dFOXO Activates Large and Small Heat Shock Protein Genes in Response to Oxidative Stress to Maintain Proteostasis in Drosophila*

Maintaining protein homeostasis is critical for survival at the cellular and organismal level (Morimoto, R. I. (2011) Cold Spring Harb. Symp. Quant. Biol. 76, 91–99). Cells express a family of molecular chaperones, the heat shock proteins, during times of oxidative stress to protect against proteotoxicity. We have identified a second stress responsive transcription factor, dFOXO, that works alongside the heat shock transcription factor to activate transcription of both the small heat shock protein and the large heat shock protein genes. This expression likely protects cells from protein misfolding associated with oxidative stress. Here we identify the regions of the Hsp70 promoter essential for FOXO-dependent transcription using in vitro methods and find a physiological role for FOXO-dependent expression of heat shock proteins in vivo.

The forkhead box (fox) 3 superfamily of transcription factors is defined by a DNA binding domain structurally related to the forkhead protein (2). The subfamily O (FOXO) group is distinct because of the presence of a five amino acid insert between helix 3 and helix 4 compared with other Fox family proteins. This family is conserved from Caenorhabditis elegans to mammals.
Invertebrates have a single FOXO gene (daf-16 in worms, dFOXO in flies). In mammals, the family has expanded to include four different FOXO genes (FOXO1, FOXO3, FOXO4, and FOXO6). In all organisms studied, FOXO family transcription factors play an important physiological role in protecting organisms against stress (3).
Although the best studied pathway controlling the FOXO family is insulin signaling, it is now clear that multiple stressors can activate FOXO (3). In Drosophila, dFOXO regulates the transcription of genes that promote survival under conditions of oxidative stress, changes in cellular metabolism, and unfolded proteins (4 -6). Increased dFOXO activity extends lifespan in Drosophila by changing the transcriptional landscape (7)(8)(9). Both cell autonomous and non-autonomous roles for dFOXO have been identified (10). The cell autonomous role for dFOXO is likely derived from the activation of genes with protective functions. Thus there is great interest in understanding the genes under the dFOXO regulon.
The same cellular stress conditions that activate the FOXO family often result in proteotoxicity, the accumulation of toxic protein species. Proteotoxicity results from misfolded or aggregated proteins, which can arise from acute oxidative stress, heat shock, and age (1). Unchecked, this can result in cell death, aging, and disease at the organismal level (11).
The cell uses many mechanisms to protect the proteome and maintain proteostasis. One important mechanism is the induction of expression of the heat shock proteins (Hsps) under conditions with the potential to promote proteotoxicity. The Hsps help to maintain proteostasis by acting as molecular chaperones during times of acute cellular stress or over the course of organismal aging (1,11).
There are two families of inducible Hsps with distinct protective mechanisms and functions, the small and large Hsps. The small Hsps (sHsps) are members of the Hsp20/␣-crystallin family of chaperones whose main function is to prevent the formation of denatured protein aggregates in the cell. Drosophila melanogaster species has four major small heat shock proteins Hsp22, Hsp23, Hsp26 and Hsp27 (12). The large heat shock protein family, which includes Hsp70, has ATP-dependent chaperone activity and acts to properly fold nascent polypeptide chains and improperly folded, yet soluble, proteins (13,14). Thus, both families of Hsps contribute to proteostasis using distinct mechanisms targeting different types of protein damage, so it would be advantageous to activate both pathways in response to stress.
Stress-inducible Hsps are distinct from their constitutive family members, such as the Hsp90 chaperones, in that they are specifically expressed during times of cellular stress, whereas the constitutive Hsps maintain proteostasis under normal conditions. D. melanogaster species has six stressinducible Hsp70 genes that have high sequence identity both in their promoters and within their open reading frames. The expression of the Hsps in Drosophila is necessary for proper stress resistance, and the overexpression of these genes can increase lifespan (15)(16)(17).
Heat shock transcription factor (HSF) regulates the expression of the Hsps during times of stress and during the heat shock response. HSF binds to heat shock elements (HSEs) within the promoter regions of the sHsps and Hsp70 in response to heat stress (18). To date, HSF is the best-characterized factor that influences Hsp expression, although the FOXO family member DAF-16 is known to specifically activate expression of only the sHsps in C. elegans (19).
