Set-point control of RD21 protease activity by AtSerpin1 controls cell death in Arabidopsis


For correspondence (e-mail


Programmed cell death (PCD) in plants plays a key role in defense response and is promoted by the release of compartmentalized proteases to the cytoplasm. Yet the exact identity and control of these proteases is poorly understood. Serpins are an important group of proteins that uniquely curb the activity of proteases by irreversible inhibition; however, their role in plants remains obscure. Here we show that during cell death the Arabidopsis serpin protease inhibitor, AtSerpin1, exhibits a pro-survival function by inhibiting its target pro-death protease, RD21. AtSerpin1 accumulates in the cytoplasm and RD21 accumulates in the vacuole and in endoplasmic reticulum bodies. Elicitors of cell death, including the salicylic acid agonist benzothiadiazole and the fungal toxin oxalic acid, stimulated changes in vacuole permeability as measured by the changes in the distribution of marker dye. Concomitantly, a covalent AtSerpin1–RD21 complex was detected indicative of a change in protease compartmentalization. Furthermore, mutant plants lacking RD21 or plants with AtSerpin1 over-expression exhibited significantly less elicitor-stimulated PCD than plants lacking AtSerpin1. The necrotrophic fungi Botrytis cinerea and Sclerotina sclerotiorum secrete oxalic acid as a toxin that stimulates cell death. Consistent with a pro-death function for RD21 protease, the growth of these necrotrophs was compromised in plants lacking RD21 but accelerated in plants lacking AtSerpin1. The results indicate that AtSerpin1 controls the pro-death function of compartmentalized protease RD21 by determining a set-point for its activity and limiting the damage induced during cell death.


Programmed cell death (PCD) is essential for plant development and the defense response. During the final steps of developmental cell death, the rupture of the tonoplast releases vacuolar hydrolases into the cytoplasm (Filonova et al., 2000; Avci et al., 2008; van Doorn et al., 2011). This ultimately leads to elimination of the cell remnants and contributes to the recycling of cellular contents (Groover et al., 1997). Programmed cell death can also be initiated during defense against pathogens. The ensuing cell death and additional associated defense responses contribute to limiting the growth of biotrophic-type pathogens that require living host cells for colonization. Toxins secreted by necrotrophic-type pathogens can also initiate PCD processes. However, in that case it is to the detriment of the host as the pathogen benefits from the host vestiges (Glazebrook, 2005; Spoel et al., 2007; Alkan et al., 2009).

Multiple pathways for the participation of vacuoles in defense-associated PCD have been proposed (Hatsugai et al., 2009; Hatsugai and Hara-Nishimura, 2010). One type consists of a membrane fusion pathway involving a proteasome-dependent protease activity that stimulates fusion of the plasma and vacuolar membrane to create direct funnels to the extracellular space. A second type of vacuolar-mediated cell death was described as occurring during viral infection. It is characterized by loss of vacuolar integrity and release of vacuolar contents into the cytoplasm (Hatsugai et al., 2004).

Viral-dependent enhanced vacuolar permeability requires the activities of vacuolar processing enzymes (VPEs). The VPEs have caspase-1 like activity and carry out a protein-processing function on precursor proteins, although the exact targets of VPEs are unknown (Hara-Nishimura et al., 1991; Rojo et al., 2003; Hatsugai et al., 2004; Yamada et al., 2005). Overexpression of Arabidopsis vacuolar γ-VPE-enhanced ion-leakage was related to PCD (Rojo et al., 2004) and γ-VPE mutants inhibited fumonisin-induced death (Kuroyanagi et al., 2005). Thus, VPEs are likely to be involved in activation of target proteins that stimulate vacuolar membrane breakdown but not vacuolar fusion (Hatsugai et al., 2006).

Other vacuolar proteases include the papain-like cysteine proteases (PLCPs) where RD21 (RESPONSIVE-TO-DESICCATION-21) is a dominant PLCP activity in Arabidopsis extracts (Gu et al., 2012). It contains a unique C-terminal granulin domain (Yamada et al., 2001) that shares homology to granulin-containing growth factors in animals. RD21 is present in pre-vacuole endoplasmic reticulum (ER) bodies and the vacuole (Yamada et al., 2001; Carter et al., 2004). RD21-like proteases play a role in plant immune responses in which they are thought to serve direct anti-pathogen roles. Silencing of RD21 resulted in an increase in the number of lesions in Botrytis cinerea infection (Shindo et al., 2012). In another example, the repression of the Solanaceae homolog of RD21, the apoplastic C14 protease, led to increased susceptibility to the fungus Phytophthora infestans (Bozkurt et al., 2011). Interestingly, in this hemibiotrophic interaction, multiple pathogen effectors target C14 and inhibit its activity or its secretion (Kaschani et al., 2010).

As proteases have a role in immunity, the control of their biological activity is important. We have recently shown that an evolutionarily conserved protease serpin-class inhibitor, AtSerpin1, targets RD21 (Lampl et al., 2010). AtSerpin1 was also shown to inhibit the activity of Arabidopsis AtMC9, a type-2 metacaspase of unknown function (Vercammen et al., 2006). However, other roles for AtSerpin1 have been suggested that are related to its ability to directly inhibit the proteases of herbivorous insects (Roberts and Hejgaard, 2008; Alvarez-Alfageme et al., 2011). Thus, the biological significance of protease inhibition by plant serpins remains to be established.

Serpins are relatively large proteinaceous inhibitors (340–440 amino acids) with a reactive center loop (RCL) targeted by proteases that form highly specific irreversible complexes with the serpin. Cleavage of the P1 and P1′ residues in the RCL create an acyl–enzyme intermediate transforming the serpin from a native ‘stressed’ metastable conformation to a cleaved ‘relaxed’ conformation. The released potential energy traps and inhibits the target protease breaking the catalytic cycle preventing cleavage of the acyl–enzyme intermediate (Huntington et al., 2000).

