DAF-21/Hsp90 is required for C. elegans longevity by ensuring DAF-16/FOXO isoform A function

The FOXO transcription factor family is a conserved regulator of longevity and the downstream target of insulin/insulin-like signaling. In Caenorhabditis elegans, the FOXO ortholog DAF-16A and D/F isoforms extend lifespan in daf-2 insulin-like receptor mutants. Here we identify the DAF-21/Hsp90 chaperone as a longevity regulator. We find that reducing DAF-21 capacity by daf-21(RNAi) initiated either at the beginning or at the end of larval development shortens wild-type lifespan. daf-21 knockdown employed from the beginning of larval development also decreases longevity of daf-2 mutant and daf-2 silenced nematodes. daf-16 loss-of-function mitigates the lifespan shortening effect of daf-21 silencing. We demonstrate that DAF-21 specifically promotes daf-2 and heat-shock induced nuclear translocation of DAF-16A as well as the induction of DAF-16A-specific mRNAs, without affecting DAF-16D/F localization and transcriptional function. DAF-21 is dispensable for the stability and nuclear import of DAF-16A, excluding a chaperone-client interaction and suggesting that DAF-21 regulates DAF-16A activation upstream of its cellular traffic. Finally, we show a selective requirement for DAF-21 to extend lifespan of DAF-16A, but not DAF-16D/F, transgenic daf-2 mutant strains. Our findings indicate a spatiotemporal determination of multiple DAF-21 roles in fertility, development and longevity and reveal an isoform-specific regulation of DAF-16 activity.

HSF-1 to extend nematode lifespan in response to reduced ILS 19 , in part by a cooperative induction of the longevity promoting small heat shock protein genes 19 . The above-mentioned findings both illustrate an important role of DAF-16 in proteostasis and imply further crosstalks to finely tune the DAF-16 and HSF-1 dependent responses.
In contrast to the four mammalian FOXO genes originating from gene duplication, the single C. elegans DAF-16 gene gives rise to three groups of transcripts that are transcribed from different promoters 6,20 . Isoforms A, B and D/F/H share the same Forkhead domain but have distinct N-terminal domains 2 . Further studies consistently revealed that DAF-16A and DAF-16D/F are responsible for lifespan extension 20,21 , however, their relative contribution remains controversial. These findings add another layer and a yet unknown complexity to DAF-16 regulation.
The 90-kDa heat-shock protein (Hsp90) is an essential, evolutionarily conserved eukaryotic molecular chaperone 22 . It accounts for 1-2% of the total cellular proteins. In yeast, Hsp90 is the most connected protein, affecting the function of ~20% of the proteome 22 . In mammalian cells Hsp90 is essential for the folding and activity of several hundred thermodynamically unstable 'client' proteins, predominantly kinases and transcription factors mainly involved in cell signaling and proliferation 23 . A comprehensive list of Hsp90 clients can be found at the Picard laboratory homepage (www.picard.ch/downloads/Hsp90interactors.pdf). Although Hsp90 is a promising tumor target 24 and is implicated in the evolution of new traits 22 , several pieces of evidence suggest it might play a role in aging. For example, others and we reported that Hsp90 also chaperones antiproliferative, differentiation promoting clients, such as the TGF-β receptor 25 and PPARγ 26 . Moreover, Hsp90 also regulates the heat-shock response by releasing the heat shock factor HSF-1 upon binding to denatured proteins 27 which may be utilized in age-related diseases of proteostasis and metabolism, such as diabetes 28 . Yet, the impact of Hsp90 on aging is largely unexplored.
Despite the plethora of information in yeast and mammals, little is known about the biological function and interactions of the sole C. elegans Hsp90 ortholog encoded by the daf-21 gene. Previous work using a weak gain-of-function point mutant daf-21(p673) unraveled its role in chemosensory perception and dauer development through DAF-11 29,30 , and muscle homeostasis via UNC-45 31 . The loss of DAF-21 in embryos either by large deletion mutation in the daf-21(nr2081) strain 30 or in the F1 progeny of daf-21(RNAi) treated worms 32 induced embryonic and early larval lethality and revealed a conserved function of Hsp90 in cell division through chaperoning the WEE-1 kinase [32][33][34] . Perhaps due to the essential role of daf-21 in embryogenesis, its role in C. elegans lifespan regulation is largely unexplored. In this study, using gene silencing from larval development we investigated the involvement of DAF-21 in wild-type and daf-2 induced longevity and found its specific requirement for DAF-16 isoform A activity.

daf-21/hsp90 knockdown silences DAF-21 expression and induces the heat shock response.
