SUMO-Dependent Synergism Involving Heat Shock Transcription Factors with Functions Linked to Seed Longevity and Desiccation Tolerance

A transcriptional synergism between HaHSFA9 (A9) and HaHSFA4a (A4a) contributes to determining longevity and desiccation tolerance of sunflower (Helianthus annuus, L.) seeds. Potential lysine SUMOylation sites were identified in A9 and A4a and mutated to arginine. We show that A9 is SUMOylated in planta at K38. Although we did not directly detect SUMOylated A4a in planta, we provide indirect evidence from transient expression experiments indicating that A4a is SUMOylated at K172. Different combinations of wild type and SUMOylation site mutants of A9 and A4a were analyzed by transient expression in sunflower embryos and leaves. Although most of the precedents in literature link SUMOylation with repression, the A9 and A4a synergism was fully abolished when the mutant forms for both factors were combined. However, the combination of mutant forms of A9 and A4a did not affect the nuclear retention of A4a by A9; therefore, the analyzed mutations would affect the synergism after the mutual interaction and nuclear co-localization of A9 and A4a. Our results suggest a role for HSF SUMOylation during late, zygotic, embryogenesis. The SUMOylation of A9 (or A4a) would allow a crucial, synergic, transcriptional effect that occurs in maturing sunflower seeds.


INTRODUCTION
The seed-specific heat-shock transcription factors (HSF) A9 and A4a (respectively, Almoguera et al., 2002;Tejedor-Cano et al., 2014) function in enhancing longevity and desiccation tolerance of seeds (Prieto-Dapena et al., 2006Personat et al., 2014). Among the specific effects of A9 and A4a, this pair of HSF synergically activates transcription from small Heat Shock Protein (sHSP) gene promoters . A9 and A4a directly interact with each other through their oligomerization domains, an interaction that facilitates the nuclear retention of A4a by A9, which is required for the synergism . Transcriptional activation by A9 and A4a is repressed by the Aux/IAA protein HaIAA27 Tejedor-Cano et al., 2014). Stabilized forms of HaIAA27  and a dominant-negative form of A9 that incorporated the SRDX trans-repression motif (Tejedor-Cano et al., 2010) have been used to corroborate by loss-of-function the involvement of A9 in seed-longevity. This work indirectly indicated the contribution of additional -class A-HSFs, among them A4a as confirmed by subsequent work (Tejedor-Cano et al., 2010;Personat et al., 2014;Tejedor-Cano et al., 2014).
Post-translational modification has been found important in modulating transcription factor function, which in turn has profound effects on gene expression and many developmental programs in animals and plants. Lysine modifications that include acetylation, ubiquitination, methylation and, SUMOylation -in particular-have been found to be very relevant (Freiman and Tjian, 2003;Verger et al., 2003;Hill, 2015). SUMOylation, for example, modulates the activity of transcription factors involved in abiotic stress responses in plants (Lois et al., 2003;Miura et al., 2007Miura et al., , 2009Cohen-Peer et al., 2010;reviewed by Castro et al., 2012). SUMOylation was also found to be essential for zygotic embryogenesis in seeds (Saracco et al., 2007). SUMOylation involves the covalent (and reversible) attachment of small ubiquitin-like modifier (SUMO) proteins to lysine. In the model plant Arabidopsis four different SUMO are expressed (SUMO1-3, and SUMO5). Mutational analysis in Arabidopsis has revealed that at least SUMO1/2 and SUMO3/5 do not have overlapping functions (Saracco et al., 2007; reviewed by Lois, 2011). All plants have at least one gene that encodes one form of SUMO, and the forms similar to Arabidopsis SUMO1 and SUMO2 are considered to represent the ancestral SUMO protein that is characteristic of eukaryotes. SUMOylation involves the consecutive enzyme-catalyzed steps referred to as SUMO E1 activation, E2 conjugation, and E3 ligation. De-conjugation of SUMO is catalyzed by SUMOspecific proteases. The SUMOylation and de-SUMOylation enzymes that have been more extensively studied in animal systems are conserved in plants such as Arabidopsis, tomato and rice (Kurepa et al., 2003;Novatchkova et al., 2004Novatchkova et al., , 2012Lois, 2011). Analyses of protein SUMOylation in plants have been performed mainly in Arabidopsis. This has revealed the prevalence of transcription factors and other nuclear-localized regulator proteins among the targets of SUMO (Elrouby and Coupland, 2010;Miller et al., 2010;Lois, 2011;Elrouby et al., 2013;López-Torrejón et al., 2013;Park et al., 2013). However, in plants the known SUMO-modified proteins represent only a much smaller number of SUMO targets in comparison to mammals and yeast (reviewed by Flotho and Melchior, 2013). We note that only a fraction of the potential SUMO-modified proteins has been experimentally confirmed. Furthermore, the functional consequences for reported protein SUMOylation remains unknown in most cases for plant proteins (reviewed by Lois, 2011).
The finding of potential SUMOylation sites in both A9 and A4a, as well as precedents for the importance of SUMOylation of HSF in both animal Hong et al., 2001;Anckar et al., 2006;reviewed by Björk and Sistonen, 2010) and plant systems (Cohen-Peer et al., 2010) called our attention and induced us to performing the experiments reported here. We could directly confirm the SUMOylation of A9 and also provide indirect evidence for the SUMOylation of A4a. Furthermore, we demonstrate that the modification of either A9 or A4a (respectively, at lysine residues K38 and K172) is required for their synergic transcriptional activation. Our results connect SUMOylation with HSF function during late embryogenesis in plant seeds. Thus, SUMO-modified HSFs might be involved in enhancing functions as seed-longevity and desiccation tolerance.