Here we show that AtSerpin1 accumulates in the cytosol. We demonstrate that elicitors that promote PCD affect vacuolar integrity and stimulate the accumulation of RD21 and the formation of the AtSerpin1–RD21 complex. Our results imply that AtSerpin1 can function to protect cells by limiting PCD-promoting RD21 activity that is released from the ER body or vacuole. Consistent with this role, the growth of necrotrophic but not hemibiotrophic fungi was compromised in the rd21 mutant line lacking protease activity and a line that overexpress AtSerpin1. The results suggest a role for AtSerpin1 in the control of host pro-death functions of the vacuolar protease RD21.


AtSerpin1 and RD21 are localized to different cellular compartments

In order to examine the cellular localization of AtSerpin1 in relation to RD21, expression lines of 35S:AtSerpin1-GFP and 35S:RD21-RFP were generated. Analysis of immunoblots of extracts from plants transformed with either AtSerpin1-GFP or RD21-RFP showed that the chromophores co-migrate, as fusion polypeptides of the predicted fusion size and free chromophore was not detected (Figure S1 in Supporting Information). Known identification reference probes for organelles (Nelson et al., 2007) were then transiently expressed in the 35S:AtSerpin1-GFP line. The resultant AtSerpin1-GFP signal is distinct from the ER and Golgi bodies (top images, Figure 1a). It is excluded from the vacuole and found adjacent to the plasma membrane within cytoplasmic pockets (bottom images, Figure 1a). Nuclear localization was also noted. In transient expression in the protoplast AtSerpin1 has been reported to accumulate in cytoplasm (Ahn, 2009).

Figure 1.

Confocal co-localization of AtSerpin1-GFP with RD21-RFP and mCherry tagged organelle markers. (a) Cotyledons of 35S:AtSerpin1-GFP stable lines (4 days old) were transformed with mCherry transgenes containing the following organelle markers: ER marker signal peptide was wall-associated kinase 2 with an ER retention signal HDEL; Golgi localization was soybean α-1,2-mannosidase I; tonoplast marker was γ-TIP aquaporin; plasma membrane marker was AtPIP2A plasma membrane aquaporin. Confocal images are shown of the AtSerpin1-GFP signal overlaid on the mCherry tagged organelle markers. (b) Cotyledons of 35S:AtSerpin1-GFP stable lines (4 days old) were transformed with 35S:RD21-RFP. Ruler is 10 μm. Boxes indicate regions magnified.

Previously, RD21 was shown to be localized to ER bodies and to vacuoles (Hayashi et al., 2001; Yamada et al., 2001). This observation is confirmed in imaging of a stable 35S:AtSerpin1-GFP transformed by 35S:RD21-RFP (Figure 1b). AtSerpin1-GFP fluorescence surrounds the ER bodies and is excluded from the vacuole, consistent with a cytoplasmic localization for AtSerpin1 and vacuolar and ER body localization for RD21. Similar results were obtained when the GFP reporter was fused to the N-terminal end of AtSerpin1 (Figure S2a). As a high level of protein expression may lead to mistargeting of AtSerpin1 we note that a native promoter AtSerpin1-GFP fusion shows similar results (Figure S2b). In addition, we compared overexpression AtSerpin1-HA lines with wild-type lines by subcellular fractionation (Figure S2c). In both lines, the vast majority of endogenous AtSerpin1 and AtSerpin1-HA was detected in the soluble S100 supernatant fraction and the distribution within the other fractions was similar. Thus, overexpression and wild-type AtSerpin1 are likely similarly localized.

As observed, AtSerpin1 and RD21 are present in separate cellular compartments, yet their covalent complex is readily detected in protein extracts fractionated on non-reducing denaturing gels (Roberts et al., 2011). To resolve this issue, we hypothesized that in healthy tissue the two components are separated and the formation of complex occurs rapidly only after mixing of the compartments during tissue maceration. In order to measure this process, we took advantage of the fact that RD21 belongs to the papain family of proteases that can be inhibited by the specific papain family protease inhibitor E-64. This compound irreversibly binds to the free active site cysteine to form a covalent linkage. While it cannot promote the disassociation of pre-existing complex, it will inhibit new complex from forming. To test this, extracts were prepared in the presence of freshly prepared E-64 added directly to frozen slurry at time 0 or 10 and 20 min after grinding of the seedlings or not added at all. The immunoblot was processed with α-AtSerpin1 antibody. As shown in Figure 2(a), the amount of complex increased as the addition of the E-64 inhibitor to the extract was delayed.

Figure 2.

Formation of the AtSerpin1-RD21 complex. (a) Extracts of AtSerpin1HA plants grown in B5-Gambourg's medium, were prepared with 100 μm E-64 either during grinding at 0, 10 or 20 min after grinding or with no addition of E-64 (−). The treated extracts were fractionated in a non-reducing SDS denaturing gel and the immunoblot was developed with α-AtSerpin1 antibody (upper immunoblot). The antibody was then removed and the immunoblot was developed with α-HA antibody (lower immunoblot). The lower image shows the Coomassie stained gel as loading control. (b) Accumulation of serpin-protease complex in wild-type seedlings. Non-reducing SDS denaturing fractionation of wild-type plant extracts transferred from 14 days in a long-day light regime to 2.5 days of constant light. Extracts were prepared on the second day of constant light without E-64. Time 0 is at 09:00. (c) Non-reducing SDS denaturing fractionation of wild-type extracts from plants prepared at 15:00 processed with (+) or without (−) E-64. Immunoblots were developed with α-AtSerpin1 antibody. The lower part of each figure shows Coomassie stain as a loading control.

