting to notice the change in subcellular localisation of AtMC9 from an even cytoplasmic localisation in living cells to patches or aggregates of different sizes in cells during late autolysis (Fig. 6). Based on the requirement of AtMC9 for acidic conditions (Vercammen et al., 2004), the cytoplasmic localisation probably represents the inactive zymogen and activation occurs once cytoplasm and vacuolar contents are mixed. It is not known whether the observed formation of aggregates is related to the proteolytic function of AtMC9 but, interestingly, formation of aggregates has previously been reported in connection to PCD. The yeast metacaspase Yca1 has been shown to localise to protein aggregates during heat stress and aging, and to participate in clearance of those aggregates (Zhong et al., 2010a). Also a GFP-Nitrilase fusion protein has been observed to aggregate as an early response in wound-induced cell death (Cutler & Somerville, 2005). It seems possible, therefore, that AtMC9 mediates its function during xylem cell autolysis in the same way by tethering to aggregates. What remains curious is the mechanism directing AtMC9 into such protein aggregates. AtMC9 has been reported to lack apparent aggregation domains (Tsiatsiani et al., 2011). A search for xylem proteins that interact with AtMC9 and could direct the protease to such aggregates may reveal novel targets for AtMC9 and give insight into the process of autolysis.

Acknowledgements

We thank Lenore Johansson from the Umeå Core Facility Electron Microscopy (UCEM), for technical assistance in preparation of TEM samples, Veronica Bourquin and Paige Cox for technical help, Renier van der Hoorn for the gift of DCG-04, Mikael Norberg for the vector pKGW + T35S and Taku Demura for the destination vectors pBGGUS and pBGGN, Eric Beers for seeds of the xcp1 xcp2 double mutant, Nathaniel Street for help in microarray analysis and Yves Marco for access to the Ralstonia facilities in the Laboratory of Plant Microbe Interactions. For financial support, we acknowledge the Swedish Energy Agency, the Swedish Research Council Formas (Strong Research Environment BioImprove), the Swedish Research Council (VR) and the Swedish Governmental Agency for Innovation Systems VINNOVA (UPSC Berzelii Centre), Bio4Energy – a strategic research environment appointed by the Swedish government, the Research Foundation-Flanders (FWO grant number G.0038.09N), Ghent University (Special Research Fund, BOF, and Multidisciplinary Research Partnership ‘Biotechnology for a Sustainable Economy’ project), the Academy of Finland Fellowship (decisions no. 251397 and 256073) and the University of Helsinki 3 yr research allocation program. D.G. and N.D. want to thank the French ANR (ANR-07-GPLA-014 ‘Walltalk’) for funding.