Arabidopsis Small Ubiquitin-Like Modiﬁer Paralogs Have Distinct Functions in Development and Defense

Posttranslational modiﬁcations allow dynamic and reversible changes to protein function. In Arabidopsis thaliana , a small gene family encodes paralogs of the small ubiquitin-like posttranslational modiﬁer. We studied the function of these paralogs. Single mutants of the SUM1 and SUM2 paralogs do not exhibit a clear phenotype. However, the corresponding double knockdown mutant revealed that SUM1 and SUM2 are essential for plant development, ﬂoral transition, and suppression of salicylic acid (SA)–dependent defense responses. The SUM1 and SUM2 genes are constitutively expressed, but their spatial expression patterns do not overlap. Tight transcriptional regulation of these two SUM genes appears to be important, as overexpression of either wild-type or conjugation-deﬁcient mutants resulted in activation of SA-dependent defense responses, as did the sum1 sum2 knockdown mutant. Interestingly, expression of the paralog SUM3 is strongly and widely induced by SA and by the defense elicitor Flg22, whereas its expression is otherwise low and restricted to a few speciﬁc cell types. Loss of SUM3 does not result in an aberrant developmental phenotype except for late ﬂowering, while SUM3 overexpression causes early ﬂowering and activates plant defense. Apparently, SUM3 promotes plant defense downstream of SA, while SUM1 and SUM2 together prevent SA accumulation in noninfected plants. defense gene expression via a transcriptional repressor domain. In addition, XopD contains a SUMO protease domain with SUMO protease activity toward SUM1- and SUM2-modiﬁed but also SUM3-modiﬁed proteins. Both biochemical activities (transcriptional repression and SUMO protease activity) result in enhanced virulence of bacteria expressing XopD.


INTRODUCTION
Posttranslational modifications (PTMs) set a reversible mark on specific amino acids, enabling proteins to change conformation or to recruit and interact with specific partners. These PTMinduced interactions allow, for example, dynamic responses to stress conditions or determine developmental patterns in higher eukaryotes. A prominent member of the Ubiquitin-like class of PTMs is SMALL UBIQUITIN-LIKE MODIFIER (SUMO) (Downes and Vierstra, 2005). In budding yeast (Saccharomyces cerevisiae) and humans, SUMO protein modifications are well characterized and linked to nuclear processes, such as nuclear import, transcription, chromatin remodeling, and DNA replication (Heun, 2007). It has become clear that in plants these nuclear processes are also regulated by SUMO, despite the fact that only a small number of plant SUMO targets have been identified (Miura et al., 2005(Miura et al., , 2007bGarcia-Dominguez et al., 2008;Jin et al., 2008;Budhiraja et al., 2009). Some eukaryotes (e.g., yeast [S. cerevi-siae], flies [Drosophila melanogaster], and worms [Caenorhabditis elegans]) express only a single SUMO gene, whereas vertebrates express up to four paralogs. The genome of the model plant Arabidopsis thaliana potentially encodes eight SUMO paralogs (Kurepa et al., 2003;Novatchkova et al., 2004). Currently, only four of these (SUM1, SUM2, SUM3, and SUM5) have been shown to act as functional PTMs (Colby et al., 2006;Budhiraja et al., 2009). SUM1 and SUM2 are most closely related, sharing 89% protein sequence identity, whereas SUM3 shows 48% identity and SUM5 only 35% identity to SUM1. These sequence differences suggest functional diversification of these paralogs.
The process of SUMO conjugation is largely conserved among yeast, mammals, and plants. First, SUMO precursors undergo C-terminal processing by SUMO proteases (also referred to as Ubiquitin-like proteases or Ulps) that liberate a C-terminal double-Gly motif. Subsequently, this mature form of SUMO is conjugated to target proteins via two consecutive steps catalyzed by the SUMO E1 ACTIVATING ENZYME (SAE1/SAE2 in Arabidopsis) and the SUMO E2 CONJUGATING ENZYME (SCE1 in Arabidopsis; Ubc9 in yeast and mammals). SUMO conjugation (called SUMOylation) results in the formation of an isopeptide bond between the SUMO C terminus and the side chain of an acceptor Lys in target proteins. SUMOylation is essential in Arabidopsis (i.e., null mutants of SAE1/SAE2 and SCE1 are embryo lethal) (Saracco et al., 2007). SCE1 can directly recognize and SUMOylate Lys residues embedded in a SUMOylation consensus motif (C-K-X-E/D, where C denotes a bulky hydrophobic residue and X any residue). SCE1-mediated SUMOylation can be promoted by SUMO E3 ligases (Bernier-Villamor et al., 2002;Yunus and Lima, 2006;Anckar and Sistonen, 2007). Loss of the Arabidopsis SUMO E3 ligase HIGH PLOIDY2 (HPY2) leads to severe dwarfism, and HPY2 was found to be essential for proper meristem development (Huang et al., 2009;Ishida et al., 2009). Loss of the Arabidopsis SUMO E3 ligase SAP AND MIZ1 (SIZ1) results in dwarfism, early flowering, altered responses to abiotic stresses (salt, cold, and drought and the plant hormone abscisic acid), and induction of salicylic acid (SA)-dependent disease resistance responses (Miura et al., 2005(Miura et al., , 2007b(Miura et al., , 2010Yoo et al., 2006;Catala et al., 2007;Lee et al., 2007;Jin et al., 2008;Cheong et al., 2009).
SUMOylation is reversible, and deconjugation is catalyzed by SUMO proteases. Plant SUMO proteases appear to have redundant and nonredundant functions, as exemplified by distinct phenotypes for different knockout mutants. For instance, EARLY IN SHORT DAYS4 (ESD4/Ulp1b) is essential for normal plant development (Reeves et al., 2002;Murtas et al., 2003). By contrast, OVERLY TOLERANT TO SALT1 and 2 (OTS1/Ulp1d and OTS2/Ulp1c) act redundantly in salt tolerance without exhibiting aberrant developmental phenotypes for the ots1 ots2 double mutant (Conti et al., 2008). These different biological functions can in part be explained by different biochemical activities toward SUMO paralogs and a discrete distribution of these proteases in the nucleus. For instance, OTS1, OTS2, and ESD4 each displayed different (iso)peptidase activities toward SUM1 and SUM2 conjugated proteins (Chosed et al., 2006). In addition, ESD4 was found to localize to the nuclear rim, while OTS1 and OTS2 localized to the nucleoplasm and predominantly to nuclear bodies, respectively (Murtas et al., 2003;Xu et al., 2007;Conti et al., 2008).
Phylogenetic analysis indicates that the SUMO gene family has independently expanded in mammals and plants and, like in mammals, plant paralogs appear to have acquired distinct properties (Colby et al., 2006). For instance, SUM1 and SUM2 can form poly-SUMO chains, whereas SUM3 lacks an internal SUMOylation motif required for chain formation (Colby et al., 2006;Budhiraja et al., 2009). Furthermore, several Arabidopsis SUMO proteases exhibit low isopeptidase activity toward SUM3-conjugated proteins but show high activity toward SUM1 and SUM2 conjugates (Chosed et al., 2006;Colby et al., 2006).
We examined the different biological functions of the Arabidopsis SUMO paralogs. The functions of these proteins are poorly understood due to the absence of a sum3 null mutant, embryonic lethality of the sum1 sum2 double mutants, and lack of phenotypes for the single sum1 and sum2 mutants (Saracco et al., 2007). Here, we report a viable sum1 sum2 knockdown mutant and a sum3 knockout mutant in Arabidopsis. Our data suggest that SUM3 acts downstream of SA accumulation, while SUM1 and SUM2 are involved in suppressing SA accumulation. Apparently, the Arabidopsis SUMO paralogs form a regulatory network that differentially modulates innate immune responses and flowering time.

