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′-ATGGCAGGAGTAGTAAAAGAGTTTG-3′. This PCR product was used as megaprimer for a second amplification of the same plasmid together with oligo 5′-TTGCACATTTCGACACTTCC-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′-AGAAAGAATCACACTTAgACAAGAGATC-3′ and 5′-TTGCACATTTCGACACTTCC-3′. This PCR product was used as megaprimer for a second amplification of the same plasmid together with oligo 5′-ATGGCAGGAGTAGTAAAAGAGTTTG-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.

Directed mutation of the putative SUMOylation site of HaHSFA4a was performed by Mutagenex Inc., starting from plasmid pBI221-HaHSFA4a (Tejedor-Cano et al., 2014). Mutant HaHSFA4am1 has lysine 172 (codon AAA) substituted by arginine (codon AgA) whereas in mutant HaHSFA4am2, glutamic 174 (codon GAA) was substituted by alanine (codon Gcg). Plasmids pBI221-HaHSFA4am1 and pBI221-HaHSFA4am2 were used in transient expression assays.

Transient expression assays in Figure 4B were performed in 21 dpa sunflower embryos, essentially as described (Díaz-Martín et al., 2005). The amounts of plasmid DNA (per DNA precipitate, used for five shots) were: 50 ng pBI221-A9, 15 ng pBI221-A9m1, 2.5 μg pBI221-A4a, 2.5 μg pBI221-A4am1, 2.5 μg pBI221-A4am2 (effectors), 5 μg of –1132(G4):LUC (Díaz-Martín et al., 2005) (reporter) and 1 μg of pBI221-Rluc. The total amount of plasmid DNA was adjusted (if necessary) with pBI221 to 8.5 μg.

For the assays in Figures 3B, 4C, sunflower leaves were bombarded essentially as described (Tejedor-Cano et al., 2010). The amounts of plasmid DNA in Figure 3B (per DNA precipitate, used for five shots) were: 5 μg pBI221-A9, 5 μg pBI221-A9m3, 5 μg pBI221-HaIAA27 (Carranco et al., 2010) (effectors), 5 μg of -126(G1):LUC (Díaz-Martín et al., 2005) (reporter) and 1 μg of pBI221-Rluc. The total amount of plasmid DNA was adjusted (if necessary) with pBI221 to 16 μg. The amounts of plasmid DNA in Figure 4C were as in Figure 4B except for the amounts of effectors: 10 ng pBI221-A9, 2.5 ng pBI221-A9m1, 2.5 ng pBI221-A9m2, 2.5 ng pBI221-A9m3, 2.5 μg pBI221-A4a, and 2.5 μg pBI221-A4am1.

Plasmid pRSET A-A9ΔDBD was made by replacing the XhoI to BglII fragment of HaHSFA9 from plasmid pRSET A-HaHSFA9 (construct “I” in Díaz-Martín et al., 2005) by a BglII digested PCR fragment obtained by amplification of plasmid pGBT9-HaHSFA9ΔDBD (Carranco et al., 2010) with oligos 5′-GTTCATGGCAGGAGTAGTAAAAGAG-3′ and 5′-TTGCACATTTCGACACTTCC-3′. To make plasmid pRSETA-A4a, PCR amplification of plasmid pUC-HaHSFA4a (Tejedor-Cano et al., 2014) was performed with oligos 5′-GGTATATCTTGGTCAATGATGA-3′ and 5′-GTAAAACGACGGCCAGT-3′. The SacI digested PCR product was introduced between the SmaI and SacI sites of pBI221. The amplified A4a sequence, which does not include the 5′-UTR, was then released by restriction with BamHI and KpnI and inserted into BglII and KpnI digested pRSET A vector (Invitrogen).

Proteins 6xHis:Xpress:A9ΔDBD and 6xHis:Xpress:A4a were expressed in E. coli from plasmids pRSET A-A9ΔDBD 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).

For Nicotiana benthamiana leaves infiltration assays, the HaHSFA9 and HaHSFA4a mutants were transferred from the pBI221 plasmids used in transient expression assays (described above) to the pUC SPYCE(M) vector (Waadt et al., 2008) as described for pUC SPYCE(M)-HaHSFA9 (Carranco et al., 2010) and then to the binary vector kanII-SPYCE(M) as described in Tejedor-Cano et al., 2014. Plasmids kanII-SPYCE(M)-HaHSFA9 (Tejedor-Cano et al., 2014), kanII-SPYCE(M)-HaHSFA9m1, kanII-SPYCE(M)-HaHSFA9m2, kanII-SPYCE(M)-HaHSFA9m3, kanII-SPYCE(M)-HaHSFA4a (Tejedor-Cano et al., 2014) and kanII-SPYCE(M)-HaHSFA4am1 express fusion proteins HaHSFA9:YFP^C, HaHSFA9m1:YFP^C, HaHSFA9m2:YFP^C, HaHSFA9m3:YFP^C, HaHSFA4a:YFP^C, and HaHSFA4am1:YFP^C, respectively.