Empirically, the AtSerpin1 antibody was found to readily detect the relaxed serpin form but to show marginal reactivity to the uncleaved 45 kDa native (or ‘stressed’) form. In order to better assay the amount of native unreacted serpin the blots were sequentially developed with α-HA antibody. This tag is present at the C-terminal region of the native form of AtSerpin1HA, but is removed upon cleavage of the RCL during complex formation. When the immunoblots were developed with α-HA, relatively more native AtSerpin1 was evident in extracts with early addition of E-64 (compare 0, 10, 20 and non-treated lanes in Figure 2a, lower immunoblot). Thus, the biochemical evidence using the E-64 inhibitor shows that the majority of complex detected in control healthy tissue was formed only after tissue maceration. This result is consistent with the observation that AtSerpin1 and RD21 are maintained in different cellular compartments.

Rhythmic appearance of AtSerpin1–RD21 complex

The complex between AtSerpin1 and RD21 is readily observed in AtSerpin1-HA overexpression plants and it was important to determine the level of endogenous AtSerpin1 expression. Interestingly, AtSerpin1 transcript displays rhythmic expression and shows high correlation with diurnal and circadian transcripts, whilst in contrast the expression of the RD21 transcript is non-rhythmic (Figure S3). Many Arabidopsis transcripts cycle in either a diurnal or a circadian manner (Michael et al., 2008). We therefore examined whether accumulation of the AtSerpin1–RD21 protein complex follows the fluctuating levels of the transcript. This was done by examining non-reducing fractionation of extracts prepared without the addition of E-64. The immunoblots were processed with α-AtSerpin1 serum shown above to be highly sensitive to the relaxed form of serpin present in the serpin–protease complex (Figure 2a). The results show that in plants transferred to constant light the accumulation of AtSerpin1 and its complex is indeed rhythmic, reaching a peak at 6 h and once again at 28 h (Figure 2b). The results suggest that under these conditions accumulation of AtSerpin1 protein has a component of transcriptional control and that the protein undergoes rapid cellular turnover and clearance. As expected, when the extract is prepared from wild-type seedlings at the optimum expression time, the formation of complex is abrogated in the presence of E-64 (Figure 2c). As the exposure time of Figure 2(b) is longer than that in Figure 2(a), the accumulation of native complex detected is less then in the overexpression line (compare Figure 2a,b).

AtSerpin1 and RD21 can modulate necrotroph-induced cell death

Plant vacuoles are cellular compartments that house a variety of hydrolases that participate in PCD. Therefore, we examined whether the RD21 protease functions in cell death and if the AtSerpin1 inhibitor situated in the cytoplasm can exert control over this process. Botrytis cinerea is a common necrotroph that induces mortality of host cells by the secretion of considerable amounts of oxalic acid, which impacts on vacuolar permeability (Errakhi et al., 2008). We hypothesized that if the release of a vacuolar protease like RD21 plays a role in cell death its inactivation by a serpin in the cytoplasm would modify the rate of cell death.

Detached leaves of wild type Arabidopsis and three mutant lines, overexpression AtSerpin1HA, and insertion lines atserpin1 and rd21 were inoculated with Botrytis conidia. The insertion lines do not contain immunodetectable protein products of their respective mutations (Lampl et al., 2010). Following inoculation, and over a period of 7 days, wild-type and atserpin1 mutant leaves displayed more decay around the inoculation site than the rd21 mutant and AtSerpin1HA overexpression lines (Figure 3a). We note that wild-type and atserpin1 lines showed similar infection rates. This may indicate that the level of AtSerpin1 in wild-type plants under these conditions is low. Importantly, the initial rates of germination of fungal conidia on these lines were examined and found to be identical; thus differential germination was not the cause for the distinct changes in virulence (Figure S4). Reactive oxygen species (ROS) originating from both host and pathogen are an early response to pathogen ingress and can be used to monitor the extent of pathogen ingress (Williams et al., 2011). During initial PCD and colonization monitored 36 h after inoculation a dramatic difference in the relative amounts of ROS as detected by dichlorodihydrofluorescein (DCF) staining was evident. Wild-type and atserpin1 lines showed a much higher degree of measurable ROS at the site of inoculum than did the rd21 and AtSerpin1HA overexpression lines (Figure 3b).

Figure 3.

Fungal virulence in wild-type and mutant lines. Droplet inoculation of Botrytis cinerea (7 μl of 106 spores ml−1) was carried out in a humid chamber on detached leaves of 3-week-old Arabidopsis wild-type (WT) and mutant lines: AtSerpin1HA, atserpin1 and rd21. (a) The graph shows the decay diameter measured over time in = 40 leaves in three independent experiments. A representative experiment is presented. dpi, days post-inoculation. (b) Accumulation of reactive oxygen species (ROS) as measured by dichlorodihydrofluorescein fluorescence 36 h post-incubation (hpi) in the wild type and mutants, fluorescent pixels were quantified and shown as relative fluorescence means and SE. Relative ROS is from = 15 and three independent experiments. Significance was measured by one-way anova with P-value <0.05. (c) Decay diameter in infection of Sclerotina sclerotiorum. Agar discs (5 mm2) containing hyphae of confluent grown S. sclerotiorum were placed on 3-week-old detached leaves of Arabidopsis wild type and mutants: AtSerpin1HA, atserpin1 and rd21. Leaves were incubated in a humid chamber. The S. sclerotiorum decay diameter was measured 24 hpi. Average means and SE is presented. (d) Decay diameter in infection of Colletotrichum higgisianum. Droplet incubation of C. higgisianum (7 μl of 106 spores ml−1) on detached leaves of 3-week-old Arabidopsis wild type and mutants as in (a). Decay diameter was measured 6 dpi. Average means and SE are presented. Significance was measured by one-way anova with P-value <0.05.

