Heat Shock Protein 27 Is Involved in SUMO-2&Sol

Home , Hsp27, Hsp70, Hsp90, Hyperphosphorylation, SAE1, SUMO protein

Oncogene (2009) 28, 3332–3344 & 2009 Macmillan Publishers Limited All rights reserved 0950-9232/09 $32.00 www.nature.com/onc ORIGINAL ARTICLE Heat shock protein 27 is involved in SUMO-2/3 modiﬁcation of heat shock factor 1 and thereby modulates the transcription factor activity

M Brunet Simioni1,2, A De Thonel1,2, A Hammann1,2, AL Joly1,2, G Bossis3,4,5, E Fourmaux1, A Bouchot1, J Landry6, M Piechaczyk3,4,5 and C Garrido1,2,7

1INSERM U866, Dijon, France; 2Faculty of Medicine and Pharmacy, University of Burgundy, Dijon, Burgundy, France; 3Institut de Ge´ne´tique Mole´culaire UMR 5535 CNRS, Montpellier cedex 5, France; 4Universite´ Montpellier 2, Montpellier cedex 5, France; 5Universite´ Montpellier 1, Montpellier cedex 2, France; 6Centre de Recherche en Cance´rologie et De´partement de Me´decine, Universite´ Laval, Quebec City, Que´bec, Canada and 7CHU Dijon BP1542, Dijon, France

Heat shock protein 27 (HSP27) accumulates in stressed otherwise lethal conditions. This stress response is cells and helps them to survive adverse conditions. We have universal and is very well conserved through evolution. already shown that HSP27 has a function in the Two of the most stress-inducible HSPs are HSP70 and ubiquitination process that is modulated by its oligomeriza- HSP27. Although HSP70 is an ATP-dependent chaper- tion/phosphorylation status. Here, we show that HSP27 is one induced early after stress and is involved in the also involved in protein sumoylation, a ubiquitination- correct folding of proteins, HSP27 is a late inducible related process. HSP27 increases the number of cell HSP whose main chaperone activity is to inhibit protein proteins modified by small ubiquitin-like modifier aggregation in an ATP-independent manner (Garrido (SUMO)-2/3 but this effect shows some selectivity as it et al., 2006). HSP27 has the ability to form oligomers of neither affects all proteins nor concerns SUMO-1. More- up to 1000 kDa. HSP27 oligomerization is a highly over, no such alteration in SUMO-2/3 conjugation is dynamic process modulated by the phosphorylation of achievable by another HSP, such as HSP70. Heat shock the protein. Human HSP27 can be phosphorylated on factor 1 (HSF1), a transcription factor responsible for HSP three serines (S15, S78 and S82) and the phosphoryla- expression, is one of the targets of HSP27. In stressed cells, tion provokes a shift toward small oligomers (Lambert HSP27 enters the nucleus and, in the form of large et al., 1999; Rogalla et al., 1999). The different activities oligomers, binds to HSF1 and induces its modification by of HSP27 seem to be modulated by the oligomerization SUMO-2/3 on lysine 298. HSP27-induced HSF1 modifica- pattern of the chaperone. We have shown, in vitro and tion by SUMO-2/3 takes place downstream of the in vivo, that large non-phosphorylated oligomers of transcription factor phosphorylation on S303 and S307 HSP27 are responsible for the caspase-dependent anti- and does not affect its DNA-binding ability. In contrast, this apoptotic effect of this chaperone (Bruey et al., 2000). modification blocks HSF1 transactivation capacity. These Other groups have shown that large oligomers are data show that HSP27 exerts a feedback inhibition of needed for HSP27 antioxidant activity (Rogalla et al., HSF1 transactivation and enlighten the strictly regulated 1999). Large non-phosphorylated HSP27 oligomers interplay between HSPs and HSF1. As we also show that bind to actin and the two proteins dissociate HSP27 binds to the SUMO-E2-conjugating enzyme, Ubc9, upon HSP27 phosphorylation (During et al., 2007). our study raises the possibility that HSP27 may act as a In contrast, it is proven that phosphorylated HSP27 SUMO-E3 ligase specific for SUMO-2/3. directly interacts with Daxx (Charette et al., 2000). Oncogene (2009) 28, 3332–3344; doi:10.1038/onc.2009.188; Moreover, small oligomers of HSP27 are also the form published online 13 July 2009 of the protein, which displays affinity for ubiquitin chains and accelerates the degradation of certain Keywords: HSP27; SUMOylation; HSF1; transcriptional proteins under stress conditions (Parcellier et al., activity; stress; cancer 2006). These results suggest that oligomerization/phosphorylation of the protein alters HSP27 conformation and, hence, determines its capacity to interact with its Introduction different partners. Expression of HSP genes is regulated by the family of Stress or heat shock proteins (HSPs) are induced by the heat shock transcription factors, which bind to the different stresses and help the cells to cope with these heat shock element (HSE) in the promoter region of heat shock genes and, thereby stimulate their transcription. Three members have been identified in mammals, Correspondence: Dr C Garrido, Faculty of Medicine, INSERM U866, of which heat shock factor 1 (HSF1) is the major stress- 7 boulevard Jeanne d’Arc, 21000 Dijon, Burgundy, France. E-mail: [email protected] responsive family member as no other HSF is able to Received 24 January 2009; revised 17 May 2009; accepted 29 May 2009; functionally substitute for HSF1 or to rescue the published online 13 July 2009 heat shock response in HSF1-deficient cells or mice HSP27 facilitates HSF1 sumoylation M Brunet Simioni et al 3333 (McMillan et al., 1998; Xiao et al., 1999). In response to (Gill, 2003; Hay, 2005). Many transcription factors stressful stimuli such as elevated temperatures, oxidants, such as HSF1 have been shown to be sumoylated and heavy metals, and bacterial or viral infections, HSF1 is this modification, in most cases, impairs the transactiva- activated by trimerization and hyperphosphorylation tion capacity of the substrate (Geiss-Friedlander and (Pirkkala et al., 2001; Holmberg et al., 2002). As a Melchior, 2007). result, HSF1 acquires the ability to bind to HSE and to Heat shock factor 1 has been shown to undergo stress- activate transcription of heat shock genes, which results inducible SUMO-2/3 modification on a conserved lysine, in accumulation of HSPs such as HSP27 and HSP70 K298 (Hietakangas et al., 2003). Further, it has been with, as a final outcome, cell protection (Pirkkala et al., shown that HSF1 has to be previously phosphorylated 2001; Didelot et al., 2006; Garrido et al., 2006; Schmitt on serine 303 for its subsequent sumoylation (Hietakan- et al., 2007; Anckar and Sistonen, 2007a, b). Once the gas et al., 2003). Considering the need of regulated stress conditions are over, HSP gene transcription must chaperone levels in maintaining cellular homoeostasis, immediately stop. HSF1 activity is therefore tightly repression of HSF1 activity by interaction with HSPs is regulated, in particular, through various types of post- essential. The objective of this paper was to study translational modifications. Thus, HSF1 can be phos- whether the main inducible heat shock proteins, HSP70 phorylated at different serine residues (Holmberg et al., or HSP27, have a role in the overall sumoylation process 2001, 2002) and be ubiquitinated (Lee et al., 2008). More and, more particularly, whether they could regulate recently, it has been shown that HSF1 can also be HSF1 sumoylation. We found that HSP27, in the form sumoylated (Hong et al., 2001; Hietakangas et al., 2003, of large oligomers, induced HSF1 SUMO-2/3 modifica- 2006; Anckar et al., 2006). tion and blocked its transactivation capacity. By doing Small ubiquitin-like modifier (SUMO) proteins are so, HSP27 allows a fast and precise modulation of HSP about 100 amino acids in size with a three-dimensional gene transcription, including that of its own gene. structure reminiscent to that of ubiquitin. However, SUMO and ubiquitin share o20% amino acid sequence identity and are different in their overall surface charge Results and Discussion distribution (Geiss-Friedlander and Melchior, 2007). Like ubiquitination, sumoylation results in the forma- HSP27 favours HSF1 modification by SUMO-2/3 tion of an isopeptide bond between the glycine residue To explore whether the two main stress-inducible HSPs, of the modifier protein and the e-amino group of a lysine HSP27 and HSP70, play a role in the sumoylation residue in the acceptor protein. SUMO conjugation process, we assessed their effect on HSF1 modification utilizes a multistep enzymatic pathway, in which by SUMO-2/3. To this aim, we transiently transfected proteolytically processed SUMO initially forms a HeLa cells with a vector for either Myc-tagged wild-type thioester bond with Sae1/2, the E1 SUMO-activating HSF1 (HSF1-wt) or a non-sumoylable HSF1 mutant in enzyme (Johnson, 2004). The SUMO moiety is subse- which K298 has been replaced by arginine (HSF1- quently transferred to Ubc9, the single E2 SUMO- K298R), in the presence or in the absence of vectors for conjugating enzyme, which usually binds directly the haemagglutinin (HA)-tagged HSP27 (HSP27) or HSP70 target protein at the level of sumoylated sites (Geiss- (HSP70) (Figure 1a). SUMO-2/3 modification being Friedlander and Melchior, 2007). E2-substrate interac- inducible by stress (Saitoh and Hinchey, 2000), there- tion may be facilitated by SUMO-E3 factors. The E3 fore, the absence of detection of HSF1 modified by factors increase sumoylation efficiency in a substrate- SUMO-2/3 was not surprising under our experimental specific manner, either by accelerating the transfer of conditions. The only exception for HSF1-SUMO-2/3 SUMO from Ubc9 to the substrate or merely by detection was in cells overexpressing HSP27 and HSF1- providing a scaffold (Guo et al., 2005; Geiss-Friedlander wt, as shown in Figure 1a, left panel, and confirmed on and Melchior, 2007). The SUMO family consists of four the right panel by reversing the immunoprecipitation/ members, SUMO-1, -2, -3 and -4, but SUMO-4 has only immunodetection. As expected, this SUMO-2/3 modi- been detected at the RNA level (Guo et al., 2005; Geiss- fication of HSF1 induced by HSP27 disappeared in cells Friedlander and Melchior, 2007). SUMO-2 and -3 are overexpressing the non-sumoylatable HSF1 mutant 97% identical to each other and, therefore, they cannot HSF1-K298R (Figure 1a). Therefore, overexpression be differentiated at the protein level (Saitoh and of HSP27 favours HSF1 modification by SUMO-2/3 on Hinchey, 2000). SUMO-1 and SUMO-2/3 although lysine 298 that, as shown in Figure 1a, could not be ubiquitously present in mammals, have, at least performed by an overexpression of HSP70. It should partially, distinct substrates and regulation (Saitoh and be noted that the shift observed in the size of HSF1 after Hinchey, 2000; Zhu et al., 2008). The expression of modification by SUMO-2/3 is considerably higher than SUMO-2/3 in contrast to that of SUMO-1 increases the calculated molecular mass of a SUMO molecule after different stresses (Saitoh and Hinchey, 2000; (11 kDa). This is consistent with previous studies, the Ayaydin and Dasso, 2004). Beyond this, the functional explanation given being that conjugation of SUMO in differences between the SUMO paralogues are largely the central region of HSF1 forms a branched protein unknown. In most cases, sumoylation results in the structure responsible for aberrant electrophoretic mi- addition of single SUMO entities. Sumoylation gration (Hietakangas et al., 2003, 2006). can cause changes in protein–protein interactions, Another protein target for SUMO-2/3 modification subnuclear localization and conformational changes being c-Fos (Bossis et al., 2005; Bossis and Melchior,

