SUMMARY
Cells respond to protein-damaging insults by activating heat shock factors (HSFs), key transcription factors of proteostasis. Abnormal HSF protein levels occur in cancer and neurodegenerative disorders, highlighting the importance of the tight control of HSF expression. HSF2 is a short-lived protein, but it is abundant in the prenatal brain cortex and required for brain development. Here, we reveal that HSF2 is acetylated and co-localized with the lysine-acetyl transferases CBP and EP300 in human brain organoids. Using unbiased, biochemical, cell-imaging, and in silico approaches, we show that CBP/EP300 acetylates HSF2 at specific lysine residues, which promotes HSF2 stabilization, whereas the lysine deacetylase HDAC1 catalyzes its proteasomal degradation. The CBP KIX domain and KlX-recognizing motifs in HSF2 are critical for its interaction with acetylating enzymes. The functional importance of acetylated HSF2 is evidenced in Rubinstein-Taybi syndrome (RSTS), characterized by mutated CBP or EP300. We show that RSTS patient cells exhibit decreased HSF2 levels and impaired heat shock response. The dysregulated HSF pathway in RSTS opens new avenues for understanding the molecular basis of this multifaceted pathology.
INTRODUCTION
Since their discovery three decades ago, our way to envision the regulation and roles of the Heat Shock transcription Factor family (HSFs) has been revolutionized. Originally identified and characterized due to their stress-responsiveness and ability to recognize a consensus DNA-binding site, the heat shock element (HSE), HSFs were more recently shown to perform an unanticipated large spectrum of roles under physiological and pathological conditions (Wu, 1995; Abane and Mezger 2010; Akerfelt et al., 2010; Pastor-Gomez et al., 2018). HSFs are activated by a diversity of stressors that provoke protein damage and govern the highly conserved Heat Shock Response (HSR). The HSR contributes to the restoration of proteostasis, through the regulation of genes encoding molecular chaperones, including the Heat Shock Proteins (HSPs; Hartl et al., 2011). HSFs also control immune/inflammatory pathways, metabolism, and, through dysregulation of their protein levels or activity, shape disease susceptibility to cancer, metabolic and neurodegenerative disorders. These pathophysiological roles are performed through altered expression of a broad repertoire of target genes, beyond the HSPs (Xiao et al., 1999; Inouye et al., 2007; Dai et al., 2007; Mendillo et al., 2012; Santagata et al., 2013; Anckar and Sistonen, 2011; Jin et al., 2011; Neef et al., 2011; Nakai, 2016; Pastor-Gomez et al., 2017 and 2018). The multifaceted roles of HSFs are achieved by their fascinating plasticity in terms of multi-modular structure and assembly of homo-or heterodimers or trimers, stress- and context-dependent posttranslational modifications as well as a diversity of partner networks. As a consequence, HSFs act as fine sculptors of transcriptomic and epigenetic landscapes, through dynamic interactions with other transcriptional activators or repressors and chromatin remodelling complexes (Akerfelt et al., 2010; Miozzo et al., 2015; Pastor-Gomez et al., 2018; Raychaudhuri et al. 2014).
The versatile functions of HSFs have been mostly studied with HSF1 and HSF2, two members of the mammalian HSF family, which in human comprises four additional members, i.e. HSF4, HSF5, HSFX and HSFY (Pastor-Gomez et al., 2018). The role of HSF1 in acute and severe proteotoxic stress, including exposures to elevated temperatures (42-45°C), has been extensively documented and become a paradigm for the modus operandi of the HSF family. In contrast, HSF2 appears to be responsive to stresses of relevance for chronic or pathological situations, such as fever-like temperatures at 39-41°C (Shinkawa et al., 2011), alcohol (ethanol) exposure (El Fatimy et al., 2014; Miozzo et al., 2018), and prolonged proteasome inhibition (Lecomte et al., 2010; Rossi et al., 2014). Both factors have been involved in cancer, in a dual manner. HSF1 acts as a potent, multifaceted facilitator of cancer initiation and progression (Dai et al., 2007; Mendillo et al., 2012; Santagata et al., 2013), whereas HSF2 expression can ensure a protective role, by counteracting tumor progression and invasiveness (Björk et al. 2016). The control of HSF1 protein levels is key for its role in cancer, and elevated HSF1 expression is associated with poor prognosis (Santagata et al., 2011) while its decreased expression is associated with neurodegenerative disorders (Kim et al., 2015; Jiang et al., 2013; Pastor-Gomez et al., 2017; reviewed in Pastor-Gomez et al., 2018). HSF1 and HSF2 are also important players in the physiological brain development and adult brain integrity. Furthermore, we and others have shown that the deregulation of HSF activities underlies neurodevelopmental defects (Kallio et al., 2002; Wang et al., 2003; Chang et al., 2006; El Fatimy et al., 2014; Hashimoto-Torii et al., 2014; Ishii et al., 2017; reviewed in Abane et Mezger 2010, Akerfelt et al., 2010, Pastor-Gomez et al., 2018) and neurodegenerative processes (Shinkawa et al., 2011; Pastor-Gomez et al., 2017). The complexity of the role of HSF2 and its impact on pathophysiological situations, including cancer, neurodevelopmental and neurodegenerative diseases, thus raises questions about the regulatory mechanisms of HSF2 expression.
The amount of the HSF2 protein varies in diverse cellular or embryonic contexts and conditions: both transcriptional and post-transcriptional mechanisms have been indicated to regulate HSF2 mRNA levels (Rallu et al., 1997; Björk et al., 2010). In the developing brain, the amount of HSF2 protein seems to correlate with that of Hsf2 transcripts, underlining the importance of transcriptional control (Rallu et al., 1997; Kallio et al., 2002; Wang et al., 2003). Moreover, HSF2 is a short-lived protein and its stabilization constitutes an important step controlling its DNA-binding activity (Sarge et al., 1993; Mathew et al., 1998; Kawazoe et al., 1998) and mediating its role in physiological processes and stress responses. HSF2 protein levels also fluctuate during the cell cycle and its stabilization is an important control step in fine-tuning the HSR (Elsing et al., 2014). Indeed, HSF2 modulates the stress-inducible expression of HSP genes, which is primarily driven by HSF1 (Östling et al., 2007). This transient modulatory function of HSF2 is due to the rapid poly-ubiquitination and proteasomal degradation in response to acute heat stress (Ahlskog et al., 2010). While diverse posttranslational modifications (PTMs), such as phosphorylation and acetylation, are known to control HSF1 stability (Kourtis et al., 2015; Pastor-Gomez et al., 2017; Raychaudhuri et al., 2014; reviewed in Pastor-Gomez et al., 2018), the mechanisms regulating stability of HSF2 are not well understood, and given its role in chronic stress, cancer, and physiopathological developmental processes, they are crucial to be elucidated.
In this study, using both candidate and unbiased approaches, we demonstrate a specific role for acetylation in controlling HSF2 stability, under normal growth and stress conditions. We identify the histone/lysine-acetyl transferases (HATs/KATs) CBP (CREB-binding protein; KAT3A) and EP300 (KAT3B) as key enzymes to catalyze HSF2 acetylation. We show that HSF2 is acetylated during normal brain development in human organoids and mouse cortices, where it co-localizes with CBP and EP300. We identify three main HSF2 lysine residues, whose acetylation by CBP/EP300 is critical for its stability. In contrast, HDAC1 that we identify as an HSF2 partner, by TAP-TAG and mass spectrometry, promotes HSF2-proteasomal degradation. We identify the KlX-recognition motifs in the HSF2 oligomerization domain and show that they are key for interaction with the KIX domain of the CBP protein. We also demonstrate the functional impact of HSF2 acetylation and stabilization by CBP/EP300, in a rare disease the Rubinstein-Taybi syndrome (RSTS; Lopez-Atalaya et al., 2014; Spena et al., 2015a). RSTS is characterized by mutations in the CREBBP (CBP; RSTS1, OMIM #180849) or EP300 (RSTS2; OMIM #613684) genes, with multiple congenital anomalies, neurodevelopmental defects, childhood cancer susceptibility, and vulnerability to infections. We observe that cells derived from RSTS patients exhibit marked destabilization of the HSF2 protein, which is rescued by proteasomal inhibition. These cells also display altered basal expression of HSPs and impaired HSR. The disruption of the HSF pathway in RSTS highlights the importance of the CBP/EP300 dependent regulation of HSF2 by acetylation and provides a new conceptual frame for understanding the molecular basis of this complex and multifaceted pathology.
RESULTS
HSF2 is acetylated and interacts with CBP/EP300 in the developing brain
The levels of the HSF2 protein are highly variable among cell types and organs (Abane and Mezger, 2010). In particular, HSF2 is remarkably abundant in the vertebrate developing brain, where it exhibits spontaneous DNA-binding activity, in normal non-stressed conditions (Rallu et al., 1997; Kawazoe et al. 1999; Kallio et al., 2002; Wang et al., 2003; Chang et al., 2006). As a first step to determine whether HSF2 acetylation was involved in controlling HSF2 stability similarly to HSF1 acetylation by EP300 (Raychaudhuri et al., 2014), we compared HSF2 and CBP/EP300 expression profiles, and investigated whether HSF2 acetylation could be detected in the developing mammalian cortex under physiological conditions. Indeed, CBP and EP300 are transcriptional coactivators that interact with a large number of transcription factors (TFs) and have key roles in neurodevelopment (reviewed in Chan and La Thangue 2001; Lopez-Atalaya et al., 2014).