Arabidopsis SUMO Paralogs Exhibit Distinct Gene Expression Patterns
To study the spatial and development expression patterns of the Arabidopsis SUMO paralogs, we isolated the promoters of the SUM genes and fused them to the b-glucuronidase (GUS) reporter gene. We focused our studies on the Arabidopsis SUM1, SUM2, and SUM3 paralogs, as they form a single phylogenetic clade with other plant homologs, while SUM5 falls outside this clade and has not yet been found in other plant species (see Supplemental Figure 1 and Supplemental Data Set 1 online). GUS staining of the transgenic plants revealed that SUM1 is abundantly and widely expressed in leaf and root tissues, except for in the vasculature, the lateral root primordial, and the root apex ( Figure 1). Expression of SUM2 complemented the expression pattern of SUM1 in that expression of SUM2 was high in leaf vasculature, lateral root primordial, and the root apex, while modest in mesophyl cells. The root-specific expression patterns of SUM1 and SUM2 agree with expression data for seedling root cells retrieved from the Bio-Array Resource (BAR) for plant functional genomics (Birnbaum et al., 2003;Brady et al., 2007;Winter et al., 2007).
SUM3 showed a highly tissue-specific expression pattern, as GUS staining was observed only in the hydathodes and leaf vasculature of mature leaves ( Figure 1). Hence, SUM3 expression partially overlaps with SUM2 expression in mature leaves. In roots, expression of SUM3 was restricted to specific cell types of the root vasculature and root primordia. The expression patterns of SUM3 in roots correlates with the data present in BAR. In developing flowers, SUM1 expression was observed at the base of the mature flower, in premature anthers, the stigma, and during embryogenesis in developing seeds but not in siliques. SUM2 expression was found in the vasculature of both filaments and sepals, but not petals. Furthermore, expression of SUM2 was seen at the junction of the silique and pedicel, in developing anthers, and in the style. Finally, SUM3 expression appears not to overlap with SUM1 and SUM2 expression patterns in developing flowers (i.e., SUM3 appears to be expressed exclusively in mature pollen and in developing ovaries).
We also noted that SUM1 was expressed throughout all stages of embryonic development (globule, heart, and torpedo stage) (see Supplemental Figure 2 online), while SUM2 was only expressed until the late heart stage of developing embryos. Expression of SUM3 was not detectable in seeds or in embryos. We noted that deposited microarray expression data (BAR) for Arabidopsis embryogenesis fully supported our observations of expression patterns of these SUM genes (Casson et al., 2005;Winter et al., 2007). The expression profiles of SUM1 and SUM2 are consistent with their essential role in embryogenesis (Saracco et al., 2007). Overall, these promoter-GUS studies revealed that the SUM genes have unique and largely nonoverlapping expression patterns, with SUM1 and SUM2 being widely and abundantly expressed.

The SUMO Paralogs Have Distinct Functions in Plant Development
To study the biological function of these SUMO genes, we first isolated a null allele (sum3-1) of SUM3 from a suppressor-mutator transposon insertion collection (Tissier et al., 1999) (Figure 2A). Sequencing located the transposon insertion in the third exon, 500 bp downstream of the start codon, where it disrupted the coding sequence of the mature protein (at Met-78). Loss of SUM1, SUM2, and SUM3 expression in the respective homozygous sum1-1, sum2-1, and sum3-1 plants was confirmed by RT-PCR ( Figure 2B). Homozygous sum3-1 plants showed normal plant development, as previously reported for the single sum1 and sum2 mutants (Saracco et al., 2007). When we compared the time of bolting (flowering) for individual SUM knockout lines under short-day (SD) and long-day (LD) conditions, we observed that under SDs, sum3-1 was significantly late flowering (Table 1).
As the sum1 sum2 double null mutant is embryo lethal (Saracco et al., 2007), the combined function of these two paralogs could not be studied in adult plants. To overcome this embryo lethality, we created a SUM2 knockdown mutant using artificial miRNA (amiR) gene silencing (Schwab et al., 2006). The amiR-SUM2 construct was designed to target the SUM2 transcript region that encodes the C-terminal tail extension that is removed by SUMO maturation ( Figure 2C). We identified several transgenic lines with reduced SUM2 mRNA levels (97 to 98% reduction) ( Figure 2D). These lines did not show substantially altered expression of SUM1 or SUM3, which means that amiR-SUM2 specifically targets SUM2 mRNA. None of the independent amiR-SUM2 lines showed any aberrant developmental defects, analogous to the sum2-1 null allele. Expression of amiR-SUM2 is controlled by the 35S cauliflower mosaic virus promoter, which is not expressed until the torpedo stage in plant embryos (Custers et al., 1999;Sunilkumar et al., 2002). Hence, SUM2 will not be silenced by the amiR-SUM2 construct during early embryonic development, and, based on its expression profile, this is predicted to allow embryo rescue when the amiR-SUM2 is introgressed in sum1-1. Indeed, we obtained viable homozygous F2 progeny for two independent crosses between amiR-SUM2 lines and sum1-1.
To confirm reduced accumulation of the SUM2 protein in the sum1-1 amiR-SUM2 progeny, a heat shock treatment was applied and the free and conjugated SUM1/2 protein levels were analyzed using immunoblotting ( Figure 2E). Heat shock treatment of the parental lines (sum1-1 and amiR-SUM2) resulted in an increase in SUM1/2 conjugation, as described previously (Saracco et al., 2007). Heat shock treatment of the sum1-1 amiR-SUM2 seedlings, however, resulted only in a minor increase in SUM1/2 conjugation levels, while the free SUM1/2 levels remained below the detection limit. Hence, the double knockdown mutant exhibits strong reduction of SUM1 and SUM2 conjugation. In addition, we noted that the double mutant showed increased SUM3 protein levels ( Figure 2F). The sum1-1 amiR-SUM2 double mutant was partially sterile and displayed a strong developmental phenotype (dominant for the amiR-SUM2 allele, recessive for the sum1-1 allele) including dwarfism, shortened petiole length, crooked and asymmetric leaves, leaf fusion, disturbed inflorescence internode patterning and node development, early senescence, and early flowering under both SD and LD conditions (Figures 2G to 2I, Table 1). Clearly, SUM1 and SUM2 not only act redundantly during embryogenesis, but together regulate many aspects of plant development. SUM3 does not have an apparent function in development but is involved in early flowering under SDs ( Table 1).