Oncogene HSP27 facilitates HSF1 sumoylation M Brunet Simioni et al 3334 IP SUMO-2/3 IP IgG IP IgG IP HSF1 HSP70 HSP27 HSP27 HSP70 98R 298R K2 K 1 1wt F S HSF1 K298R HSF1 K298R H HSF HSF1 Co HSF1wt HSF1wt HSF1wt HSF1wt HSF1 K298R HSP27 HSP70 WB HSF1 WB SUMO-2/3 (MYC) 175 KDa

INPUTS

HSP70 HSP27 R t w 1 F S SF1 K298 o Co H HSF1 K298R H HSF1wt Co Co HSF1wt HSF1 K298R Co C HSP70 70 KDa HSP27 27 KDa HSF1 83 KDa

IP SUMO-2/3 IP SUMO-2/3 IP IgG

c-Fos ++++– siRNA siRNA Scr HSP27 HSP27 ––++– WB HSF1 SUMO-2/3 – ++– – 175 KDa INPUTS WB c-Fos 97 KDa HSP27 27 KDa INPUTS

HSF1 83 KDa c-Fos ++++-

HSP27 --++- 14-3-3 30 KDa

SUMO-2/3 - ++ --

c-Fos 66 KDa

HSC70 70 KDa

Figure 1 Heat shock protein 27 (HSP27) induces modiﬁcation by SUMO-2/3 of heat shock factor 1 (HSF1) but not of c-Fos. (a) Left panel, immunoprecipitation (IP) carried out with an anti-SUMO-2/3 antibody was followed by immunodetection of HSF1 in HeLa cells transiently transfected with either a control vector (Co) or a vector for haemagglutinin (HA)-tagged HSP27-HA, HSP70-HA, HSF1 wild type (Myc-tagged HSF1-wt) or the non-sumoylatable Myc-tagged HSF1-K298R mutant. When indicated, HSF1-wt-Myc and HSF1-K298R-Myc-expressing cells were co-transfected with HSP27-HA or HSP70-HA (last four lanes). Right panel, the reverse immunoprecipitaion/immunodetection was carried out: IP with anti-HSF1, western blot (WB) with anti-SUMO2/3. IP IgG, immunoprecipitation with a non-relevant antibody. (b) HeLa cells transfected with c-Fos with or without HSP27-HA expression vector were immunoprecipitated with a SUMO-2/3 antibody followed by immunodetection of c-Fos. (c) IP carried out with an anti-SUMO-2/3 antibody was followed by immunodetection of HSF1 in 48-h serum-depleted HeLa cells transiently transfected with an HSP27 small interfering RNA (siRNA) or a scrambled (Scr) siRNA. INPUTS: the content of the indicated proteins in the cellular extracts is shown.

2006), we also tested whether HSP27 could facilitate its HSP27 increases general protein sumoylation in cells sumoylation. We found that c-Fos modification by We next analysed whether HSP27 or HSP70 affected the SUMO-2/3 in contrast to that of HSF1 was not affected general profile of proteins modified by SUMO-2/3. by HSP27 (Figure 1b). This indicated some substrate Mouse embryonic fibroblasts (Figure 2a) and HeLa cells selectivity in the action of HSP27. (not shown) were transiently transfected with HSP27- Thus, in the absence of stress other than the cell HA or HSP70-HA expression vectors, and we favoured transfection per se, HSP27 can induce SUMO-2/3 SUMO-2/3 modification by stress (48-h serum deple- modification of HSF1 but not of c-Fos. The fact that tion, Figure 2a, or 100 nM staurosporine for 24 h, not HSP27 depletion by means of a small interfering RNA shown). After denaturing cell lysis followed by immu- (siRNA) decreases HSF1 modification by SUMO-2/3 noblotting analysis with antibodies against SUMO-2/3, induced by stress (that is, serum depletion for 48 h, the signal intensity of the sumoylated proteins profile Figure 1c) further confirms a role for HSP27 in protein was stronger in HSP27-overexpressing cells compared sumoylation. with control-transfected cells. In contrast, HSP70

Oncogene HSP27 facilitates HSF1 sumoylation M Brunet Simioni et al 3335

siRNA CoHSP27 Scr MW Recovery time (h) MW o MW HSP70-wt HSP27-wt C KDa NT SDNT SD NT SD KDa 0 0.5 1 2 6 12 KDa 95 95 95