Transient Expression Assays in Sunflower
Directed mutation of the putative SUMOylation sites of HaHSFA9 were made by megaprimer PCR-mutagenesis (see Chen and Przybyla, 1994;Carranco et al., 2010). In HaHSFA9m1, lysine 38 was mutated to arginine. Mutations were introduced by PCR of the plasmid pBI221-HaHSFA9 (Almoguera et al., 2002) with the mutagenic oligo 5 -GGTTCCTCTcTAATCTTCATCATC-3 and 5 -ATGGCAGG AGTAGTAAAAGAGTTTG-3 . This PCR product was used as megaprimer for a second amplification of the same plasmid together with oligo 5 -TTGCACATTTCGACAC TTCC-3 . This final PCR product, digested with StyI and BglII, replaced the corresponding wild type fragment in pSK-HaHSFA9 (Almoguera et al., 2002) to obtain pSK-HaHSFA9m1. In HaHSFA9m2, lysine 206 was mutated to arginine. Mutations were introduced by PCR of the plasmid pBI221-HaHSFA9 with the mutagenic oligo 5 -AGA AAGAATCACACTTAgACAAGAGATC-3 and 5 -TTGCACAT TTCGACACTTCC-3 . This PCR product was used as megaprimer for a second amplification of the same plasmid together with oligo 5 -ATGGCAGGAGTAGTAAAAGAGT TTG-3 . This final PCR product, digested with StyI and BglII, replaced the corresponding wild type fragment in pSK-HaHSFA9 (Almoguera et al., 2002) to obtain pSK-HaHSFA9m2. HaHSFA9m1 and HaHSFA9m2 were introduced into pBI221 vector for transient expression assays as described for pBI221-HaHSFA9. The double mutant HaHSFA9m3 has both lysines 38 and 206 mutated to arginine. To make pBI221-HaHSFA9m3, the EcoRI-EcoRI wild type fragment from pBI221-HaHSFA9m1 was replace for its mutant version obtained from pBI221-HaHSFA9m2.
Proteins 6xHis:Xpress:A9 DBD and 6xHis:Xpress:A4a were expressed in E. coli from plasmids pRSET A-A9 DBD and pRSET A-A4a, respectively, and purified with resin IMAC Sepharose TM 6 Flast Flow, GE Healthcare. In vitro SUMOylation assays were performed as described in García-Domínguez et al. (2008). Hundred nanogram of purified protein was used as the target. Reactions were started with 2 mM ATP and stopped with Laemmli buffer. Proteins were detected by Western blot with antibodies against 6xHis (GE Healthcare).

Western Blot Assays
Proteins from agroinfiltrated N. benthamiana leaves or E. coli cells were extracted with 2x Laemmli's buffer. Total protein samples of E. coli (5 µg) or N. benthamiana leaves (40 µg) were run in SDS-PAGE: 8% acrylamide gels for GFP-fusion proteins and 10% for the rest. Anti-6xHis antibody (GE Healthcare) at 1/1000 dilution was used to detect 6xHis-tagged proteins. Anti-HA-Peroxidase antibody (Roche) at 1/1000 dilution was used to detect YFP C -fusion proteins (detection did not need a secondary antibody). Anti-GFP antibody (Clontech) at 1/2000 dilution was used to detect GFP-fusion proteins. Anti-mouse IgG-Peroxidase (Oncogene TM ) at 1/5000 dilution was used as secondary antibody for Western blots with anti-6xHis. Antirabbit IgG-Peroxidase (GE Healthcare) at 1/50000 dilution was used as secondary antibody for Western blots with anti-GFP. The ECL Prime system (GE Healthcare) and X-ray films were used for detection of the recombinant proteins.