Accumulation of Conjugation-Deficient SUM1 and SUM2 Mutants Mimics a siz1 Phenotype
To further investigate the extent to which the SUMO paralogs affect different biological processes, we created transgenic Arabidopsis lines that overexpress either a mature (wild type [WT]) SUMO or a conjugation-deficient mutant (DGG) of each paralog. For SUM1(WT) and SUM2(WT), we identified several highly expressing Arabidopsis lines, as evidenced by the high levels of free (nonconjugated) tagged SUMO in the total protein extracts ( Figure 3A, aAtSUM1/2and aHIS-specific antibodies). These lines showed increased accumulation of a high molecular weight (HMW) smear representing SUMOylated proteins. Tagged SUM1 and SUM2 are present in this HMW smear as they coprecipitated with the HIS-tagged protein fraction under protein denaturing conditions (6 M GuCl) ( Figure 3A, IP:NI2+). The HMW smear was SUMO protease sensitive, as in the absence of the SUMO protease inhibitor N-ethylmaleimide (NEM) (Lois et al., 2003;Murtas et al., 2003) it partially disappeared and the protein levels of free SUM1/2 increased (see Supplemental Figure 3 online). To further show that the tagged SUMO(WT) proteins are conjugated to SUMO targets in planta, we induced mass SUMOylation using heat shock treatment. After heat shock, the tagged mature SUM1, SUM2, and SUM3 proteins were conjugated to other proteins and the SUMOylation levels exceeded those observed in the nontransgenic controls ( Figure 3B). This observation implies that in wild-type plants, the levels of free SUMO protein are limiting during the mass SUMOylation triggered by a heat shock.
We also generated transgenic plants that overexpress the conjugation-deficient mutants (DGG) of SUM1, SUM2, or SUM3. Surprisingly, high expressors of SUM1(DGG) and SUM2(DGG) also showed increased SUM1/2 conjugation levels ( Figure 3A, total protein extract). This indicates that their overexpression either inhibits SUMO deconjugation and/or stimulates SUMO conjugation, as the HMW smear did not coprecipitate with the His-tagged protein fraction in plants expressing the DGG mutant ( Figure 3A, IP:NI2+). Affinity precipitation confirmed that the DGG mutants are not conjugated to SUMO targets. Notably, pulldown of SUM1(DGG) or SUM2(DGG) revealed an additional protein (complex) of ;31 to 35 kD, which we also detected when precipitating SUM1(WT) or SUM2(WT) ( Figure 3A, IP:NI2+, 2xSUM1/2). As SUM1 and SUM2 can form poly-SUMO chains via an internal SUMOylation site (Colby et al., 2006), this larger protein (complex) likely signifies SUMOylation of the DGG mutant by endogenous SUM1 or 2, resulting in SUMO dimers. SUM3 does not contain an internal SUMOylation site, and, in agreement with this, we did not observe a SUMO dimer when we precipitated tagged SUM3(DGG) (neither by exposing the blot to aSUM1/2-specific antibodies or to aHIS antibodies that recognize the tagged SUM3) ( Figures 3A and 3B).
High expressors of SUM1(DGG) or SUM2(DGG) showed a strong developmental phenotype in comparison to plants overexpressing SUM1(WT) or SUM2(WT) (Figures 3C to 3E; see Supplemental Table 1   (A) To scale diagram of the SUM3 gene and the location of the T-DNA insertion (arrowhead) that interrupts the reading frame at Met-78 in sum3-1. Exons and introns are represented by boxes and bent lines, respectively. White, black, and gray boxes reflect the 59/39-untranslated regions, the coding region of the mature SUM3 protein, and the C terminus removed during maturation, respectively. (B) RT-PCR analysis confirmed the null alleles for the sum1-1, sum2-1, and sum3-1 genotypes (wild-type Col-0 was included as positive control, n = 4). Left, target gene amplified by PCR; top, genotype tested. PCR was performed for 30 cycles. UBQ10 mRNA was amplified as a positive control for PCR amplification.
(C) To scale diagram of the SUM2 gene with the amiR-SUM2 target site indicated () in the C-terminal extension, which is removed during maturation. Schematic organization as in (A). Bottom, the sequence of amiR and the target sequence in SUM2. (D) Introduction of amiR-SUM2 in Arabidopsis resulted in SUM2 silencing without silencing of SUM1 or SUM3, as shown using qRT-PCR. Wild-type (Col-0) and sum2-1 were included as controls for SUM2 expression. Depicted are the mean expression levels 6 SD. The biological samples (n = 4) were normalized using TUB4 expression, and the y axis shows the relative expression levels of the SUM genes in the different lines compared with the level found in wild-type plants.
(E) Homozygous sum1-1 amiR-SUM2 seedlings showed reduced levels of both free SUM1/2 and SUM1/2 conjugates in response to heat shock (HS). SUMO conjugation levels were detected in total protein extracts using antibody against SUM1/2 (aAtSUM1/2). PonceauS stain and the aUPGase immunoblot (WB: aUPGase) are shown as a control for equal protein loading. +, HS; À, no HS. (F) The sum1-1 amiR-SUM2 double knockdown showed increased accumulation of SUM3 protein (arrowheads). Other conditions the same as for (E). (G) to (I) SUM1 and SUM2 act redundantly in plant development, as independent sum1 amiR-SUM2 lines showed strong dwarfism (G), leaf crooking, occasional leaf fusion, shortened petioles, early leaf senescence combined with early flowering under SDs (H), reduced apical dominance, partial sterility (I) (arrowheads pointing left), and disrupted inflorescence patterning leading to up to five developing siliques at single node positions (arrowheads pointing right). without exhibiting other visible growth phenotypes ( Figure 3C). In addition, high expressors of SUM1(DGG) and SUM2(DGG) were partially sterile, exhibited disturbed internode patterning of the inflorescence stem, and showed pronounced senescence at the leaf periphery ( Figures 3C to 3E). In fact, the growth phenotype of the high expressors of SUM1(DGG) and SUM2(DGG) closely resemble the phenotype of the siz1 mutant, whereas intermediate expressors resemble a mild sum1-1 amiR-SUM2 knockdown mutant (line G). Possibly, accumulation of SUM1(DGG) and SUM2(DGG) confers a dominant-negative effect on SIZ1 function. This effect was not found for SUM3(DGG) expressing lines.
In agreement with this, we noted that SUM3 protein levels never reached the high levels found for SUM1 or SUM2 when overexpressing the corresponding genes ( Figure 3B, aHIS). We also observed that overexpression of SUM3(WT) or SUM3(DGG) did not alter SUM1/2 conjugation in any of the transgenic plants tested ( Figure 3A, total protein extract; see Supplemental Figure 5 online). Likewise, we obtained a normal increase in SUM1/2 conjugation after heat shock in plants overexpressing SUM3, although the tagged SUM3 became conjugated to other proteins as a result of the heat shock ( Figure 3B, compare lanes 2 and 8, middle panel). The reduced SUM3 protein levels correspond with reduced transcript levels of the SUM3 transgene in comparison to the expression levels of SUM1 and SUM2 transgenes (see Supplemental Figure 6 online). This suggests that the absence of a clear phenotype for the SUM3(DGG)-overexpressing plants may at least in part be due to relatively low SUM3 protein levels.