72 72 72 SUMO-2/3 SUMO-2/-3 SUMO-2/3 26 26 17 26 17 17 HSP27 27 HSC70 HSP27 27 HSP70 70 70 HSC70 70 HA 27 HSC70 70 p a s l wt A - - -A 7 7 7 2 2 7891011121314 15 16Fractions 2 MW MW SP KDa KDa Co Co HSP H HSP HSP27-Asp HSP27-Ala Co HSP27-wt 95 95 HSP27-wt 72 72 HSP27-Ala SUMO-2/3 SUMO-1 HSP27-Asp 26 26 0 0 5

00 6 8 9 KDa 7 4 0 2 182 1 17 17 HSC70 70 HSC70 70 Figure 2 Heat shock protein 27 (HSP27) promote proteins modification by SUMO-2/3. (a) Total cell extracts from HeLa cells transfected with a haemagglutinin-tagged HSP27, HSP70 or control vector were stressed (serum-depleted for 48 h) and analysed by immunoblotting with SUMO-2/3 antibody. (b) HeLa cells were heat shocked at 42 1C for 1 h. At the indicated recovery time at 37 1C, the content of proteins modified by SUMO-2/3 and of HSP27, HSP70 and HSC70 were assayed by immunoblotting. (c) Proteins modified by SUMO-2/3 were determined in serum-depleted HeLa cells transiently transfected with an HSP27 small interfering RNA (siRNA) or a scrambled (Scr). Co, mock transfected cells. (d) Extracts from 48-h serum-depleted control (Co), HSP27-wild type (wt)-, HSP27-Ala- and HSP27-Asp-transfected REG colon carcinoma cells were size-fractionated through a Superose-6 column. HSP27 presence in the fractions was monitored by immunoblotting. (e) The profile of proteins modified by SUMO-2/3 (left panel) or by SUMO-1 (right panel) was determined by western blot in cells described in (d). MW, molecular weight. overexpression did not alter the amount of protein Large oligomers of HSP27 are needed to favour modified by SUMO-2/3 (Figure 2a). As cell transfection sumoylation is an artificial way to obtain an accumulation of HSPs, Heat shock protein 27 (HSP27) has been shown to form we induced the overexpression of endogenous HSPs by oligomers of up to 1000 kDa. This oligomerization is heat shock and studied the repercussion in the total modulated by the phosphorylation status of the protein, cellular amount of proteins modified by SUMO-2/3. phosphorylation provoking a shift towards small HeLa cells were exposed to 42 1C for 1 h and then oligomers. Accordingly, gel filtration experiments cultured at 37 1C for different times to allow HSP showed that an HSP27 mutant in which the three expression. As expected, HSP70 was early induced phosphorylable serines have been replaced by alanines reaching a peak 30 min–1 h after the heat shock and, (HSP27-Ala) abundantly yields large oligomers, whereas by 2 h HSP70 level decreased (Figure 2b). In contrast, an HSP27 mutant in which the serine residues have been HSP27, a late inducible HSP, accumulated in the cells replaced by the phosphomimetic aspartates (HSP27- only at 6–12 h after the heat shock (Figure 2b). A heat Asp) exclusively yields small oligomers (Lambert et al., shock also induced an increase in general protein 1999; Bruey et al., 2000; Parcellier et al., 2006) modification by SUMO-2/3 and this increase was (Figure 2d). For these studies we used colon cancer coincidental with HSP27 accumulation in the cells REG cells stably transfected with the different HSP27 (Figure 2b). Further, SUMO-2/3 modification induced mutants because these cells do not express any by 48-h serum depletion was strongly decreased in HeLa detectable level of endogenous HSP27 (Bruey et al., cells in which HSP27 levels were depleted by means of a 2000) (Figure 2d). Using the same experimental specific siRNA (Figure 2c). Altogether, these results conditions as in Figure 1, we found that although indicate that HSP27 can affect general protein sumoyla- the HSP27-Ala mutant favoured overall SUMO-2/3 tion, whereas HSP70 cannot. protein modification more efficiently than HSP27-wt,

Oncogene HSP27 facilitates HSF1 sumoylation M Brunet Simioni et al 3336 the HSP27-Asp mutant was unable to induce this post- large oligomers, can affect the cellular content of translational modification (Figure 2e, left panel). This proteins modified by SUMO-2/3 and that of HSF1 in effect of HSP27 and HSP27-Ala favouring general particular. protein sumoylation is specific for SUMO-2/3 as it does To gain insight into the mechanism whereby HSP27 not concern SUMO-1 (Figure 2e, right panel). SUMO-1 favours HSF1 modification by SUMO-2/3, we studied is constitutively expressed, whereas SUMO-2/3 is whether HSP27 could interact with the Ubc9 SUMO E2 induced by different stresses (Saitoh and Hinchey, enzyme. It has been previously shown that HSP27 can 2000). However, neither wild-type HSP27 nor its associate with Drosophila Ubc9 (Joanisse et al., 1998) phosphorylation/oligomerization mutants modified the and we confirmed this result for mammalian Ubc9. transcription level of the sumo-2/3 genes. (Supplemen- As shown in Figure 3b, wild-type HSP27 and HSP27- tary Figure 1 and data not shown) Ala associate with Ubc9, whereas HSP27-Asp, which is Next, in these same cells, we analysed the effect of the unable to induce SUMO-2/3 modification, did not HSP27 phosphorylation/oligomerization mutants on interact with Ubc9. It is therefore possible that HSP27 HSF1 modification by SUMO-2/3. We observed that, might act as a scaffold to strengthen the interaction although HSP27-Ala induced a more efficient sumoyla- between Ubc9 and HSF1. tion of HSF1 than HSP27-wt, the non-oligomerizing HSP27-Asp mutant was inefficient at inducing HSF1 HSF1 and large oligomers of HSP27 associate SUMO-2/3 modification (Figure 3a, left panel). Again, Co-immunoprecipitation experiments carried out with this modification specifically concerned SUMO-2/3 an equal amount of antibody in HeLa cells (Figure 4a) because HSP27 and HSP27-Ala did not alter the showed that endogenous HSP27 can associate with amount of HSF1 modified by SUMO-1 (Figure 3a, HSF1. The interaction already observed in non-stressed right panel). We conclude that HSP27, in the form of cells increased after stresses inducing HSP27 expression

IP SUMO-2/3 IP SUMO-1 27-Ala 27-Ala HSP27-wt HSP27-wt HSP27-Asp Co HSP27-Asp HSP HSP Co

WB HSF1 WB HSF1 175 KDa 175 KDa

INPUTS HSP27-wt HSP27-Asp Co HSP27-Ala HSP27-Ala Co HSP27-wt HSP27-Asp HSP27 27 KDa HSF1 83 KDa 14-3-3 30 KDa

IP IP INPUTS IgG HA Co HSP70-wt HSP27-wt HSP27-Ala HSP27-Asp IgG Co HSP70-wt HSP27-wt HSP27-Ala HSP27-Asp 70 KDa WB Ubc9 20 KDa HA

27 KDa

14-3-3 30 KDa

Figure 3 Large oligomers of heat shock protein 27 (HSP27) favour heat shock factor 1 (HSF1) modiﬁcation by SUMO-2/3 and associate to Ubc9. (a) Immunoprecipitation (IP) carried out with an anti-SUMO-2/3 antibody or an anti-SUMO-1 antibody was followed by immunodetection (western blot (WB)) of HSF1 in the REG cells described in Figure 2d. The endogenous level of HSF1 in the cells is shown (INPUTS). (b) IP with an haemagglutinin (HA) antibody in extracts from HeLa cells transfected with a HA-tagged empty vector, HA-HSP70, HA-HSP27, HA-HSP27-Ala or HA-HSP27-Asp, was followed by immunodetection of Ubc9.