In planta Protein Localization Assays
Nicotiana benthamiana leaves were infiltrated with Agrobacterium harboring plasmids expressing the fusion proteins GFP:HaHSFA4am1 alone or together with HaHSFA9m3:YFP C and analyzed with a confocal laser-scanning Olympus FV1000 microscope as described in Tejedor-Cano et al. (2014).

Statistics
Detailed procedures for ANOVA analyses have been described previously (see Prieto-Dapena et al., 2006, and references therein).

SUMOylation of A9 and A4a: Direct Detection of SUMOylated A9
Using the SUMOplot 1 and SUMOsp 2.0 (Ren et al., 2009) programs, we identified two putative SUMOylation sites in A9 (K38, K206), while a single site (K172) was present in A4a (Table 1A). Site-directed mutagenesis of these sites (changing K to R) was used to analyze their potential SUMOylation. Diverse SUMOylation assays were accomplished using the WT HSFs and 1 http://www.abgent.com/sumoplot  The same result was obtained in the absence (-A9) or presence (+A9) of coexpressed A9; however, the coexpressed A9 was SUMOylated (right panel). Anti-HA antibody was used to detect the fusion proteins in B and the right panels of C and D. Anti-GFP antibody was used to detect the fusion proteins in the left panels of C and D; for further details, see Section "Materials and Methods". Rest of symbols as in Figure 1.
different mutant proteins ( Figure 1A). For example, in vitro SUMOylation assays performed with the Arabidopsis SUMO enzymes E1, E2, and E3, plus SUMO1 and the recombinant A9 protein easily detected a band with retarded mobility consistent with the SUMOylation of A9. In contrast, parallel analyses using the WT A4a protein failed to detect SUMO1-modified forms of A4a ( Figure 1B). To confirm in vivo the SUMOylation of A9, to identify the SUMOylated residue(s), and to further attempt detection of A4a SUMOylation, additional assays were made in Nicotiana benthamiana (Figure 1C). When plasmids encoding the A9 and Arabidopsis SUMO1 proteins were co-infiltrated in leaves, retarded mobility of the WT A9 protein was observed. A similar result was obtained after co-infiltration of SUMO1 with the A9m2 mutant form of A9. In contrast, SUMO-modified forms of A9 were not detected when the A9m1 (or A9m3) mutant proteins were similarly co-expressed. These results confirmed SUMOylation of the A9 protein, also suggesting that SUMOylation in planta occurs mainly at position K38, even without co-expressed SUMO1 (Figure 1D). The equivalent analyses using SUMO1 and the WT A4a protein did not reveal hints of SUMOylation ( Figure 1C). We also unsuccessfully tried to detect A4a SUMOylation in N. benthamiana with different strategies that were designed to cope with several possible limiting steps either individually or combined (Figure 2). For example, we explored if SUMO3 is required instead of SUMO1, or if the SUMOylation of A4a needs A9 (Figures 2B,C, respectively).
In addition, to examine the possibility that a very efficient SUMO de-conjugation is what prevents detection of modified A4a, we coexpressed A4a and a SUMO1 mutant form impaired in de-conjugation (SUMO1 Q90A, Budhiraja et al., 2009); again not achieving success ( Figure 2D). We also used A4amNES a mutant form of A4a that is not exported from the nucleus ). This form was tested with Arabidopsis E2 (SCE1) and SUMO1, SUMO3, or SUMO1 Q90A (Figure 2D, results for SUMO1 Q90A). We thus ruled out that nuclear localization of A4a is required for its SUMOylation. Additional SUMOylation assays were performed in E. coli. We used Arabidopsis SUMO1 or SUMO3, further attempting detection of the SUMOylation of A4a. Only A9 was SUMOylated in E. coli, and SUMOylation was observed using either the SUMO3 or SUMO1 form ( Figure 1E). The results presented so far confirmed that A9 is SUMOylated in planta (at least at K38 and perhaps also at K206); while the SUMOylation of A4a (at K172), if real, it would be more elusive.