To the best of our knowledge, the expression of the HSF2 protein in the human developing cortex has not been reported. By generating brain organoids from human induced pluripotent stem cells (iPSCs; Lancaster et al., 2014), we found that HSF2 was expressed at different stages, from day 20 (embryoid bodies) to day 60 of differentiation (D20 – D60), as shown in mRNA and protein analyses (Figure 1A and S1A). By immunofluorescence on D60 organoids, we found that HSF2 was expressed in neural progenitor cells (NPCs; located in areas of dense DAPI-staining, as verified by SOX2 staining, Figure S1B) and in neurons, expressing beta-III tubulin, in regions displaying a cortical-like morphology (Figure 1B and S1B). In addition, we observed co-labeling of EP300 and HSF2 in NPCs (arrowheads) and neurons (arrows) (Figure 1C). Thus, HSF2 and EP300 (also CBP; see Figure S1C) thus exhibited similar expression territories (NPCs and neurons; Figure S1B,C). In addition, HSF2, CBP and EP300 were expressed, in a concomitant manner, in the mouse cortex from E11 to E17 (Figure S1D; note that CBP and EP300 expression patterns have been previously reported in the mouse embryonic brain, mostly at early stages; Kawasaki et al. 1998; Yao et al., 1998; Partanen et al., 1999; Bhattacherjee et al; 2009). The similarity of their expression patterns supported a possible interaction between HSF2 and the CBP and EP300 in the developing cortex. Accordingly, HSF2 was co-immunoprecipitated with CBP and EP300 (Figure 1D and Figure S1E, left panels). We also detected acetylated HSF2 in the developing mouse cortex (Figure 1E and Figure S1E, right panels) and in D40 human brain organoids (Figure 1F). Similarly, HSF2 was found acetylated in SHSY-5Y neuroblastoma cells, a human cancer cell line of neural origin (Figure S1F). Altogether, these results show that, HSF2 is acetylated and interacts with CBP and EP300 in the developing brain.
Analysis of CBP/EP300-mediated HSF2 acetylation in mammalian cell lines
In order to explore the mechanism and function of HSF2 acetylation, we first verified whether HSF2 was a novel substrate for acetylation by CBP/EP300, in human in HEK 293 cells, coexpressing CBP-HA or EP300-HA and tagged HSF2. Using GFP-or Myc-Trap (Rothbauer et al., 2008) assay, we found that the immunoprecipitated exogenous HSF2-YFP or HSF2-Myc protein was acetylated by EP300 or CBP (Figure 2A) and that no acetylation was observed in cells transfected by dominant-negative CBP (Figure 2B).
To identify the acetylated lysine residues in HSF2, we co-expressed Flag-HSF2 with EP300-HA in HEK 293 cells. HSF2 was immunoprecipitated and the acetylation of lysines was detected by mass spectrometry (MS). Among the 36 lysine residues of HSF2, we identified eight acetylated lysines: K82, K128, K135, K197, K209, K210, K395, and K401 (Figure 2C, Figure S2A, and Table S1). Single point mutations (K82, K128, K135, and K197), or mutation of the doublet K209/K210 (Figure S2B) to arginine (R), which prevent acetylation, did not abolish HSF2 acetylation, suggesting that, in line with our MS data, the acetylation of HSF2 occurs on more than one lysine residue. Indeed, the mutation to either arginine (R) or glutamine (Q) of three or four lysines (K82, K128, K135 and K197), located within the DNA-binding domain and the first hydrophobic heptad repeat (HR-A/B), dramatically reduced HSF2 acetylation (Figure 2D and Figure S2C). Using an in vitro acetylation assay coupled with HPLC, we found that synthetic HSF2 peptides containing K135 and K197 residues were readily acetylated by recombinant purified CBP-HA, in an acetyl-CoA-dependent manner (Figure 2E and Figure S2D-F), but not a peptide containing the K82 lysine. Note that we could not perform this experiment on the HSF2 peptide containing K128, due to its insolubility. Taken together, our data suggest that HSF2 is acetylated by CBP/EP300 at three major lysine residues, residing in the HR-A/B domain: K128, K135 and K197.
Interaction of HSF2 with the core catalytic domain of CBP
Prompted by the finding that a catalytically active CBP is necessary for HSF2 acetylation, we examined whether HSF2 could bind to the core catalytic domain of CBP. The CBP central core catalytic region (“Full-HAT”; Figure 3A) contains the Bromodomain BD, the cysteine/histidine-rich region CH2, and the HAT domain and allows the coupling of substrate recognition and histone/lysine acetyltransferase (HAT) activity (as in EP300; Delvecchio et al., 2013; Dancy and Cole, 2015; Dyson and Wright 2016). The CH2, in particular, contains a RING domain and the PHD (plant homeodomain, a non-canonical zinc finger; Park et al., 2013). With Biolayer interferometry, we observed that the recombinant Full-HAT domain directly interacted with immobilized biotinylated recombinant full-length HSF2 (Figure 3B). Within this region, the recombinant PHD domain, but not the HAT, RING or BD domain, was able to interact with HSF2, in a similar manner as the “Full-HAT” domain (Figure 3B). Interestingly, the interaction of HSF2 with the Full-HAT or the PHD domain was more efficient than with HSP70, which has been reported to interact with HSF2 (Huttlin et al., 2015; Tang et al., 2016). As expected, it is likely that the interaction between HSF2 and the catalytic HAT domain was too transient to be captured in these experiments, because the structural integrity of the HAT domain is dependent on the BD and CH2 domains. The latter includes the PHD domain which is involved in protein interaction and in the case of CBP is considered an integral part of the enzymatic core HAT domain, being critical for HAT activity (Aasland et al., 1995; Bordoli et al., 2001; Kalkhoven et al, 2002). We determined that the KD of HSF2 interaction with the CBP Full-HAT domain was 1.003E-09 M (+/-2.343E-11; R2= 0.988488; Figure S3A). The interaction of HSF2 with the Full-HAT domain, and the PHD within, together with our data on HSF2 acetylation strongly suggest that HSF2 is a bona fide substrate of CBP, and potentially of EP300 since their HAT domains display 86% identity.
HSF2 interacts with CBP and EP300 in cellulo
Next we investigated whether HSF2 can interact with CBP or EP300 in cellulo. Using HEK 293 cells in GFP-Trap assay, we showed that CBP-HA or EP300-HA co-immunoprecipitated with HSF2-YFP (Figure 3C,D). We also observed that endogenous CBP and EP300 co-precipitated with endogenous HSF2 in the murine neuroblastoma N2A cells (Figure S3B). We confirmed these interactions by an independent technique, the fluorescent three-hybrid assay (F3H; Figure 3E-G; Herce et al., 2013), using GFP-binder and HSF2-YFP, together with CBP-HA or EP300-HA. In negative control experiments, GFP-binder, which was recruited to the LacOp array locus, was unable to recruit endogenous CBP, CBP-HA, or EP300-HA (Figure S3C-E). Similarly, HSF2-YFP was unable to locate at the LacOp array locus in the absence of GFP-binder (Figure S3C-E). Co-transfection of HSF2-YFP with CBP-HA or EP300-HA resulted in the formation of a red spot in the nucleus, showing co-recruitment to the HSF2-YFP focus, in 68.4 % and 48.6 % of the cells, respectively (Figure 3F, upper panels and 3G). The abundance of CBP in the BHK cells allowed us to detect the co-recruitment of HSF2-YFP with endogenous CBP in 43.9% of these cells (Figure 3F, lower panels).
The HR-A/B domain of HSF2 is key for interaction with CBP and acetylation
To determine which HSF2 domains were important for its interaction with CBP, we expressed Flag-HSF2, either full-length or deletion mutants of different domains, in HEK 293 cells. The deletion mutants of different domains, we showed that the deletion of HR-A/B (or DBD plus HR-A/B, but not DBD only) led to a marked decrease in HSF2 acetylation (Figure S4A-C; see red arrow) and interaction with CBP (Figure S4D). These results were in line with our findings that the major acetylated lysine residues reside within the HR-A/B domain. The deletion of the TAD (transcription activation domain; Jaeger et al., 2016) also resulted in decreased acetylation of the Flag-HSF2 (Figure S4B-C) and was associated with decreased interaction with CBP (Figure S4D). This domain may also be important for interaction between CBP and HSF2, since the multivalent interactions between CBP/EP300 and many TFs generally involve their TADs (Lee et al., 2009; Wang et al., 2012; reviewed in Thakur et al., 2013).