Oncogene HSP27 facilitates HSF1 sumoylation M Brunet Simioni et al 3337 IP IP IgG HSF1 IP IP HSP27 HSP27 HSP27 IgG HSP27 Co wt Ala Asp NT SD NT SD NT SD NT SD NT SD 83 KDa WB HSF1 WB HSP27 27 KDa

INPUTS

INPUTS HSP27 HSP27 HSP27 MW Co wt Ala Asp IgG NT SD KDa NT SD NT SD NT SD NT SD HSF1 83 HSP27 27 KDa HSP27 27 83 KDa HSF1 HSC70 70 14.3.3 30 KDa

NT Eluted fractions (size) 12345

KDa 1000 480 760 295 182 HSP27 27 KDa

HSF1 83 KDa IP SUMO-2/3

WB HSF1 175 KDa

SD Eluted fractions (size) 12345

KDa 182 295 1000 760 480 HSP27 27 KDa

HSF1 83 KDa IP SUMO-2/3

WB HSF1 175 KDa

Figure 4 Heat shock protein 27 (HSP27) large oligomers associate to heat shock factor 1 (HSF1). (a) Immunoprecipitation (IP) carried out with an anti-HSP27 antibody was followed by immunodetection (western blot (WB)) of HSF1 in HeLa cells either untreated (NT) or stressed by 48-h serum depletion (SD). (b) Immunoprecipitation of HSF1 from control-, HSP27-wild type (wt)-, HSP27-Ala- or HSP27-Asp-transfected REG cells was followed by immunodetection of HSP27 in NT and SD cells). (c) HeLa cell extracts from NT and SD cells were fractionated through a Superose-6-column to separate the HSP27 oligomers by their size (as in Figure 2d). After concentration of the cell fractions, extracts containing a same amount of HSP27 were immunoprecipitated with SUMO-2/3 and the presence of HSF1 was analysed by western blot.

such as staurosporine treatment (100 nM, for 24 h, not SUMO-2/3 modiﬁcation. A same amount of HSP27 shown) or serum starvation for 48 h (Figure 4a). To was immunoprecipitated from the cell fractions further characterize HSP27 and HSF1 interaction, we obtained (though the HSP27 oligomerization status used REG cells transfected with a control vector or varied, as in Figure 2d). Then, we determined by vectors for either HSP27-wt, HSP27-Ala or HSP27-Asp immunoblotting the amount of HSF1 in the precipitates. (Figure 4b). Immunoprecipitation of endogenous HSF1 We observed that only the fractions corresponding to with the exogenously expressed HSP27 proteins showed large multimers of HSP27 in the serum-depleted cell that HSF1 interacted with HSP27-wt and HSP27-Ala extracts contained SUMO-2/3-HSF1 (Figure 4c). Here, mutant (Figure 4b). Interestingly, the mutant HSP27- it is interesting to note that HSF1 immunoprecipitated Asp, which is unable to form large oligomers and induce with the HSP27 antibodies exclusively showed an HSF1 sumoylation, did not interact with HSF1 apparent molecular weight of 83 kDa (size of HSF1) (Figure 4b). The ability of large oligomers of HSP27 with no detectable HSF1 at 175 kDa, which would to interact with HSF1 was further studied by gel correspond to HSF-1-SUMO-2/3 (Hietakangas et al., ﬁltration analysis of extracts from HeLa cells either left 2003, 2006). The simplest scenario explaining this untreated or serum depleted for 48 h both in order to data would be that, after interaction with and sumoyla- increase HSP27 endogenous content and to induce tion of HSF1, HSP27 would dissociate from its

Oncogene HSP27 facilitates HSF1 sumoylation M Brunet Simioni et al 3338 sumoylated substrate once having played its sumoyla- HSP27 inhibits HSF1 transcriptional factor activity tion-promoting part. without affecting upstream HSF1 activation steps We next studied whether HSP27 affected hsf1 gene transcription and/or the abundance of the HSF1 Large oligomers of HSP27 and HSF1 colocalize protein. Neither HSP27-wt nor its mutants HSP27-Ala in the nucleus with SUMO-2/3 or HSP27-Asp affected the transcription level of the hsf1 Microscopic analysis of HSP27 and HSF1 cellular gene (Supplementary Figure 2). At the protein level, distribution before and after stress (48-h serum deple- HSF1 content was the same in cells either overexpres- tion) showed that although HSP27 was mainly cytosolic, sing HSP27 or in those where HSP27 was depleted by HSF1 could be found both in the nucleus and cytosol means of a specific siRNA (Supplementary Figure 3a). (Figures 5a). After stress, HSP27 nuclear localization is Furthermore, the amount of HSF1 was the same in cells greatly potentiated (Figure5a and inputs of Figure 5c). expressing HSP27-wt or in the mutants, HSP27-Ala and The nuclear localization of HSP27 after stress was more HSP27-Asp (Supplementary Figure 3b). Altogether, evident in the case of HSP27-Ala (Figure 5a). In these results indicate that the effect of HSP27 on contrast, HSP27-Asp, which only gives small oligomers, HSF1 seems to be exclusively at the post-translational did not accumulate within the nucleus under the stress level. conditions used in this study (48-h serum depletion, Heat shock factor 1 phosphorylation on S303 is Figure 5a and inputs of Figure 5c or staurosporine needed for HSF1 sumoylation and for its nuclear 100 nM, not shown). These results suggest that the co-localization with SUMO (Hietakangas et al., 2003). formation of large oligomers is necessary either for Therefore, we tested whether the HSP27 effect on HSF1 HSP27 nuclear translocation or for its retention within sumoylation is the result of an increase in the the nucleus. The latter hypothesis seems more plausible phosphorylation of HSF1 on S303. To study this, we as small oligomers of HSP27 although in small used an antibody specifically recognizing phosphory- quantities can also be found in the nucleus (mutant lated S303 and S307 (Hietakangas et al., 2003). The HSP27-Asp, Figure 5a and inputs of Figure 5c). There- amount of HSF1 phosphorylated on S303 and S307 fore, it is likely that large oligomers of HSP27 form was compared in control-transfected REG cells and within the nucleus (after import of either HSP27 REG cells expressing either HSP27-wt or HSP27-Ala monomers or small oligomers) and can no longer exit (Figure 6a). We found that the amount of HSF1 due to their size. phosphorylated on these serine residues was the same Heat shock factor 1 has already been reported to be in all transfected cells, even though only those expres- found in the nucleus together with SUMO proteins sing HSP27-wt and HSP27-Ala had an increase in the (Hietakangas et al., 2003). Here we found that after content of HSF1 modified by SUMO-2/3. This result stress SUMO-2/3 strongly accumulates in the nucleus of suggests that HSP27 favours HSF1 sumoylation down- HSP27-Ala-expressing cells or in the nucleus of cells stream of these phosphorylation events. expressing HSP27-wt (Figure 5b). We therefore con- Heat shock factor 1 DNA-binding and transactiva- clude that large oligomers of HSP27 accumulate in the tion activities have been shown to be regulated nucleus together with HSF1 and SUMO-2/3. independently (Anckar et al., 2006). We therefore first To show that HSP27 and HSF1 associate within the wondered whether HSP27-induced modification of nucleus where they accumulate with SUMO-2/3 under HSF1 by SUMO-2/3 could affect HSF1 DNA-binding stress conditions, we carried out co-imunoprecipitations activity. Binding of HSF1 to DNA was studied by experiments using nuclear and cytosolic fractions from means of a pull-down approach (Anckar et al., 2006). REG cells transfected with vectors for HSP27-wt, In this experiment, we used biotinylated double- HSP27-Ala or HSP27-Asp (Figure 5c). We detected no stranded oligonucleotides containing the consensus interaction between HSP27-wt and HSF1 in the nucleus HSE as bait. As shown in Figure 6b, HSF1 was equally of non-stressed cells. In contrast, a strong interaction pulled down whether extracts were from REG-stressed could be detected in the nuclear fraction of serum- (serum-depleted) control-transfected cells, or from depleted cells (Figure 5c). Similarly, a strong nuclear HSP27-wt-, HSP-27-Ala- or HSP27-Asp-transfected interaction was also observed in serum-depleted cells. These results suggest that HSP27 does not affect HSP27-Ala-expressing cells (data not shown). As HSF1 DNA-binding ability. This is in accordance with expected, no interaction was observed in the HSP27- previous results showing that modification by SUMO Asp fractions (Figure 5c). In the non-stressed cells, a did not significantly alter HSF1 DNA binding (Hieta- quantitatively less important interaction was observed in kangas et al., 2006). the cytosolic fractions from HSP27-wt- and HSP27-Ala- We next examined whether HSP27 could affect HSF1 expressing cells (Figure5c and not shown), but not from transcription factor activity. REG cells were transfected the HSP27-Asp-expressing cells (Figure 5c). Thus, in with either a control empty vector or vectors for HSP27- non-stressed cells, the interaction with cytosolic wt, HSP27-Ala or HSP27-Asp. These transfections were (inactive) HSF1 also involves large oligomers of carried out together with wild-type HSF1 or the HSP27. Although we still do not know the function sumoylation-deficient HSF1-K298R mutant and a of this cytosolic interaction, it is tempting to speculate luciferase reporter plasmid containing the proximal that HSP27 might help maintain HSF1 inert in HSE of human hsp70 promoter. The transfection per se non-stressed cells. was enough stress to induce HSF1 transactivation