The necrotrophic fungal phytopathogen Sclerotina sclerotiorum is in a manner similar to Botrytis, dependent on oxalic acid (OA) for its pathogenicity (Kim et al., 2008). Inoculation of detached leaves of wild type Arabidopsis and mutant lines with Ssclerotiorum showed reduced rates of expansion of the decay diameter during infection of the rd21 mutant and AtSerpin1HA overexpression lines, similar to that obtained for Botrytis (Figure 3c).

Necrotrophic interactions thrive on cell death while hemibiotrophic interactions benefit from initial delayed cell death. To examine the interaction in a fungus with a different lifestyle, detached leaves of wild-type and mutant Arabidopsis lines were infected with the hemibiotrophic fungus Colletotrichum higgisianum. This pathogen invades the leaf via appressoria followed by the development of bulbous hyphae that grow in intimate biotrophic contact with living epidermal cells. In later stages, it develops a necrotrophic lifestyle (O'Connell et al., 2004). Significantly, a reciprocal pattern of interaction to that observed for Botrytis and Sclerotina was obtained. Serpin overexpression AtSerpin1HA and rd21 lines displayed enhanced susceptibility while atserpin1 was similar to the wild type (Figure 3d). Thus, in the necrotrophic interaction of Botrytis and Sclerotina the presence of excess protease inhibitor or the absence of RD21 activity inhibited fungal ingress compared with the wild type or lines without AtSerpin1. In contrast, in the hemibiotrophic C. higgisianum interaction the reciprocal trend was observed where the reduced host cell death afforded by rd21 and AtSerpin1HA lines was conducive to the initial establishment of this pathogen. The results are consistent with the hypothesis that the release of the vacuolar protease activity can modify the rate of pathogen-induced cell death.

Oxalic acid as a chemical elicitor of cell death

Oxalic acid is the major chemical effector stimulating the demise of host cells in Botrytis and Sclerotina infections. Fungal mutants that have reduced OA secretion show a loss of pathogenicity (Godoy et al., 2008). Oxalic acid toxicity was shown to require activation of host anion channels that destabilize vacuole integrity during OA-induced PCD (Errakhi et al., 2008). We examined whether treatment with OA could phenocopy the results observed with infection. Leaf disks prepared from the wild type and various mutant lines were incubated with 20 mm OA, with or without the anion channel inhibitor niflumic acid (Errakhi et al., 2008). Cell death was estimated by measuring the absorption of the Evans blue dye extracted from stained dead cells. As shown in Figure 4(a), cell death was highest in OA-treated wild-type and atserpin1 mutant lines and significantly lower in the AtSerpin1HA overexpression and rd21 mutant lines. In addition, RD21 overexpression lines showed enhanced cell death (Figure S5). The presence of niflumic acid significantly reduced cell death, consistent with the known involvement of the anion channels in vacuolar collapse during cell death (Errakhi et al., 2008). Thus, the levels of AtSerpin1 and RD21 determine cell fate in the presence of OA.

Figure 4.

Cell death induced by oxalic acid (OA). Cell death in leaf disks of 3-week-old Arabidopsis wild type (WT) and mutant plants incubated with double-distilled water (control), 20 mm OA or 20 mm OA with 200 μm niflumic acid (Nif). All control and treatments were carried out at pH 5.0 for 48 h. Lines used were: wild type, AtSerpin1HA, atserpin1 and rd21. (a) Relative cell death was measured with Evans blue. Average and SE are presented. Significance was estimated by one-way anova, P-value <0.05. (b) AtSerpin1HA plants were treated with 20 mm of oxalic acid. Protein was extracted from leaf disks after 24 h in the presence of E-64. Extracts were fractionated on non-reducing denaturing gels and gels were immunoblotted with α-AtSerpin1 (left) or α-RD21 (right) antibody. The lower image shows the Coomassie stained gel as a loading control.

To monitor for possible formation of the AtSerpin1–RD21 complex, proteins were extracted in the presence of E-64. As shown in Figure 2(a), E-64 added during the extraction process will inhibit the in vitro formation of complex. However, the detection of serpin–protease complex, despite the addition of E-64, is an indication of pre-formed complex that was induced in the cell as a result of treatment. Serpin–protease complex together with cleaved serpin products and the concomitant appearance of RD21 was observed 24 h after treatment of AtSerpin1HA tissue with OA (Figure 4b). The complex is not noted in the immunoblot with α-RD21 due to the fact that the protease is distorted in the complex and reacts less well to the antibody, as has been noted before (Lampl et al., 2010). The pre-formed complex indicates an OA-induced loss of membrane integrity. In all, the lower level of death in OA-treated plants in rd21 and AtSerpin1HA lines and enhanced level of death in atserpin1 mutant lines matches the results of Botrytis infection. Thus, the results are consistent with RD21 serving as a host pro-death function that can be modified by the presence of host serpin.