The presence of KIX motifs in the HSF2 HR-A/B domain promotes binding to the CBP KIX domain
The interaction between HSF2 and the core catalytic domain is likely facilitated/stabilized via specific domains of these two proteins. CBP/EP300 interacts with many transcription factors via different binding sites, including the KIX domain (kinase-inducible domain interacting domain; Figure A, left). The KIX domain contains two distinct binding sites that are able to recognize the “ΦXXΦΦ” KIX motif, where “Φ” is a hydrophobic residue, and “X” is any amino acid residue “X”. Importantly, we found that the HR-A/B domain of HSF2 contains several conserved, overlapping and juxtaposed KIX motifs (Figure 4A). We verified, using recombinant proteins that HSF2 was co-immunoprecipitated with the CBP KIX domain, in in vitro GST-pull-down experiments (Figure 4B). We then modeled the interaction between the HSF2 HR-A/B KIX motifs and the CBP KIX domain. Based on sequence similarities between the HR-A/B domain, lipoprotein Lpp56, and the transcription factors GCN4, ATF2, and PTRF (Figure S4E; see Experimental procedures), we first developed a structural model of the HSF2 trimeric, triple coiled-coil, HR-A/B domain (Jaeger et al., 2016). Next, we evaluated the possibility of interactions of the KIX recognition motifs in the HR-A/B region with the CBP KIX domain. Best poses suggested that the HR-A/B KIX motif region contacted the c-Myb surface (Figure 4C), thereby proposing a close interaction of the HSF2 KIX motifs with tyrosine Y650 (Figure 4D). We assessed the impact of in silico mutations of the Y650 amino-acid of the CBP KIX domain or K177, K180, F181, V183 residues in the HSF2 KIX motifs, involved in the contact with CBP (Figure 4E). Interestingly, the in silico mutation of CBP residue Y650A profoundly decreased the probability of interaction of the HSF2 KIX motifs with the KIX domain (Figure 4F, upper panel; and Figure S4F (b). We also found that the Y650A mutation disrupted HSF2 and KIX interaction, in GST-pull-down experiments (Figure 4F, lower panel). Interestingly, this mutation has been identified in RSTS, and associated with a severe neurodevelopmental phenotype (Spena et al., 2015b). In addition and in constrast to F181A or V183A mutation (Figure S4F (c,d)), the K177A or Q180A mutation within the HSF2 KIX motifs also disrupted HSF2-KIX interaction (Figure S4F (e,f)), highlighting the importance of these residues in these interactions
The acetylation of HSF2 governs its stability in non-stress conditions
We then explored the functional impact of the CBP/EP300-mediated acetylation of HSF2. Endogenous HSF2 protein levels were decreased upon pharmacological inhibition of CBP/p300 activity in N2A cells, using the specific inhibitor C646 (Bowers et al., 2010; Dancy and Cole, 2015; Figure 5A and Figure S5A). This decrease was abolished by treatment with the proteasome inhibitor, MG132, suggesting that the decrease in the HSF2 protein levels was dependent on the proteasomal activity, and that HSF2 was degraded when CBP/EP300 activity was inhibited (Figure 5A). This observation provided the first evidence for acetylation playing a regulatory role in HSF2 stability. To further investigate the role of acetylation in the regulation of HSF2 protein levels, we generated CRISPR/Cas9 Hsf2KO U2OS cell lines (2KO; Figure S5B-F) and measured the protein levels of wild type HSF2 and HSF2 acetylation mutants, which mimic either constitutively acetylated (3KQ) or non-acetylated (3KR) HSF2 (Figure 5B-D). We first verified that the HSF2 WT, and HSF2 3KQ and HSF2 3KR were expressed at comparable levels (Figure S5D), and capable of binding DNA in in vitro and its binding sites on repeated pericentromeric Satellite III DNA (Sat III), in ex vivo experiments (Figure S5E,F). To monitor the decay of a pre-existing pool of HSF2 molecules, we performed pulse-chase experiments using the SNAP-TAG technology (Bodor et al., 2012). A pool of SNAP-HSF2 molecules was covalently labeled by adding a fluorescent substrate to the cells. At t0, a blocking non-fluorescent substrate was added, quenching the incorporation of the fluorescent substrate to newly synthesized HSF2 molecules (Figure 5B), allowing us to measure the decay in the fluorescence intensity of the corresponding labeled HSF2 bands. When 2KO cells were transfected with wild-type SNAP-HSF2 (SNAP-HSF2 WT), the decay in fluorescence intensity of the corresponding bands was observed within 5 hours (Figure 5C and D). Preventing HSF2 acetylation in cells transfected with SNAP-HSF2 3KR resulted in a similar decay (Figure 5C and D). In contrast, mimicking acetylation in cells expressing SNAP-HSF2 3KQ protected HSF2 from decay (Figure 5C and D). Importantly, proteasome inhibition with MG132, prevented the decrease in SNAP-HSF2 WT and SNAP-HSF2 3KR fluorescent intensity, strongly suggesting that HSF2 decay was dependent on the proteasomal activity (Figure 5E).
HDACl/2 is involved in destabilization of the HSF2 protein under non-stress and stress conditions
To identify the enzymes that could function as deacetylases for HSF2, we performed an unbiased screen for HSF2 binding partners, using a double-affinity TAP-TAG approach (Bürkstümmer et al., 2006; Figure S6A-C). For this purpose, we generated a HeLa-S3 cell line expressing doubletagged HSF2 (or transfected with the empty vector as a negative control) and analyzed nuclear extracts by MS. We identified HDAC1 as one of the protein partners of HSF2 (Figure 6A). In addition to HSF2, we found nucleoporin Nup62 (Figure 6A), an earlier identified HSF2 partner that provided a positive control for the quality of our TAP-TAG/MS analysis (Yoshima et al., 1997). We also performed immunoprecipitation of HSF2 in extracts from the mouse E17 cortices, a stage at which HSF2 protein levels are decreased (see Figure S1D), followed by MS analysis. We identified both HDAC1 and HDAC2 as HSF2 partners (Figure S6D). Using the F3H approach (Figure 6B and Figure S6E), and co-immunoprecipitation in GFP-Trap assays (Figure 6C), we confirmed the interaction between HSF2 and HDAC1 in mammalian cell lines. We then observed that preventing the acetylation of HSF2 by expressing Myc-HSF2 3KR increased the polyubiquitination of HSF2, compared to HSF2 WT or HSF2 3KQ (Figure 6D). For evaluating the impact of HDAC1 and other Class I HDACs on the acetylation of HSF2, we expressed tagged HDAC1, HDAC2, HDAC3, or HDAC8, concomitantly with CBP-HA and Flag-HSF2 in HEK 293 cells (Figures 6E and S6F). HDAC1 overexpression resulted in marked reduction in HSF2 acetylation levels, whereas HDAC2 and HDAC3 had only limited effects, and HDAC8 no obvious impact (Figures 6E and S6F).
Because heat shock (HS) provokes the degradation of the HSF2 (Ahlskog et al., 2010), we also examined the role of acetylation on the turnover of the HSF2 protein upon HS. The impact of acetylation on the decay of HSF2 upon HS analyzed using the SNAP-TAG technology. Mimicking the acetylation of the three major acetylated lysine residues mitigated the decay of fluorescence intensity of SNAP-HSF2 3KQ induced by HS, compared to SNAP-HSF2 WT or 3KR (Figure 6F). The impact of HDAC inhibition on endogenous HSF2 was investigated in neural N2A cells. We first verified that HS induced decay in HSF2 protein levels also in neural N2A cells, although it occurred at a slower rate (Figure S6G) when compared to HeLa or HEK 293 cells (Ahlskog et al., 2010). Treatment with 1 mM of the Class I inhibitor VPA dampened the decline in HSF2 protein levels in N2A cells upon HS (Figure 6G). This suggested that Class I HDAC activity participates to the degradation of HSF2 upon HS, likely through HSF2 deacetylation (Figure 6G). We then verified that HS increased HSF2 polyubiquitination in HEK293 cells, as previously reported (Ahlskog et al. 2010; Figure 6H, Mock). To investigate whether HDAC1 could favor HSF2 polyubiquitination, likely through HSF2 deacetylation, we examined the impact of overexpression of a dominant-negative form of HDAC1 on HSF2 ubiquitination. We showed that, indeed, the increase in HSF2 polyubiquitination upon HS was mitigated in HEK 293 cells transfected with dominant-negative HDAC1 (Figure 6H). Altogether, these results support a role of HDAC1 (and possibly other Class I HDACs) in the destabilization of HSF2 under normal and stress conditions, through HSF2 polyubiquitination and proteasomal degradation. In summary, the acetylation status of HSF2 thus appears to play a key role for the stabilization of the HSF2 protein, by protecting it from proteasomal degradation.
Declined HSF2 protein levels in the Rubinstein-Taybi Syndrome (RSTS)
To evaluate the impact of CBP and EP300 on HSF2 levels, in a pathological context, we compared the amounts of HSF2 protein in cells derived from either healthy donors (HD) or RSTS patients, which are characterized by autosomal-dominant (heterozygous) mutations in the CBP or EP300 genes, leading to neurodevelopmental defects and intellectual disabilities (see Figure S7A for a description of the mutations). We used human primary skin fibroblasts (hPSFs), at early passages (see Experimental Procedures), to avoid the putative compensation processes during ex vivo culture. We observed that HSF2 protein levels were markedly decreased in RSTS patients mutated in either CBP or EP300 (RSTSCBP, patient 1 [PI] and RSTSEP300, patient 2 [P2]; Figure 7A-C). These levels were restored to the levels that are comparable in healthy donors (HD) when hPSFs had been treated with the proteasome inhibitor MG132 (Figure 7A and B). Note that we verified the impact of CBP or EP300 mutations in RSTS hPSFs, by showing that the amount of acetylated lysine residue K27 in histone H3 (AcH3K27) was reduced in both cases (Figure 7A and C; right panels). Interestingly, Class I HDAC inhibition by VPA could not restore HSF2 levels in RSTSCBP or RSTSEP300 hPSFs, although HD and RSTS cells displayed similar levels of HDAC1 (Figure S7D). This suggests that, in RSTS hPSFs, CBP and EP300 are key for the stabilization of HSF2 (Figure 7A; Figure S7B). Notably, the stability of HSF2 was impaired both in the RSTSCBP patient carrying a mutation in the catalytic domain of CBP and in the RSTSEP300 patient carrying a deletion of the KIX domain of EP300, which was shown to be a binding site for HSF2 as shown by our results (Figure 4C). This suggests that both domains are required for the regulation of HSF2 stability, in line with our results (Figure 3 and 4). Altogether these results demonstrate that the proteasomal turnover of HSF2 is increased in RSTS hPSFs, carrying mutated EP300 or CBP, thereby strongly suggesting that EP300 and CBP are key regulators of HSF2 protein stability.