Oncogene HSP27 facilitates HSF1 sumoylation M Brunet Simioni et al 3339 overlay overlay overlay overlay HSP27 HSF1 HSP27 HSP27-HSF1 nucleus HSP27 HSF1 nucleus nucleus HSF1 -nucleus

NT H S P 27 wt SD

H NT S P 27 A l SD a

H NT S P 27 A s p SD

SUMO-2/3 IP IgG IP HSP27 nucleus SUMO-2/3 nucleus HSP27-wt HSP27-Asp H cyto cyto nucleus cyto nucleus S NT P NT SD NT SD NT SD NT SD 2 WB HSF1 83 KDa 7 w SD INPUTS t IgG HSP27-wt HSP27-Asp H cyto cyto nucleus cyto nucleus S NT P NT SD NT SD NT SD NT SD 83 KDa 2 HSF1 7 A HSP27 27 KDa l SD a HSC70 70 KDa HSP60 60 KDa H Lamin B 70 KDa S NT P 2 7 A s SD p

Figure 5 HSP27 and heat shock factor 1 (HSF1) colocalize in the nucleus under stress conditions. (a) Fluorescence microscopy analysis of heat shock protein 27 (HSP27) (green) and HSF1 (red) in non-treated (NT) or 48-h serum-depleted (SD) REG cells expressing either HSP27- wild type (wt) or the mutants HSP27-Ala or HSP27-Asp. (b) Fluorescence microscopy analysis of SUMO-2/3 (green) in the cells and conditions described in (a). Nuclei, labelled with Hoechst 33342, are stained in blue. Magnification Â 300. (c) Immunoprecipitation with anti- HSP27 was followed by HSF1 immunodetection in nuclear and cytoplasmic extracts from control (NT) or 48-h SD REG cells expressing either HSP27-wt or HSP27-Asp. INPUTS: content of HSP27 and HSF1 in the cell extracts. Cytosolic/mitochondrial HSP60 and nuclear Lamin B were also immunodetected. The colour reproduction of the figure is available on the html full text version of the paper. A full colour version of this figure is available at the Oncogene journal online.

Oncogene HSP27 facilitates HSF1 sumoylation M Brunet Simioni et al 3340 IP IP HSF1 IgG

Co HSP27-wt HSP27-Ala Co

NT SD NT SD NT SD SD HSF1 P S303/7

HSF1 83 KDa t sp A - 7 a 2 Al P27-w - S SP 7 H HSP27-Ala Co H Scr

SP2 HSF1 83 KDa HSP27-Asp Co H HSP27-wt Scr

HSF1 HSC70 70 KDa

BOUND INPUTS Figure 6 Heat shock protein 27 (HSP27) expression does not affect heat shock factor 1 (HSF1) phosphorylation or DNA-binding ability. (a) HSF1 was immunoprecipitated in non-treated (NT) or 48-h serum-depleted (SD) REG cells control (Co) or expressing HSP27-wild type (wt) or HSP27-Ala. The amount of HSF1 phosphorylated in S303 or S307 in the precipitates was determined. (b) Extracts from 48-h serum-depleted REG cells expressing either a control vector (Co) or a vector for HSP27-wt, HSP27-Ala or HSP27-Asp were incubated with biotinylated double-stranded oligonucleotide containing the consensus heat shock element as bait. The amount of precipitated HSF1 after pull-down was immunodetected (BOUND). Scr, scrambled.

activity in control cells (transfected with an empty Concluding remarks vector, a luciferase reporter vector and either HSF1-wt or HSF1-K298R). HSF1 activity was increased when After stress, cells transiently express HSPs in a precise the cells were submitted to further stress (serum and coordinated manner: first, the early inducible HSPs depletion, Figure 7). These levels of transcriptional such as HSP70 and HSP90 and, then, the late inducible activity were not significantly altered in the cells ones like HSP27. HSPs, by tightly controlling their own expressing non-sumoylatable HSF1-K298R, indicating production at different levels, can act as sensors to that under these experimental conditions (low HSP27 determine the intensity and duration of the stress levels) sumoylation did not significantly regulate HSF1 response. HSP70 and HSP90 have already been reported transcriptional factor activity. In contrast, under condi- to control HSF1 activation. In non-stressed cells, HSP70 tions of expression of large oligomers of HSP27 (in cells and HSP90 bind to monomeric HSF1 and render it inert transfected with HSP27-wt and, still more, in cells in the cytosol (Shi et al., 1998; Voellmy and Boellmann, transfected with the HSP27-Ala mutant), HSF1 tran- 2007). During attenuation of the stress response, the scriptional activity was strongly reduced (Figure 7). transcriptional activity of HSF1 is repressed by direct The effect of HSP27-wt and HSP27-Ala decreasing binding of HSP70 and HSP90 to the HSF1 transactiva- HSF1 transcription factor activity was lost when we tion domain (Shi et al., 1998; Voellmy and Boellmann, expressed the HSF1 sumoylation mutant HSF1-K298R 2007). Here, we show for the first time the involvement (Figure 7). This shows that the regulation of HSF1 of a heat shock protein in the sumoylation process and transcription activity by HSP27 involves the sumoyla- particularly in HSF1 modification by SUMO-2/3. tion of the transcription factor. The expression of the HSP27, in the form of large oligomers, by inducing HSP27-Asp mutant did not have any measurable action HSF1 sumoylation also regulates the stress response in HSF1 transcription factor activity (Figure 7), again (Figure 8). Immediately after stress, it has been reported showing the absence of effect for the small multimers that HSP27 is transiently phosphorylated, which results of HSP27. in a shift towards small oligomers (Mehlen et al., 1995a) Altogether, these results indicate that accumulation in that, as it is shown in this study, are unable to induce the nucleus of large oligomers of HSP27 induces an SUMO-2/3 modification of HSF1. As a consequence, increase in HSF1 modification by SUMO-2/3, which HSP27 is freely transcribed right after stress. Later on, results in a decrease in its transactivating capacity. This likely when the amount of HSPs in general, and that of allows HSP27 to regulate its own content in the cells HSP27 in particular, is sufficiently high, HSP27 is with a negative feedback mechanism and to control the de-phosphorylated (Mehlen et al., 1995b) and re-forms stress response (Figure 8). The kinetics of HSF1 large oligomers. As shown here, these large forms of sumoylation and that of HSP27 accumulation after HSP27 accumulate within the nucleus, bind to HSF1 stress like heat shock confirm this model (Supplemen- and favour its modification by SUMO-2/3, which tary Figure 4). ultimately blocks the transcription of HSP genes

Oncogene HSP27 facilitates HSF1 sumoylation M Brunet Simioni et al 3341 2500 ** * * 2000 **

1500

1000

Luciferase activity (a.u.) 500

0 NT SD NT SD NT SD NT SD NT SD NT SD NT SD NT SD

Co HSP27-wt HSP27-Ala HSP27-Asp Co HSP27-wt HSP27-Ala HSP27-Asp

HSF1-wt HSF1 K298R Figure 7 Heat shock protein 27 (HSP27) affects heat shock factor 1 (HSF1) transcription factor activity. REG cells control (white bars) or expressing HSP27-wild type (wt) (black bars), HSP27-Ala (light grey bars) or HSP27-Asp (dark grey bars) were transfected either with vectors for HSF1-wt or HSF1-K298R. After transient transfection with an HSF1 luciferase reporter plasmid, the cells were either left untreated (NT) or serum-depleted (SD) for 48 h. Co-transfection with a thymidine kinase Renillea luciferase plasmid was used to normalize transfection efﬁciencies. a.u., arbitrary unit. Results are the means of three independent experiments. Bars indicate s.d. Statistic analysis: Student’s t-test. *Po0.05, **Po0.005.

