Arabidopsis Type I Metacaspases Control Cell Death

See allHide authors and affiliations

Science  03 Dec 2010:
Vol. 330, Issue 6009, pp. 1393-1397
DOI: 10.1126/science.1194980

The Yin and Yang of Plant Caspases

The function of plant metacaspases, identified by limited sequence homology to the animal caspases that control cell death, has remained elusive. Coll et al. (p. 1393) have now elucidated the actions of two metacaspases in the small plant Arabidopsis. One metacaspase, AtMC1, promoted cell death, and the other, AtMC2, acted antagonistically to stall cell death. The results help to elucidate the mechanisms by which plants control cell survival during development and defend against pathogen attack.


Metacaspases are distant relatives of animal caspases found in protozoa, fungi, and plants. Limited experimental data exist defining their function(s), despite their discovery by homology modeling a decade ago. We demonstrated that two type I metacaspases, AtMC1 and AtMC2, antagonistically control programmed cell death in Arabidopsis. AtMC1 is a positive regulator of cell death and requires conserved caspase-like putative catalytic residues for its function. AtMC2 negatively regulates cell death. This function is independent of the putative catalytic residues. Manipulation of the Arabidopsis type I metacaspase regulatory module can nearly eliminate the hypersensitive cell death response (HR) activated by plant intracellular immune receptors. This does not lead to enhanced pathogen proliferation, decoupling HR from restriction of pathogen growth.

Programmed cell death is essential for plant development (1). Metacaspases in plants, fungi, and protozoa are distant homologs of caspases in the CD cysteine protease superfamily (caspases, paracaspases, gingipains, clostripains, legumains, and separin). These evolutionarily related proteases share the caspase-hemoglobinase fold, and common catalytic-site domains (2, 3). The Arabidopsis thaliana genome encodes three type I and six type II metacaspases (AtMCs) (3). Both types contain a conserved putative catalytic domain and plausible sites for autocatalytic processing but differ in their N-terminal domains. Despite many reports of caspase-like activities in plants (49), no experimental data exist defining functions or substrates for plant type I metacaspases. Type II metacaspase functions are nearly as enigmatic: Recombinant plant type II metacaspases AtMC4 and AtMC9 can undergo in vitro autocatalytic processing (10), and the Tudor staphylococcal nuclease, cleaved during programmed cell death in pine, is the only in vivo type II metacaspase substrate defined in plants (11). Although classic animal caspases cleave after an aspartate at the P1 position, metacaspases cleave after a basic amino acid residue such as lysine or arginine (9).

LSD1 is a negative regulator of cell death initiated by localized superoxide production occurring during the hypersensitive response (HR), a cell death that often accompanies pathogen recognition. HR sites form normally in lsd1 (12), but the typical sharp boundary between dead and live cells subsequently breaks down and “runaway cell death” ensues throughout the leaf (12). This phenotype requires proteins that are also necessary for pathogen recognition and salicylic acid accumulation (1216). Hence, lsd1 is a sensitized genetic background for studying the control of oxidative stress–dependent cell death (1720).

All three type I Arabidopsis metacaspases, AtMC1 (At1g02170), AtMC2 (At4g25110), and AtMC3 (At5g64240), possess a conserved, plant-specific LSD1-like zinc-finger N-terminal motif (CxxCRxxLMYxxGASxVxCxxC) (21) (fig. S1A). The LSD1 zinc-finger domain can function in protein interactions (22). In yeast, LSD1 and AtMC1 interact via their zinc-finger domains (fig. S1, B to D). In contrast, AtMC2 interacts only very weakly with either AtMC1 or LSD1 in this assay (fig. S1, B to D). We substantiated the AtMC1-LSD1 interaction using in vivo coimmunoprecipitation in transgenic Arabidopsis (Fig. 1B) and in transient expression assays in Nicotiana benthamiana (23) (fig. S2A). The predicted N-terminal prodomain of AtMC1 is required for interaction with LSD1 (fig. S2A). AtMC2 does not coimmunoprecipitate with either LSD1 (fig. S2B) or AtMC1 (fig. S2C) under these conditions.