Impaired heat shock response in RSTS cells
HSF1 is the essential driver of the acute heat shock response (HSR) in mammals (McMillan et al., 1998). Although dispensable for the HSR in most cellular contexts (McMillan et al., 1998), HSF2 acts as a fine tuner of the HSR (Östling et al., 2007; Elsing et al., 2014), which governs the magnitude to which HS induces HSP gene expression. Therefore, we evaluated the ability of RSTS cells to mount a HSR. In the absence of heat stress, we observed that RSTS hPSFs displayed lower amounts of HSP70 and HSP90 than HD counterparts (Figure 7D). Furthermore, RSTSEP300 hPSFs exhibited limited capacity in inducing HSP70 accumulation upon HS and during the recovery phase from heat stress (Figure 7D). Notably, this limited induction did not result from impairment of HSF1 activation, since HSF1 seemed activated by HS in RSTSEP300 hPSFs, as assessed by its mobility shift in SDS-PAGE (see arrow head and arrow in Figure 7D). This shift is a hallmark of HSF1 hyperphosphorylation, which, although not required for HSF1 activation, accompanies the induction of HSF1 transactivation potential (Sarge et al., 1993; Budzyński et al., 2015; reviewed in Anckar and Sistonen, 2011;). HSF1 and HSF2 do not only control the transcription of the Hsp genes, but they also upregulate the transcription of Sat III (Sat III) pericentromeric heterochromatin regions, in response to acute heat stress. This process results in the accumulation of non-coding RNAs, predominantly at the Satlll 9ql2 locus, which participates to the formation of specific subnuclear structures, called the nuclear stress bodies (nSBs; Jolly et al, 1997, 1999, 2004; Rizzi et al., 2004; Sandqvist et al., 2009). We therefore used nSBs as read-out for the HSR integrity in RSTS cells and observed that the stress-inducible formation of nSBs was reduced by more than 50% in RSTSEP300 hPSFs when compared to their HD counterparts (Figure 7E). A similar reduction in the formation of nSBs was observed in RSTSEP300 or RSTSCBP lymphoblastoid cells (LBs; Figure S7F).
DISCUSSION
HSFs exert a wide spectrum of key pathophysiological roles, whose unravelling has aroused an exciting wave of reawakened interest during the last decade and emphasized the importance of the tight control of HSF protein levels: excessive or insufficient HSF expression favors the development or progression of devastating diseases, including cancer, neurodevelopmental, and neurodegenerative disorders.
In this study, we have unraveled a novel mechanism that regulates the stability of the labile HSF2 protein, in different cellular contexts: the KATs CBP and EP300, control the acetylation of three highly conserved lysine residues K128, K135, and K197, located in the HR-A/B oligomerization domain, which are mainly responsible for HSF2 acetylation, and thereby contribute to the stability of the HSF2 protein. Moreover, we demonstrate the importance of this regulation which operates both in normal and stress conditions, in the pathological conditions of a rare disease.
HSF2 is acetylated by CBP/EP300 in various contexts including brain development
Based on our analysis, HSF2 is acetylated by the overexpression CBP or EP300 in cell systems. Moreover, we observe the acetylation of the endogenous HSF2 protein in human and murine neural embryonic tissues and cell lines (Figure 1 and S1), showing that HSF2 acetylation is not restricted to one specific cell context. CBP and EP300 and, more generally, HATs play significant roles during neurodevelopment (reviewed by Lopez-Atalaya et al., 2014). HSF2, CBP and EP300 exhibit very similar expression patterns along the organoid differentiation of human brain organoids, as they do in the developing mouse cortex. This is, to our knowledge, the first assessment of the expression profiles of HSF2, CBP, and EP300 in such models of the developing human brain and of interactions between HSF2 and CBP/EP300. These results point out the putative importance of HSF2 acetylation in physiological contexts and suggesting that the acetylation of HSF2 might participate to its high levels of expression and putatively role in cortical development. In line with this, note that we identified HDAC1 by MS analysis as an HSF2 partner in E17 cortices, a stage at which the HSF2 protein levels are markedly lowered (Figure S7; Rallu et al., 1997; El Fatimy et al., 2014).
Mode of HSF2 interaction with CBP/EP300
Endogenous or exogenous HSF2 interacts with CBP and EP300 in cell systems and in the developing brain cortex, as mentioned above. Our study unravels domains important for HSF2 acetylation per se, and for the anchorage of HSF2 to CBP/EP300. First, the full-length HSF2 protein interacts with the CBP core catalytic domain in vitro, confirming that HSF2 is a bona fide substrate of CBP. Interestingly, HSF2 also strongly interacts with the PHD domain, located in the catalytic core of the CBP/EP300 proteins (Delvecchio et al., 2013). Interestingly, mutations in the PHD domain have been identified in RSTS patients (Kalkhoven et al., 2003).
Second, we find that the HR-A/B oligomerization domain, but not the DBD, is necessary for HSF2 interaction with CBP and acetylation. We show that the HR-A/B domain specifically interacts with the KIX domain of CBP/EP300 in in silico analysis.
The KIX domain present in CBP and EP300 serves as a docking site for the binding of many transcription factors (TFs) that contributes to the properties of CBP/EP300 to act as a molecular bridge, stabilizing the interactions between TFs and the transcription apparatus (Parker et al., 1996; reviewed in Thakur et al., 2014). We find that the HR-A/B oligomerization domain contains a series of KlX-binding motifs that are necessary for HSF2 interaction with CBP and HSF2 acetylation. We show that these motifs specifically interact with the KIX domain of CBP/EP300 by in silico analysis. The in silico analyses indicate that the HR-A/B KIX motifs in HSF2 bind to the c-Myb site of the KIX domain. Indeed, the CBP or EP300 KIX domain can simultaneously - and in a cooperative manner – binds two polypeptide ligands, on two distinct surfaces, which have been historically called the “c-Myb” and the “MLL” (Mixed Lineage Leukemia protein) sites (reviewed in Thakur et al., 2014). For example, the activation domains of MLL and CREB, MLL and c-Myb, or Jun and CREB, can concomitantly bind to KIX, each TF contacting a different KIX surface, namely, the c-Myb or the MLL site (Goto et al., 2002; Campbell and Lumb, 2002). This suggests that the binding of another TF via the “MLL site” might potentially modulate HSF2 and CBP interaction, via the KIX domain.
In support of the close interaction between the HSF2 KIX recognition motifs and the CBP KIX domain, we show that the mutation of the tyrosine residue Y650 in the CBP KIX domain, which contacts the HSF2 KIX motifs, disrupts HSF2-CBP interaction in in silico and in vitro experiments.
We also identify the HSF2 TAD (Jaeger et al., 2016) as potentially important for HSF2 interaction with CBP. This result is reminiscent of other TF TADs that bind the KIX domain (reviewed in Thakur et al., 2014). Moreover, two different domains of the same TF (like p53 or FOXO3a; Lee et al., 2009; Wang et al., 2012) can simultaneously bind the KIX domain. It renders possible that HSF2 could simultaneously interact with CBP (and most likely EP300), through two distinct domains, the HR-A/B and the TAD domains. In addition, the di-or trimeric coil-coiled structure of HSF2 might also broadens the possibility of establishing multiple contacts with CBP, through KIX or other CBP domains, known to interact also with TADs, which paves the way for future studies.
The KIX domain of CBP/EP300 has a key role in the formation and stabilization of vital multiprotein complexes during transcriptional activation. Our data mining on two data sets generated in K562 cells: our HSF2 ChlP-Seq data (Vihervaara et al., 2013) and publically available EP300 ChlP-Seq data (ENCODE UCSC Accession: wgEncodeEH002834)) indicate that HSF2 and EP300 share common binding regions on the genome (data not shown), including two known HSF2 target genes: HSP70 and UBB. We have already validated UBB as an HSF2 target gene in K562 cells (Vihervaara et al., 2013) and EP300 is recruited to HSP70B, in stress and brain development contexts (Xu et al., 2008; Marinova et al., 2009; Westerheide et al., 2009; Takii et al., 2014). This suggests that the interaction of HSF2 and CBP/EP300, which is important for the acetylation and stabilization of HSF2, can also occur on DNA, and might also contribute to the sculpting of the epigenetic and transcriptomic landscape of HSF2 target genes.