Latent monomeric HSF1 HSP27 Stress insult

HSF1 HSP27 trimer large oligomers

HSP27-HSF1 interaction

Nucleus SUMO-2/3

hsp genes

HSE

Figure 8 Scheme representing how heat shock protein 27 (HSP27) may affect heat shock factor 1 (HSF1) activity. Under conditions of increased HSP27 expression (that is stress conditions), large oligomers of HSP27 accumulates in the nucleus. Most probably, these large multimers of HSP27 form within the nucleus after the import of small oligomers of HSP27 and are temporary trapped because of their size. Nuclear large oligomers of HSP27 interact with HSF1. This interaction favours HSF1 sumoylation by SUMO-2/3. As a result, HSF1 transcription factor activity is impaired. Therefore, HSP27 can modulate its own production with a feedback mechanism.

(Figure 8). However, it should be noted that, although SUMO-activating enzyme (Johnson, 2004). The SUMO less important, we can detect an interaction between moiety is subsequently transferred to Ubc9, the single cytosolic HSF1 and large oligomers of HSP27 under E2 SUMO-conjugating enzyme, which usually binds non-stressed conditions. This interaction should not directly to the target protein at the level of sumoylated result in increased HSF1 sumoylation as phospho- sites (Geiss-Friedlander and Melchior, 2007). Although rylation on S303, needed for HSF1 nuclear localization the enzymatic activity required for substrate modifica- and activation, is required for HSP27 to induce HSF1 tion can be carried out by the E1 and E2 enzymes alone modification by SUMO-2/3. HSP27 might bind to (Anckar et al., 2006), the E2-substrate interaction may inactive cytosolic HSF1 to maintain it inert, as it does be facilitated by SUMO E3 factors. The E3 factors for HSP70 or HSP90 (Shi et al., 1998; Voellmy and increase sumoylation efficiency in a substrate-specific Boellmann, 2007). manner, either by accelerating the transfer of SUMO SUMO conjugation utilizes a multistep enzymatic from Ubc9 to the substrate or merely by providing a pathway, in which proteolytically processed SUMO scaffold (Johnson, 2004). Therefore, our data raise the initially forms a thioester bond with Sae1/2, the E1 interesting possibility that the chaperone HSP27 might

Oncogene HSP27 facilitates HSF1 sumoylation M Brunet Simioni et al 3342 be an E3 factor as it facilitates HSF1 modification by 0.2% NP40 was added. Each sample was homogenized by SUMO-2/3 with some specificity (as HSP27 does not inversion and incubated on ice for 2 min. Cell lysate was affect sumoylation of a protein like c-Fos). The fact that centrifuged for 4 min at 400 g. The pellet was washed another HSP27 binds to mammalian Ubc9 (this paper) and to time in 1 ml buffer I, centrifuged for 4 min at 400 g. The pellet Drosophila Ubc9 (Joanisse et al., 1998) further suggests was recovered in buffer II (Tris 20 mM pH7.4; Na4P2O7 40 mM; that HSP27 may act like an E3 SUMO protein to permit MgCl2 5mM; NaF 50 mM;Na3VO4 100 mM; EDTA 10 mM; Triton 1%, SDS 1%), sonicated and centrifuged for 10 min at stronger interaction between Ubc9 and HSF1. Another 20 000 g. The supernatant was kept and immunoprecipitation interesting observation of this study is the fact that experiments were carried out as described below. HSP27 concerns modification by SUMO-2/3 but not For cytoplasmic extracts, 106 cells were carefully resus- that by SUMO-1. The mammalian SUMO protein pended in 400 ml of buffer I (Tris 10 mM pH7.4, NaCl 10 mM, family includes four members, of which only SUMO- MgCl2 3mM, phenylmethylsulphonyl fluoride 0.5 mM, dithio- 2/3 are conjugated in a stress-dependent manner. threitol 2 mM) and incubated on ice for 10 min. Then 0.2% A recent report in which the authors used a proteomic NP40 were added, Each sample was homogenized by return- approach to determine the proteins modified by SUMO- ment and incubated on ice for 2 min. Cell lysate was 2/3 shows that most targets do not contain a consensus centrifuged for 2 min at 10 000 g. The supernatant was sumoylation site (Blomster et al., 2009). Stress-inducible centrifuged another time for 2 min at 10 000 g and immunoprecipitation experiments were carried out as described above. chaperones might be involved in this consensus site- independent recognition mechanism of SUMO-2/3. A tight regulation of inducible HSP expression is Immunoprecipitation essential as an abnormal accumulation of these proteins Cellular proteins were extracted using immunoprecipitation can have pathological consequences. Clinical and buffer (50 mM HEPES pH 7.6; 150 mM NaCl; 5 mM EDTA; 0.1% Nonidet P-40 (NP-40), 20 mM N-ethyl maleimide). In all, experimental studies have showed that a constitutive 500 mg of proteins were incubated with 4 mg of the indicated accumulation of inducible HSPs (mainly HSP27 or antibody (anti-green fluorescent protein, anti-HSF1 and anti- HSP70) is associated with a poor prognosis and an HSP27 (TebuBio, Le Perray en Yvelines, France), anti- increase in the tumorigenic and metastatic potential of SUMO-2/3 and anti-SUMO-1 (Cell Signaling, St Quentin en tumour cells (Jaattela, 1995; Garrido et al., 1998; Bruey Yvelines, France)) with constant agitation at 4 1C overnight. et al., 2000; Gurbuxani et al., 2001; Westerheide et al., Then, immunocomplexes were precipitated with protein A/G- 2006; Didelot et al., 2007). It would be interesting to Sepharose (GE Healthcare, Orsay, France). determine whether defaults in the modulation of HSF1 sumoylation by HSP27 are involved in this abnormally Immunoblotting high HSPs expression characteristic of cancer cells. Cell were lysed in RIPA buffer (50 mM Tris–HCl pH 8, 150 mM NaCl, 0.5% sodium deoxycholate, 0.1% SDS, 1% NP40, protease inhibitor cocktail (Roche, Fontenay, France) and 20 mM N-ethyl maleimide) or in immunoprecipitation buffer. Materials and methods In all, 10–40 mg proteins were used for immunoblotting with the following antibodies: anti-HSP27, -HSP70, -HSF1 (Tebu- Cell culture and treatment Bio), -S303/7-phospho-HSF1 (kind gift of L Sistonen (Hieta- HeLa cells were cultured in Dulbecco’s modified Eagle’s kangas et al., 2003)), -SUMO-2/3, -SUMO-1 (Cell Signaling) medium (Sigma, St Quentin, Fallavier, France) containing and -HSC70 (Santa Cruz, Tebu Bio, Le Perray en Yvelines, 10% fetal calf serum (FCS). REG cells were grown in Ham’s France). F-10 medium (Sigma) supplemented with 10% FCS (Caignard et al., 1990). staurosporine was purchased from Sigma. Gel filtration Cells were lysed in gel filtration buffer (25 mM HEPES pH 7.5, Transfections, plasmids and siRNA dithiothreitol 0.2 mM, CHAPS 1% and 20 mM N-ethyl REG cells stably transfected with HSP27 phosphorylation maleimide). One mg protein of the supernatant was fractio- mutants were already described (Bruey et al., 2000). pcDNA3- nated by fast protein liquid chromatography using a Superose- HA-HSP27-wt plasmid codes for wild-type HA-tagged HSP27. 6 column (Pharmacia, Uppsala, Sweden) at a flow rate of pcDNA3-HA-HSP27-Ala and pcDNA3-HA-HSP27-Asp cor- 0.3 ml/min. Fractions of 1 ml were collected on ice and respond to the substitutions of HSP27 serine residues (serines concentrated using Amicon Ultra (Millipore, Saint Quentin 15, 78 and 82) by alanine or aspartate, respectively (provided en Yvelines, France). by M Gaestel, Hannover Medical School, Germany). HSP27 siRNA (50-AAAATCCGATGAGACTGCCGC-30) and luciferase siRNA (50-AACTTACGCTGAGTACTTCGATT-30) Immunofluorescence staining were purchased from Ambion (Applied Biosystems, Courta- Cells were fixed with 4% phosphate-buffered saline (PBS)– boeuf, France). Transfections were carried out following the paraformaldehyde and permeabilized by incubation with 3% manufacturer’s instructions with jetPEI reagent (Ozyme, St PBS–Triton X-100. After washing with PBS, samples were Quentin Yvelines, France) or Interferine (Ozyme) for siRNA. saturated with 3% PBS–bovine serum albumin (BSA) before overnight incubation at 4 1C with HSP27, HSF1 or SUMO-2/3 antibodies. After washes with 1% PBS–BSA, appropriate Cell fractionation secondary antibodies coupled with fluorochromes (Alexa Nuclear extracts were prepared from 106 cells carefully 486 nm and 568 nm; Molecular Probe, Leiden, the Netherland) resuspended in 800 ml of buffer I (Tris 10 mM pH7.4; NaCl were added. The nucleus was labelled with Hoechst 33342. 10 mM; MgCl2 3mM; phenylmethylsulphonyl fluoride 0.5 mM; Images were acquired using the Cell Observer Station (Zeiss, dithiothreitol 2 mM) and incubated on ice for 10 min. Then Le Pecq, France).