DAF-16 has a role in maintaining proteostasis in C. elegans by transcriptionally up-regulating a subset of sHsp genes (20). A number of these genes play a role in DAF-16-dependent lifespan extension and contains DAF-16 recognition sequences within their promoter (19).
Consistent with the results from C. elegans, here we show a direct role for Drosophila dFOXO in the expression of the inducible sHsps. The sHsps are activated when dFOXO activity is increased by oxidative stress. In contrast to the results in C. elegans, we also establish the large Hsp, Hsp70, as a direct dFOXO target and determine the promoter sequence elements required for dFOXO activation. We show a physiological role for dFOXO-induced transcription of Hsp70 in the oxidative stress response. In Drosophila, dFOXO activates transcription of both classes of protein chaperones, providing a broader network of transcriptional targets to better protect cells against stress and proteotoxicity.

Results
Heat Shock Proteins Are Targets of dFOXO-We recently identified genomic targets of constitutively active Drosophila FOXO (dFOXO CA ) by chromatin immunoprecipitation (ChIP) followed by microarray analysis (21). Gene ontology analysis showed the genes involved in the heat shock response were significantly enriched (p ϭ 2.8e Ϫ2 ). Specifically, HSF and the stress-inducible Hsps have dFOXO bound at their promoters

B.
A. (Figs. 1A and 2A) (21). Interestingly, and contrasting work done in C. elegans, we identified the stress-inducible Hsp70 family of protein chaperones as a dFOXO target. Consistent with results in C. elegans and early work in Drosophila, we also saw enrichment of dFOXO binding to the Drosophila sHsps (Hsp22, Hsp23, Hsp26, and Hsp27) promoter regions ( Fig. 1A) (4,19).
To validate these results, we measured dFOXO CA binding to the Hsp promoters by performing ChIP-qPCR. As seen with the microarray experiments, there is enrichment of dFOXO binding at HSF and all of the stress-inducible Hsps that we examined (Figs. 1B and 2C). We utilized RT-qPCR to measure the effect of dFOXO CA overexpression on HSF and Hsp transcript levels. Both Hsp70 and the sHsps are induced by dFOXO CA expression (Fig. 1C). By contrast, under these conditions, dFOXO does not affect HSF transcript or protein levels despite binding to the promoter of HSF ( Fig. 2D; see Fig. 5B). These findings demonstrate that the inducible sHsp genes are a conserved set of FOXO family targets between C. elegans and D. melanogaster. In addition, the stress-inducible Hsp70 genes are dFOXO targets in Drosophila.
These results identified Hsps as a direct target of dFOXO, so we looked for potential FOXO-response elements (FREs) within the promoter regions. All of the inducible Hsp promoters contain multiple FREs (Fig. 3A). To help determine the promoter regions that respond to dFOXO, we put firefly luciferase under the control of the promoter regions of Hsp70Bb (Ϫ419 to ϩ63) or Hsp22 (Ϫ730 to ϩ98) (representative of the sHSPs) and co-transfected dFOXO CA in Drosophila cells. Although controlling for luciferase transfection efficiency using a consti-tutive Renilla reporter, we measured the effect of dFOXO CA expression on the activity of these promoters. We found Hsp70 and Hsp22 are both activated by dFOXO CA in the culture model, indicating these regions contain FOXO responsive sequences (Fig. 3B).
ChIP data identify Hsp70 genes and sHsps as direct targets of dFOXO, and we observed that expression of dFOXO CA increases expression of both Hsp70 and sHsps ( Fig. 1, B and C). Because in cell-based assays it is difficult to differentiate direct from indirect effects, we then performed in vitro transcription assays using the promoter regions we have identified as responding to dFOXO.
We tested dFOXO's ability to directly activate the transcription of the Hsp70 and Hsp22 promoters. The addition of recombinant dFOXO to Drosophila nuclear extract activates transcription of both the Hsp70 and Hsp22 constructs (Fig. 3C). The response is comparable to a synthetic reporter containing four copies of a consensus FRE (4ϫFRE) described previously (5). The addition of dFOXO to the extract does not effect the transcription from the histone H.4 promoter (Fig. 3C), indicating the response is specific to the Hsp promoters. These results show dFOXO is capable of directly activating the Hsp70 and Hsp22 promoters.
dFOXO Binds the Hsp70 Promoter Upstream of the HSEs-The heat shock elements to which HSF binds in the Hsp70 promoter are well characterized (22,23); however, the required sequences for dFOXO activation are unknown. The Hsp70 promoter contains four HSEs as well as four putative FREs (TGTTTT or TGTTTAT). The FREs contained in this pro- moter are variants of the sequences enriched in dFOXO-bound regions we previously identified (21) (1-4 in Fig. 4A). These sequences are very similar to those found upstream of 4E-BP, which is one of the best-characterized dFOXO targets (5,24).