Dynamics of HSF2 acetylation by CBP/EP300, deacetylation by HDAC1, and degradation
Based on our data, the acetylation of HSF2 by CBP/EP300 limits its proteasomal degradation, as it does for other TFs, like p53, STAT3, or HIF1alpha (Grossman, 2001; Jain et al., 2012; reviewed in Yang and Seto, 2008; Geng et al., 2012). Acetylation does not seem to act on HSF2 stability by preventing the polyubiquitination of the three lysine residues, K128, K135, and K197. Indeed, only combined mutations of these lysines to 3KQ, but not to 3KR, prevent HSF2 proteasomal degradation. In addition, 3KR mutations increase HSF2 polyubiquitination, whereas 3KQ does not (Figure 6D). Previous proteome-wide quantitative analyses of the ubiquitin-modified protein have revealed that HSF2 ubiquitination occurs on multiple residues spanning over the HSF2 protein, in addition to K128, K135, and K197, including K51, K151, K210 and K420. Most of them reside in the HR-A/B domain or its vicinity, suggesting a crosstalk between acetylation and ubiquitination (Kim et al., 2011; Wagner et al., 2011; Akimov et al., 2018; www.Dhosdhosite.org; though, note that these studies have not investigated the functional impact of these ubiquitination events on HSF2 turnover; reviewed by Gomez-Pastor et al., 2018). In addition, we showed that the acetylation of the lysine residues K128, K135, and K198 also limits the degradation of HSF2, in heat shock conditions.
We identify HDAC1 as one major HDAC involved in HSF2 deacetylation and the control of HSF2 proteasomal degradation, in basal and HS conditions, which may provide an explanation for the quick degradation of HSF2 upon HS by APC/C (Ahlskog et al., 2010). In particular, the expression of a dominant-negative HDAC1 prevents the induction of HSF2 polyubiquitination by HS. The other Class I HDACs, HDAC2 and HDAC3 might also be at play. We do not exclude that Class III HDACs, in particular SIRT1, which deacetylates HSF1, might be involved in HSF2 deacetylation, but treatment with NAM, a sirtuin inhibitor, gave inconsistent results in our hands. Finally, the interaction of HSF2 with CBP/EP300, or with HDACs is likely highly dynamics, as is its acetylation status.
We also do not exclude that other HATs might be involved in the regulation of HSF2 stability. For example, HSF1 proteasomal turnover and DNA-binding activity are regulated by EP300 or GCN5 (Westerheide et al., 2009; Zelin et al., 2012; Raychaudhuri et al., 2014). Notably, different HSF1 lysine residues are involved in these two distinct functions (reviewed by Miozzo et al., 2015). We do not know whether HSF2 acetylation might potentially regulate HSF2 activity. We did not observe overt modification in the DNA-binding activity of HSF2 3KR and 3KQ mutants, which prevent or mimic HSF2 acetylation on K128, K135, and K198, respectively (Figure S5E and F). However, the acetylation of these residues might have subtler effects, and that of others might influence HSF2 DNA-binding and function, including interaction with other partners (transcriptional co-activators or repressors), which would deserve future studies.
HSF2 destabilization and impaired heat shock response in RSTS syndrome
The accelerated turnover of the HSF2 protein in RSTS primary cells, which is counteracted by proteasome inhibition, confirms the importance of HSF2 interaction with CBP and EP300, and constitutes another indication for the role of acetylation on its stability.
RSTS is a rare genetic, autosomal dominant NDD, characterized by intellectual disability, multiple congenital anomalies, including heart and skeleton malformations, and elevated susceptibility to infections and childhood cancers (Spena et al., 2015). The two identified genes mutated of deleted in RSTS respectively represent 60% for CBP and 8-10% for EP300, of clinically diagnosed cases. The presence of one mutated allele of CBP or of EP300 is sufficient to provoke this very disabling disease. This is surprising given that, not only a wild-type copy of the affected gene, but also the two wild-type alleles encoding the other protein (CBP or EP300) are present in the patients. The mutations in RTS patients are extremely diversified and can affect the protein expression, the protein-protein binding or the catalytic activity, which results in loss of specificity toward interacting partners and/or the substrates. The mutated allele via a rupture of this subtle equilibrium could thereby exerts a “dominant-negative” effect and and compromises compensation by the wild-type copies of the other KAT3 (Merk et al., 2018; Lopez-Atalaya et al., 2014). Accordingly, even though we neither observe reduced levels of CBP in our RSTSEP300 hPSFs, nor of EP300 in our RSTScsp hPSFs (Figure 7A-C), the presence of a catalytic inactive CBP or deleted KIX domain EP300 allele is sufficient to impact HSF2 proteasomal turnover in our system (Figure 7A-C). Likewise, several lines of evidence suggest that the presence of a dominant-negative CBP or EP300 allele is sufficient to impact HSF2 proteasomal turnover, in these cells. Indeed, the mutated CBP or of EP300 allele, despite the presence of the other CBP (or, conversely, EP300) wild-type alleles, seems sufficient to compromise HSF2 stability and lead to its proteasomal degradation. This suggests that the presence of the dominant-negative mutated allele impairs HSF2 acetylation. In addition, VPA is unable to restore HSF2 levels in RSTS cells, likely because HSF2 is not acetylated in these cells. This reinforces the hypothesis that the integrity of EP300 and CBP function is key to the control of HSF2 levels in these cells.
HSF2 is a modulator of the HSR by fine-tuning the expression of HSP genes and the formation of nSBs (Östling et al., 2007; Sandqvist et al., 2009). This occurs through co-binding of HSF2 and HSF1, via the formation of heterotrimers, to the regulatory regions of the HSP genes and Satlll loci (Alastalo et al., 2003; Östling et al., 2007; Sandqvist et al., 2009).
First, in the RSTS context, HSF2 seems majorly deregulated (HSF1 availability being only slightly affected), and we find that the basal HSP70 and HSP90 protein levels are markedly reduced. Second, in response to HS, the magnitude of induction of HSP70 is impaired, and HSF2 thus seems to play an important role in the HSR in the PSFs cells, in line with the work of Östling et al. (2007). Moreover, we also observe that the reduction of HSF2 levels in RSTS cells affects the formation of nSBs. It is therefore possible that the HSR could be regulated at two different levels by the proteasome, providing an exquisite and sophisticated way to control cell proteostasis, via the stabilization of: 1) HSF1 in an EP300-dependent manner (at least in some cell systems; Raychaudhuri et al., 2014); and/or 2) HSF2, in CBP- and EP300-dependent manner (here in hPSFs). The balance between these two arms of regulation (HSF1 or HSF2-driven) could be tipped by the cell context, and potentially the HSF1/HSF2 ratio (Östling et al., 2007; Sandqvist et al., 2009).
In conclusion, equipment in chaperones is altered in RSTS cells in unstressed conditions, in a chronic manner, and in stressed conditions, this alteration might confer those cells vulnerability to proteostasis challenge.
The dysregulation of the HSF pathway in RSTS, which includes marked reduction of HSF2 levels, might have several implications for this multifaceted disease. Indeed, RSTS is a neurodevelopmental disorder and HSF2, as a TF involved in neurodevelopment both in normal and stress conditions, might contribute to RSTS neurodevelopmental defects, through reduction of its availability in RSTS patient cells. Second, RSTS patients show of extreme vulnerability to airway infections, which is mainly due to deficit in mounting a response to polysaccharides (Naimi et al., 2006; Herriot et al., 2016) and because the HSF pathway is involved in response to polysaccharides, inflammatory and immune response and lung protection against stress, its deregulation could also contribute to this deficit (Xiao et al., 1999; Inouye et al., 2007; Wirth et al., 2003). Therefore, any imbalance in the delicate composition of the HSP repertoire in normal conditions and the triggering of HSF-driven stress-responses could participate to this vulnerability. Globally, the deregulation of the HSF pathway, which controls brain development, tumor initiation and progression might constitute an interesting novel reading key for this complex disease.
EXPERIMENTAL PROCEDURES
CONTACT FOR REAGENT AND RESOURCE SHARING
More detailed information and requests for resources and reagents should be directed to and will be fulfilled by the co-corresponding authors: Aurélie de THONEL (jrelie.dethonel{at}univ-paris-diderot.fr), Lea SISTONEN (lea.sistonen{at}btk.fi), and Valérie MEZGER (valerie.mezger{at}univ-apris-diderot.fr).
Reagents and treatments
Proteasome inhibitor MG132 was used for 6h (N2A, U2OS) at a final concentration of 20μM. HDACs inhibitor VPA (Interchim, AYJ060) was used at 1 mM for 3 h. The HAT inhibitor C646 (Sigma-Aldrich; SML0002) was used at a final concentration of 20 or 40 μM for 4h. For all chemicals, DMSO was used as vehicle (control).
Heat shock treatments were performed in water bath at 42 or 43°C for the indicated times.
Plasmid Table and constructs
The human HSF2-Snap (WT/mutants) were constructed from the HSF2-Myc (WT/mutants) plasmid after digestion of the inserts by EcoRI and Kpnl and cloning into the EcoRI and EcoRV sites in frame with the C-terminal Flag tag in pSNAPf plasmid using In-Fusion Kit (Clontech). The human HSF2-YFP was constructed by PCR and cloned into the Xhol and Sail sites in frame with the N-terminal YFP tag in EGFp-C1 plasmid using In-Fusion Kit (Clontech). All PCR-amplified products for both plasmids were sequenced to exclude the possibility of second site mutagenesis. The cDNA coding for the acetyltransferase domain of murine CBP (1097-1774) was a kind gift of Pr. Ricardo Dalla-Favera (Columbia University, New York) and was used to generate cDNA coding for key domains of CBP: Full HAT (1096-1700), HAT (1322-1700), RING (1205-1279), PHD (1280-1321), Bromodomain (1096-1205), later sub-cloned in pet28a plasmid (Invitrogen) in order to produce 6 His-tagged proteins.