Accumulation of SUMO1 or SUMO2 Paralogs Promotes SA-Dependent Defense Responses
The developmental defects of the siz1 mutant are largely caused by constitutive activation of SA signaling upstream of PHYTO-ALEXIN DEFICIENT4 (PAD4) that results in increased disease resistance Jin et al., 2008;Cheong et al., 2009;Miura et al., 2010). Other phenotypes that are generally associated with constitutive SA signaling are spontaneous cell death, constitutive expression of pathogenesis-related (PR) genes, increased resistance to infection by Pseudomonas syringae pv tomato DC3000 (PstDC3000), and reduced elicitor-dependent hypersensitive response (HR) (Rate and Greenberg, 2001;Devadas and Raina, 2002;Heil and Baldwin, 2002;Lorrain et al., 2003). We therefore examined whether overexpression of the SUMO variants resulted in constitutive activation of SA signaling. Overexpression of SUM1 or SUM2, but also of SUM3 (both WT and DGG) resulted in enhanced resistance to PstDC3000 (maximum 100-fold reduced bacterial growth) and correspondingly reduced disease symptoms, such as chlorosis and water-soaked lesions ( Figure 4).
We also examined whether overexpression of these SUMO variants reduced avrRpm1-mediated HR (the used ecotype Columbia-0 [Col-0] carries the cognate resistance gene RPM1 required for recognition of avrRpm1) ( Figure 5). High expressors of mature SUM1, SUM2, and SUM3 showed a strong reduction in HR as exemplified by the reduced ion leakage from the leaf discs upon infiltration of Pst expressing avrRpm1 ( Figure 5A). Similarly, high expressors of SUM1(DGG) and SUM2(DGG) also showed a strong reduction in ion leakage ( Figure 5B). Interestingly, expression of SUM3(DGG) resulted in ;20% lower levels of ion leakage than control plants (Col-0). The reduction in ion leakage correlated in each case with a reduction in the number of leaves showing extensive trypan blue staining at 7 h after infiltration of Pst expressing avrRpm1 (see Supplemental Figure 7 online), which confirms the reduced levels of cell death induction.
To further link these SUMO overexpression phenotypes to activation of SA-dependent defense responses, we determined the SA levels in these plants ( Figure 5C). Several SUMO overexpression lines showed increased SA levels, similar to the siz1-2 mutant. We noted a maximal 10-fold increase in SA-2-O-b-Dglucoside (SAG) levels in these lines. Concomitantly, expression of the SA marker gene PR1 was >100-fold increased in the 35S-SUM1 and 35S-SUM2 lines (both WT and DGG), while 35S-SUM3 lines showed up to 100 times increased expression of PR1 (both WT and DGG) ( Figure 5D). We could also show accumulation of the PR1 protein in these plants ( Figure 5E). Enhanced exposure of the blot revealed accumulation of the PR1 protein also in the 35S-SUM3(DGG) lines ( Figure 5F). Under SD conditions, accumulation of the PR1 protein was first observed in 4-week-old 35S-SUM2 (WT or DGG) plants ( Figure 5G), whereas growth defects were already visible in 2-to 3-week-old plants. This suggested that the SUMO-dependent upregulation of PR1 is in part age dependent, possibly involving progressive accumulation of free SUMO or SUMOylated proteins. The 35S-SUM1 (DGG)-and 35S-SUM2(DGG)-expressing plants also developed macroscopically visible lesions 6 to 8 weeks after germination under SD conditions (see Supplemental Figure 8A online). Trypan blue staining revealed that not only the DGG-expressing plants showed spontaneous cell death, but also the plants that overexpress mature SUMO (SUM1, SUM2, or SUM3) (see Supplemental Figure 8B online). Notably, overexpression of SUM3 (DGG) did not result in spontaneous cell death. In conclusion, hyperaccumulation of each of the three SUMO paralogs resulted in activation of SA-dependent defense responses.  (A) Overexpression of both mature (WT) and a conjugation-deficient (DGG) mutant of HN-tagged SUM1 and SUM2 resulted in increased SUM1/2 conjugation levels in 5-week-old plants. By contrast, overexpression of SUM3 (both WT and DGG) did not lead to increased conjugation of SUM1 or SUM2. SUM1(DGG) and SUM2(DGG) are not conjugated to target protein, but they are themselves SUMOylated by endogenous SUMO, resulting in a covalent SUMO dimer (2xSUM1/2). The wild type (Col-0) was included as control.  (Wildermuth et al., 2001). ICS1 is essential for SA synthesis following pathogen recognition, and the sid2-1 allele allows normal plant growth (Wildermuth et al., 2001;Strawn et al., 2007;Garcion et al., 2008;Yaeno and Iba, 2008). The homozygous F2 progeny of the sid2-1 35S-SUM2(WT) and the sid2-1 35S-SUM2(DGG) crosses showed a partial reversion of the SUMO-dependent growth defects (i.e., leaf curling was lost but the plants remained dwarfed) ( Figure 6A). In addition, the leaves did not fully elongate, resulting in shorter and more rounded leaves. We next examined the expression of several SA marker genes (PR1, PR2, PR5, and PAD4) in these crosses and in their parental lines ( Figure 6B). Loss of a functional allele of ICS1 reverted expression of these SA marker genes back to normal levels despite the overexpression of SUM2 (WT or DGG). These genetic data confirm that overexpression of SUM2 (WT or DGG) results in ICS1-dependent SA synthesis and subsequent expression of defense genes. Furthermore, not all 35S-SUM2-related developmental phenotypes can solely be attributed to SA accumulation.