We wished to further explore the potential changes in membrane integrity by employing the acetoxymethylester (AM) of BCECF [2′,7′-bis-(2-carboxyethyl)-5-(6)-carboxyfluorescein] a widely used fluorescent viability and pH indicator stain. In non-stressed cells, BCECF accumulates in the cytoplasm where cellular esterases hydrolyze the acetoxymethyl ester group and trap the nascent negatively charged molecule between the plasmalemma and tonoplast membranes (Duan et al., 2007). It was found that BCECF rapidly accumulated in the cytoplasm of epidermal cells in control tissue (Figure 5a, left panel). However, upon addition of OA some of the cells began to show BCECF accumulation in the whole cell/vacuole or in vesicular structures of dead or dying cells (Figure 5a, middle and right panels). Quantitative analysis of the cell-staining types was carried out in the wild type and the mutant lines. Tissue of atserpin1 treated with OA for 12 h showed a mixed population of cells with a staining distribution of 10% cytoplasm, 87% vacuole and 2% vesicles, respectively (Figure 5b, top). In contrast, tissue from AtSerpin1HA and rd21 plants showed a significantly reduced proportion of vacuolar staining of 12 and 19%, respectively, that was only marginally higher than control tissue. This trend continued after 24 h of OA treatment with a comparatively larger proportion of cells showing vesicle-like staining in atserpin1 tissue (23%; Figure 5b, bottom). This stage correlates with cell death as verified by Evans blue stain (Figure 4a). In contrast, in AtSerpin1HA and rd21 lines a majority of the cells (70 and 63%, respectively) retained cytoplasmic-type staining of tissue. Thus, the BCECF pattern of staining highlights that changes in membrane integrity in response to OA that are exacerbated in atserpin1 lines. The results are consistent with a scenario whereby AtSerpin1 regulates the protease activity of released RD21 and thereby ultimately controls cell viability.

Figure 5.

Distribution of 2′,7′-bis-(2-carboxyethyl)-5-(6)-carboxyfluorescein (BCECF) fluorescence dye in response to oxalic acid (OA) treatment of Arabidopsis wild-type and mutant lines. Leaf disks of 10 mm diameter from 3-week-old Arabidopsis wild-type and mutant plants, AtSerpin1HA, atserpin1 and rd21, were incubated in double-distilled water at pH 5.0 as a control or 20 mm OA at pH 5.0. The disks were stained for 10 min with 1 μm BCECF, 12 and 24 h post-incubation. (a) Major forms of BCECF staining in epidermal cells. Three types of staining were defined, cytoplasmic (left), vacuolar (middle) and vesicular (right). (b) Distribution of the BCECF staining types; blue bars represent detection of BCECF in the cytoplasm, brown bars represent detection of BCECF in the vacuole and yellow bars represent vesicle formation. Micrographs (= 30) were analyzed for each of the plant lines and calculated for staining distribution. Average and SE are presented. Significance was measured by one-way anova, P-value <0.05. (c) Confocal localization of AtSerpin1-GFP and RD21-RFP on water control, pH 5.0. (d) Confocal co-localization of AtSerpin1-GFP/RD21-RFP 24 h after the addition of 20 mm OA. Images are from the surface of the epidermis. The box shows the selected area of enlargement. Scale bar is 10 μm.

The cellular distribution of AtSerpin1 and RD21 was examined in seedlings of stable 35S:AtSerpin1-GFP/35S:RD21-RFP lines treated with OA. As shown in Figure 5(c), in non-treated control tissue, AtSerpin1-GFP surrounds the ER bodies and the vacuole that contain RD21-RFP. When the seedlings were treated with OA, RD21-RFP accumulation became notable in many, but not all, ER bodies, in accordance with the results shown in Figure 4(b). Furthermore, a yellow halo at the ER body–cytoplasmic interface appeared, which is indicative of a degree of chromophore overlap (Figure 5d). This observation is consistent with loss of membrane integrity and coincides with the appearance of AtSerpin1HA–RD21 complex after treatment with OA (Figure 4b).

Formation of serpin–protease complex during induced cell death

Changes in membrane integrity may also play a role in the way that other elicitors induce cell death. As an example, the salicylic acid agonist benzothiadiazole (BTH) is known to induce moderate levels of cell death (Coll et al., 2010). To test the possible role of serpin–protease interaction, BTH was applied to seedlings and the initiation of cell death was monitored with BCECF. As shown in Figure 6(a), in atserpin1 lines at 24 h, the application of BTH resulted in a high proportion of cells displaying vacuolar localization of BCECF as opposed to cytoplasmic localization (64 and 36%, respectively). At 48 h the majority of cells in the same mutant line showed the appearance of staining in vesicular structures (55%). In contrast, the serpin overexpression line AtSerpin1HA showed only 2% vesicle staining at 48 h whereas rd21 and wild-type lines showed 8 and 4%, respectively. Concomitant with cell death, formation of E-64-resistant serpin complex was detected as shown for wild-type seedlings (Figure 6b). When 35S:AtSerpin1-GFP/35S:RD21-RFP seedlings were treated with BTH, RD21-RFP accumulation became notable in many, but not all, ER bodies (compare Figure 6c,d). Furthermore, the ER bodies were surrounded by a halo of diffuse yellow fluorescence indicative of chromophore mixing. The observation of such mixing is consistent with the observed formation of serpin–protease complex.

Figure 6.