Patient material
Informed consent for skin biopsy and culture of PSFs were obtained from the RSTS patients’parents (Patients 1 and 2, [P1] AND [P2]; CHU, Robert Debré Hospital, Paris, France; Drs S; Passemard and A. Verloes), in accordance with the Declaration of Helsinki and approved by the local ethics committee, CPP lie de France, agreement n° P10012 8. PSFs from healthy donors have been described in Yehezkel et al., 2008). PSFs were grown in HAM’s F10 supplemented with 12% FBS in humidified atmosphere with 5 % CO2 at 37°C. See Figure S7A for a description of the deletion or mutation carried by the patients.
Cell lines
Cell culture, transfections and treatments: Murine N2A (N2A, DSMZ # ACC 148), Hamster BHK, Human fibroblasts (HD), HEK293T (ATCC®, CRL-11268™), U2OS (osteosarcoma, ATCC®, HTB-96™), U2OS-CrisprH5F2K0 (2KO), BHK cells (kindly provided by Dr. Leonhardt H and cultured as described (Herce et al., 2013) and HeLa-S3 (kindly provided by Dr. Slimane Ait-Si-Ali and cultured as described ((Yahi et al., 2008)) cells were grown in DMEM (Lonza Group Ltd.) supplemented with 4,5 g/L glucose and 10% fetal bovine serum (FBS, Life technology) in humidified atmosphere with 5 % C02 at 37°C. RSTS lymphoblastoid cells were obtained from Dr. D Lacombe (CHU Bordeaux; patient 3 (P3)) and from CRB-lnstitut Médical Jérôme Lejeune, CRB-BioJeL (BB-0033-00016; May 05, 2018, Paris; patients 4 and 5 (P4 and P5)). Lymphoblastoid cells were grown in RPMI (Life technology) supplemented with 4,5 g/L glucose and 10% FBS with L-glutamine (Life technology) in humidified atmosphere with 5 %
Mice model
Specific pathogen-free C57BL/N female mice were purchased from Janvier (Lyon, France) and maintained in sterile housing in accordance with the guidelines of the Ministère de la Recherche et de la Technologie (Paris, France). Rodent laboratory food and water were provided ad libitum. Experiments were performed in accordance with French and European guidelines for the care and use of laboratory animals. The project has been approved by the Animal Experimentation Ethical Committee Buffon (CEEA-40) and recorded under the following reference with the Ministère de I’Enseignement Supérieur et de la Recherche (#2016040414515579).
Patients’ primary fibroblasts
For fibroblasts primary cells, The study was conducted in accordance with the Declaration of Helsinki with an approved written consent form for each patient (CPP ESTI: 2014/39; N°ID: 2014-A00968-39), and approval was obtained from the local ethics committee ESTI (license: NCT02873832).
Generation of CRISPR/Cas9 Hsf2KO U2OS cells
Guide RNAs (gRNA) targeting the exon 1 of HSF2 were designed using our own software (http://crispor.tefor.net/) and cloned into pMLM3636 guide RNA expression plasmid (Addgene 43860). U2OS cells were transfected with Cas9 and guide RNA expression plasmids using Amaxa electroporation as recommended by the manufacturer (Lonza). Cas9 expression plasmid was from the Church lab (Addgene 41815). One week after transfections, cells were seeded at single cell density. Clones were genotyped by DNA sequencing of PCR products spanning the targeted region of the HSF2 gene. The selected U2OS clone presented 3 different outframe mutations on HSF2 exon 1, each corresponding to a different allele (Supplementary Table 2). Guide RNA sequence targeting the 1st AUG on HSF2 exon 1: 5’-UGCGCCGCGUUAACAAUGAA-3’. Primers used for PCR cloning for validation: forward (hHSF2_Cr_ATG_F): 5’-AGTCGGCTCCTGGGATTG-3’ and reverse (hHSF2_Cr_ATG_R): 5’-AGTGAGGAGGCGGTTATTCAG-3’. The genomic sequences of the mutated hHSF2 alleles and the resulting lack in HSF2 protein and HSF2 DNA-binding activity are described in Table in Figure S5B. The selected U2OS clone presented 3 different out-frame mutations in the HSF2 exonl, each one corresponding to a different allele (see Figure S5B).
Production of KIX-GST, His-CBP domain and HSF2 proteins
Escherichia coli 21 (DE3) were transformed with the different 6His-tag CBP constructs for production of the different CBP domains as previously described (Duval et al, 2015). All proteins were stored in 20 mM Tris-HCl, 150 mM NaCl, pH 7.5 and kept at -80°C until use. E. coli BL21 bacteria were transformed with the different GST-KIX constructs (Cook et al., 2011) and grown in presence of ampicilline and chloramphenicol at 37°C (4-6h). Bacteria were then grown with 1 mM Isopropyl β-D-1-thiogalactopyranoside over-night. After centrifugation (4400 rpm at 4°C), bacteria were lysed in PBS pH8, 300mM NaCl, Triton X100 1%, lmg/mL lysozyme, protease inhibitors under stirring at 4°C for 30min. Bacteria were sonicated (BRANSON sonicator, power 20%, 10” ON/20’?FF) and centrifugated at 4°C, 16000g for 30 min. Gluthatione sepharose 4B beads (G4510-10ML Sigma-Aldrich) were added to the cleared supernatants and subjected to rotation for lh30 min at 4°C. The mixture was loaded into a column (Sigma-Aldrich) and washed with PBS/NaCl 300mM pH8, then Tris 50 mM/NaCl 150 mM pH8. Proteins were eluted with 5mL of elution buffer (50mM Tris HCl pH8, 150 mM NaCl, 10 mM GSH). The protein concentration was measured using Bradford method. In vitro transcription and translation reactions were performed using a TNT T7 coupled reticulocyte lysate system as recommended by the supplier (Promega, Charbonnieres-les-Bains, France). 1 μg of the plasmid DNA template was transcribed and the protein was translated at 30°C for 90 min.
Immunoprecipitation and Western blotting
Protein extracts from cells were prepared using a modified Laemmli buffer (5% sodium dodecyl sulphate, 10 % glycerol, 32.9 mM Tris-HCl pH 6.8) supplemented with protease inhibitors (Sigma-Aldrich). Brain tissues were prepared with a lysis buffer (Hepes 10 mM ph 7,9; NaCl 0.4M, EGTA 0.1 M; glycerol 5 %, DTT 1 mM, PMSF 1 mM, protease inhibitor (Sigma-Aldrich), phosphatase inhibitor (Roche)). Then, 30 μg of proteins from lysates were subjected to migration on 8-12% acrylamide gels and transferred on to polyvinylidene difluoride membranes (GE Healthcare Europe GmbH) in borate buffer (50 mM Tris-HCl and 50 mM borate) for 1h45 at constant voltage (48 V). The membranes were incubated with primary antibodies overnight at 4 °C, then washed in Tris-buffered saline–Tween 0.1% and incubated for 1 h with horseradish peroxidase (HRP)-coupled secondary antibody (Jackson Immunoresearch). The signal was revealed using a chemiluminescent reagent (Pierce® ECL Plus Western Blotting Substrate, Thermo Scientific) and was detected using hyperfilm (HyperfilmTM ECL, Amersham Biosciences) and a film processor (Konica Minolta). Poly-ubiquitinated HSF2 was detected as described in Ahlskog et al., 2010.
Biolayer Interferometry
For in vitro protein-protein interaction experiments, we used biolayer interferometry technology (Octet Red, Forté-Bio,USA). Recombinant HSF2 (TP310751 Origen) was desalted (ZebaTM Spin Desalting Columns, 7K molecular-weight cutoff, 0.5 ml (1034–1164, Fisher Scientific, Germany)) and biotinylated at a molar ratio biotin/protein (3:1) for 30 min at room temperature (EZ-Link NHS-PEG4-Biotin (1189-1195, Fisher Scientific, Germany)). Excess Biotin was removed using ZebaTM Spin Desalting Columns. Biotinylated recombinant HSF2 was used as a ligand and immobilized at 100 nM on streptavidin biosensors after dilution in phosphate-buffered saline (PBS; 600s). Interactions with desalted analytes diluted in PBS at 100 nM (recombinant CBP domains 6His-tag Full HAT, Bromodomain (BD), RING), or HSP70 as a positive control (ADI-SPP-555, Enzo-Life Sciences)) were analysed after association (600 s). All sensorgrams were corrected for baseline drift by subtracting a control sensor exposed to running buffer only.
