The Plant Journal (2010) 64, 71–85
doi: 10.1111/j.1365-313X.2010.04310.x
The AAA-type ATPase AtSKD1 contributes to vacuolar maintenance of Arabidopsis thaliana Mojgan Shahriari1, Channa Keshavaiah1,†, David Scheuring2, Aneta Sabovljevic1, Peter Pimpl2,‡, Rainer E. Ha¨usler1, Martin Hu¨lskamp1,* and Swen Schellmann1,* 1 Biozentrum Ko¨ln, University of Cologne, Zu¨lpicher Street 47 b, 50674 Cologne, Germany, and 2 Heidelberg Institute for Plant Science, Abteilung Zellbiologie, Im Neuenheimer Feld 230, 69120 Heidelberg, Germany Received 27 May 2010; revised 1 July 2010; accepted 9 July 2010; published online 19 August 2010. * For correspondence (fax +49 221 470 5062; e-mail
[email protected] or
[email protected]). † Present address: Genetics and Biotechnology Lab, Department of Biochemistry, University College Cork, Cork, Ireland. ‡ Present address: ZMBP, Developmental Genetics, University of Tu¨bingen, Auf der Morgenstelle 3, 72076 Tu¨bingen, Germany.
SUMMARY The vacuole is the most prominent organelle of plant cells. Despite its importance for many physiological and developmental aspects of plant life, little is known about its biogenesis and maintenance. Here we show that Arabidopsis plants expressing a dominant-negative version of the AAA (ATPase associated with various cellular activities) ATPase AtSKD1 (SUPPRESSOR OF K+ TRANSPORT GROWTH DEFECT1) under the control of the trichome-specific GLABRA2 (GL2) promoter exhibit normal vacuolar development in early stages of trichome development. Shortly after its formation, however, the large central vacuole is fragmented and finally disappears completely. Secretion assays with amylase fused to the vacuolar sorting signal of Sporamin show that dominant-negative AtSKD1 inhibits vacuolar trafficking of the reporter that is instead secreted. In addition, trichomes expressing dominant-negative AtSKD1 frequently contain multiple nuclei. Our results suggest that AtSKD1 contributes to vacuolar protein trafficking and thereby to the maintenance of the large central vacuole of plant cells, and might play a role in cell-cycle regulation. Keywords: VPS4, endosomal sorting complexes required for transport, Arabidopsis, vacuole, SKD1, MVB.
INTRODUCTION In yeast, mutant screens have identified a large number of genes responsible for vacuolar protein sorting (Rothman and Stevens, 1986; Robinson et al., 1988; Rothman et al., 1989). According to their cellular morphology they have been grouped into six subclasses (A–F; Raymond et al., 1992). In recent years the so-called ‘class E’ genes have become the focus of intense research (reviewed in Hurley et al., 2009). Mutations in all of the 19 class E genes cause the formation of an abnormal endomembrane structure, the class E compartment, which contains endosomal and vacuolar markers (reviewed in Babst, 2005; Hurley and Emr, 2006). Class E genes are responsible for recognition and sorting of mono- and multi-ubiquitylated cargo into distinct endosomal membrane regions that later form internal vesicles of multivesicular bodies (MVBs). These MVBs eventually fuse with the vacuole and release their content into its lumen (reviewed in Babst, 2005; Hurley and Emr, 2006; Hurley et al., 2009). ª 2010 The Authors Journal compilation ª 2010 Blackwell Publishing Ltd
Sorting into the inner vesicles of the MVBs is accomplished by the sequential action of four multiprotein complexes, ESCRT-0, -I, -II, -III (endosomal sorting complexes required for transport) (Katzmann et al., 2001, 2003; Babst et al., 2002a,b). ESCRT-0 recognizes ubiquitylated cargo and recruits ESCRT-I that takes over the cargo via Vps23p (Katzmann et al., 2001, 2003). Cargo is passed on to ESCRT-II with its ubiquitin-binding component Vps36p (Babst et al., 2002b). ESCRT-II in turn recruits ESCRT-III that is responsible for cargo concentration (Babst et al., 2002a). After deubiquitylation, cargo is internalised and delivered to the vacuole (Amerik et al., 2000). A key event and prerequisite for budding of internal vesicles is the ATP-driven disassembly of the ESCRT components from the endosomal membrane mediated by the AAA-type (ATPase associated with various cellular activities) ATPase Vps4p (mammalian SKD1; Babst et al., 1997, 1998; Finken-Eigen et al., 1997; Shirahama et al., 1997; Scheuring et al., 1999; Yoshimori et al., 2000). In yeast Vps4 mutants, 71