Fig. 1

AtMC1 interacts with LSD1. (A) Scheme of AtMC1 and AtMC2 proteins (not to scale). Z1, LSD1-like zinc finger; PP, proline-rich region; p20 and p10, putative caspase-like catalytic domains (2); numbered H and C, putative catalytic residues (2); atmc1 and atmc2, T-DNA insertion sites. (B) AtMC1 and LSD1 coimmunoprecipitate. Protein extracts from leaves of lsd1 atmc1 [pLSD1::LSD1-myc] × lsd1 atmc1 [Dex-AtMC1-HA] F1 plants were immunoprecipitated with anti-myc–coupled magnetic beads. Crude extract (input) and eluate (elution) were analyzed by SDS–polyacrylamide gel electrophoresis (SDS-PAGE) with anti-HA (top) or anti-myc (bottom) immunoblot. The eluate is 8× concentrated as compared to the input. The asterisk denotes a nonspecific cross-reacting band; arrowheads are ~34 kD, the expected apparent molecular mass of LSD1-myc. WB, Western blot.

We obtained transfer DNA (T-DNA) insertional mutant alleles of atmc1, atmc2, and lsd1 in the Col-0 accession (Fig. 1A and fig. S1B). atmc1, atmc2, and the atmc1 atmc2 double mutant exhibited no signs of enhanced production of reactive oxygen species and had no obvious phenotypes (fig. S3). We generated lsd1 atmc1 and lsd1 atmc2 double mutants, and the lsd1 atmc1 atmc2 triple mutant to determine whether atmc1 or atmc2 modifies the lsd1 cell death phenotype. lsd1 runaway cell death can be induced with benzo(1,2,3)thiadiazole-7-carbothioic acid S-methyl ester (BTH), a salicylic acid analog (16). lsd1 exhibited maximum ion leakage (a cell death proxy) 96 hours after BTH treatment. The Col-0 wild type, atmc1, atmc2, and atmc1 atmc2 did not display significant increases in ion leakage (Fig. 2A). lsd1 atmc1 exhibited suppressed BTH-induced ion leakage (Fig. 2A, left). In contrast, lsd1 atmc2 displayed accelerated BTH-induced ion leakage (Fig. 2A, right). Ion leakage in lsd1 atmc1 atmc2 was similar to that in lsd1 atmc1 (Fig. 2A, right). Consistent with these data, morphological lsd1 phenotypes were abolished in the absence of AtMC1, and more pronounced in the lsd1 atmc2 mutant (fig. S3). Hence, AtMC1 is a positive mediator of lsd1 runaway cell death, and AtMC2 acts genetically as a negative regulator of AtMC1.

Fig. 2

AtMC1 and AtMC2 antagonistically control lsd1 runaway cell death. (A and B) Four-week-old homozygous transgenic plants (genotypes, right) were sprayed with 150 μM BTH, and conductivity was measured at the time points indicated. Repeated three times in (A) and once in (B). Error bars indicate two times the standard error, calculated from four replicate measurements per genotype and data point. Letters a to d at right represent groups with significant differences [P < 0.05, Tukey’s honest significant difference (HSD) test]. hpt, hours post treatment. (C) Four-week-old plants of the indicated genotypes were sprayed with 150 μM BTH. Tissue was harvested at the indicated time points, and 50 μg of total protein were analyzed by SDS-PAGE and anti-HA immunoblot.

Hemagglutinin (HA)–epitope-tagged AtMC1 and AtMC2 constructs controlled by their native promoters (pAtMC1::AtMC1-HA and pAtMC2::AtMC2-HA) complemented the respective mutant phenotypes for BTH-induced ion leakage in lsd1 (Fig. 2B). AtMC1-HA protein was detectable 1 day after BTH spray, before plants displayed any visible sign of cell death, and accumulated thereafter (Fig. 2C). AtMC2-HA was undetectable until 2 days after BTH treatment and accumulated to lower levels than AtMC1-HA (Fig. 2C). We observed no HA-tagged low-molecular-weight products, suggesting either a lack of processing during activation or instability of a putative C-terminal HA-tagged fragment.