To investigate the impact of reduced DAF-21 capacity on longevity we employed a daf-21(RNAi) construct described earlier 35 . In agreement with an essential role of Hsp90 in cell proliferation and development 22 daf-21(RNAi) treatment of the parental F0 generation induced embryonic and early larval lethality in F1 in one 32 , and sterile F1 worms exhibiting developmental phenotypes in another study 35 . To investigate the impact of DAF-21 during larval development we started daf-21 silencing from hatching. daf-21(RNAi) efficiently reduced daf-21 mRNA and protein levels in young adults compared to empty vector (EV) control ( Fig. 1a and b). Consistent with its reported role in vulval development and muscle function 35,36 , daf-21(RNAi) fed worms exhibited a protruding vulva in ~90% of the population and a mild hypomotility, but no other developmental phenotypes (Fig. S1). Likewise, daf-21 silencing from hatching neither arrested nor significantly delayed development (Fig. S1). These observations indicate that daf-21 knockdown employed from hatching does not induce larval lethality and developmental abnormalities 32 , except aberrant vulval morphogenesis during the L4 phase 37 .
To differentiate between the effects exerted by DAF-21 during larval development and adulthood, respectively, we also treated worms with daf-21(RNAi) from the midst of L4 stage. As expected, no protruding vulva phenotype was detected. Reflecting an essential role in oogenesis and embryonic development, daf-21 silencing from hatching caused sterility accompanied by a lack of oocytes in the gonad, whereas that from the L4 stage caused reduced brood size and sometimes embryonic lethality ( Supplementary Fig. S1d,e). daf-21 knockdown employed from hatching or from the L4 stage similarly induced an hsf-1 dependent hsp-16.2 and hsp-70 mRNA expression in young adults ( Fig. 1c-f), consistent with a compensatory activation of the heat shock response upon the reduction of available DAF-21 protein 27 . We also observed a modest, but significant elevation of survival during heat shock in young adults fed by daf-21(RNAi) from hatching (Fig. 1g), reflecting an increased stress resistance associated with longevity and healthspan 38 . Hence, daf-21(RNAi), employed either during or after larval development is a safe approach to reduce DAF-21 capacity without compromising develoment and health.
Reduction in DAF-21/Hsp90 capacity limits normal lifespan. Next, we addressed how reducing DAF-21 capacity starting at the beginning or at the end of larval development, respectively, might affect the natural lifespan of wild-type nematodes. First, we measured the lifespan of N2 worms fed by either empty vector or daf-21(RNAi) from hatching throughout the entire life. Our results show that daf-21(RNAi) shortened lifespan by ~27% compared to EV ( Fig. 2a and Table S1). The reduction in lifespan caused by daf-21(RNAi) was less pronounced, but still significant, if RNAi treatment was administered from the L4 stage throughout adulthood (Fig. 2b). The daf-21(p673) gain-of-function allele appears to decrease lifespan mainly via bagging (death resulting from a deficiency in egg-laying and internal hatching of the progeny) 31 . We did not observe bagging and premature death in the daf-21 silenced population. The comparable lifespan reduction by daf-21 knockdown during and after development, respectively, suggests that daf-21 affects longevity independent of egg laying and fertility ( Supplementary Fig. S1d,e). The decreased lifespan, despite an increased heat shock response, which predicts and induces longevity 39,40 might be the consequence of different longevity promoting mechanism(s) impaired in the absence of an optimal DAF-21 capacity. daf-21/hsp90 is required from larval development for longevity conferred by reduced ILS. Increased longevity of animals with reduced ILS has been established as one of the most robust effects that prolong lifespan 5 . Therefore we compared the lifespan of daf-2(e1370) and daf-2(e1370); daf-16(mgDf50) single and double mutant animals, respectively, grown on plates with bacteria harboring empty vector or daf-21(RNAi). In accordance with the literature daf-2 mutation caused an increase in lifespan compared to wild type which was entirely abrogated by daf-16 loss of function ( Fig. 2c and Table S1) 20,41,42 . daf-21(RNAi) treatment from hatching significantly reduced the lifespan of daf-2(e1370) worms showing a requirement for DAF-21 to fully manifest the increased longevity conferred by reduced ILS. The partial inhibitory effect of daf-21(RNAi) might be due to the unsuccessful penetration of dsRNA into neurons which do not express SID-1 protein 43 . To address this possibility, we used RNA interference to silence the expression of daf-2 combined with daf-21 knockdown. The lifespan of Figure 1. DAF-21/Hsp90 knockdown reduces daf-21 mRNA and protein expression and induces the heat shock response. daf-21(RNAi) reduces daf-21 mRNA (a) and protein (b) levels, respectively. Protein levels for DAF-21 and actin presented here were derived from the same membrane cut in two and incubated independently using their respective antibodies. Full-length blots are presented in Supplementary Fig. S1. daf-21(RNAi) employed either from hatching (c,e) or from the L4 stage (d,f) induces hsp-16.2 (c,d) and hsp-70 (C12C8.1) (e,f) mRNA levels at non-heat shock conditions in an hsf-1 dependent manner and increases survival during heat shock (g). mRNA expression was assayed by qRT-PCR, normalized to β-actin mRNA and log2 transformed fold change values (mean ± SEM) were expressed relative to the respective EV control values of one. Please note that the asymmetric length of SEM error bars are a consequence of the logarithmic scale. Data shown are from three independent experiments. qRT-PCR statistics were analyzed by ANOVA and are given in Supplementary Table S3. Survival data were compared using Student's t-test. EV: empty vector RNAi. *p < 0.05; **p < 0.01; ***p < 0.001. wild-type animals upon daf-2(RNAi)/EV showed a ~100% increase compared to empty vector (Supplementary Fig. S2 and Table S1) which was abolished in daf-16(mu86) mutant background. Again, we found a 20-50%, partial decrease of the lifespan of daf-2 silenced nematodes upon daf-21(RNAi) treatment comparable to that observed in daf-2 mutants, indicating a longevity supporting effect of DAF-21 in non-neuronal tissues.