72 Mojgan Shahriari et al. vacuolar, endocytic and late-Golgi markers accumulate in the class E compartments together with members of the ESCRT-III complex (Babst et al., 1998). Vps4/SKD1 activity is regulated by its oligomeric state. In its ADP-bound state, it forms a dimer that is localised in the cytoplasm; upon ATP binding, Vps4p/SKD1 oligomerises and translocates to the endosomal membrane. Oligomerisation also promotes the Vps4p ATPase activity. Oligomeric state and membrane association are regulated by the two class E proteins Vta1 and Vps46p that both bind to Vps4p/SKD1 in an ATP-dependent manner (Azmi et al., 2006; Lottridge et al., 2006). Vta1p interacts on endosomal membranes with Vps60p, an additional Vps4p regulator (Shiflett et al., 2004). Two mutations of Vps4p have been described that are deficient either in ATP binding or in ATP hydrolysis. They produce dominant-negative class E phenotypes when transformed into wild-type yeast cells or mammalian cell cultures (Babst et al., 1997; Fujita et al., 2003). In plants, it has been shown that homologues of the core ESCRT proteins are present and that the mutant of the ESCRT-I subunit elch (Vps23) leads to defects in plant cytokinesis (Spitzer et al., 2006; Winter and Hauser, 2006). A plant homologue of Vps4p (mcSKD1) has been described for the ice plant (Mesembryanthemum crystallium), an inducible halophyte (Jou et al., 2004). It is localised mainly in the endoplasmic reticulum-Golgi network and facilitates K+ uptake (Jou et al., 2006). In a recent publication, Haas and colleagues characterized the Arabidopsis homologue of Vps4p, AtSKD1. They showed that AtSKD1 is involved in the biogenesis of MVBs. Cultured cells expressing a dominant-negative version of AtSKD1 contained enlarged endomembrane structures reminiscent of the class E compartments described for yeast and animals (Haas et al., 2007). Analysis of AtSKD1 in intact Arabidopsis plants was limited by the fact that no atskd1 loss-of-function mutants exist. Dominant-negative AtSKD1 leads to cell-death and does not allow constitutive expression under the 35Spro-promoter. Therefore Haas and colleagues used an ethanolinducible promoter that, however, conferred lower and more variable expression than 35Spro. The observed effects were weaker in these transgenic plants than in cell culture experiments (Haas et al., 2007). Nevertheless, in TEM studies of induced plants they could show that AtSKD1 is involved in MVB formation. In our analysis, we have followed a different approach by expressing dominantnegative versions of AtSKD1 under the strong cell typespecific promoter GL2pro that restricts expression to leaf trichomes and non-hair-cell files in the root epidermis (Szymanski et al., 1998). In these lines, we found defects in vacuolar maintenance and vacuolar trafficking of soluble cargo and multinucleated trichome cells resembling the elch mutant (Spitzer et al., 2006).
RESULTS Generation and characterization of dominant-negative AtSKD1 constructs To analyse the function of AtSKD1 in plants, we used constructs carrying mutations in AtSKD1 that have been shown in yeast, mammals and ice plant to render the respective VPS4 proteins dominant-negative. We created three versions: changing Lys178 to Ala (K178A) inhibits ATP binding and Glu232 to Gln (E232Q) substitution inhibits ATP hydrolysis. Additionally we combined both mutations (AQ). To test the functionality of the constructs, we expressed the three mutant and the wild-type versions in Escherichia coli, purified the recombinant protein (Supplementary Figure S1a–d) and analysed ATPase activities (Figure 1). For the wild-type protein, ATPase activity was linearly correlated with the protein concentration, yielding an apparent specific activity of 1.18 U mg)1 protein (Figure 1a). The ATPase activity of the mutated proteins AtSKD1(K178A), AtSKD1(E232Q) and AtSKD1(AQ) was diminished to 3.2, 1.8 and 