Analysis of benzothiadiazole (BTH) and flagellin treatment. (a) Distribution of 2′,7′-bis-(2-carboxyethyl)-5-(6)-carboxyfluorescein (BCECF) dye in response to BTH treatment of Arabidopsis wild-type and mutants. Discs of 10 mm diameter of 3-week-old Arabidopsis wild-type and mutant plants, AtSerpin1HA, atserpin1 and rd21, were incubated in double-distilled water at pH 5.0 as a control or in 0.5 mm BTH. Leaf disks were stained for 10 min with 1 mm BCECF 24 and 48 h post-incubation and analyzed as in Figure 5. Blue bars represent detection of BCECF fluorescence in the cytoplasm, brown bars represent detection in the vacuole and yellow bars represent appearance in vesicles. Micrographs (= 30) were analyzed for each line or treatment. Average and SE are presented. Significance was measured by one-way anova, P-value <0.05. (b) Formation of serpin complex in BTH or control treated wild-type seedlings in the presence or absence of E-64. Arrow indicates complex. The lower image shows the Coomassie stained gel as loading control. (c) (d) Confocal localization of AtSerpin1-GFP and RD21-RFP on water control and BTH treatment. Box shows selected area of enlargement. Images were collated from a Z-series starting from the surface of the epidermis (c and d, upper images) or are from the central section of the epidermal cell (d, bottom images). Scale bar is 10 μm. In (d) seedlings treated with 0.5 mm BTH. The box shows the selected area of enlargement. Scale bar is 10 μm. (e) Formation of serpin–protease complex in flagellin-treated wild-type seedlings. Arrow indicates complex. The lower image shows the Coomassie stained gel as a loading control.

Full-length flagellin polypeptide has been reported to induce cell death (Naito et al., 2008). After application of flagellin to wild-type seedlings induction of cell death was readily observed (Figure S6). In parallel, E-64-resistant complex accumulated (Figure 6e). Thus, cell death and the concomitant appearance of AtSerpin1–RD21 complex were obtained after the application of full-length flagellin. Taken together, these results are consistent with a scenario in which RD21 sequestered in the ER bodies and the vacuole is released during plant–pathogen interactions where it then interacts with cytoplasmic AtSerpin1.


The results here show that RD21 contributes to cell death functions, which are regulated by AtSerpin1. Mechanistically, the control is achieved by maintaining these components in different cellular compartments. AtSerpin1 resides at the cytoplasmic side of ER bodies while RD21 accumulates in constitutive or stress-induced ER bodies and in the vacuole (Hayashi et al., 2001; Matsushima et al., 2002; Ogasawara et al., 2009). Overexpression lines of AtSerpin1 and knockout lines of RD21 showed significantly reduced PCD. Reciprocally, when the amount of RD21 was raised in overexpression lines, the rate of cell death was enhanced. This is consistent with a scenario whereby PCD activates the release of death-promoting RD21; however, even at the final execution step, the cell reserves a degree of death-reprieve capacity in the form of protease control by AtSerpin1. The AtSerpin1 knockout was similar to the wild-type infection rate indicating that in the cases shown here it is likely that the level of AtSerpin1 in the wild type is low.

The results here differ significantly from those recently published by Shindo et al. (2012). In that case, the rate of infection of B. cinerea at a singular time point was measured and larger spreading lesions were noted in mutant rd21 lines. The authors suggested that RD21 has an antifungal function. In our case, by using multiple time points we show consistently that mutant rd21 is more resistant in two different pathovar systems; B. cinerea and S. sclerotiorum. The resistance of the rd21 genotype to cell death and the hypersensitivity to cell death of overexpression RD21 was further confirmed using the OA toxin. This evidence shows that RD21 has a pro-death function. The differences in these two reports may arise from use of different B. cinerea isolates or the use of a whole plant (5 weeks old, short day, 10-h light regime) as opposed to detached leaves (3 weeks old, long day, 16-h light regime) used here.

Interestingly, RD21 localization differs from that reported for its Solanaceae homolog, C14, which shows apoplastic localization (Kaschani et al., 2010; Bozkurt et al., 2011). This variance may be a result of species-specific differences or represent the expression of non-orthologous genes; indeed, multiple RD21 homologs are present in Arabidopsis. Furthermore, C14 is thought to contribute to immunity by degrading pathogen effector or pathogen structural molecules or be necessary for cellular signaling; hence, reduction in C14 activity was shown to result in significantly enhanced susceptibility to the hemibiotroph P. infestans (Bozkurt et al., 2011). RD21 may also play a role in activating further cellular processes that lead to PCD. However, it is probable that our results place RD21 in a role of executor of host PCD, probably downstream from early cellular signaling events that launch PCD, for the following reasons: (i) multiple pathogen species, elicitors and toxins that induce PCD converge on RD21 and all are impacted by AtSerpin1; (ii) the levels of RD21 are induced during pathogenesis; and (iii) RD21 is a major cellular cysteine protease. In the light of this, the inhibition of C14 activity brought about by effectors of P. infestans may be related to the benefit that a hemibiotroph obtains in maintaining host cells alive rather than the prevention of processing of pathogen targets; in a manner similar to that demonstrated here for Arabidopsis–C. higgisianum interactions.

Endoplasmic reticulum bodies have been described in the Brassicales but not in other species. Two types of ER bodies have been noted in Arabidopsis that accumulate different β-glucosidases: a constitutive type containing a PYK10 and a wound-induced type, BGLU18 (Ogasawara et al., 2009). Similarly, RD21 accumulation was detected in some but not all ER bodies. In Arabidopsis and other cruciferous plants, ER bodies have been proposed to be part of a specialized glucosinolate defense system. It is thought that the membranes of the ER body separate the innocuous glycosylated glucosinolate precursor substrates present in the cytoplasm from the activating myrosinases present in the ER bodies. During stress or herbivory, the destruction of the cellular compartments will bring these components together (Nagano et al., 2005; Yamada et al., 2011). In an analogous manner, RD21 and AtSerpin1, which show vacuole/ER and cytosol localization, respectively, undergo stress-related interactions after the membrane of the vacuolar or ER body is compromised.