SUM1 and SUM2 Jointly Suppress Activation of SA-Dependent Defense Responses
Considering that both the sum1 amiR-SUM2 knockdown mutant and high expressors of SUM1(DGG) and SUM2(DGG) mimic the developmental phenotype of the siz1 mutant, we examined whether SUM1 and SUM2 are involved in the regulation of SAdependent innate immunity in noninfected plants. To this end, the different sum mutants were examined in Pseudomonas disease assays. None of the single sum knockout lines showed altered resistance to PstDC3000 or exhibited reduced avrRpm1mediated HR (Figures 7A and 7B). The sum1-1 amiR-SUM2 knockdown, however, did show significantly increased resistance to PstDC3000 ( Figure 7C). In corroboration, we found that the sum1-1 amiR-SUM2 lines showed fourfold to eightfold increased SAG levels in comparison to the parental lines, while the SA and SAG levels were not increased for the single sum knockout lines analyzed in parallel ( Figure 7D). These sum1-1 amiR-SUM2 double mutants showed 10-fold increased PR1 expression in comparison to wild-type Arabidopsis (Col-0), whereas the corresponding single mutants did not show increased PR1 expression ( Figure 7E). In agreement with this, trypan blue staining revealed extensive microscopic cell death in 6-to 8-week-old rosette leaves (SD) of noninfected sum1-1 amiR-SUM2 plants ( Figure 7F). Cell death was not restricted to individual cells but occurred in larger clusters of mesophyl cells (lesions) and in the vasculature. This microscopic cell death was not observed in control plants (Col-0, sum1-1, and sum2-1).
Hence, SUM1 and SUM2 are together essential to prevent increased SA accumulation and induction of SA-dependent gene expression in noninfected plants.
Expression of SUM3 but Not of SUM1 or SUM2 Is Highly Induced by SA Treatment Considering that the sum1 amiR-SUM2 plants showed both increased SA levels and increased SUM3 protein levels ( Figures  7D and 2F, respectively), we examined whether SA accumulation causes induction of SUM3. Analysis of deposited transcriptome  Total protein extracts were prepared from 17-d-old seedlings grown under LD conditions. PonseauS (PonS) staining of the blot confirmed equal protein loading. At least two independent transgenic lines were examined per construct.
(F) Similar as in (E), except that enhanced exposure of the blot revealed that transgenic lines overexpressing 35S-SUM3(DGG) also showed increased PR1 protein accumulation. PR1 protein levels in these lines were lower than those observed for 35S-SUM1 and 35S-SUM2 (both WT and DGG), as shown in Figure 3E.
(G) Similar as in (E), except that total protein extracts were prepared from plants grown under SD conditions. Top: genotype tested (top). Accumulation of the PR1 protein is age dependent in plants that overexpress SUM2 variants.
[See online article for color version of this figure.] data indicated that expression of SUM3 is indeed induced by the SA analog benzothiadiazole (Thibaud-Nissen et al., 2006;Wang et al., 2006). Moreover, SUM3 expression is also increased in the siz1 mutant (Catala et al., 2007). These same microarray data sets did not indicate altered expression of the SUM1 or SUM2 genes in either the siz1 mutant or after SA application. We therefore examined the SUM3 expression profile after SA treatment ( Figure 8A). Using quantitative RT-PCR (qRT-PCR), we found that SA treatment strongly induces SUM3 expression, but only transiently, peaking 3 h after SA application. At this time point, SUM3 expression was 80-fold increased compared with mock-treated plants. The SA-dependent induction of SUM3 appeared to be paralog specific, as expression of SUM2 was hardly induced by either mock or SA treatment, and expression SUM1 was only modestly increased by SA treatment. An analysis using the SUM3 promoter-GUS lines confirmed the SA sensitivity of the SUM3 promoter ( Figures 8B and 8C). Moreover, expression of SUM3 was no longer restricted to the leaf vasculature and hydathodes, as GUS staining could be detected in mesophyl cells after SA treatment. In parallel, we confirmed that GUS staining was not increased in the SUM1 and SUM2 promoter-GUS reporter lines after SA treatment. Subsequently, we tested whether SUM3 can be induced by the Flg22 peptide, a 22residue peptide that acts as general elicitor of plant defense responses and is derived from bacterial flagellin (Zipfel et al., 2004). Infiltration with 10 mM Flg22 peptide resulted in a transient induction of SUM3 within 6 h, while infiltration with the negative control (Flg22 Atum peptide) triggered a modest induction of SUM3 at 9 h ( Figure 8D). Upregulation of SUM3 required ICS1, confirming strict regulation of the expression of SUM3 downstream of SA after perception of the Flg22 peptide by the FLS2 receptor. By contrast, SUM1 and SUM2 expression was not substantially altered by infiltration with the Flg22 peptide. As an additional control, we confirmed that PR1 was only highly induced in wild-type plants (Col-0) and not in sid2-1 plants after infiltration with the Flg22 peptide ( Figure 8D). These data are supported by microarray data showing that SUM3 is induced after infection by Pseudomonas (6 to 16 h after infection by Pst DC3000 and P. s. pv maculicola strain ES4326) (Journot-Catalino et al., 2006;Thilmony et al., 2006;Underwood et al., 2007;Winter et al., 2007). We also examined SUM3 protein levels after SA treatment (see Supplemental Figure 9 online). We found that the SUM3 conjugation profiles did not change in 35S-SUM3(WT) plants after SA treatment or treatment with the proteasome inhibitor MG132. However, we noted that free HN-tagged SUM3 became subject to an apparent posttranslational modification (hnSUM3*) as a result of the SA treatment. This implies a functional link between SUM3 expression, SUM3 protein accumulation, and SA signaling. We conclude that induction of SUM3 appears to be part of the endogenous defense signaling cascade in Arabidopsis and, since induction of SUM3 preceded expression of PR1, SUM3 conjugation might regulate (part of) the SAdependent defense responses.

SUM1 and SUM2 Regulate Plant Development and Jointly Suppress Activation of SA-Dependent Signaling
Here, we establish distinct biological functions for the Arabidopsis SUMO paralogs in development and innate immunity. The sum1 amiR-SUM2 knockdown mutant showed that SUM1 and (A) SUM null alleles (sum1-1, sum2-1, and sum3-1) did not show significantly different resistance to PstDC3000 than the control (wild-type Col-0), as determined by analysis of variance of the mean log bacterial count at t = 0 d (group a) or t = 3 d (group A). Depicted are mean log bacterial counts 6 SE (n = 8 samples). Data are from one representative experiment (n = 3).
(B) Loss of a single SUMO paralog did not impair ion leakage (HR induction) triggered by Pst expressing avrRpm1. Means 6 SD were determined using four leaf discs per genotype. Symbols represent the genotypes indicated. As control, we infiltrated leaf discs with buffer only (10 mM MgCl 2 ). Data are from one representative experiment (n = 3).
(C) The sum1-1 amiR-SUM2 double mutant shows enhanced resistance to PstDC3000. Depicted is mean log bacterial growth 6 SE (n = 8) for two independent crosses of sum1-1 amiR-SUM2 (lines B and G) and the parental lines. The asterisk above the bars indicates significant differences in mean log bacterial count at *P = 0.05 and **P = 0.01 (pair-wise Student's t test with the parents). The assay was repeated three times with similar results.
(D) The sum1-1 amiR-SUM2 double mutant showed increased accumulation of free SA and SAG in comparison to the parental lines and wild-type plants (Col-0). Bars represent means 6 SD (n = 4). Five-week-old SD-grown plants were sampled. The assay was repeated three times with similar results. FW, fresh weight.
(E) The sum1-1 amiR-SUM2 double mutant showed at least 10-fold increased PR1 expression levels compared with control plants. RNA was extracted from 17-d-old seedlings grown under LD conditions. Individual biological samples were normalized against TUB4 expression, and the mean expression in the wild-type plants (Col-0) was set to 1. SUM2 act as basal regulators of normal plant development, prevent activation of defense responses in noninfected plants ( Figure 9A), and prevent premature flowering. These functions of SUM1 and SUM2 are consistent with the constitutive expression of the corresponding genes. Our expression studies suggested a more generic role for SUM1 and a more specialized role for SUM2 in the leaf vasculature. Nevertheless, the individual sum1 or sum2 null mutants do not have obvious growth abnormalities (Saracco et al., 2007). Expression of SUM3 is generally very low and restricted to a few cell types, but its expression is highly, transiently, and widely induced by SA. Thus, expression of these three SUM genes is differentially regulated in both time and space.