2.6%, respectively compared with the wild type (WT; Figure 1b–d). The substrate dependency of AtSKD1(WT) obeyed Michaelis–Menten kinetics (Figure 1e) with an apparent Km value for ATP of 266 164 lM and a Vmax of 1.77 0.27 U mg)1. The Km is in a similar range to that reported for VPS4p (600 lM; Babst et al., 1998). The assessment of kinetic constants for the mutated proteins was hampered by their low specific activity (Figure 1f–h). Compared with AtSKD1(WT), the Vmax declined to 0.035 0.005 and 0.07 0.03 U mg)1 in AtSKD1(K178A) and AtSKD1(E232Q), respectively, with accompanying apparent Km values for ATP of 52.2 39 and 544 566 lM. The reaction velocity of AtSKD1(AQ) was linearly correlated with the ATP concentration in the analysed range and thus did not allow the determination of kinetic constants (Figure 1h). Dominant-negative AtSKD1 causes abnormally enlarged late endosomes AtSKD1 is ubiquitously expressed; the highest protein levels being detectable in leaves (Haas et al., 2007). In leaf trichomes AtSKD1 expression is upregulated 1.83-fold compared with the whole leaf (Jakoby et al., 2008). To analyse the effects of malfunction of AtSKD1 in Arabidopsis plants we expressed the dominant-negative versions under the GL2 promoter (GL2pro), which restricts expression to trichomes and non-root-hair files (Szymanski et al., 1998). This approach avoids two problems that have been described for the use of the ethanol-inducible system in this case: low overall expression and expression variability (Haas et al., 2007). To analyse the subcellular localisation and the cellbiological phenotypes of dominant-negative AtSKD1 we expressed C-terminal YFP fusions of AtSKD1(AQ) and
ª 2010 The Authors Journal compilation ª 2010 Blackwell Publishing Ltd, The Plant Journal, (2010), 64, 71–85
Vacuolar maintenance by AtSKD1 73 localisation (data not shown). In about 10% of the cells AtSKD1:YFP(WT) is localised on dotted structures that are also positive for the MVB marker ARA7 (Figure 3c) but not for G-rb and SYP61-YFP (Figure 3a,e). The 35Spro:AtSKD1(K178A) mutant showed the same localisation as the wild-type version (data not shown). In trichomes, the localisation of the wild-type fusion protein varied depending on the age of the observed cell (Figure 2a–d,i). In young trichome cells that had not developed a large vacuole yellow fluorescent protein (YFP) fluorescence was present in the cytosol and in the nucleus (Figure 2a–c) whereas in older, fully vacuolated cells (Figure 2d,i) and in elongated non-root-hair cells (Figure 2m,o) fluorescence could be observed in punctae. This mirrors the two localisation types found in the protoplast assays. Thus, the localisation of AtSKD1 in an individual protoplast is probably dependent on the age of the cultured cell that the protoplast was derived from. Fluorescence in non-root-hair-cells and both young and old trichomes expressing GL2pro:AtSKD1(AQ) was detectable only in large speckles, again mirroring the protoplast experiments and the cell culture results obtained by Haas and colleagues (Figure 2e–h,n,p; Haas et al., 2007). Interestingly, the nuclear localisation observed with AtSKD1:YFP(WT) was absent in the AtSKD1:YFP(AQ) line (Figure 2j–l). Dominant-negative AtSKD1 induces trichomes with multiple nuclei Figure 1. Enzyme kinetics of AtSKD1 and its putative dominant-negative versions. (a)–(d) Dependences of the AtSKD1 ATPase activity on the protein content: (a) AtSKD1(WT), (b) AtSKD1(K178A), (c) AtSKD1(E232Q), (d) AtSKD1(AQ) (WT, wild type; AQ, both mutations combined). The curves were constructed with data sets from three (a) or two independent protein preparations (b)–(d). (e)–(h) The ATP dependence of the reaction velocity of recombinant AtSKD1 protein: (e) AtSKD1(WT), (f) AtSKD1(K178A), (g) AtSKD1(E232Q), (h) AtSKD1(AQ). The curves were constructed from datasets obtained from two independent protein preparations. Note the different scales of the y-axis of the wild type (a, e) and mutant kinetic curves (b–d, f–h). All data are the mean values of two independent measurements. Correlation coefficients (R2) for the curve fits are given in each graph.