Functional Consequences of SUMOylation of A9 and A4a. Indirect Detection of SUMOylated A4a
To further investigate the occurrence and relevance of A9 and A4a SUMOylation, we performed transcriptional assays using the WT and mutant HSF forms (in separate or combined). Most precedents in literature link SUMOylation with the regulation of repression in both animal and plant systems (reviewed by Verger et al., 2003;Gill, 2005;García-Domínguez and Reyes, 2009). Thus, we started by using the A9 and A9m3 forms in transient repression assays by IAA27 . We found that in bombarded sunflower leaves, the K38R and K206R substitutions in A9m3 moderately augmented the transcriptional activation of the HaHsp17.6 G1 (G1, Carranco et al., 1999) promoter, about 1.2-fold compared to what observed with WT A9 (Figure 3). However, the A9 and A9m3 proteins accumulated to similar levels in infiltrated leaves of N. benthamiana (Figure 1C). The moderately enhanced transactivation capacity of A9m3 (Figure 3B), and that of other mutant forms of A9 (see Figures 4B,C), would fit precedent work performed with other mammalian and plant HSFs analyzed in separate Hietakangas et al., 2006;Tateishi et al., 2009;Björk and Sistonen, 2010;Cohen-Peer et al., 2010). We remark that the mutated lysines in A9m3 did not affect transcriptional repression by IAA27, which was observed to the same extent using either A9 or A9m3. (Figure 3; statistical analyses from these and other experiments reported here are included in the Supplementary Table S1). Thus, SUMOylation of A9 would not affect transcriptional repression of A9 by IAA27. IAA27 not only represses activation by A9, but also coactivation by A9 and A4a . Thus, we also performed additional transient experiments originally designed to explore the potential effects of SUMOylation on repression by IAA27 of the synergism between the two HSFs in bombarded sunflower embryos. These experiments where performed using the Hahsp17.7 G4 promoter (G4, Almoguera et al., 1998). In these experiments (Figure 4B), the mutant A9m1 form activated transcription with higher efficiency than the WT form; this effect was compensated by adjusting the amounts of these HSFs (see Materials and Methods). A statistically significant synergistic effect was still observed when the mutant form of one HSF was combined with the WT form of the other HSF ( Figure 4B). However, and surprisingly, the transcriptional synergism between A9 and A4a was fully abolished when the mutant forms of both HSF proteins were combined ( Figure 4B). These results strongly suggest that SUMOylation of A9 at K38 (Figure 1) or modification of A4a at K172 is required for the synergism; because of this unexpected result it was not necessary further testing IAA27 in Figure 4. Similar effects of the mutant proteins on the synergism were observed by transient expression in bombarded sunflower leaves, where additional lysine substitutions (A9m2 and A9m3) in A9 were also analyzed. These experiments confirmed the major effect of SUMOylation at K38, also indicating a minor contribution of SUMOylation at K206 (Figure 4C). We did not directly detect SUMOylated A4a, but the similar lack of synergism with A9m3, obtained with the A4am1 (with a substitution of the lysine residue) and A4am2 mutants, provide a strong, although indirect, evidence for the SUMOylation of A4a at K172 (Figure 4B). The E174A substitution in A4am2 would impair interaction of A4a with the SUMO-E2-conjugating enzyme Ubc9 in the vicinity of K172 (see for example, Sampson et al., 2001). E174A is not expected to affect other modifications of K172 (as acetylation or ubiquitination). We therefore infer that the posttranslational modification at K172 in A4a that is required for the transcriptional synergism is, most likely, also a SUMOylation.

SUMOylation Does Not Affect Nuclear Retention of A4a
The transcriptional synergism involving A9 and A4a requires the mutual interaction of both HSF in the nuclei. In absence of A9, A4a is mostly localized in the cytosol; the interaction of A4a with A9 hinders a NES motif localized in A4a leading to nuclear retention of A4a . Therefore, we investigated if the lysine residues identified as SUMOylation sites in A9 and A4a are necessary for nuclear retention of A4a. The results in Figure 5 show that the K172R mutant form of A4a showed a mostly cytosolic localization similar to what reported for WT A4a . Co-infiltration of the mutant forms of A4a and A9 ( Figure 5) enhanced the nuclear localization of A4am1, again in a similar way as described for the co-expression of the two WT HSF . Therefore, major effects of the analyzed mutations on HSF hetero-oligomerization would be unlikely. We conclude that SUMOylation at the residues mutated in the two HSF proteins used in the experiments of Figure 5 would not be required for their mutual interaction and for the subsequent nuclear retention of A4a.