Next, we determined the effect of daf-21(RNAi) employed from the L4 stage on the lifespan of daf-2 mutants. In all 4 trials it extended the longevity of daf-2(e1370) animals, however, in two of them the effect was non-significant ( Fig. 2d and Table S1). These observations indicate a longevity supporting effect of DAF-21 primarily during, but not after, development, in both wild-type and daf-2 nematodes. We asked if daf-21 affected the fertility of daf-2 mutants. Consistent with earlier findings 44 daf-2 mutation delayed reproduction which was similarly abolished by daf-21(RNAi) employed either from hatching or from the L4 stage ( Supplementary Fig. S1f,g). Thus, the differential effect of the daf-21(RNAi) treatment on daf-2 lifespan appears to be independent of fertility.
In the majority of experiments, daf-21(RNAi) further shortened the lifespan of daf-16 and daf-2; daf-16 mutants, however, it was significantly less than its lifespan shortening effect in the strains harboring wild-type daf-16 (Figs 2, S2 and S3 and Supplementary Table S1). Moreover, daf-16 eliminated the longevity increasing tendency of daf-21(RNAi) in early adulthood of daf-2 strains (Fig. 2d). Thus, it appears that the longevity action of DAF-21 might possess both a daf-16 dependent as well as an independent component in both wild-type worms and those with reduced ILS. daf-21/hsp90 knockdown in neurons promotes dauer development of wild type, that in non-neuronal tissues does not affect dauer formation of daf-2 mutants. Nematodes with lowered ILS have a high tendency to initiate an alternative developmental pathway and form dauer larvae under stressful conditions. Likewise, the daf-21(p673) point mutant also promotes dauer formation in C. elegans 29 by inhibiting the DAF-11 guanylcyclase in chemosensory neurons 30 . Hence, we investigated the interference between daf-2 and daf-21 in dauer development. As shown before 45 , daf-2(e1370) mutants grown on 25 °C almost exclusively turned into dauer larvae, while having a secondary mutation in daf-16 abrogated this effect (Fig. 3a). In agreement with previous reports 20, 21 it appeared that daf-16a isoform per se is sufficient to manifest the dauer program while having only daf-16d/f resulted in a smaller rate of dauer formation. Dauer decision by lowered ILS is made at the L1 stage and requires predominantly neuronal daf-2 and daf-16 44,46 . Consistent with this, silencing daf-21 from hatching in non-neuronal cells did not affect dauer formation in any of the strains tested (Fig. 3a). Therefore, we employed the TU3335 strain that expresses SID-1 in each cell including neurons 47 and confirmed and extended earlier findings about the neuronal requirement of sufficient DAF-21 function to bypass dauer arrest 29,30 (Fig. 3b). Dauer decision by DAF-11 is also made at late L1 phase 48 , showing that daf-21 is already silenced in (late) L1. The dauer arrest by neuronal daf-21(RNAi) and the embryonic/early larval lethality caused by the embryonic loss of DAF-21 30,32 prevented us to test its impact on daf-2 induced dauer formation. However, our findings together with the lifespan data indicate that DAF-21 in peripheral tissues affects wild-type and daf-2 longevity that is spatiotemporally separated from its effect on development.

DAF-21/Hsp90 facilitates DAF-16A nuclear translocation. The fact that the impact of DAF-21 on
daf-2 lifespan is dependent on DAF-16 suggested a functional link between DAF-21 and DAF-16. Hence, we asked how daf-21 knockdown affects DAF-16 function. An important step in DAF-16 activity in response to lowered ILS, as well as stresses including heat shock, is its nuclear translocation 49 . First, we monitored the intracellular localization of DAF-16 in the TJ356 strain containing a daf-16a/b::GFP transgene 16 in response to daf-2 and daf-21 knockdown employed from hatching. To ensure similar RNAi dosages, animals were fed with bacteria containing empty vector (EV), daf-2(RNAi)/EV, daf-21(RNAi)/EV and daf-2(RNAi)/daf-21(RNAi) in 1:1 ratios. In daf-2 silenced nematodes a large proportion of DAF-16A/B was localized in the nuclei of both intestinal as well as muscle cells. We found that silencing daf-21 partially inhibited the daf-2 knockdown-induced nuclear translocation of DAF-16A/B::GFP ( Fig. 4a and b). This result indicates that DAF-21 facilitates DAF-16 translocation in response to lowered ILS. To address if this phenomenon might be the consequence of a process altered during development or it is present in adulthood, we repeated this measurement by placing L4 larvae on RNAi plates. The previously observed increase in DAF-16 nuclear localization upon daf-2(RNAi) was confirmed under these conditions as well as daf-21(RNAi)'s ability to inhibit such translocation ( Fig. 4c and d).