Oncogene HSP27 facilitates HSF1 sumoylation M Brunet Simioni et al 3343 Oligonucleotide pull-down assay Reverse transcriptase–PCR analysis In all, 300 mg of cell lysates containing 40 mM N-ethylmalei- RNA extraction was carried out with the Nucleospin RNA II mide were incubated with 0.5 mM annealed oligonucleotide in kit (Macherey-Nagel, Hoerd, France), and the OneStep reverse binding buffer (20 mM Tris–HCl pH 7.5, 100 mM NaCl, 2 mM transcriptase–PCR kit (Qiagen, Courtaboeuf, France) was EDTA, 10% glycerol). Salmon sperm DNA was added (0.5 mg/ used according to the manufacturer’s instructions, and the ml), and proteins were allowed to bind to the oligonucleotides following specific primers for the hsf1 gene (forward, for 30 min at room temperature. Samples were pre-cleared with 50-GGTCAAGCCAGAGAGAGACG-30; reverse, 50-CTCAT CL-4B Sepharose (Sigma) for 30 min at 4 1C, and the GCTTCATGGCCAGGA-30), the sumo-2 gene (forward, remaining DNA was precipitated with 15 ml of 50% slurry of 50-GAAAAGCCCAAGGAAGGAGT-30; reverse, 50-GTCT UltraLink streptavidin gel (Pierce, Rockford, IL, USA) for 1 h GCTGTTGGAACACATCA-30) with the human hprt gene as at 4 1C. After washing, DNA-bound proteins were eluted with a control (forward, 50-GGACAGGACTGAACGTCTTGC-30; denaturing buffer, followed by SDS-polyacrylamide gel reverse, 50-CTTGAGCACACAGAGGGCTACA-30). electrophoresis and immunoblotting using a polyclonal HSF1 antibody. The HSE-containing oligonucleotides used were: 50-biotin-AACGAGAATCTTCGAGAATGGCT-30 and 50-AGCCATTCTCGAAGATTCTCGTT-30, and the corre- Conflict of interest sponding scrambled control oligonucleotides 50-biotin-AAC GACGGTCGCTCCGCCTGGCT-30 and 50-AGCCAGGCG The authors declare no conflict of interest. GAGCGACCGTCGTT-30 (Invitrogen, Cergy Pontoise, France).

Luciferase reporter assays Acknowledgements Cells were harvested at 24 h after transfection and assayed using Promega’s Luciferase Assay System (Promega, Charbonnieres, We thank L Sistonen and J Anckar (Turku Centre for France). Cells were washed once in PBS, scraped into 100 mlof Biotechnology, Finland) for HSF1 and SUMO-2/3 tools, their reporter lysis buffer and rocked 15 min at 4 1C. A total of 20 mlof helpful advices and discussions. We thank M Gaestel and cell lysates were combined with 100 ml assay reagent and A Vertii for sharing with us HSP27 phosphorylation mutants’ luciferase activity was measured using a luminometer (LUMAT constructions. This work was supported by grants from the LB 9507, Berthold Technologies, Thoiry, France). To normalize ‘Ligue Nationale Contre le Cancer’ and its committees in the for transfection efficiency, the activity of co-transfected ‘Nie` vre’ and ‘Saône et Loire’. MB and ALJ are recipients of a b-galactosidase was also assayed. In all, 20 ml of cell lysates were doctoral fellowship from the ‘Ligue Nationale contre le combined with 100 ml of lysis buffer and 100 mlof2Â Cancer’, ADT has a postdoctoral fellowship from ‘L’Associa- b-galactosidase assay buffer. Samples were incubated at 37 1C tion pour la Recherche contre le Cancer’, and EF has an INCa for 30 min. The reaction was terminated by the addition of 1 M of financing. CG and MP lead teams ‘Labelliseés’ from the ‘Ligue Na2CO3 and the absorbance at 420 nm was recorded. Nationale contre le Cancer’.