Mutant HaHSFA4am1 was also fused to GFP and introduced in the binary vector pRCS2-nptII (Tzfira et al., 2005) as described for HaHSFA4a in Tejedor-Cano et al., 2014. Plasmids pRCS2-nptII-EGFP:HaHSFA4a, pRCS2-nptII-EGFP:HaHSFA4a (NESmut) (Tejedor-Cano et al., 2014) and pRCS2-nptII-EGFP:HaHSFA4am1 express fusion proteins GFP:HaHSFA4a, GFP:HaHSFA4amNES and GFP:HaHSFA4am1, respectively.

For AtSUMO1 expression, pUNI-AtSUMO1 (ABRC; Stock # U17495) was amplified by PCR with oligos 5′-GGAGATCTAGACATGTCTGCAAACCAGG-3′ and 5′-ATGAGAGCTCAGGCCGTAGCACCAC-3′ and cloned into the SacI (blunted with T4 DNA polymerase) and XbaI (Klenow filled) sites of pBI221 vector. For AtSUMO3, pUNI-AtSUMO3 (ABRC, Stock # U83833) was amplified by PCR with oligos 5′-GGAGATCTAGGCATGTCTAACCCTCAAG-3′ and 5′-ATGAGAGCTCAAAGCCCATTATG-3′ and cloned into vector pBI221 as described for AtSUMO1. For the stabilized mutant AtSUMO1 Q90A the transgene was released from plasmid pSK-Tag3-SUMO1 Q90A (Budhiraja et al., 2009) by digestion with NdeI and XbaI, filled with Klenow and cloned into vector pBI221 as described for AtSUMO1. AtSCE1 was released from plasmid, pGST-AtSCE1 (García-Domínguez et al., 2008) by SmaI and NcoI digestion, filled with Klenow and cloned into the SacI (blunted with T4 DNA polymerase) and SmaI sites of pBI221 vector. AtSUMO1, AtSUMO1 Q90A, AtSUMO3, and AtSCE1 were excised from plasmids pBI221-AtSUMO1, pBI221-AtSUMO1 Q90A, pBI221-AtSUMO3, and pBI221-AtSCE1, respectively, as HindIII-EcoRI fragments and introduced in the corresponding sites of the vector pBI121.

Nicotiana benthamiana leaves infiltration assays were described in Carranco et al. (2010). In Figures 1C,D, 2B–D (right panels), Agrobacterium harboring plasmids expressing the fusion proteins HaHSFA9:YFP^C, HaHSFA9m1:YFP^C, HaHSFA9m2:YFP^C, HaHSFA9m3:YFP^C, HaHSFA4a:YFP^C, or HaHSFA4am1:YFP^C were co-infiltrated with plasmid for expression of AtSUMO1, AtSUMO1 Q90A, AtSUMO3, or AtSCE1 as indicated in the corresponding figures. In Figures 2C,D (left panels), Agrobacterium harboring plasmids expressing the fusion proteins GFP:HaHSFA4a, GFP:HaHSFA4amNES, or GFP:HaHSFA4am1 were co-infiltrated with plasmid for expression of HaHSFA9:YFP^C, AtSUMO1, AtSUMO1 Q90A, or AtSCE1 as indicated in the corresponding figures.

For HSFs expression in E. coli, HaHSFA9 was amplified from pBI221-A9 by PCR with oligos 5′-aaaaagcaggcttcATGGCAGGAGTAGTAAAAG-3′ and 5′-agaaagctgggTCAGCTTTTGGGATCAACTC-3′. HaHSFA4a was amplified from pBI221-A4a with oligos 5′-aaaaagcaggcttcATGATGAATGATGTTCATG-3′ and 5′-agaaagctgggTCACTTCTCTCTACTGAAG-3′. The resulting DNA fragments were cloned into pDONR201 and then transferred to pER32b-GW (that introduces N-terminal Trx and 6xHis tags), as described in Elrouby and Coupland (2010). Each resulting plasmid was transformed into E. coli BL21 (DE star) together with pCDF-SAE and pACYC-SCE-SUMO3 (Elrouby and Coupland, 2010) or pCDFDuet-AtSUMO1-AtSCE1 and pACYCDuet-AtSAE1-AtSAE2 (Okada et al., 2009). As negative control E. coli BL21 (DE star) was also transformed with similar plasmid combinations lacking the SUMO harboring plasmids. All the genes where induced over night with 0.5mM IPTGat 28°C.

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. Anti-rabbit 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.

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).

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

Using the SUMOplot¹ 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 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 A9Δ 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 (Tejedor-Cano et al., 2014). 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.

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 (Carranco et al., 2010). 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 (Anckar et al., 2006; 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 (Tejedor-Cano et al., 2014). 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 post-translational modification at K172 in A4a that is required for the transcriptional synergism is, most likely, also a SUMOylation.

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 (Tejedor-Cano et al., 2014). 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 (Tejedor-Cano et al., 2014). 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 (Tejedor-Cano et al., 2014). 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.