We asked whether Hsp90 might act on DAF-16 function in an isoform-specific manner. Two out of the three DAF-16 isoform groups, A and D/F/H, are targeted by ILS in longevity regulation 6,20,21,50,51 . Therefore, we employed two strains that express different fluorescently tagged DAF-16 isoforms in a daf-16(mgDf50) null mutant background: daf-16a::rfp and daf-16d/f::gfp and their respective daf-2(e1370) variants 20 . The transcripts and proteins structures of the two isoforms are depicted in Fig. 5a and b. RNAi treatment was employed from hatching, and DAF-16 localizaton was visualized on day 1 of adulthood. Consistent with previous reports DAF-16A::RFP showed an explicit nuclear localization in both intestinal and muscle cells while DAF-16D/F::GFP remained largely cytosolic in response to daf-2(e1370) compared to the respective strains with intact daf-2 alleles  DAF-21/Hsp90 is required for DAF-16A dependent transcriptional function. Next we studied how daf-21 silencing affects the expression of various DAF-16 target genes. We looked into the expression of genes that are thought to be regulated by both isoforms (sod-3 and old-1), some that are reportedly DAF-16A-regulated (scl-20 and gst- 20) and genes that are the targets of DAF-16 isoform F (lea-1, scl-1, col-183, R05D8.7). In the first series of experiments, in order to clearly isolate the undesired cross-talk from the other DAF-16 isoforms, we took use of the above daf-16a::rfp and daf-16d/f::gfp strains. The induction of sod-3 and old-1 mRNA expression -caused by the daf-2(e1370) mutation -were both diminished by daf-21(RNAi) in the daf-16a::rfp strain ( Fig. 6a and b), while they were unaffected in the daf-16d/f::GFP strain ( Fig. 6e and f). The daf-21-dependent nature of sod-3 and old-1 induction was reinforced by using daf-2(RNAi) on wild-type animals ( Supplementary Fig. S7). Moreover, old-1 was only upregulated in the daf-2(e1370) mutant variant of the daf-16a::rfp strain ( Fig. 6b and f). These results corroborate previous data on sod-3 21 and propose old-1 as a preferential DAF-16A target. We also selected scl-20 and gst-20, identified as specific DAF-16A targets 21 . scl-20 encodes a putative p53 target that functions to regulate both lifespan and tumor cell proliferation 52 . gst-20 is an ortholog of human hematopoietic prostaglandin D synthase and is involved in the daf-2 induced, diet dependent extension of adult lifespan 21 . Our experiments confirmed the upregulation of both genes by the daf-2(e1370) allele in the daf-16a::rfp strain and showed an efficient inhibition of gst-20 expression by daf-21(RNAi) (Fig. 6c and d). To address the influence of DAF-21 on the transcriptional activity of the DAF-16D/F isoform, lea-1 and scl-1 were selected for our purposes as selective targets 20,21 . lea-1 encodes a protein that is predicted to be hydrophilic and heat-resistant, and that might participate in anhydrobiosis 53 , while scl-1 encodes a predicted secretory protein that is a member of the cysteine-rich secretory protein (CRISP) family 54 . In contrast to DAF-16A target genes, these two transcriptional targets in a daf-16d/f::GFP transgenic background were not inhibited by daf-21(RNAi) (Fig. 6g and h).