For immunoprecipitation of exogenous proteins, using GFP-Trap
GFP-Trap®-A (ChromoTek) contains a small recombinant fragment of alpaca anti-GFP-antibody, covalently coupled to the surface of agarose beads. It enables purification of any protein of interest fused to GFP, eGFP, YFP, CFP or Venus. HEK 293 cells were transfected by a combination YFP-or Myc-tagged hHSF2 and HA-tagged EP300, CBP (WT or DN) or GFP-tagged HDAC1, or mock vector, with XtremGENE HP Reagent (Sigma-Aldrich) following manufacturer’s instructions. Cells were lysed in Lysis buffer (50 mM Hepes pH 8, 100 mM NaCl, 5 mM EDTA, Triton X-100 0.5 %, Glycerol 10 %, VPA (1 mM), DTT 1 mM, PMSF 1 mM, proteases inhibitors, phosphatase inhibitors (Roche)) and then, HSF2 was immunoprecipitated using anti-GFP-trap antibody. Immunoprecipitated proteins were run on a 8 % SDS-polyacrylamide gel, followed by an immunodetection of CBP or EP300 protein using anti-HA antibody. The amount of immunoprecipitated HSF2 was determined after reblot of the IP membrane with an anti-GFP or anti-Myc antibody. The amount of HSF2 and CBP or EP300 proteins, in the input samples, were detected with anti-GFP or Myc and anti-HA antibodies, respectively.
For immunoprecipitation of endogenous proteins
Brain cortices or organoids, or cells (SHSY5Y, N2a) were lysed 30min in Lysis buffer A (25 mM Hepes pH 8, 100 mM NaCl, 5 mM EDTA, Triton X-100 0,5%, VPA (1 mM), PMSF 1 mM, proteases inhibitors, phosphatase inhibitors (Roche)). After centrifugation (15 min, 12 000 g) and preclearing, cell lysates were subjected to immunoprecipitation overnight using an anti-mouse HSF2 (Santa-Cruz) and a non relevant IgG (Sigma-aldrich) as a negative control that were pre-incubated 1h at RT with Protein G UltraLink Resin beads (53132, Pierce). Protein complexes were then washed 4 times in wash buffer (25 mM Tris HCl pH 7.5,150 mM NaCl, 1 mM EDTA, Triton X-100 0.1 % Glycerol 10 %, 1 mM VPA, 1 mM PMSF, proteases inhibitors, phosphatase inhibitors (Roche)), and suspended in 2x Laemmli buffer. After boiling, the immunoprecipitates were resolved in 8 % SDS-PAGE and immunoblots were performed using an antirabbit pan acetyl-Lysine, anti-mouse HSF2 (Santa-Cruz), EP300 (Santa-Cruz) and CBP (CST). The amount of HSF2 and CBP or EP300 proteins in the input samples were detected with anti-mouse HSF2 and anti-rabbit CBP (CST) or EP300 (Santa-Cruz) antibodies.
Purification of HSF2 from HEK293 cells and mass spectrometry analysis of acetylated lysines
The protocole used is the same as for HSF1, as described by Westerheide et al., 2009. HEK 293 cells were transfected with mouse HSF2-beta Flag with or without CMV-EP300, treated with trichostatin A or nicotinamide as indicated 18 hours prior harvesting and lysed in RIPA buffer. HSF2-Flag was immunoprecipitated using α-Flag M2 affinity gel beads (Sigma F2426) and eluted with Flag peptide. Purified mHSF2-Flag was separated by SDS-PAGE, excised from the gel, digested with trypsin, and subjected to tandem mass spectrometric analysis by a hybrid quadrupole time-of-flight instrument (QSTAR, Applied Biosystems, Foster City, CA) equipped with a nanoelectrospray source. MS/MS spectra were searched against the IPI mouse sequence database (68,222 entries; version 3.15) using Mascot (Matrix Science, Boston, MA; version 1.9.05) and X! Tandem (www.thegpm.org; version 2006.04.01.2) database 4 search algorithms. Mascot and X! Tandem were searched with a fragment and precursor ion mass tolerance of 0.3 Da assuming the digestion enzyme trypsin with the possibility of one missed cleavage. Carbamidomethylation of cysteine was included as a fixed modification whereas methionine oxidation, N-terminal protein and lysine acetylation were included as variable modifications in the database search. Peptide identifications were accepted at greater than 95.0% probability as determined by the Peptide Prophet algorithm (7) and validated by manual inspection of the MS/MS spectra, as shown in Table S2. Related to Figure 2C and S2A and Table 1.
Tandem Affinity Purification (TAP) and M/S identification of HSF2 partners in HeLa cells
We performed retroviral transduction to establish HeLa-S3 cell lines expressing double-tagged HSF2 proteins. Hsf2-alpha and Hsf2-beta cDNA, cloned from E16 mouse brain were inserted in vector PCEMM-CTAP (Euroscarf P30536; Bürckstümmer et al., 2006) with CMV-driven expression of insert and GFP used as an indirect reporter (IRES; Fig S6A). PCEMM-CTAP allows sequential immunoprecipitation of the tagged protein through two tags: protein G (able to bind IgG) and the streptavidin-binding peptide (GS-TAP; Fig S6A). These two tag-modules are separated by two clivage sites for the Tobacco Etch virus (TEV) protease. Retrovirus production (Moloney murine leukemia virus type) and cell transduction with CTAP-empty (no insert), CTAP-Hsf2alpha and CTAP-Hsf2beta were performed as described (Sandrin and Cosset, 2006). GFP-positive Hela-S3 clones were then isolated by clonal dilution, selected by FACS (INFLUX 500, BD BioSciences, IMAGOSEINE Platform, Jacques Monod Institute, Paris) and amplified. By FACS we could isolate four cell sub-populations according to the intensity of the GFP signal to identify a population of cells expressing the recombinant tagged protein at levels similar to that of the endogenous HSF2 protein. GFP-positive Hela-S3 clones stably expressing GS-HSF2alpha, GS-HSF2beta or empty vector, were grown in floating cultures (spinners) and 10 G of cells (3 x 109 cells) per cell line was collected as described (Fristch et al. 2010). Nuclear extracts were prepared as described (Fristch et al. 2010; except the DNase 1 treatment was omitted). Total nuclear extracts were incubated with IgG-agarose beads overnight (Sigma; A2909). Then beads were incubated twice with TEV enzyme (Invitrogen, ref. 12575023)) for 45 minutes transferred to Poly-prep columns (Bio-rad). Eluates were collected in TEV buffer. These eluates were then incubated with Dynabeads My-one streptavidin for 1 h, washed once, and eluted using 5 mM D-biotin. Proteins were concentrated using TCA/acetone precipitation and dissolved in Laemmli sample buffer. Samples from three independent experiments were sequentially sent to TAPLIN Biological Mass Spectrometry Facility (Harvard Medical School, Boston, MA, USA) for MS (LC/MS/MS) analysis as in =ristch et al. (2010).
Immunoprecipitation and MS analysis of the endogenous HSF2 protein from E17 mouse cortices
We also analyzed by MS the partners of the endogenous protein HSF2 from E17 fetal cortices. Brain cortices were mechanically lysed into 4.5 volumes of the following buffer: 10 mM Hepes (pH 7.9), 400 mM NaCl, 5 %, glycerol, 100 mM EGTA) and two cycles of freezing and thawing in liquid nitrogen, and centrifuged at 20,000 g for 30 min. Supernatants were used to immunoprecipitate HSF2 using a monoclonal antibody (Abeam) or PBS as a negative control, and protein G agarose (Roche). Immunoprecipitates were analyzed on SDS-PAGE gel bands, containing the HSF2 protein (staining with colloidal blue) were sent to TAPLIN Biological Mass Spectrometry Facility (Harvard Medical School, Boston, MA, USA) for MS (LC/MS/MS) analysis.
SNAP-Tag labelling of HSF2 molecules and analysis of protein decay
CRISPR/Cas9 Hsf2KO U2OS cells were transfected with SNAP-tagged-HSF2 WT, -HSF2 3KQ or -HSF2 3KR constructs (Xtrem-Gen HP, Sigma-Aldrich), incubated in the presence of the cell-permeable SNAP-Cell® Oregon green fluorescent substrate (1,25 mM) and then with SNAP-Cell® Block (0,5mM) during the pulse chase according to the manufacturer’s instructions (New England Biolabs). Cells were lysed in modified Laemmli buffer (5 % SDS, 10% glycerol, 32.9 mM Tris-HCl pH 6.8) supplemented with 1 mM DTT (Sigma-Aldrich), and their extracts (15 μG) run on precast 4-20% Mini-Protean TGX gel (BioRad) and run as for western blots. Gels were scanned on a Typhoon Trio imager (GE Healthcare; excitation 532 nm, emission 580 nm, PMT 700 V) for determination of signal intensity of the covalently-bound fluorescent products as described in (Sanial et al., 2017).
Fluorescence three-hybrid assay
Fluorescence three-hybrid assay was performed according to Herce et al. (2013). BHK cells were transfected with constructs expressing expressing YFP-HSF2, CBP-HA, or EP300-HA, and GBP-Laci, using different combinations (ratio 1:1.5:2) at 70-80 % confluency using reverse transfection by Lipofectamine 2000 (ThermoFisher Scientific), as indicated. Medium was changed after 4 hours for all transfection. After 24 h, the cells were fixed in 4 % paraformaldehyde on coverslip and stained with mouse anti-HA (Covance) or rabbit anti-CBP antibody (Santa-Cruz), followed by a staining with mouse or rabbit Cy3 secondary antibody (Jackson Immunoresearch), respectively. Confocal microscopy images were taken on a confocal microscope Leica TCS SP5 (IMAGOSEINE Imaging Platform in Institut Jacques Monod) and images were analyzed using Fiji software.