SUM1 and SUM2 Exert Their Functions in Development and Defense via the SUMO E3 Ligase SIZ1
The functions of SUM1 and SUM2 in development and SA signaling mimic that of the gene encoding the SUMO E3 ligase SIZ1. SIZ1 suppresses SA-dependent defense responses upstream of PAD4 in noninfected plants . The siz1 mutant and the sum1 amiR-SUM2 knockdown mutants show similar growth defects. Furthermore, the siz1 mutant has largely lost mass conjugation of SUM1 and SUM2 proteins in response to heat shock, linking SIZ1 protein function to SUM1 and SUM2 conjugation (Cheong et al., 2009). We found that hyperaccumulation of the conjugation-deficient mutants SUM1(DGG) and SUM2(DGG), but not SUM3(DGG), results in a dose-dependent growth phenotype, again closely resembling the siz1 mutant, and, similar to siz1 (Miura et al., 2010), these growth defects were largely SA dependent ( Figure 6). Together, these data suggest that both SUM1(DGG) and SUM2(DGG) can inhibit SIZ1 function in vivo. Such an inhibition of SIZ1 function by binding of nonfunctional SUMO variants is conceivable, as the in planta SIZ1 protein levels are low and SUM1(DGG) can interact with SIZ1 in the yeast split-ubiquitin system (Cheong et al., 2009). This SUM1-SIZ1 protein interaction does not depend on the major protein domains of SIZ1 (SAP, PHD, SP-RING, or PINIT domains), but likely involves binding of SUMO to the SIM of SIZ1 Cheong et al., 2009). Complementation studies of the siz1 mutant revealed that point muta-tions in the SIM motif could not fully complement the SAdependent phenotypes, and a functional SUMO E3 ligase domain (SP-RING) was essential to complement these phenotypes (Cheong et al., 2009). Based on these observations, we hypothesize that SIZ1 promotes conjugation of SUM1 or SUM2 to unknown negative regulator(s) of SA signaling that need(s) to be SUMOylated in noninfected plants ( Figure 9A). Apparently, SUMOylation of this regulator depends on the SIM in SIZ1. SIZ1 is the plant prototype of yeast SIZ1 and the mammalian PIAS proteins, which regulate gene expression via chromatin remodeling (Miura et al., 2007a;Garcia-Dominguez et al., 2008;Rytinki et al., 2009). Also in yeast and mammals, the function of the SIZ1 homologs depends on their capability to noncovalently interact with SUMO via their SIM (Reindle et al., 2006;Rytinki et al., 2009). Based on this analogy, we have proposed before that SUMOylation of chromatin modifying enzymes and/or transcription factors inhibits transcription of a regulator(s) of the SA signaling pathway (van den Burg and Takken, 2009). Further studies are needed to elucidate the exact role of these SUMO paralogs and SIZ1 in the regulation of SA-sensitive gene expression.

Is Tight Regulation of SUM1 and SUM2 Expression Important for Proper Plant Development and Repression of Innate Immunity?
Proper plant development and repression of innate immunity in noninfected plants likely require tight regulation of SUM1 and SUM2 genes in planta, as hyperaccumulation of both WT and DGG resulted in activation of SA-dependent innate immunity rather than suppression. In addition, we and others have shown that overexpression of SUM1 and SUM2 variants results in enhanced SUM1/2 conjugation in plants (this study; Lois et al., 2003;Murtas et al., 2003). The other studies did not report SUM1-or SUM2-related growth defects related to overexpression (Lois et al., 2003;Murtas et al., 2003). However, when we reanalyzed plants that overexpress His-tagged SUM1 (Lois et al., 2003), we found increased expression of the SA marker gene PR1 (see Supplemental Figures 10A to 10C online), showing that overexpression of SUM1 induces host defenses. We cannot exclude that some of the mild growth defects observed in our study can be attributed to the tag used, as N-terminal labeling of SUMO with relatively large peptide tags can interfere with SUMO protein function (Wohlschlegel et al., 2004;Budhiraja et al., 2009). However, our HN-tagged SUMO proteins were successfully used as substrates for SUMOylation in planta (Figure 3), and increased expression resulted in increased SUMOylation levels, demonstrating their functionality. Therefore, we propose that the SIZ-related phenotypes induced by overexpression of either the mature form or the DGG mutant are likely SIM dependent without inhibiting SIM-independent SUMO conjugation by SCE1 (as evidenced by the increased SUMOylation levels). This implies that SIZ1 function in SA signaling can be inhibited in planta by an excess of free SUMO. This requires that the levels of free SUM1 and SUM2 in wild-type plants be relatively low, which is indeed the case (Saracco et al., 2007). Alternatively, the increase in SUM1/2 conjugation levels due to overexpression of SUM1 and 2 variants could be caused by product inhibition of SUM1/2-specific SUMO proteases (like ESD4, OTS1, and OTS2). Product inhibition has been reported as a result of overexpression of WT and DGG mutants of mammalian SUMO paralogs (Shen et al., 2006;Ihara et al., 2007;Mukhopadhyay and Dasso, 2007). Whereas SUM3(DGG) overexpression did not increase SUM3 conjugation in plants, SUM3(WT) overexpression did induce a modest increase. In both cases, the overall SUM1/2 conjugation levels remained unaffected, which can be explained in two ways. First, SUM3 did not accumulate to the same high protein levels as the SUM1 and SUM2 proteins; hence, SUM3 might not reach SUMO protease inhibitory concentrations. Second, none of the known SUM1-and SUM2-specific SUMO proteases has isopeptidase activity toward SUM3 conjugates (Chosed et al., 2006;Colby et al., 2006). This means that SUM3 is unlikely to inhibit these SUM1-and SUM2-specific proteases. In summary, the increased SUMOylation levels upon SUM1 and SUM2 overexpression could be due to product inhibition of SUMO proteases that are not inhibited by the relatively low levels of SUM3. Additional studies should reveal which SUMO proteases are important for SA-dependent signaling.