To further examine a potential isoform-specific regulation we measured the expression of DAF-16A and DAF-16D/F isoform specific targets in a wild-type and mutant daf-16 background in response to daf-21 silencing from the L4 stage. Comparing the mRNA expression in a daf-2(e1370) and daf-16(mgDf50); daf-2(e1370) strains, respectively, showed that the DAF-16A specific sod-3, old-1, gst-20 and scl-20 were induced by the daf-2 mutation in a daf-16 dependent manner and their expression was inhibited by silencing daf-21 with the exception of scl-20 (Fig. 7a-d). DAF-16D/F target genes were supplemented by two additional DAF-16D/F specific targets: col-183, a predicted structural constituent of cuticle 55 and R05D8.7, an ortholog of human hydroxysteroid 17-beta  56 . We found that all mRNAs were efficiently induced in the daf-2(e1370) mutant but not inhibited by daf-21(RNAi) (Fig. 7e-h). Some of the findings were confirmed by using an independent daf-21(RNAi) sequence 35 (Supplementary Fig. S8). Thus, our findings provide compelling evidence that DAF-21 specifically regulates the transcriptional activity of DAF-16A. DAF-21/Hsp90 is not necessary for DAF-16A stability and acts upstream of its nuclear import. Mammalian Hsp90 is a specific chaperone that stabilizes the conformation of a large number of client proteins including various transcription factors 22 . If Hsp90 function is compromised, destabilized clients are polyubiquitinylated by ubiquitin conjugating enzyme complexes and subsequently degraded by the proteasome 57 . Several E3 ubiquitin ligases target mammalian FOXOs to degradation 18 . In C. elegans, the null mutation of the RLE-1 E3 ubiquitin ligase has been shown to lead to DAF-16 protein stabilization and DAF-16 dependent lifespan extension 58 . We reasoned if DAF-21 stabilized DAF-16 conformation, then reduced DAF-21 capacity would result in DAF-16 aggregation and disrupt lifespan extension in rle-1 mutants. Although we did not detect a substantial increase of DAF-16A::RFP in rle-1 background, daf-21(RNAi) neither interfered with DAF-16 distribution nor caused DAF-16 aggregation (Supplementary Fig. S9). This might be due to the action of other, yet unidentified ubiquitin ligase(s). However, DAF-21 did not influence the turnover of DAF-16 protein: we neither observed   Similarly to mammalian cells, when ILS signaling is ample, AKT-1 and AKT-2 kinases phosphorylate DAF-16/FOXO which prevents its accumulation in nuclei by anchoring it to cytosolic 14-3-3 scaffold proteins 49 . In response to reduced ILS, DAF-16 translocates into the nucleus. To assess if DAF-21 acts upstream or at the level of DAF-16 nuclear traffic, we employed a strain harboring the daf-16a AM ::gfp (AM: "AKT site mutant") transgene, in which serines and threonines in the AKT phosphorylation sites were changed to alanines 49 . In agreement with previous work 49 , DAF-16A AM ::GFP was nuclearly localized in animals possessing wild-type daf-2 ( Supplementary Fig. S10). If DAF-21 was necessary for DAF-16A to achieve its native functional conformation, then reducing DAF-21 capacity would result in an unstable DAF-16A AM ::GFP unable to enter the nucleus and would be degraded. This was not the case, because neither the quantity, nor the localization of DAF-16A AM ::GFP was modified by daf-21(RNAi) (Supplementary Fig. S10), providing further evidence for the conformational independence of DAF-16A from DAF-21 and indicating that the nuclear import of unphosphorylated DAF-16A does not require DAF-21. Also, these findings made unlikely a DAF-21-dependent inhibition of nuclear export of DAF-16A. Instead, DAF-21 appears to influence DAF-16A activation upstream of its nuclear traffic. This result is supported by our measurements of DAF-16A-specific target genes in the DAF-16A AM strain. Of the previously examined four specific DAF-16A targets only scl-20 and sod-3 were induced in the DAF-16A AM strain when compared to the daf-16(mu86) mutant ( Supplementary Fig. S10), although only scl-20 reached the level of statistical significance. However, their expression was not altered by daf-21(RNAi), suggesting that nuclear DAF-16 is able to exert a partial transcriptional activity which is independent of DAF-21. These findings corroborate the idea that DAF-21 acts upstream of DAF-16A nuclear traffic.

DAF-21/Hsp90 ensures daf-16a dependent longevity. Our findings identified DAF-21 as an
isoform-specific modulator of DAF-16 activity. Further, we observed that a reduction in DAF-21 capacity from larval development limits daf-2 induced longevity. Hence, we investigated how daf-21(RNAi) employed from hatching affects longevity specified by individual DAF-16A and D/F isoforms in the context of reduced ILS by using strains expressing single isoforms in a daf-2(e1370); daf-16(mgDf50) mutant background. In concordance with published data 20,21 both daf-16a::rfp and daf-16d/f::gfp transgenic strains exhibited longer lifespan compared to daf-2(e1370); daf-16(mgDf50) background ( Fig. 8a and b and Table S1) indicating that both isoforms are involved in daf-2 induced longevity. daf-21(RNAi) further shortened the lifespan of daf-2(e1370); daf-16(mgDf50) double mutants suggesting DAF-21 might also target longevity promoting factors other than DAF-16. Importantly, daf-21 knockdown consistently diminished the lifespan of the DAF-16A::RFP expressing strain in all four biological replicates, while in two out of four trials it failed to modify that of the DAF-16D/ F::GFP transgenic worms (Fig. 8a,b and Table S1). These results indicate that an optimal DAF-21 capacity from larval development plays a role in longevity through selectively ensuring DAF-16A function.