Homology searches with AtSerpin1 sequence showed it to be most similar to vertebrate intracellular Clade B serpins (Fluhr et al., 2011) that monitor intracellular protease activity released from lysosomes (Silverman et al., 2004). For example, the serpin Spi2A (serine protease inhibitor 2A) promotes the survival of cytotoxic T lymphocytes by inhibiting executioner proteases released from the lysosome, thus allowing progenitors cells to differentiate into memory CD8 T-cells (Liu et al., 2004). In nematodes, the cytoplasmic localized serpin SRP-6 can inhibit calpains and lysosomal cysteine peptidases (cathepsins K, L, S and V). When nematodes are subjected to osmotic shock, lysosomal content is released to the cytoplasm. If sufficient serpin-derived anti-peptidase activity exists within the cell the damage is contained and the animals survive. In srp-6 null mutants that lack serpin such stress results in necrotic cell death (Luke et al., 2007). Thus in animals, intracellular serpin inhibitors control lysosomal peptidases from exacerbating apoptosis or necrosis. In an analogous manner, AtSerpin1 affords protection by serving a pro-life function during elicitor or biotic-induced PCD.

For a mechanism by which serpin can protect cells, the initial release of RD21 should set off a signal that would encourage further release. What could the nature of that signal be? In animals, an amplifying feedback loop initiated by lysosomal rupture and subsequent release of proteases is thought to stimulate mitochondrial stress, cytochrome c release and ROS production that stimulates further lysosomal rupture (Zhao et al., 2003). In plants, a clear hierarchical relationship of events leading to vacuolar collapse has yet to be established. Here we have used the differential distribution of the vital stain BCECF to accentuate the changes in vacuolar integrity and compartmentalization. What motivates these changes? Anion channel effluxes play a role and are probably activated by calcium signals. Thus, niflumic acid, which blocks anion effluxes, protects tissue from OA as shown here and from cryptogein and harpin-induced PCD (Wendehenne et al., 2002; Gauthier et al., 2007; Errakhi et al., 2008; Reboutier and Bouteau, 2008). It has been reported that anion efflux promotes the accumulation of VPE-related transcripts (Gauthier et al., 2007) and VPEs were reported to have a role in the disruption of vacuole integrity during virus- or harpin-induced hypersensitive response (Hatsugai et al., 2004; Lam, 2005; Zhang et al., 2010). How VPEs induce vacuolar destruction is not known, but it is not directly related to a role in the proteolytic maturation of pre-proRD21 as this function was shown to be independent of VPE activity (Gu et al., 2012). Finally in plants, ROS that lead to PCD are characterized in some cases by mitochondrial stress and release of cytochrome c from mitochondria (Tiwari et al., 2002; Krause and Durner, 2004; Yao et al., 2004; Scott and Logan, 2008). The difficulty in establishing a clear hierarchical relationship between the phenomena described above may be due to the fact that, as noted for animal vacuolar–mitochondrial interplay, death-related pathways autoamplify each other.

Serpins are uniquely adapted to serve as a surveillance system that limits protease activity and governs PCD as they would facilitate removal of protease from the cytosol in a molecule-by-molecule manner. Such protease clearance would be due to the formation of an irreversible covalent complex, which may hasten protease turnover by other endogenous proteases or result in its permanent denaturation (Gettins, 2002). Consequently, the level of cytosolic serpins such as AtSerpin1 in Arabidopsis, or of SRP-6 in Caenorhabditis elegans constitutes a set-point for control of homeostatic cellular processes and PCD. It is notable that AtSerpin1 shows rhythmic appearance at the transcript and protein level. Rhythmic expression of defense genes was shown in resistance against Hyaloperonospora arabidopsidis (Wang et al., 2011). Similarly, metabolites that mediate the plant's responses to herbivores accumulate in diurnal rhythms in Nicotiana attenuata (Kim et al., 2011). Presumably, plants are programmed to ‘anticipate’ the likelihood of biotic stress that would require the triggering of PCD.

Significantly, RD21-like cysteine proteases that contain granulin and homologs of AtSerpin1 that contain a conserved ‘LR’ amino acid sequence in the predicted P2-P1 region of the RCL are highly conserved in both monocots and eudicots, suggesting an essential function (Roberts and Hejgaard, 2008; Fluhr et al., 2011). It is tempting to speculate that features of control of pro-death functions are similarly conserved. Surprisingly, AtSerpin1 and RD21 null mutants have no obvious phenotypes under normal growth conditions (Lampl et al., 2010; Gu et al., 2012). Hence, broad conservation and ubiquitous tissue distribution of AtSerpin1 and RD21-like genes may argue for additional fundamental roles in cellular metabolism that are likely masked by the functional cellular redundancy afforded by multigene serpin and protease families. This work presents molecular tools for understanding the control of cell death as stimulated by, but possibly not limited to, biotic interactions.

Experimental Procedures

Mutant lines and antibody

The construction of AtSerpin1HA and the origin of the insertion lines atserpin1, rd21 and antibodies for AtSerpin1, RD21 and for the HA epitope have been described (Lampl et al., 2010). Construction of AtSerpin1-GFP and RD21-RFP are detailed in Methods S1.

Transient transformation of the organelle marker lines (Nelson et al., 2007) and the RD21-RFP construct was done by using the fast agro-mediated seedlings transformation (FAST) assay (Li et al., 2009).