SA-Dependent SUM3 Expression Promotes Plant Defense Responses and Flowering
Interestingly, activation of plant defense responses lifts the SUM1-and SUM2-dependent inhibition of SA signaling via an unknown mechanism and allows accumulation of SA, which subsequently triggers SUM3 expression and SUM3 protein accumulation ( Figure 9B). Free SUM3 protein appears to be subject to an unknown posttranslational modification as a result of SA signaling. It has already been shown that mammalian SUMO1 is subject to phosphorylation in the N-terminal tail, but the conditions that induce this modification are unknown (Matic et al., 2008). Overall, we find that SUM3 appears to act downstream of SA synthesis, but upstream of plant defense responses, since overexpression of SUM3 resulted in PR1 expression and increased resistance.
The autonomous pathway promotes flowering during the juvenile-to-adult transition by downregulating FLOWERING LO-CUS C (FLC) expression (Baurle and Dean, 2006). Repression of the autonomous flowering pathway appears to involve SIZ1dependent SUMOylation because SIZ1 is required for full FLC expression (Jin et al., 2008). In agreement, we find that SUM1 and SUM2, like SIZ1, suppress flowering under both SD and LD conditions. In part, SIZ1 inhibits flowering by promoting SU-MOylation of FLOWERING LOCUS D, a transcriptional repressor of FLC. More importantly, SIZ1-dependent inhibition of flowering appears to involve suppression of SA accumulation in adult plants (Jin et al., 2008). In wild-type adult plants, SA accumulation stimulates the autonomous pathway by repressing FLC expression via an unknown mechanism (Martinez et al., 2004). This age-dependent accumulation of SA might also allow expression of SUM3. In agreement, SUM3 was not expressed in juvenile leaves, but SUM3 became increasingly expressed in the leaf vasculature of adult plants (6-to 8-week-old plants). Moreover, the sum3 mutant displayed delayed flowering, while SUM3 overexpression promoted flowering under SDs. This signifies that the function of SUM3 differs from that of SUM1 and SUM2 in flowering and SA-dependent responses. However, the sum3 mutant did not exhibit increased susceptibility to PstDC3000, which indicates that the function of SUM3 in plant defense could be redundant.
SUM3 is normally not extensively conjugated to other proteins, and its conjugation does not increase after a heat shock (Kurepa et al., 2003). However, transient overexpression of SUM3 does increase overall SUM3 conjugation (Budhiraja et al., 2009). This implies that SA-dependent accumulation of SUM3 will result in a transient increase in SUM3 conjugation. It remains to be determined whether this SA-dependent induction of SUM3 affects a common set of SUMO targets or concerns SUM3-specific targets. In support of the latter, a proteomics screen established that the set of SUM3 conjugates only partially overlapped with SUM1 and SUM5 conjugates identified (Budhiraja et al., 2009).
A prediction from the proposed function of SUM3 in pathogen defense is that inhibition of SUM3 conjugation should result in enhanced susceptibility to plant pathogens. Indeed, the effector protein XopD (Xanthomonas outer protein D) from the tomato (Solanum lycopersicum) pathogen Xanthomonas campestris pv vesicatoria is known to act as a SUMO isopeptidase in planta, and this activity enhances virulence of this pathogen on tomato (Hotson et al., 2003;Kim et al., 2008). Interestingly, XopD appears to have SUMO isopeptidase activity not only toward SUM1 and SUM2 conjugates, but also toward SUM3 conjugates (Colby et al., 2006). It is tempting to speculate that XopD isopeptidase activity might be targeted in vivo toward SUM3 conjugates to dampen host innate immune responses ( Figure  9C). Identification of these SUM3 targets and investigation of their role in plant defense can put this model to the test. In conclusion, SUM1 and SUM2 together inhibit activation of SAdependent responses in noninfected plants, whereas SUM3 appears to potentiate these responses following recognition of pathogen infection.

GUS Assay
GUS assays (Jefferson et al., 1987) were performed on nonfixed Arabidopsis tissue at different developmental stages. GUS localization was observed after staining with 0.5 mg/mL 5-bromo-4-chloro-3-indolyl-b-Dglucoronide. The observed expression patterns were based on at least four independent transgenic lines showing similar GUS staining patterns.

Measurement of SA Levels (Free and 2-O-b-D-Glucoside)
SA extraction was performed according to Verberne et al. (2002). Leaf material (200 to 300 mg) was homogenized (FastPrep,23 45s) in 90% methanol spiked with 200 ng SA-d4 (internal control for extraction efficiency), followed by 100% methanol extraction. The organic phases were combined, vacuum-dried (Speed-Vac), and resuspended in warm water. The samples were buffered to a pH of 4.5 in 0.1 M NaAcetate and split in half. To one sample, almond b-glucosidase (Sigma-Aldrich) was added to hydrolyze SAG to SA. Both samples were incubated at 378C overnight, followed by two extractions with an organic phase (50:50:1 mixture of cyclohexane/ethyl acetate/isopropanol). The organic phases were combined, dried (Speed-Vac without heating), and redissolved in 70% methanol for liquid chromatography-mass spectrometry analysis. Liquid chromatography-mass spectrometry settings were as described (Diezel et al., 2009). The ratios of ion intensities were determined for m/z 137 (daughter ion m/z 93) and m/z 141 (daughter ion m/z 97) for SA and SA-d4, respectively.

Constructs
For the promoter-GUS reporter assays, an ;2-kb DNA fragment upstream (59) of the start codon was cloned and fused to the GUS reporter in pEpiGreenB (pGreenII 0229 derivative; http://www.pgreen.ac.uk/). The resulting promoter-GUS fusions were used for stable transformation of Arabidopsis.
The protein tag consisted of the HIS tag and the streptavidin binding peptide NANO15-tag (HN-tag) fused to the free N terminus (Lamla and Erdmann, 2004). Expression of the chimaeras was driven by the 35S promoter. The tandem-affinity tag was created by overlap PCR amplification and directly cloned in the ClaI-BamHI sites of pEpiGreenB. The DNA fragments corresponding to mature SUMO protein (i.e., SUM1 [2 to 93 amino acids], SUM2 [2 to 92 amino acids], and SUM3[2 to 93 amino acids]) or SUMO (DGG) were PCR amplified using cDNA from Arabidopsis Col-0 as template and cloned in the BamHI and XbaI sites of pEpiGreenB. This created a 59/39 Pro35S-His/Nano15-SUMOx-NOS terminator fusion construct.
The amiR-SUM2 was engineered and PCR amplified from the miR319a backbone (pRS300) according to the instructions of Web MicroRNA Designer (Schwab et al., 2006). The resulting PCR fragment was subcloned in the ClaI-XbaI of pEpiGreenK (pGreenII0029 derivative), creating a Pro35S-amiR-SUM2-NOS terminator fusion construct. All final fragments were sequenced. Transgenic plants were generated by floral dipping using Agrobacterium tumefaciens strain GV3101 (Clough and Bent, 1998).