Discussion
We have found that a reduction of DAF-21 capacity by RNA interference in non-neuronal tissues from the beginning of larval development shortens wild-type and extended lifespan conferred by lowered ILS. This effect persists in wild-type worms if RNAi treatment is initiated at the L4 stage and appears to operate both via daf-16 dependent and independent routes. Lifespan reduction happens without apparent developmental problems, even despite an increased heat stress tolerance and decreased fertility. Our findings therefore propose a direct role for DAF-21/Hsp90 as a longevity regulator. We have also shown that dauer formation, in contrast, requires neuronal silencing of daf-21. Finally, we have observed that daf-21 knockdown compromises oogenesis and embryonic development of the F1 generation. Thus, a spatiotemporal specificity governs the pleiotropic effects of DAF-21 to support fertility, larval development and longevity (Fig. 9). Such spatiotemporal function is also characteristic to other regulatory mechanisms. For instance, ILS during early larval development primarily affects dauer arrest, whereas during adulthood extends lifespan 44 . Likewise, neuronal daf-16 regulates dauer development in larvae and intestinal daf-16 is required for longevity during early adulthood 46 .
Our observations also suggest that the novel pro-longevity function of DAF-21 may be distinct from those controlling development although it already appears to operate during larval development. Though the identification of the mechanism(s) require further studies, potential candidates might be a less ordered organization of protein-protein and signaling networks 22 , a faster deterioration of proteostasis 59 especially in muscle, where lack of DAF-21 impairs filamental structure integrity resulting in a decline in motility relative to wild type 31 . Likewise, an important aging-associated phenomenon is immune failure and bacterial infection, which is enhanced in hsf-1, daf-16, skn-1 and also in daf-21 nematodes 60-63 .
All the proteostatic defense, tissue integrity and immunity are more robust in daf-2 mutants 19,60 , in which we have observed that daf-21 silencing during adulthood does not shorten, but marginally extend lifespan. We speculate that the reduction of DAF-21 in early adulthood acting in concert with the mild metabolic stressor daf-2 activates a pro-longevity response that antagonizes the daf-21 induced lifespan shortening effect. Potential mechanisms might involve the HSF-1 dependent heat shock response 19,27 , the differential regulation of the respective DAF-16 isoforms 20,21 (this study) or a daf-21 independent pathway. However, the validation of this idea and the identification of the real mechanism(s) await further research.
Our findings provide evidence that DAF-21 contributes to longevity via specifically ensuring DAF-16A function, reflected by the specific requirement of DAF-21 for DAF-16A nuclear translocation, transcriptional function and lifespan extension mediated by DAF-16A as the sole DAF-16 isoform. Importantly, DAF-21 was necessary for the induction of three out of four DAF-16 target genes, each of them extends lifespan: sod-3 64 , old-1 65 , scl-20 52 and gst-20 21 , the latter two identified as DAF-16A specific targets 21 (Fig. 6a-d). sod-3 has already been identified as a shared DAF-A and DAF-16D/F target 20,21 and our findings propose old-1 as a DAF-16A preferential target. In contrast, DAF-16D/F function was unaffected by daf-21 knockdown. Although DAF-16B might also be influenced by DAF-21, its involvement in the DAF-21 dependent effects is highly unlikely, because it does not regulate lifespan and is mainly expressed in neurons and pharynx that are refractory to RNAi 20,49,51,66 .
This study confirms the involvement of both DAF-16A and D/F in lifespan control when ILS is lowered 20,21 . Our experiments using transgenes made by Kwon et al. 20 do not permit us to draw a conclusion regarding the importance of the individual isoforms. A recent excellent study using single isoform mutants has shown that DAF-16D/F is only required for longevity when DAF-16A is not expressed and DAF-16A orchestrates a much more comprehensive gene expression pattern relative to DAF-16D/F (93% vs 30% of DAF-16 regulated mRNA-s) 21 . Based on this study, we propose a model: DAF-21 affecting the DAF-16A-regulated genes is required to elicit daf-2-induced daf-16-mediated longevity. If DAF-21 capacity is scarce, DAF-16A activity is compromised, DAF-16D/F may still evoke a compensatory response resulting in lifespan extension. Perhaps, this might underlie the lifespan extension and partial lifespan decrease, respectively, of daf-2 worms treated by daf-21(RNAi) in adulthood and from hatching. Other lifespan extending signals that selectively activate DAF-16D/F operate independently of DAF-21 availability. What molecular mechanism(s) may be involved in the functional interaction between DAF-21 and DAF-16A? The fact, that all DAF-21, DAF-16A and DAF-16D/F are pleiotropically expressed in almost every tissue excludes the possibility of a merely tissue-specific interaction 20,31,59 . We observed a DAF-21 dependent DAF-16A nuclear translocation in muscle, hypodermis and intestine. Previous work identified that the intestinal, but not neuronal or muscle-specific expression of DAF-16 is responsible for the longevity effect of lowered ILS 46 . Furthermore, both muscle and intestinal DAF-16 combat reproductive aging 67 . Based on this, we speculate that the longevity promoting DAF-21-DAF-16A interaction may predominantly occur in the muscle and intestine.