To determine which promoter regions dFOXO is capable of binding, we performed electrophoretic mobility shift assays where regions of the promoter were labeled and incubated in the presence recombinant dFOXO. We plotted the fraction bound against the concentration of dFOXO in the reaction (Fig.  4B) to determine an apparent disassociation constant (K d ). We fit the data to a nonlinear regression, assuming one binding site. Probe A has the lowest apparent K d followed by Probe C, whereas Probe B is bound only slightly better than nonspecific DNA. These results indicate that the sequences contained in Probes A and C can be bound by dFOXO directly, potentially resulting in FOXO-dependent activation of Hsp70. We decided to test whether these regions also correlate with FOXO-dependent transcriptional activity.
To better characterize whether the FOXO-dependent activity requires the identified dFOXO binding sites, we carried out a promoter deletion analysis and compared these to the fulllength Hsp70 reporter containing all HSEs and putative FREs. Deletion 1 removes FRE 1 and 2. Deletion 2 contains only FRE 4, and deletion 3 does not contain any FRE-containing sequences (Fig. 4C).
We performed a dual luciferase assay after transfection of Drosophila cells with these constructs and dFOXO CA or a vector-only control. We normalized the response to a construct that contains the minimal Hsp70 core promoter element lacking all HSEs and FREs (Ϫ67 to ϩ1). We found that deletions 1 and 2 have less FOXO-dependent activity than the full-length Hsp70 promoter. The construct that contains no FREs, deletion 3, has no FOXO-dependent expression (Fig. 4D). These results indicate that the putative FRE-containing sequences that dFOXO bound in vitro are necessary for FOXO-dependent activation of the Hsp70 promoter.
Oxidative Stress Activates Endogenous dFOXO Resulting in Hsp Transcription-Our in vitro data and the data collected using the expression of constitutively active dFOXO identified Hsp70 as a dFOXO target. The question remains of what conditions result in endogenous dFOXO activating transcription of the Hsps. A likely candidate is oxidative stress. dFOXO is activated in response to cellular oxidative stress by JNK (4). Hsp expression increases in response to oxidative stress, but this   transcriptional activation has previously been attributed exclusively to HSF activity (25).
To test the role of dFOXO in Hsp expression in response to oxidative stress, we first used the compound paraquat. Paraquat creates intracellular reactive oxygen species by undergoing reduction into the superoxide free radical in the mitochondria. It has been used previously to activate FOXO-mediated transcription (4,26,27). To confirm our ChIP data using constitutively active dFOXO, we also performed ChIP for endogenous dFOXO in response to paraquat treatment. dFOXO is enriched ϳ12-fold at the Hsp70 promoter with paraquat treatment (Fig.  5A). Thus, oxidative stress results in dFOXO binding to the Hsp70 promoter.
To determine the relative contribution of HSF or dFOXO to Hsp70 activation, we knocked down dFOXO, HSF, or both in cultured cells using RNA interference (Fig. 5B). In addition, we treated cells with compounds that create intracellular oxidative stress through different mechanisms. We used paraquat, the mitochondrial uncoupler 2,4-dinitrophenol (DNP), the first compound shown to induce the heat shock response (28), and diethyl maleate, which depletes glutathione.
We incubated cells with nonspecific double-stranded RNA or double-stranded RNA directed against dFOXO, HSF, or both. After 72 h, the cells were incubated with paraquat, DNP, or diethyl maleate. RT-qPCR was performed to quantitate the relative expression of Hsp70 for each condition. Both dFOXO and HSF are required for the full activation of Hsp70 in response to oxidative stress created by paraquat or DNP treatment (Fig. 5C). However, HSF is solely responsible for the Hsp70 induction in response to diethyl maleate (DEM; Fig. 5C). These results show endogenous dFOXO and HSF are responsible for activating transcription of Hsp70 in response to intracellular reactive oxygen species.