AtSKD1(WT) under the control of the GL2 promoter (Figure 2) and performed colocalisation experiments with the wild type and all dominant-negative AtSKD1 mutants in Arabidopsis protoplasts (Figure 3). In the protoplast system, we observed large membrane structures with 35Spro:AtSKD1:YFP(AQ), and similarly with 35Spro:AtSKD1:YFP(E232Q) (not shown) that are also positive for ARA7:CFP as a MVB marker (Figure 3d). No colocalisation with the Golgi marker G-rb or the TGN marker SYP61-YFP (Niemes et al., 2010) could be observed (Figure 3b,f). In the AtSKD1:YFP(WT)transfected protoplasts we found two kinds of localisations. The majority of the cells show a largely cytosolic and nuclear
Transformants of untagged versions of the three dominantnegative constructs and wild-type AtSKD1 under GL2pro control were analysed for trichome phenotypes. On first inspection, the overexpression of the dominant-negative constructs apparently led to bigger trichomes with additional branches. Statistical analysis (Fisher exact test, http://www.quantitativeskills.com/sisa/statistics/fiveby2.htm; Table 1) showed that the distribution of branch numbers in all dominant-negative lines is different (P < 0.001) from the GL2pro:AtSKD1(WT) and the Col-0 controls. However, branching is deregulated rather than increased, resulting in a broader distribution of the number of branches per trichome. The average number of three branches per trichome is not changed (Table 1). As regulation of branching in Arabidopsis trichomes is often correlated with endoreduplication levels (Perazza et al., 1999) we analysed the DNA content of trichome nuclei. We did not find any deviations from wild type in either of the transgenic lines (Table 1). Instead, we found multinucleated trichomes in all dominant-negative versions (Figure 4a,b, Table 1). The phenotype was most pronounced in the GL2pro:AtSKD1(AQ) line with 5.8% of the trichomes containing up to three nuclei per cell (Table 1). Altogether, the three dominant-negative lines show an increase in phenotypic strength (branching deregulation and multiple
ª 2010 The Authors Journal compilation ª 2010 Blackwell Publishing Ltd, The Plant Journal, (2010), 64, 71–85
74 Mojgan Shahriari et al.
(a)
(b)
20 µm
(e)
20 µm
(f)
20 µm
(i)
(c)
80 µm
(g)
20 µm
(j)
(d)
80 µm
(h)
80 µm
(k)
80 µm
(l)
Figure 2. Subcellular localisation of GL2pro:AtSKD1(WT):YFP and GL2pro:AtSKD1(AQ):YFP in Arabidopsis trichome and non-root-hair cells (WT, wild type; AQ, both mutations combined). Developmental row of trichomes expressing GL2pro:AtSKD1(WT):YFP (a–d) and GL2pro:AtSKD1(AQ):YFP (e–h). Trichomes were analysed in the following stages: unbranched (a), (e); twobranched (b), (f); three-branched (c), (g); mature (d), (h). (i) Close-up of a mature trichome expressing GL2pro:AtSKD1(WT):YFP. Note the dotted structures. (j)–(l) Colocalisation of GL2pro:AtSKD1(AQ):YFP with 4¢,6-diamidino-2-phenylindole (DAPI): (j) AtSKD1(AQ):YFP; (k) DAPI; (l) overlay (arrowheads in (k) and (l) point to the nucleus). (m)–(p) GL2pro:AtSKD1(WT):YFP (m) and GL2pro: AtSKD1(AQ):YFP (n) in non-root-hair cells. (o), (p) Brightfield images. Scale bars = 20 lm.
80 µm
(m)
(n)
(o)
(p) 20 µm
nuclei) from GL2pro:AtSKD1(K178R) to GL2pro:AtSKD1(AQ) that is correlated with the strength of the biochemical and cell biological defects. AtSKD1 can interact with the Arabidopsis ESCRT system The trichome phenotype of the dominant-negative AtSKD1 lines resembles the elch mutant in which about 2% of the trichomes show multiple nuclei (Spitzer et al., 2006). A functional connection of both proteins is not surprising as ELCH encodes the homologue of yeast Vps23p, a class E protein. We used yeast two-hybrid (Y2H) and bimolecular fluorescence complementation (BiFC) assays to determine the interactions of AtSKD1 with Arabidopsis ESCRT core components (Spitzer et al., 2006) and homologues of ESCRT-III-associated proteins (Winter and Hauser, 2006; Figures 5, 6, S2 and S3, Table 2). AtSKD1 interacts with both paralogs of each of VPS20 and VPS32 and with VPS24-1. All interactors are small coiled-coil proteins that are putative members of the ESCRT-III subcomplex (Figure 5a,b; Winter and Hauser, 2006). AtSKD1 displayed no interactions with ESCRT-I and -II members. The Arabidopsis genome encodes two proteins with similarities to yeast Vps46p and Vps60p (Figure 5a; Winter