Modeling of the HR-A/B domain and KIX domain of CBP
Prediction of secondary structure of the HSF2 HR-A/B domain was performed using Psipred (http://bioinf.cs.ucl.ac.uk/psipred/) and nps@ (https://npsa-prabi.ibcp.fr/). The tertiary structure of the same domain was predicted using http://petitjeanmichel.free.fr/itoweb.petitjean.freeware.html. Sequence similarity between human HSF2 HR-A/B (Sandqvist et al., 2009), lipoprotein Lpp56 of E. coli (Shu et al., 2000), yeast transcriptional factor GCN4 (mutated on some residues to stabilize heptad repeats; Shu et al., 1999) and murine PTRF (Polymerase I and Transcript-Release Factor) and human ATF2 (Activating Transcription Factor 2; a member of the ATF/CREB family) that is known to interact with CBP/EP300 (Bordoli et al., 2001) was explored using Uniprot (Pundir et al., 2017; https://www.uniprot.org/) and ClustalW (Thompson et al., 1994; https://www.ebi.ac.uk/Tools/msa/clustalw2/). Step 1: Based on this sequence similarity a sequence alignment of HR-A/B was developed (using Uniprot and ClustalW; see Figure S3C), a structural model of the monomer HR-A/B was generated using using Modeller (v9.19 (Eswar et al., 2006; https://salilab.org/modeller/), verified by ERRAT (Colovos et al., 1993; http://servicesn.mbi.ucla.edu/ERRAT/) and RESprox (Berjanskii et al., 2012; http://www.resprox.ca/) and ramachandran plot (Ramachandran et al., 1963) followed by the development of the trimer using SymmDock (Schneidman-Duhovny et al., 2005; v Beta 1.0; http://bioinfo3d.cs.tau.ac.il/SymmDock/). Step 2: the interaction between HR-A/B and the KIX domain (pdb: 2LXT; Brüschweiler et al., 2013) was simulated and compared using Zdock (v 2.1; Pierce et al., 2014) and Firedock (Mashiach et al., 2008; Odoux et al., 2016). More precisely, the ten best results generated by Zdock and Firedock, and scored according to their Root Mean Square Deviation (RMSD), which were compared thanks to a visualization program ICM (Fernandez-Recio et al, 2005). The mutation of the key residuse involved in the interaction between HR-A/B and the KIX domain have been performed using PyMOL (v2.0) and the docking was done as described above.
RP-UFLC-based separation and quantification of CBP substrate peptides (HSF2) and their acetylated forms
For acetylation assays, we synthetized several 5-fluorescein amidite (5-FAM)-conjugated peptide substrates based on the human HSF2 sequence and containing various lysine residues of interest (Proteogenix):
5-FAM-SGIVK82QERD-NH2, referred to as K82 peptide
5-FAM-SSAQ135VQIR-NH2, referred to as K135 peptide
5-FAM-SLRRK197RPLL-NH2, referred to as K197 peptide
NB: the synthesis of the HSF2 K128 peptide failed repeatedly because of its insolubility (Manufacturer’s information).
We also synthesized acetylated versions of these HSF2 peptides as standards. Samples containing HSF2 peptides and their acetylated forms were separated by RP-UFLC (Shimadzu) using Shim-pack XR-ODS column 2.0 x 100 mm 12 nm pores at 40°C. The mobile phase used for the separation consisted of 2 solvents: A was water with 0.12 % trifluoacetic acid (TFA) and B was acetonitrile with 0.12 % TFA. Separation was performed by an isocratic flow depending on the peptide:
80 % A/20 % B, rate of 1 ml/min for K82 and K135
77 % A/23 % B, rate of 1 ml/min for K195 and K197
HSF2 peptide (substrate) and their acetylated forms (products) were monitored by fluorescence emission (λ= 530 nm) after excitation at λ= 485 nm and quantified by integration of the peak absorbance area, employing a calibration curve established with various known concentrations of peptides.
In vitro acetyltransferase assay
To determine the activity of recombinant CBP-Full HAT on HSF2 peptides, we used 96-wells ELISA plate (Thermofisher) and assays were performed in a total volume of 50 μL of acetyltransferase buffer (50 mM Tris pH8, 50 mM NaCl) with 500 nm CBP-Full HAT, 50 μM HSF2 peptides and 1 mM dithiothreitol (DTT). Reaction was then started with the addition of 100 μM Acetyl-CoA (AcCoA) and the mixture was incubated 20 min at room temperature. 50 μL of HClO4 (15 % in water, v/v) was used to stop the reaction and 10 μL of the mixture were injected into the RP-UFLC column for analysis. For time course studies, aliquots of the mother solution were collected at different time points and quenched with 50 μL of HClO4 prior to RP-UFLC analysis.
SUPPLEMENTAL EXPERIMENTAL PROCEDURE
Related to-Figure S3A:
For in vitro protein-protein interaction experiments, we used biolayer interferometry technology (Octet Red, Forté-Bio). Recombinant HSF2 (TP310751 Origen) was desalted (ZebaTM Spin Desalting Columns, 7K molecular-weight cutoff, 0.5 ml (1034–1164, Fisher Scientific Germany)) and biotinylated at a molar ratio biotin/protein (3:1) for 30 min at room temperature (EZ-Link NHS-PEG4-Biotin (1189-1195, Fisher Scientific Germany)). Excess Biotin was removed using ZebaTM Spin Desalting Columns. Biotinylated recombinant HSF2 was used as a ligand and immobilized at 100 nM on streptavidin biosensors after dilution in phosphate-buffered saline (PBS; 600 s). Interactions with desalted analytes diluted in PBS at 100 nM (recombinant CBP domains 6His tag (Full HAT) or HSP70 as a positive control (ADI-SPP-555, Enzo-Life Sciences)) were analysed after association (600 s) and dissociation (600 s) steps at 26 °C. All sensorgrams were corrected for baseline drift by subtracting a control sensor exposed to running buffer only. Each KD was determined with a 1:1 stoichiometry model using a global fit with Rmax unlinked by sensor (FortéBio, Data analysis software version 7.1.0.89).
Acknowledgments
We warmly thank the patients and their families for their participation in this study. We thank Slimane AIT-SI-ALI for helpful discussions and comments on the manuscript, Anne PLESSIS (Jacques Monod Institute, Paris, France) for helpful discussions on setting the SNAP-TAG technology, Anne VANET (Jacques Monod Institute, Paris, France) for helpful discussions on the HSF2 structural modelling. We thank Lauriane FRITSCH and Slimane AIT-SI-ALI (UMR7216, for HeLa-S3 cells and growth conditions for TAP-TAG analyses). We thank the Imaging Platform IMAGOSEINE and especially N. BOGETTO for her help in sorting the GFP-positive HeLa-S3 cells. We are grateful to Heinrich LEONHARDT (Ludwig-Maximilians University, München, Germany) for F3H cellular and molecular tools and Pierre-Antoine DEFOSSEZ and Laure FERRY (UMR7216) for helpful guidance in F3H and GFP-Trap experiments, Isabelle LEMASSON (East Carolina University, USA) for the KIX-GST constructs, and Sophie POLO (UMR7216) for SNAP-TAG vectors. We thank Isabelle COUPRY and Benoit ARVEILER (CHU de Bordeaux, France) for primary skin fibroblasts from healthy donors. We thank the Institut Medical Jérôme Lejeune for the gift of Lymphoblastoid cells (patients 4 and 5). We are grateful to Delphine BOHL, and Stéphane BLANCHARD from Pasteur Institute (Retrovirus et Transfert Génétique, INSERM U622) for their help in producing the retroviruses for TAP-TAG experiments in Hela-S3 cells. We thank Clara GIANFERMI (UMR7216) for microscopy pictures of organoids and nSBs. We thank Isabelle Le PARCO and the staff from the Buffon animal housing facility at the Jacques Monod Institute (Paris Diderot University, Paris, France) and the Bioprofiler Platform at the UMR8251 Biologie Fonctionnelle et Adaptative for in vitro acetylation assays.
Footnotes
Funding information: VM was funded by the CNRS (Projet International de Coopération Scientifique PICS 2013-2015) for her collaboration with LS and by the Short Researcher Mobility France Embassy/MESRI-Finnish Society of Sciences and Letters; the Agence Nationale de la Recherche («NeuroHSF», Programme Neurosciences, Neurologie and Psychiatrie ANR-06-NEURO-024-01 and « HSF-EPISAME », SAMENTA ANR-13-SAMA-0008-01). LS was funded by the Academy of Finland, Sigrid Jusélius Foundation, Magnus Ehrnrooth Foundation and Cancer Foundation Finland.
RA was supported by PhD Fellowships from Neuropôle lle-de France and Fondation ARC, FM by PhD Fellowships from the CNRS and Fondation pour la Recherche Médicale and from a Postdoc Fellowship from SAMENTA ANR-13-SAMA-0008-01), AD by a Ministère de I’Enseignement supérieur, de la Recherche et de I’lnnovation (MESRI) Doctoral Fellowship, and ADT, DSD, and DB by the PICS travel grant, and GP and MH by Master 2 Intership Fellowships from SAMENTA ANR-13-SAMA-0008-01. JKA was supported by Magnus Ehrnrooth foundation. MCP was supported by the Turku Doctoral Network in Molecular Biosciences and Magnus Ehrnrooth Foundation. The supporting bodies played no role in any aspect of study design, analysis, interpretation or decision to publish this data.