Because both dFOXO and Hsp70 are required for adult flies to survive paraquat treatment (24,29), we used paraquat to test if dFOXO activation of Hsp70 is relevant in vivo. We subjected a fly line containing a disruption of the dFOXO gene (Ϫ/Ϫ dFOXO ⌬94 ) and the isogenic parental line (w DAH ) to paraquat

B.
A.    treatment and measured the response of the Hsp genes. We starved 7-day-old adult male flies for 5 h, then provided them Schneider's media with or without paraquat, and the flies were allowed to consume the solution for 24 h. Total RNA was extracted from the whole fly, and RT-qPCR was performed to measure the relative Hsp transcripts. In response to intracellular reactive oxygen species production, the dFOXO-null flies had less expression of Hsp22, Hsp23, Hsp26, Hsp27, and Hsp70 (Fig. 5D). This result indicates that dFOXO mediates expression of sHsps and the large heat shock protein Hsp70 in response to oxidative stress in vivo and that HSF alone is insufficient for the response.

Discussion
In the work described above we have identified an expanded set of Hsp transcriptional targets for Drosophila FOXO. The FOXO-dependent expression of the Hsps expands the role for dFOXO in maintaining proteostasis in response to stress and identifies a new transcriptional activator for the Hsp70 genes. Hsp70 properly folds nascent proteins and re-folds soluble proteins, whereas sHsps disassemble aggregates; both work to correct non-native protein interactions. Thus, dFOXO activation of both chaperones protects against proteotoxicity at two separate stages of protein stabilization. Specifically, in response to oxidative stress, dFOXO is a necessary contributor to increased Hsp transcript levels. Furthermore, our data indicate that dFOXO is the major contributor to the response for the sHsps. Together with HSF, dFOXO is able to mount a transcriptional response that allows cells to survive acute stress. We propose that having two transcription factors that can activate stress response genes is advantageous to the cell because maintaining proteostasis is imperative for cell survival when the accumulation of oxidized, misfolded, or aggregated proteins results in proteotoxicity (Fig. 6).

B.
In Drosophila, dFOXO binds the promoters of more than a thousand genes, and identifying the subset of genes that reduces cellular damage is ongoing within the field (21, 30 -33). Drosophila FOXO was previously shown to influence the expression of a sHsp-like gene, and there have been previous attempts to show Drosophila FOXO-dependent Hsp70 expression; however, the FREs disrupted in that study do not match those identified here and oddly included the TATA core, abrogating promoter function and complicating the interpretation of the results (4,9). This work defines a comprehensive set of Hsps that are direct targets of Drosophila FOXO that help to maintain protein homeostasis.
The differential requirements for HSF and dFOXO in response to specific oxidative stress inducing agents was an unexpected result. The Hsp70 transcriptional response to paraquat and DNP requires dFOXO, and contrast, diethyl maleate, which depletes glutathione from the cells, does not require dFOXO for full activation of Hsp70 (Fig. 5C). Because of the differential response, it is plausible that they respond to differ-ent signals; HSF directly senses misfolded proteins and drives expression of Hsps, whereas dFOXO may sense the stress indirectly (34,35). A possible explanation implicates the mitochondria as a dFOXO-specific sensor. Paraquat undergoes reduction within and damages the mitochondria (36), and DNP is a mitochondrial uncoupler. Our work suggests that HSF responds to all oxidative stress, but FOXO responds to selective types of oxidative stress, perhaps through the mitochondria.
We also identified the HSF gene as bound by dFOXO. However, we could find no effect on HSF transcription or HSF protein levels under the experimental conditions used here despite the fact that dFOXO was found reproducibly bound to the HSF promoter. This might indicate there is another signaling event required for FOXO-mediated regulation of HSF that is missing from our experimental approach. This is intriguing because a close relationship between HSF and FOXO in stress responses and lifespan regulation has been proposed in C. elegans (19,37). daf-16 is required for hsf-1 to extend lifespan (19). There is evidence to support that HSF and DAF-16 affect each other's activity, and DAF-16 and HSF have a set of overlapping targets but do not require always require each other for transcription of their target genes (19,(37)(38)(39). Future work should be aimed at identifying the pathways whose cross-talk is required for the connection between FOXO and HSF regulation.