Plant growth conditions, fractionation, immunoblots and microscopy

Arabidopsis thaliana ecotype Col-0 plants were grown under white light in a 16-h light/8-h dark cycle at 21°C. Subcellular fractionation of tissue is described in Methods S1. Extraction of proteins and their fractionation on reducing and non-reducing denaturing gels has been described in Roberts et al. (2011). For inhibition by E-64, overexpression AtSerpin1HA plants were grown in controlled environment chambers at 21°C under a 16-h light regime on solid agar medium that contained B5 Gamborg's nutrients in 0.8% (w/v) phytoagar (Invitrogen, Experiments were performed on 14-day-old seedlings. Whole plants were extracted with extraction buffer [20 mm 2-amino-2-(hydroxymethyl)-1,3-propanediol (TRIS), pH 8.0, 1 mm EDTA and 50 mm NaCl). When E-64 (E-64c, Cayman Chemical, was added, it was added directly to the frozen slurry or after 10 or 20 min. The extracted tissue was centrifuged for 15 min at 17,000 g and gel fractionation of 80 μg of protein from the extract was carried out on 10% SDS-PAGE. Fractionated proteins were transferred onto polyvinylidene fluoride membrane (Bio-Rad, with a wet transfer for 1.5 h at constant voltage (100 V). The membrane was developed with AtSerpin1 antibodies (1:1000) and secondary anti-guinea pig horseradish peroxidase (1:3000) or with RD21 antibodies (1:1000), and secondary anti-rabbit horseradish peroxidase (1:3000).

Confocal microscopy analysis was carried out on 1-week-old transgenic stable AtSerpin1-GFP/RD21-RFP plants treated with elicitors. All images were taken with the 60 × plan-apo oil immersion objective on an Olympus IX81 FV1000D Spectral Type. For GFP, excitation was at 488 nm using an argon laser and emission was at 500–545 nm. For examination of RFP and Cherry, excitation was at 559 nm and emission was at 575–620 nm.

Pathogenicity assays

Botrytis cinerea, isolate B.O.05 (Quidde et al., 1999) and C. higgisianum, isolate IMI349063 (O'Connell et al., 2004) infection assays on Arabidopsis were performed by droplet inoculations of conidia. Sclerotina sclerotiorum, isolate 1980 (Godoy et al., 1990) infection assays on Arabidopsis were performed by plug disk inoculation.

Hyphal growth of B. cinerea was monitored by staining with bromophenol blue followed by fluorescence microscope observation (Nikon Eclipse E800, 15 h post-infection. Ten images were taken for each line. Growth efficiency was quantified by measuring the germinated spores versus the total number of spores. An average growth value for each leaf was used to generate the histograms. Decay diameter was scored during 7 days of inoculation as lesions of necrotic growth. The assay was repeated at least three times, and a representative dataset is used. Botrytis cinerea infected leaves were evaluated for cell death by staining with Evans blue. Cell death was quantified 48 h post-inoculation in the following way: treated leaf disks were submerged in 0.2% (w/v) Evans blue at 25°C for 60 min and rinsed five times with PBS. Disks were placed in 1% (w/v) sodium dodecyl sulfate (SDS) solution and incubated for 10 min at 37°C. Solutions were centrifuged at 13,000 g for 5 min. The quantity of remaining dye was measured spectrophotometrically at 600 nm. Measurements were expressed as relative values, with 1.0 corresponding to maximum cell death (Kim et al., 2008).

The accumulation of ROS was detected with DCF (D-399, Molecular Probes, Invitrogen) 36 h post-inoculation by immersing the tissue for 10 min in 10 μm DCF in the dark at room temperature. The leaves were rinsed with PBS five times, and the DCF fluorescence was imaged in a fluorescence microscope (Nikon Eclipse E800) using 465-nm and 535-nm excitation and emission filters, respectively. Fifteen images for each line were analyzed by Image J software in three independent experiments to obtain the average green fluorescence. Relative fluorescence means and SD were calculated.

Treatments with elicitors

For oxalic acid treatment, leaf discs of 10 mm diameter were prepared from 3-week-old Arabidopsis wild-type and mutant lines and incubated with 20 mm sodium-oxalate (pH 5.0) with or without 200 μm niflumic acid (pH 5.0). Disks were placed in the dark at room temperature. Control leaf disks were incubated with double-distilled water at pH 5.0. For BTH treatment 2-week-old plants were sprayed with 0.5 mm BTH in 0.1% Tween-20. For flagellin treatment, recombinant full-length polypeptide was purified as described (Naito et al., 2007). Five-day-old seedlings grown on MS agar plates were transferred into liquid MS medium containing 3 μg ml−1 flagellin. Cell death was analyzed after 4 days with Evans blue.

The BCECF viability assay

Viability was determined and quantified by using the acetoxymethylester of BCECF (B-1150, Molecular Probes, Invitrogen). Disks of 10 mm diameter from 3-week-old Arabidopsis wild-type and mutant lines were incubated in DDW at pH 5.0 as a control or in 20 mm sodium-oxalate at pH 5.0 or in 0.5 mm BTH for the times indicated. The incubated disks were washed twice with PBS and loaded with 10 μm BCECF for 10 min in the dark at room temperature and then rinsed several times with incubation buffer to remove excess dye. Fluorescence intensity was measured at 520 nm after excitation at 488 nm using a fluorescence or confocal microscope (Nikon Eclipse E800 or Olympus IX81 FV1000D Spectral Type). For each mutant and treatment 25 images were analyzed for cell type staining.

Statistics analysis

Significance was measured for all plots by one-way anova with a P-value of 0.05. All pairwise multiple comparisons were performed using the Student–Newman–Keuls method.


This work was supported by the Israel Science Foundation (grant no. 1008/11). We thank Professor Yuki Ichinose, Okayam University Graduate School of Natural Science and Technology for recombinant full-length flagellin polypeptide. Vladimir Kiss provided excellent technical assistance for confocal microscopy. We thank Professor Ikuko Hara-Nishimura, Kyoto University, for kindly providing RD21 antibody and Dr Renier A. L. van der Hoorn, Max Planck Institute for Plant Breeding Research, for kindly providing the RD21-ko line. We are grateful for the excellent comments on this manuscript provided by Dr Thomas H. Roberts, University of Sydney. The authors have no conflicts of interest to declare.