Pseudomonas Disease Assays
Fresh cultures of PstDC3000 strains were streaked from glycerol stocks onto selective King's B plates containing kanamycin (25 mg/mL) and rifampicin (25 mg/mL) and grown at 288C. For plant inoculations, bacteria were grown overnight on King's B plates. Bacterial cells were scraped from the plates and suspended in 10 mM MgCl 2 -100 mM sucrose to a final OD 600 of 0.1. Dilutions were made from this solution for the different inoculations. For disease inoculations, 4-week-old Arabidopsis plants (SD conditions) were syringe infiltrated with PstDC3000 at 10 5 colonyforming units (cfu) mL 21 (OD 600 =0.0002) in 10 mM MgCl 2 and 0.02% v/v silwet L77. Leaf disc samples from eight plants (n = 8) were taken directly after infiltration (t=0 d) and three days post infiltration (t=3 d). Leaf discs were macerated (Mixer Mill, Retsch) in 10 mM MgCl 2 , serially diluted and plated for colony counting. Bacterial growth in the transgenic and sum knock-out lines was compared with that of the wild-type (Col-0). To assess the development of disease symptoms caused by Pst infection, plants were infiltrated with 5 3 10 6 cfu mL 21 (OD 600 =0.005) of PstDC3000 and placed under a LD regime. In all cases, plants were placed under a translucent cover for 24 h prior to and post infection to increase humidity.

HR Assays (Ion Leakage and Trypan Blue Staining)
HR was examined using the ion leakage assay and trypan blue staining. HR was triggered by infiltrating PstDC3000 carrying avrRpm1 (hereafter, Pst+avrRpm1), which induces an RPM1-dependent HR in Arabidopsis (Col-0) within 7 h of infiltration (Dangl et al., 1992). For ion leakage assays, 4-week-old Arabidopsis plants (SD conditions) were infiltrated with Pst +avrRpm1 at 5 3 10 7 cfu mL 21 (OD 600 = 0.1) in 10 mM MgCl 2 . Three leaf disc samples were taken from the infiltrated zones, extensively washed for 1 h in 50 mL water, and then placed in 3 mL water at ambient temperature with gyratory agitation. The conductivity was measured for 11 h per four leaf discs per genotype (B-173 conductivity meter; Horiba). Mean conductivity and SE were calculated for three repetitions per experiment. At least two additional independent transgenic lines were tested per construct. For trypan blue staining, 4-week-old Arabidopsis plants (SD conditions) were syringe infiltrated with Pst+avrRpm1 (Grant et al., 1995) at 2.5 3 10 7 cfu mL 21 (OD 600 = 0.05) in 10 mM MgCl 2 . Seven hours after infiltration, the rosette leaves were stained with a 1:1 mixture (v/v) of ethanol and lactic acid-phenol-trypan blue solution (2.5 mg mL 21 trypan blue, 25% [v/v] lactic acid, 25% phenol, 25% glycerol, and water) and boiled for 5 min. For destaining, the trypan blue solution was replaced with a chloral hydrate solution (2.5 g/mL in water).

RNA Analysis
Total RNA was extracted using Trizol reagent (Invitrogen) according to the manufacturer's recommendations. For cDNA synthesis, 2 mg RNA was used in combination with Superscript III (Invitrogen). Real-time PCRs were performed in an ABI 7500 real-time PCR system (Applied Biosystems) using a Platinum SYBR Green qPCR SuperMix-UDG Kit (Invitrogen). The cycling program was set to 2 min, 508C; 10 min, 958C; 40 cycles of 15 s at 958C; and 1 min, 608C, and a melting curve analysis was performed at the end of the PCR. Primer pairs were tested for specificity and for amplification efficiency with a standard cDNA dilution curve. PCR amplification efficiencies were determined for each primer pair per plate using LinRegPCR, and the mean PCR amplification efficiency for each primer pair per plate was subsequently used to determine the gene expression levels for the different samples (Ruijter et al., 2009). The different biological samples were subsequently normalized against expression of b-Tubulin4 (TUB4). Quantitative expression analysis was conducted for at least three independent experiments including at least three independent biological replicates per sample in each experiment.

SA and flg22 Treatment
Four-week-old SD plants were placed in high humidity chambers for 24 h prior to treatment. Plants were spray inoculated with 1 mM SA in 0.015% (v/v) Silwet-L77 or mock treated (0.015% [v/v] Silwet-L77) and returned to high humidity. Active Flg22 peptide and inactive Flg22 Atum peptide were synthesized by Genscript and dissolved at 10 mM prior to infiltration. Samples for RNA extraction were taken at the indicated time points, and cDNA was prepared as outlined above.

Protein Work
For aPR1 immunoblot analysis, seedlings were homogenized in liquid nitrogen, thawed on ice in extraction buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM EDTA, 2% [w/v] polyvinylpyrrolidone K25, 13 protease inhibitors, 0.5% Nonidet P-40, and 2 mM DTT), and centrifuged for 10 min at 13,000g. The supernatant was mixed 1:1 with 23 SB (125 mM Tris-HCl, pH 6.8, 4% SDS, 20% glycerol, and 100 mM DTT), and the samples were boiled for 10 min. Proteins were separated on 15% SDS-PAGE and blotted onto polyvinylidene fluoride membrane. Secondary immunoglobulins conjugated to horseradish peroxidase were visualized using ECL Plus (GE Healthcare). Primary antibodies against Arabidopsis SUM1/2, SUM3 (Abcam) (Kurepa et al., 2003), His-tag (a-pentaHIS; Qiagen), and PR (aPR1; March-Diaz et al., 2008) are described elsewhere. For heat shock-induced hyper-SUMOylation, Arabidopsis seedlings were grown in liquid culture for 14 d under continuous light at 228C, exposed to a 30min heat shock at 378C, and then returned to 228C for 1 h (Kurepa et al., 2003). Total protein was extracted as described above (the extraction buffer included in addition 20 mM NEM and 8 M Urea) and separated on SDS-PAGE gradient gels (6 to 20%) for immunoblotting. For Ni 2+ resin affinity precipitations, plant material was dissolved in extraction buffer that included 20 mM NEM and 6 M GuCl, and, after clarification, the protein samples were incubated with Ni 2+ resin for 2 h at 48C, followed by three washes of the resin with wash buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 2 mM EDTA, 0.5% Nonidet P-40, and 8 M urea). The bound fraction was eluted by boiling in 13 SB for 10 min and loaded on SDS-PAGE. The total protein fraction containing GuCl was precipitated with 10% (w/v) trichloroacetic acid and washed twice in ice-cold 80% acetone prior to loading on SDS-PAGE. The protein precipitate was redissolved in 13 SB prior to loading.

Phylogeny
Phylogenetic analyses were conducted in MEGA4 using Minimum Evolution. The percentage of replicate trees in which the associated taxa clustered together in the bootstrap test (1000 replicates) is shown next to the branches. Taxa that were supported by <50% of the replicates are collapsed. There were a total of 80 positions in the final data set. The ME tree was searched using the Close-Neighbor-Interchange algorithm at a search level of 3. The neighbor-joining algorithm was used to generate the initial tree. All positions containing gaps and missing data were eliminated from the data set (complete deletion option).

Supplemental Data
The following materials are available in the online version of this article.