Much of the work on the FOXO family has focused on its role in modulating aging and lifespan. We suggest another physiological role for FOXO-dependent transcription of Hsp70 may occur as the organism ages. Because acute oxidative stress has a transcriptional profile similar to aging (25), we propose that dFOXO may play a role in Hsp transcription over the course of aging. The accumulation of free radicals as well as aging results in an increase in Hsp22 and Hsp70 transcription (25). Increased expression of these genes is also predictive of improved survival rate in response to stress (40). Previously, the aging-dependent expression of Hsp70 was attributed solely to HSF. However, it seems likely that dFOXO's contribution was overlooked (41). Based on our current results, we propose a role for FOXO-dependent activation of the large heat shock protein family during aging. Both dFOXO and HSF can potentially activate the expression of both families of heat shock protein genes in D. melanogaster in response to both oxidative stress and aging, increasing survival. Further work will be required to determine if this role is conserved in higher animals.

Experimental Procedures
Fly Lines, Constructs, and Antibodies-The w DAH and dFOXO-null (⌬94/⌬94) fly lines have been previously described (21). pAc5V5-dFOXO CA and pGL4xFRE were previously described (21). Promoter regions from genomic Hsp70Bb, Hsp22, and histone H4 were cloned into pGLbasic vector, and these were used in dual-luciferase assays and as templates for in vitro transcription. pGLHsp70Bb was used for cloning deletion constructs and band shift probes. For Western blotting and ChIP, antibodies against tubulin (DSHB Hybridoma Product E7), dFOXO (21), and HSF (42) Table 1.
ChIP-The 321 stable line was induced with 500 M CuSO 4 for 16 h. ChIP was performed as previously described (20). The immunoprecipitated DNA was assayed by qPCR using primers for the control rRNA gene and the promoter regions of HSF, Hsp22, Hsp23, Hsp26, Hsp27, and Hsp70 (Table 1).
In Vitro Transcription and His-tagged dFOXO Purification-Nuclear extracts were prepared from Drosophila embryos as previously described (43). Promoter templates (pGLHsp70Bb, pGLHsp22, pGL4xFRE, and pGLH4), recombinant His-dF-OXO, and recombinant NTPs were added to the extracts. dFOXO purification and primers extension assays were performed as previously described (5).
Transient Transfection and Dual-luciferase Assay-S2C1 cells were plated at a 0.5 ϫ 10 6 cells/ml in 24-well plates. The cells were transfected with reporter plasmid and expression plasmid at a ratio of 1:10 using the Effectene protocol (Qiagen). Twenty hours post-transfection the cells were harvested according to the passive lysis protocol for the Promega Dual-Luciferase reporter assay system. The expression of firefly and renilla luciferase was measured either using the dual-luciferase reporter assay system (Promega), or firefly luciferase was measured in 75 mM HEPES, pH 8.0, 20 mM DTT, 5 mM MgSO 4 , 530 M ATP, 500 M coenzyme A, 500 M D-luciferin, and 100 M EDTA, and Renilla luciferase was measured by adding an equal volume of 1.0 M NaCl, 0.5 M Na 2 SO 4 , 25 mM Na4PPi, 15 mM EDTA, 10 mM NaOAc, and 0.1 mM coelenterazine.
Band-shift Assay-Regions of the Drosophila Hsp70 promoter were cloned into pBC (Stratagene), and the probes were made by PCR with primers labeled with Dylight 680 fluorophore (Thermo). The band-shifts were done as previously described (5). Recombinant dFOXO was incubated with labeled probes and separated on a 5% acrylamide, 1ϫTGE (25 mM Tris, 190 mM glycine, 1 mM EDTA), 4 mM MgCl 2 , 2.5% glycerol gel. Gels were imaged using LI-COR Odyssey. The data were plotted as the fraction of probe bound to final concentration of recombinant dFOXO in the binding reaction. The data were fit to a nonlinear regression with the assumption of a single, specific binding site, and the apparent disassociation constant (K d ) was calculated.
Paraquat Feeding-Adult fly feeding protocol was described previously (26). RT-qPCR was performed to determine the relative levels of transcription of both the control rp49 and Hsps.
Gene Ontology Analysis-Genes previously described as enriched for dFOXO binding (21) were analyzed using the DAVID Bioinformatics Resources available online (david.ncifcrf.gov).