We monitored AtMC1 or AtMC2 expression in transgenic plants carrying promoter β-glucuronidase (GUS) reporter genes (pAtMC1::GUS and pAtMC2::GUS, respectively) in either Col-0 or lsd1. AtMC1 expression was confined to the leaf veins. However, 24 hours after BTH treatment, AtMC1 was expressed in a narrow zone of several cells adjacent to cell death sites in lsd1 (fig. S4A). Cells expressing AtMC1 are destined to die as runaway cell death expands over time (15). AtMC2, in contrast, was expressed at low levels, except for a halo of nonstained cells outlining cells that are committed to die (fig. S4B).

We infected Col-0 pAtMC1::GUS and pAtMC2::GUS transgenic plants with either the obligate biotrophic oomycete Hyaloperonospora arabidopsidis (Hpa; isolate Emwa1), or the hemibiotrophic bacteria Pseudomonas syringae pv. tomato [Pto; strain DC3000(avrRpm1)] to trigger HR via the specific intracellular Toll–interleukin-1 receptor (TIR) and coiled-coil (CC)–nucleotide-binding–leucine-rich repeat (NB-LRR) (CC-NB-LRR) immune receptors RPP4 (24) and RPM1 (25), respectively. AtMC1 was activated upon NB-LRR–based pathogen recognition and subsequently expressed in a spatially restricted manner in cells destined to undergo HR (fig. S5). In contrast, AtMC2 was expressed in a relatively more diffuse zone around infection sites (fig. S5).

We generated transgenic plants conditionally expressing either a full-length or an N-terminal truncated version of AtMC1 lacking the LSD1-like extension (AtMC1-HA and AtMC1-ΔN-HA, respectively). High levels of AtMC1-HA accumulation resulted in cell death in Col-0 transgenics; lower levels did not (Fig. 3A). Col-0 AtMC1-ΔN-HA plants accumulating relatively low levels of protein nevertheless exhibited enhanced cell death (Fig. 3A). We failed to recover transgenics expressing AtMC1-ΔN-HA in lsd1 either by direct transformation or from crosses of two independent Col-0 AtMC1-ΔN-HA-expressing transgenic lines to lsd1 [see methods in supporting online material (SOM)]. These results suggest that the LSD1-like putative prodomain of AtMC1 negatively regulates its pro–cell death activity.

Fig. 3

AtMC1 function is negatively regulated by its LSD1-like putative prodomain and requires predicted metacaspase catalytic residues. (A) Three-week-old plants from sets of three independent homozygous transgenics of indicated genotypes in Col-0 were sprayed with 20 μM dexamethasone (Dex). Tissue was harvested 1 day after treatment; total protein was analyzed by SDS-PAGE with anti-HA immunoblot (upper panel). Dex-treated plants photographed 3 days after treatment are shown (lower panel). (B) As in (A), but using two independent transgenics of the genotypes listed above the blot. (C) AtMC1-mediated ion leakage requires the putative catalytic cysteine residue C220. Plants treated as described in (A) were harvested at 46 hpt and processed as described in the SOM. Error bars represent two times the standard error. Letters a to c represent experimental groups with significant differences (P < 0.05, Tukey’s HSD test). The experiment was repeated twice. (D and E) AtMC1 function in cotyledons requires additional signals. Ten-day-old seedlings were pre-treated with 20 μM Dex (+Dex). One day later, half of the seedlings were inoculated with 50,000 spores/ml of Hpa isolate Emwa1 (Dex + Hpa) to activate RPP4. All seedlings were harvested 3 days after inoculation and stained with Trypan blue to visualize cell death. (D) Pictures of representative Trypan blue–stained cotyledons. (E) 20 cotyledons were evaluated per genotype and treatment. CD, cell death; <50% of cot., extensive cell death covering less than 50% of cotyledon; >50% of cot., extensive cell death covering more than 50% of cotyledon.