References

Anckar J, Hietakangas V, Denessiouk K, Thiele DJ, Johnson MS, Didelot C, Lanneau D, Brunet M, Joly AL, De Thonel A, Chiosis G Sistonen L. (2006). Inhibition of DNA binding by differential et al. (2007). Anti-cancer therapeutic approaches based on sumoylation of heat shock factors. Mol Cell Biol 26: 955–964. intracellular and extracellular heat shock proteins. Curr Med Chem Anckar J, Sistonen L. (2007a). Heat shock factor 1 as a coordinator of 14: 2839–2847. stress and developmental pathways. Adv Exp Med Biol 594: 78–88. Didelot C, Schmitt E, Brunet M, Maingret L, Parcellier A, Garrido C. Anckar J, Sistonen L. (2007b). SUMO: getting it on. Biochem Soc (2006). Heat shock proteins: endogenous modulators of apoptotic Trans 35: 1409–1413. cell death. Handb Exp Pharmacol 172: 171–198. Ayaydin F, Dasso M. (2004). Distinct in vivo dynamics of vertebrate During RL, Gibson BG, Li W, Bishai EA, Sidhu GS, SUMO paralogues. Mol Biol Cell 15: 5208–5218. Landry J et al. (2007). Anthrax lethal toxin paralyzes actin- Blomster HA, Hietakangas V, Wu J, Kouvonen P, Hautaniemi S, based motility by blocking Hsp27 phosphorylation. EMBO J 26: Sistonen L. (2009). Novel proteomics strategy brings insight into 2240–2250. the prevalence of SUMO-2 target sites. Mol Cell Proteomics 8: Garrido C, Brunet M, Didelot C, Zermati Y, Schmitt E, Kroemer G. 1382–1390. (2006). Heat shock proteins 27 and 70: Anti-apoptotic proteins with Bossis G, Malnou CE, Farras R, Andermarcher E, Hipskind R, tumorigenic properties. Cell Cycle 5: 2592–2601. Rodriguez M et al. (2005). Down-regulation of c-Fos/c-Jun AP-1 Garrido C, Fromentin A, Bonnotte B, Favre N, Moutet M, Arrigo AP dimer activity by sumoylation. Mol Cell Biol 25: 6964–6979. et al. (1998). Heat shock protein 27 enhances the tumorigenicity Bossis G, Melchior F. (2006). Regulation of SUMOylation by of immunogenic rat colon carcinoma cell clones. Cancer Res 58: reversible oxidation of SUMO conjugating enzymes. Mol Cell 21: 5495–5499. 349–357. Geiss-Friedlander R, Melchior F. (2007). Concepts in sumoylation: Bruey JM, Paul C, Fromentin A, Hilpert S, Arrigo AP, Solary E et al. a decade on. Nat Rev Mol Cell Biol 8: 947–956. (2000). Differential regulation of HSP27 oligomerization in tumor Gill G. (2003). Post-translational modiﬁcation by the small ubiquitin- cells grown in vitro and in vivo. Oncogene 19: 4855–4863. related modiﬁer SUMO has big effects on transcription factor Caignard A, Genne P, Olsson NO, Oriol R, Martin F. (1990). F11C activity. Curr Opin Genet Dev 13: 108–113. antigen: A membrane marker able to distinguish two regressive and Guo D, Han J, Adam BL, Colburn NH, Wang MH, Dong Z et al. progressive variants from a rat colon adenocarcinoma. Int J Cancer (2005). Proteomic analysis of SUMO4 substrates in HEK293 cells 46: 633–639. under serum starvation-induced stress. Biochem Biophys Res Charette SJ, Lavoie JN, Lambert H, Landry J. (2000). Inhibition of Commun 337: 1308–1318. Daxx-mediated apoptosis by heat shock protein 27. Mol Cell Biol Gurbuxani S, Bruey JM, Fromentin A, Larmonier N, Parcellier A, 20: 7602–7612. Jaattela M et al. (2001). Selective depletion of inducible HSP70

Oncogene HSP27 facilitates HSF1 sumoylation M Brunet Simioni et al 3344 enhances immunogenicity of rat colon cancer cells. Oncogene Mehlen P, Kretz-Remy C, Briolay J, Fostan P, Mirault ME, Arrigo 20: 7478–7485. AP. (1995a). Intracellular reactive oxygen species as apparent Hay RT. (2005). SUMO: a history of modification. Mol Cell 18: 1–12. modulators of heat-shock protein 27 (hsp27) structural organization Hietakangas V, Ahlskog JK, Jakobsson AM, Hellesuo M, Sahlberg and phosphorylation in basal and tumour necrosis factor alpha- NM, Holmberg CI et al. (2003). Phosphorylation of serine 303 is a treated T47D human carcinoma cells. Biochem J 312(Pt 2): 367–375. prerequisite for the stress-inducible SUMO modification of heat Mehlen P, Mehlen A, Guillet D, Preville X, Arrigo AP. (1995b). shock factor 1. Mol Cell Biol 23: 2953–2968. Tumor necrosis factor-alpha induces changes in the phosphoryla- Hietakangas V, Anckar J, Blomster HA, Fujimoto M, Palvimo JJ, tion, cellular localization, and oligomerization of human hsp27, a Nakai A et al. (2006). PDSM, a motif for phosphorylation- stress protein that confers cellular resistance to this cytokine. J Cell dependent SUMO modification. Proc Natl Acad Sci USA 103: Biochem 58: 248–259. 45–50. Parcellier A, Brunet M, Schmitt E, Col E, Didelot C, Hammann A Holmberg CI, Hietakangas V, Mikhailov A, Rantanen JO, Kallio M, et al. (2006). HSP27 favors ubiquitination and proteasomal Meinander A et al. (2001). Phosphorylation of serine 230 promotes degradation of p27Kip1 and helps S-phase re-entry in stressed cells. inducible transcriptional activity of heat shock factor 1. EMBO J FASEB J 20: 1179–1181. 20: 3800–3810. Pirkkala L, Nykanen P, Sistonen L. (2001). Roles of the heat shock Holmberg CI, Tran SE, Eriksson JE, Sistonen L. (2002). Multisite transcription factors in regulation of the heat shock response and phosphorylation provides sophisticated regulation of transcription beyond. FASEB J 15: 1118–1131. factors. Trends Biochem Sci 27: 619–627. Rogalla T, Ehrnsperger M, Preville X, Kotlyarov A, Lutsch G, Hong Y, Rogers R, Matunis MJ, Mayhew CN, Goodson ML, Park- Ducasse C et al. (1999). Regulation of Hsp27 oligomerization, Sarge OK et al. (2001). Regulation of heat shock transcription chaperone function, and protective activity against oxidative stress/ factor 1 by stress-induced SUMO-1 modification. J Biol Chem 276: tumor necrosis factor alpha by phosphorylation. J Biol Chem 274: 40263–40267. 18947–18956. Jaattela M. (1995). Over-expression of hsp70 confers tumorigenicity to Saitoh H, Hinchey J. (2000). Functional heterogeneity of small mouse fibrosarcoma cells. Int J Cancer 60: 689–693. ubiquitin-related protein modifiers SUMO-1 versus SUMO-2/3. Joanisse DR, Inaguma Y, Tanguay RM. (1998). Cloning and J Biol Chem 275: 6252–6258. developmental expression of a nuclear ubiquitin-conjugating Schmitt E, Gehrmann M, Brunet M, Multhoff G, Garrido C. (2007). enzyme (DmUbc9) that interacts with small heat shock proteins Intracellular and extracellular functions of heat shock proteins: in Drosophila melanogaster. Biochem Biophys Res Commun 244: Repercussions in cancer therapy. J Leukoc Biol 81: 15–27. 102–109. Shi Y, Mosser DD, Morimoto RI. (1998). Molecular chaperones as Johnson ES. (2004). Protein modification by SUMO. Annu Rev HSF1-specific transcriptional repressors. Genes Dev 12: 654–666. Biochem 73: 355–382. Voellmy R, Boellmann F. (2007). Chaperone regulation of the heat Lambert H, Charette SJ, Bernier AF, Guimond A, Landry J. (1999). shock protein response. Adv Exp Med Biol 594: 89–99. HSP27 multimerization mediated by phosphorylation-sensitive Westerheide SD, Kawahara TL, Orton K, Morimoto RI. (2006). intermolecular interactions at the amino terminus. J Biol Chem Triptolide, an inhibitor of the human heat shock response that 274: 9378–9385. enhances stress-induced cell death. J Biol Chem 281: 9616–9622. Lee YJ, Kim EH, Lee JS, Jeoung D, Bae S, Kwon SH et al. Xiao X, Zuo X, Davis AA, McMillan DR, Curry BB, Richardson JA (2008). HSF1as a mitotic regulator: Phosphorylation of et al. (1999). HSF1 is required for extra-embryonic development, HSF1 by Plk1 is essential for mitotic progression. Cancer Res 68: postnatal growth and protection during inflammatory responses in 7550–7560. mice. EMBO J 18: 5943–5952. McMillan DR, Xiao X, Shao L, Graves K, Benjamin IJ. (1998). Zhu J, Zhu S, Guzzo CM, Ellis NA, Sung KS, Choi CY et al. (2008). Targeted disruption of heat shock transcription factor 1 abolishes Small ubiquitin related modifier (SUMO) binding determines thermotolerance and protection against heat-inducible apoptosis. substrate recognition and paralog-selective SUMO modification. J Biol Chem 273: 7523–7528. J Biol Chem 283: 29405–29415.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Oncogene