Apart from tissue expression the isoform-specificity of the DAF-21-DAF-16A interaction may be due to differences in the respective sequences of the DAF-16A and D/F isoforms. It is established that the two isoforms share most of their 3′ exons including the DNA binding Forkhead domain, but the 5′ region corresponding to the N-terminal of the protein products are unique 2,20 (Fig. 5a,b). A possible scenario might be that similarly to several transcription factors DAF-16A has an intrinsic instability which requires the chaperone function of DAF-21 22,23 . However, DAF-21 does not seem to stabilize DAF-16 because when daf-21 was silenced, DAF-16A did not show aggregation, preserved its protein and mRNA level and biological function in the null mutant strain of rle-1, the E3 ubiquitin ligase responsible for the polyubiquitinylation of DAF-16 58 . Moreover, silencing daf-21 did not prevent the nuclear import and transcriptional activity of the constitutively nuclear AKT-phosphorylation mutant DAF-16A AM ::GFP 49 indicating a proper folding of DAF-16A. These results also suggest that DAF-21 regulates DAF-16A activation upstream of its nuclear traffic. Promoter swapping between DAF-16A and D/F verified that the N-terminal segment of DAF-16A is responsible for the efficient nuclear entry of DAF-16 20 . Importantly, out of the three consensus AKT phosphorylation RxRxxS/T motifs conserved in human FOXOs and in DAF-16A and B, the N-terminal site of DAF-16D/F is replaced by a different QxRxxS which is probably the reason behind an asymmetrical regulation of the two isoforms by AKT-1 and AKT-2 20 . We speculate that the mechanism of the DAF-21-dependent differential interaction lies within the primary sequence of DAF-16 isoforms, subject to various regulatory modifications. One such isoform specific input has been published: TORC1 inhibition extended lifespan by inducing the nuclear translocation of DAF-16D/F without influencing DAF-16A 68 . Their and our findings lend support for an isoform-specific integration of distinct regulatory inputs by DAF-16/FOXO in longevity regulation. The biochemical nature of these inputs and the exact mechanisms of action is an important challenge for future studies.
DAF-21 is not the only heat shock protein that regulates DAF-16 nuclear traffic. HSF-1 and the constitutive HSP70 isoform HSP-1 are required for the nuclear export of DAF-16A/B in the daf-16a/b::gfp overexpressing TJ356 strain which, in turn, protects from the detrimental effect of DAF-16 hyperactivation on immunity 61 . Excessive activation of SKN-1/Nrf also led to compromised immune responses in C. elegans 62 indicating a general importance to optimize the activity of stress inducible regulators. Likewise, an adequate orchestration of the activities of distinct master regulators are crucial for efficient organismal responses, such as the collaboration of HSF-1 with DAF-16 to extend lifespan in dampened ILS mutants 19 . Intriguingly, DAF-21/Hsp90 is a prime sensor of proteostasis, a stabilizer of signaling networks 22,26 as well as a regulator of both stress-responsive transcription factors: it suppresses HSF-1 27,31,59 while supports DAF-16A activation (this study). Thus, the reciprocal regulations through DAF-21 co-ordinating DAF-16 (and HSF-1) activity may ensure an integrated organismal response sensing both the steady state proteostatic milieu and nutrient availability, which promote proper self-maintenance and on the long term, longevity.
In summary, our work uncovers a previously unappreciated role of DAF-21 in C. elegans longevity which provides a crosstalk between the proteostasis and nutrient signaling networks via an isoform specific regulation of DAF-16 activity. Considering the strong structural and functional conservation of DAF-16/FOXO isoforms 3,18 and DAF-21/Hsp90 22 , a similar regulatory mechanism might operate in mammals.
The HT1888 and KB6 strains were crossed together and the daf-16a::rfp transgene was monitored by fluorescence microscopy in the F2 generation while the presence of rle-1(cxTi510) mutation was tested using allele-specific PCR.
RNA interference. HT115(DE3) E. coli strains producing dsRNA against daf-21 (hsp90) 35 were kindly provided by Eileen Devaney (University of Glasgow, UK) and were created by cloning a 74 bp and a 294 bp region of hsp90 into the L4440 vector. daf-2(RNAi) 70  Supplementary Table S7. RNAi treatment was performed using standard RNAi feeding method as described 71 : RNAi feeding E. coli clones were grown overnight in LB medium containing 100 μg/ml ampicillin. Worms were grown on plates containing 1 mM IPTG, 50 μg/ml ampicillin and 6.25 μg/ml tetracyclin and seeded with E. coli HT115 strains harboring the L4440 empty vector (EV) control and specific RNAi vectors, respectively, from hatching -or they were transferred onto RNAi plates after reaching the L4 larval stage on OP50. Measurements were made after 2 days on RNAi bacteria. To control for proper RNAi dosage in double RNAi treatments, overnight cultures of the RNAi bacterial strains were mixed in a 1:1 ratio, mixing in the empty vector harbouring strain in single RNAi controls.