Typical caspase active sites include a histidine-cysteine catalytic dyad, also found in metacaspases (3). The predicted catalytic cysteine residue was essential for autoprocessing of the pine metacaspase mcII-Pa (26), of Arabidopsis AtMC4 and AtMC9 in bacteria (10), and of AtMC1 and AtMC5 in yeast (27). In Trypanosoma brucei metacaspase TbMCA4, a second, metacaspase-specific cysteine residue was catalytic, and the corresponding residue compensated for mutation of the classic cysteine residue in AtMC9 (28, 29). We generated Col-0 transgenic lines expressing high or low levels of either a conditional AtMC1 expression construct containing a cysteine-to-alanine mutation (AtMC1-C220A-HA) or a double mutation including the potentially compensatory cysteine (AtMC1-C99A-C220A-HA) (Fig. 3B). Conditional overexpression of AtMC1-HA resulted in a dose-dependent increase in ion leakage, whereas overexpression of either AtMC1-C220A-HA or AtMC1-C99A-C220A-HA did not (Fig. 3C). Hence, AtMC1-C220 is required for function, and AtMC1-C99 cannot compensate for its loss.

Overexpression of AtMC1 did not induce cell death in cotyledons before infection. However, massive cell death occurred after infection with Hpa and activation of RPP4 (Fig. 3, D and E). Hence, AtMC1 activation in seedlings, in contrast to adult plants (Fig. 3A), requires an additional signal, which is consistent with the fact that lsd1 runaway cell death also does not occur in cotyledons.

We generated similar transgenics in lsd1 and Col-0 that conditionally expressed full-length AtMC2 (AtMC2-HA), an analogous N-terminal truncated version of AtMC2 (AtMC2-ΔN-HA) and either AtMC2 or AtMC2-ΔN with cysteine-to-alanine mutations in the predicted catalytic (C256) and potentially compensatory (C135) cysteine residues (AtMC2-C135A-C256A-HA and AtMC2-ΔN-C135A-C256A-HA). BTH-induced lsd1 runaway cell death was suppressed in transgenic lines expressing full-length AtMC2-HA (fig. S6, A and B). Cell death suppression was nearly complete in AtMC2-ΔN-HA–expressing lines (Fig. 4A and figs. S6C and S7). Surprisingly, the predicted catalytic cysteines were not required for cell death suppression by either AtMC2 or AtMC2-ΔN-HA (Fig. 4A and fig. S6B).

Fig. 4

AtMC2 negatively regulates AtMC1. Seedlings of the depicted genotypes were sprayed with 1 μM Dex 6, 11, and 16 days after germination or left unsprayed. CACA denotes constructs where both putative catalytically active cysteine residues (C135A C256A for AtMC2) were mutated to alanine. (A) For ion leakage measurements, plants were sprayed with 150 μM BTH 1 day after the last treatment with Dex. Tissue was harvested at 43 hpt and processed as described in the SOM. Blue lines, Dex-treated genotypes; yellow lines, non–Dex-treated controls. Error bars represent two times the standard error. Letters a and b at right represent experimental groups with significant differences (P <0.05, Tukey’s HSD test). The experiment was repeated twice. (B) Plants were vacuum-infiltrated with 250,000 colony-forming units (CFU)/ml of Pto DC3000(avrRpm1). Twelve hours later, plants were harvested and stained with Trypan blue to visualize cell death. To quantify cell death, all dead cells in one field of vision (10× magnification) were counted. Average and two times the standard error were calculated from 20 leaves per genotype and treatment. The experiment was repeated. (C) Ten hours after the last treatment with Dex, plants of indicated genotypes were treated as in (B). (D) Seedlings were dip-infiltrated with Pto DC3000(avrRpm1) at 2.5 × 107 CFU/ml. The average and two times the standard error were calculated from four samples per genotype. The experiment is representative of three independent replicates. (E) Three hours after the last Dex spray, plants were treated as in (D). (F) One day after the last Dex treatment, plants were inoculated with 50,000 spores/ml of Hpa Emwa1. Three days later, plants were stained with Trypan blue, and interaction sites per field of vision were counted. The experiment was repeated three times. TN, trailing necrosis; FH, free hyphae.