DISCUSSION
The reported direct evidence for SUMOylation of A9 at K38 (Figure 1), and indirect evidence for SUMOylation of A4a at K172 (Figure 4B), adds these two HSFs to the -yet small-set of plant proteins that are known to be modified by SUMO (Budhiraja et al., 2009;Elrouby and Coupland, 2010;Miller et al., 2010Miller et al., , 2013Lois, 2011;Castro et al., 2012;Elrouby et al., 2013;López-Torrejón et al., 2013;Park et al., 2011Park et al., , 2013. HSFA9s from Arabidopsis and other dicot plants present potential SUMOylation sites at positions close to that of K38 and K206 in sunflower A9 (Table 1B). This does not occur for the sunflower A4a site and other A4a HSFs. Precedent studies in animal systems showed that, in general, SUMOylation enhances the function of a variety of repressor complexes (reviewed by Verger et al., 2003;Gill, 2005;García-Domínguez and Reyes, 2009; for example, see Kang et al., 2010). In contrast, the conjoint analysis of A9 and A4a uncovered a novel and unexpected positive effect for the SUMOylation of these two HSFs. Their synergic co-activation did not occur when the SUMOylated lysines were mutated in A9 and A4a ( Figure 4B). This would fit less usual reports, where SUMOylation enhances transcription, also only in animal systems (Kotaja et al., 2002;Wang et al., 2004Wang et al., , 2007Choi et al., 2011;Yang et al., 2012;Huber et al., 2013); even one of these reports showed that SUMOylation could enhance synergic interactions between transcription factors (Kotaja et al., 2002). We explored simple mechanistic explanations for how the analyzed SUMOylations are required for the synergism. For example, the results in Figure 5 showed that an effect of SUMOylation on the interaction between A4a and A9 is unlikely. SUMOylation might still enhance the interaction of A4a with A9, but this effect would be too-transient or subtle for it to be detected under the conditions in the experiments of Figure 5. The synergism would rather be affected by the analyzed mutations at a subsequent stage after mutual HSF interaction and nuclear co-localization . More complex, alternative explanations of two types would be compatible with the results reported here. SUMOylation has been shown to induce conformational changes in proteins as for example thymine DNA glycosylase (Baba et al., 2005;Steinacher and Schär, 2005). Protein-protein interactions, which in some instances enhance transcriptional activation, have been also shown to be induced by SUMOylation (Ishov et al., 1999;Kotaja et al., 2002;Wang et al., 2004;Choi et al., 2011). Thus, SUMOylation of A9 (or A4a) may facilitate a conformational change required for the synergism, and (or) interaction with a still non-identified coactivator. HSF coactivator proteins have been identified mostly in animals, as for example DAXX, ASC-2, and CHIP (Boellmann et al., 2004;Hong et al., 2004;Kim et al., 2005). In plants, HSF-coactivator studies include only work on HSFB1, which is involved in heat stress responses in tomato (Bharti et al., 2004). However, and as far as we know a connection between SUMOylation and HSF coactivation has not been explored besides the results reported here.
Our results are consistent with an activation model in which SUMOylated A9/A4a complexes bound to DNA sequentially recruit transcriptional coactivator(s) and (or) chromatin remodeling factor(s). We cannot exclude that A9/A4a SUMOylation may also induce conformational changes that, as shown for HSF1 and CHIP1 (Kim et al., 2005), could contribute to the proposed recruitment. We showed that HSF SUMOylation is required for a transcriptional synergism that is involved in the enhancement of two crucial functions in seeds , and references therein): longevity and desiccation tolerance, both acquired in maturing zygotic embryos (reviewed by Dekkers et al., 2015;Sano et al., 2016;Leprince et al., 2017). Precedent studies in plants have only indicated functional connections of SUMOylation with non-embryonic development, for example with the control of flowering (Xu and Yang, 2013;Elrouby et al., 2013;Elrouby, 2014). From the results reported here, we propose that seed HSF SUMOylation may also contribute to explaining the essential role of SUMO in seed development that was inferred from a previous study (Saracco et al., 2007). Our results reveal that, as in animals (see for example, Kang et al., 2010), SUMOylation is involved in the modulation of transcriptional activity in embryos.