Lifespan assays. All lifespan assays were performed at 20 °C. Animals were synchronized by allowing gravid adults to lay eggs for 4 hours and using the next generation for experiments after they reached young adult stage. Approximately 35 animals were transferred to each of 3 plates containing 5-fluorodeoxyuridine (FUDR) (Sigma-Aldrich) to a final concentration of 51 µM. Day 1 is defined as the day the worms were placed on the FUDR plates. Every second day animals were scored by tapping with a platinum worm pick starting from day 7. Worms that crawled into the agar, onto the wall of the plate or died from vulval bursting were censored.
Fluorescence microscopy. After treatments indicated in the figure legends at least 50 worms per condition were placed on a 2% agarose pad, and immobilized by adding 25 mM NaN 3 in M9 buffer. Pictures were taken by a Leica DMI6000B epifluorescence microscope with a DFC480 camera (Figs 4 and S4) or by Nikon Eclipse E400 microscope with Diagnostic Instruments SPOT model 1.5.0 camera (Figs 5, S5 and S6) using GFP or RFP fluorescent filters, respectively. Epifluorescent microscopic images are representatives of at least 3 independent experiments. In the case of DAF-16 localization studies, animals were sorted into three categories: 'nuclear' refers to animals that showed exclusively nuclear localization, 'intermediate' labels animals that had both nuclear and cytosolic fluorescence and 'cytosolic' refers to animals with solely cytosolic GFP or RFP expression 72,73 . In case of RNAi treatment the animals were kept on RNAi feeding bacteria for two days after reaching L4 stage. mRNA expression analysis. mRNA from well-fed synchronized population of adult worms was isolated using GeneJET RNA Purification Kit (Thermo Scientific). The mRNA was then transcribed into cDNA by RevertAid ™ Premium Reverse Transcriptase (Thermo Scientific). qPCR measurements were performed in an ABI 7300 Real-time PCR machine using Maxima ™ SYBR Green/ROX qPCR Master Mix (Thermo Scientific).
Primer sequences are listed in Supplementary Table S7. Relative amounts of mRNAs were determined using the Comparative Cycle Treshold Method for quantitation and normalized to beta-actin mRNA levels. Each experiment was repeated three times. In case of RNAi treatment the animals were kept on RNAi feeding bacteria for two days after reaching L4 stage. Thermotolerance assay. 30 young adult animals from a synchronized population were transferred to each of 3 plates for every condition. Plates were put in an incubator preheated to 35 °C for 6 hours. Then, plates were placed back into a 20 °C incubator. Following a 5 hour recovery time the animals were scored every 24 hours by tapping with a platinum worm pick. Worms that crawled into the agar, onto the wall of the plate or died from vulval bursting were censored. All thermotolerance measurements were repeated at least three times. Statistical analysis was performed by comparing different conditions and using two-tailed Student's T-test to determine the level of significance.
Dauer assay. 10 gravid hermaphrodites were allowed to lay eggs on each plate for 4 hours. Then the plates were transferred to 25 °C and animals were allowed to grow until the third day, when adults and dauer larvae were scored from each strain and condition. Phenotypic characterization. 10 gravid hermaphrodites were allowed to lay eggs on EV and daf- 21(RNAi) plates for 1 hour in order to get a highly synchronous population of animals. The plates were kept at 20 °C and progeny was scored for phenotypic differences after three days. Pictures were taken using Nikon Eclipse E400 microscope with Diagnostic Instruments SPOT model 1.5.0 camera. Fertility assay. 10 L4 hermaphrodites from each condition of each strain were placed onto treatment plates individually. All plates were incubated at 20 °C. Every 24 hours mothers were transferred onto new plates. Progeny was scored on the plates 48 hours after the removal of the mother. The assay was carried on till the last animal stopped laying eggs.
Western Blotting. Synchronized population of animals were grown on 10 cm NGM plates with IPTG seeded with either daf-21(RNAi) or empty vector (EV). Worms were washed three times using M9 buffer and frozen at −80 °C. After thawing, 200 µl of lysis buffer (50 mM This-HCl, 0.25% SDS, 1% IGEPAL CA-630, 150 mM NaCl, 1 mM EDTA, 2x Complete (Roche)) was added to the samples. After three freeze-thaw cycles samples were sonicated 6 times for 10 seconds and centrifuged for 10 minutes at 10000 g at 4 °C. Supernatant was transferred into new tubes and was stored at −20 °C or at −80 °C for longer storage. Western blotting was carried out as previously described 74 . Samples were run in 12% poly-acrylamide gel (Bio-Rad) and transferred to nitrocellulose membrane (Bio-Rad). Blocking was done by incubation in TBS-T with 5% skim milk powder for 1 hour at room temperature. Primary antibodies were Actin (monoclonal Anti-β-Actin, Sigma-Aldrich, 1:10000 in TBS-T with 5% BSA) and Hsp90 75 (polyclonal Anti-Hsp90, 1:2000 in TBS-T with 5% BSA). Secondary antibodies were HRP-labelled anti-mouse (1:2000 in TBS-T with 5% skim milk powder) and anti-rabbit (1:3000) IgG, respectively (Dako). Membranes were incubated with ECL reagent (GE Healthcare) for 1 min and developed.