and Hauser, 2006). Vps46p and Vps60p interact with Vta1 (the yeast orthologue of AtLIP5, LYST INTERACTING PROTEIN5) and VPS4p regulating its endosomal association and ATP hydrolysis rate. Vps4p, Vta1p, Vps60p and Vps46p have recently been renamed ESCRT-III-associated proteins (Azmi et al., 2006; Lottridge et al., 2006; Leung et al., 2008). The Arabidopsis homologues of Vps46p, CHMP (charged MVB proteins) 1A and B, affect endosomal sorting and localisation of the auxin efflux carrier PIN2 (Spitzer et al., 2009). We found interactions of AtSKD1 with AtLIP5 and VPS60-1 and of AtLIP5 with VPS60-1 and CHMP1A in both Y2H and BiFC (Figures 5b, 6, S2 and S3). Homo-dimerization (AtLIP5, CHMP1A, VPS60-1) and the interactions between CHMP1A and -B, and between CHMP1B and AtLIP5 were detectable with the Y2H system only. All detected interactions are summarized in Figure 5b. To determine the subcellular localisation of the interactions we performed colocalisation experiments in Arabidopsis protoplasts transformed with the respective BiFC constructs of interacting proteins in combination with markers for the Golgi and MVBs (Figure 6). We found that BiFC signals were located on punctuate structures that were positive for the MVB marker ARA7 (Figure 6e–h). By
ª 2010 The Authors Journal compilation ª 2010 Blackwell Publishing Ltd, The Plant Journal, (2010), 64, 71–85
Vacuolar maintenance by AtSKD1 75 Figure 3. Colocalisation of AtSKD1 (WT) and AtSKD1 (AQ) with markers for Golgi, multivesicular bodies (MVBs), and TGN in Arabidopsis protoplasts (WT, wild type; AQ, both mutations combined). For each cotransfection four images are shown: localisation of the respective AtSKD1 version (green), localisation of the respective marker (red), an overlay and a brightfield picture. Colocalisation in the overlay appears in yellow. Colocalising MVBs in (c) and (d) are indicated by arrowheads. Scale bar = 10 lm. (a) AtSKD1:YFP(WT) G-rb (Golgi). (b) AtSKD1:YFP(AQ) G-rb. (c) AtSKD1:YFP(WT) ARA7:CFP (MVB). (d) AtSKD1:YFP(AQ) ARA7:CFP. (e) AtSKD1:CFP(WT) AtSYP61:YFP (TGN). (f) AtSKD1:CFP(AQ) AtSYP61:YFP.
(a)
(b)
(c)
(d)
(e)
(f)
contrast, no colocalisation was found with the Golgi marker G-rb (Nelson et al., 2007; Figure 6a–d). In summary, AtSKD1 can interact with ESCRT-III components and putative ESCRT-III-associated proteins on MVBs. Therefore it is likely that the function of AtSKD1 in the ESCRT pathway is conserved between plants, yeast and mammals. Dominant-negative AtSKD1 induces vacuolar defects in Arabidopsis trichomes About 5 days after leaf initiation, trichomes of the dominantnegative transgenic lines were collapsing (Figure 4c,d). To test the viability of the collapsed trichomes we performed a dead/live staining with fluorescein-diacetate (FDA) and
propidium iodide (Figure 7). To our surprise most of the collapsed cells were still giving rise to cytosolic FDA fluorescence, indicating that the cells were alive (Figure 7i). Only later did they lose the staining and instead accumulate the dead stain propidium iodide (Figure 7j–l). Staining with FDA is based on enzymatic cleavage of FDA, releasing free green-fluorescing fluorescein. This reaction occurs in the cytosol of the cell, leading to a negative staining of the vacuole. As we could not observe dark regions in the collapsed trichome cells (Figure 7i) we concluded that the cells had lost their vacuole and had collapsed because of reduced turgor. To further substantiate this hypothesis, we crossed GL2pro:AtSKD1(WT) and
ª 2010 The Authors Journal compilation ª 2010 Blackwell Publishing Ltd, The Plant Journal, (2010), 64, 71–85
76 Mojgan Shahriari et al. Table 1 Statistics of branch numbers, DNA content, cluster frequencies and multiple nuclei of the wild type (Col-0) and dominant-negative AtSKD1versions Trichomes with a given branch number (%)
Trichomes with a given number of nuclei (%)
Genotype
1
2
3
4
5
n
DNA content (C)
SD (C)
n
Cluster (%)
n
1
2
3
4
n
Col-0 pGL2:AtSKD1 pGL2:AtSKD1(K178A) pGL2:AtSKD1(E232Q) pGL2:AtSKD1(AQ)
0 0