To assess whether AtMC2 suppressed NB-LRR–mediated HR, we infected Col-0, atmc1, atmc2, atmc1 atmc2, and transgenic plants expressing AtMC2-ΔN-HA or AtMC2-ΔN-C135A-C256A-HA with Pto DC3000(avrRpm1). We observed enhancement of RPM1-mediated HR in atmc2 (Fig. 4B). Conversely, we observed suppression of RPM1-mediated HR in atmc1 and in atmc1 atmc2 at low bacterial inocula resembling natural infections (Fig. 4B), but not at artificially high doses. Conditional expression of either AtMC2-ΔN (Fig. 4C and fig. S6F) or full-length AtMC2 (fig. S6, D and E) suppressed RPM1-mediated HR, phenocopying atmc1.The suppression of HR did not result in increased susceptibility to Pto DC3000(avrRpm1) (Fig. 4, D and E). RPP4-mediated HR was also suppressed by conditional expression of AtMC2-ΔN and in atmc1 (Fig. 4F). Thus, AtMC2 inhibits at least RPM1- and RPP4- mediated HR, via direct or indirect regulation of AtMC1. The putative catalytic sites of AtMC2 are not required for HR suppression (Fig. 4, B, C, and F, and fig. S6E). atmc2 did not exhibit increased RPP4-mediated HR (30). Although Hpa Emwa1 is an obligate biotrophic oomycete, AtMC2-ΔN–mediated suppression of HR did not abolish RPP4-dependent pathogen growth restriction (Fig. 4F). These surprising observations are consistent with previous results showing that HR can occur in the absence of pathogen growth restriction and that HR can be inhibited by caspase inhibitors (6, 31). The pathogen-stress marker PR1 is not induced by AtMC2-ΔN overexpression, suggesting that these results are not a general consequence of stress. Finally, AtMC2 control of AtMC1 is posttranscriptional (fig. S7).

Our data establish AtMC1 as a pro-death caspase-like protein required for both superoxide-dependent cell death in a reactive oxygen–sensitized state and for full HR mediated by intracellular NB-LRR immune receptor proteins. AtMC2 antagonizes these functions. AtMC1 and AtMC2 contain a plant-specific N terminus shared with LSD1. The functions of both AtMC1 and AtMC2 are enhanced by removal of this domain and, at least for AtMC1, this domain is required for interaction with LSD1. In the absence of LSD1, AtMC1-dependent cell death requires additional, unknown signals and is developmentally regulated. Hence, activation of AtMC1 is complex and probably context-dependent. The inhibitory function of AtMC2 does not require classic cysteine catalytic residues. The AtMC1-AtMC2 antagonism is reminiscent of animal caspase-12, which negatively regulates caspase-1 to dampen the inflammatory response to bacteria and in colitis-associated colorectal cancer (32, 33); these caspase-12 functions do not require catalytic function (33). Caspase-12 also inhibits NOD-like receptor-mediated innate immunity independent of caspase-1 (34), providing a striking parallel to our observation that AtMC2 inhibits AtMC1-dependent cell death controlled by analogous plant NB-LRR innate immune receptors. These results suggest an ancient link between cell death control by divergent metacaspase/caspase proteases and innate immune receptor function governed by NB-LRR or NLR proteins.

Supporting Online Material

Materials and Methods

Figs. S1 to S7


References and Notes

  1. Single-letter abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; Y, Tyr; and x, any amino acid.
  2. Materials and methods are available as supporting material on Science Online.
  3. Because this assay counts interaction sites, the number of spores that infect each leaf determines the maximum number of interactions that can occur per leaf. This cannot be altered by an increase in RPP4-dependent responses.
  4. We thank J. McDowell (Virginia Tech, Blacksburg, VA) and M. Nishimura (UNC) for critical reading of the manuscript. This work was funded by NIH RO1 GM057171 to J.L.D.; a Swiss National Science Foundation Fellowship (PBEZA-115173) to N.S.C.; and the Research Fund, Ghent University (grant 12051403), and Fonds voor Wetenschappelijk Onderzoek-Vlaanderen to F.V.B. and D.V., respectively. We thank D. Baltrus for statistical analysis, T. Perdue of the UNC Biology Microscope Facility for patient and expert assistance, E. Washington and T. Eitas for pMDC7 vectors, B. van de Cotte for technical assistance, and K. Overmyer for early contributions to this work.
View Abstract

Stay Connected to Science

Navigate This Article