Curcumin activates Nrf2 through PKCδ-mediated p62 phosphorylation at Ser351

Curcumin, a phytochemical extracted from Curcuma longa rhizomes, is known to be protective in neurons via activation of Nrf2, a master regulator of endogenous defense against oxidative stress in cells. However, the exact mechanism by which curcumin activates Nrf2 remains controversial. Here, we observed that curcumin induced the expression of genes downstream of Nrf2 such as HO-1, NQO1, and GST-mu1 in neuronal cells, and increased the level of Nrf2 protein. Notably, the level of p62 phosphorylation at S351 (S349 in human) was significantly increased in cells treated with curcumin. Additionally, curcumin-induced Nrf2 activation was abrogated in p62 knockout (−/−) MEFs, indicating that p62 phosphorylation at S351 played a crucial role in curcumin-induced Nrf2 activation. Among the kinases involved in p62 phosphorylation at S351, PKCδ was activated in curcumin-treated cells. The phosphorylation of p62 at S351 was enhanced by transfection of PKCδ expression plasmid; in contrast, it was inhibited in cells treated with PKCδ-specific siRNA. Together, these results suggest that PKCδ is mainly involved in curcumin-induced p62 phosphorylation and Nrf2 activation. Accordingly, we demonstrate for the first time that curcumin activates Nrf2 through PKCδ-mediated p62 phosphorylation at S351.


Results
Curcumin activates Nrf2. To examine whether curcumin activates Nrf2 in neuronal cells, cells were treated with curcumin for 12 h. Treatment of neuronal cells with 10 μM curcumin did not alter their viability ( Supplementary Fig. 1). As shown in Fig. 1A, the mRNA levels of genes downstream of Nrf2 such as GST-mu1, HO-1, NQO1, p62, and NDP52 were significantly increased in cells treated with curcumin, compared to untreated controls. The expression levels of these proteins were also increased (Fig. 1B). Additionally, Nrf2 protein levels increased highly in the presence of curcumin (Fig. 1B). Immunohistochemical staining using a Nrf2-specific antibody revealed that Nrf2 was mainly localized in the nuclei of cells treated with curcumin ( Fig. 1C). Nrf2 levels also increased in nuclear fractions of cells treated with curcumin, compared to control cells not treated ( Supplementary Fig. 2). The transcriptional activity of the promoter containing the triple ARE sequences increased 2.3-fold in cells treated with curcumin (Fig. 1D). Together, these results suggest that Nrf2 is activated in neuronal cells by curcumin.
Nrf2 activation is dependent on p62. A few studies have suggested that curcumin induces autophagic cell death or apoptosis by inducing endogenous reactive oxygen species (ROS) production 26,27 . We sought to confirm whether increased endogenous ROS in curcumin-treated neuronal cells elicited Nrf2 activation. To do this, we analyzed the cellular fluorescence intensity of DCF-DA, a cell-permeable indicator for ROS, following the treatment of neuronal cells with curcumin. There was no significant difference between cells treated with curcumin and untreated control cells (Fig. 2), thereby indicating that curcumin itself did not induce cellular ROS production in neuronal cells. As shown in Fig. 1A,B, curcumin significantly increased both the mRNA and protein levels of p62, and increased Nrf2 protein levels. As p62 phosphorylation at S351 can help stabilize and activate Nrf2 protein 21 , we examined the level of phosphorylated p62 at S351 using a S351 phospho-specific p62 antibody. Results showed that the level of p62 protein phosphorylated at S351 was dramatically increased after 6 h or 12 h of curcumin treatment (Fig. 3A). Therefore, we postulated that increased phosphorylation of p62 at S351 by curcumin could stabilize and activate Nrf2 in neuronal cells.
To determine the extent to which p62 was involved in curcumin-mediated Nrf2 activation, we examined the effect of curcumin treatment on the level of Nrf2 protein in p62 knockout (−/−) MEFs. As shown in Fig. 3B, the previously observed increase in Nrf2 protein level was abated in curcumin-treated p62 (−/−) MEFs. Moreover, the NQO1 protein was not significantly induced in the p62 (−/−) MEFs, in contrast to the wild-type MEFs.
To further examine whether curcumin-induced increase of p62 phosphorylation at S351 could affect the association between Keap1 and Nrf2, we investigated their interaction using the co-immunoprecipitation test. As shown in Fig. 4A,B, the protein amount of Keap1 co-immunoprecipitated with Nrf2 in the cells treated with curcumin was about 38% compared to that in control cells not treated. Also, curcumin-induced increase in the promoter activity containing ARE elements was ameliorated in the presence of p62 mutant (S349A) (Fig. 4C). Thus, our results strongly suggest that curcumin-mediated Nrf2 activation is highly dependent on p62 and its phosphorylation at S351.
PKCδ phosphorylates p62 at S351. p62 has been reported to be phosphorylated at S351 by several kinases such as mTORC1, Tak1, CK1, and VPS34-dependent PKCδ under several conditions 17 . To determine which kinase was involved in curcumin-induced p62 phosphorylation at S351, we examined the levels of p62 phosphorylation at S351 following pretreatment with several kinase-specific inhibitors, including Torin1 (for mTORC1), PP242 (for mTORC1), (5Z)-7-Oxozeanol (for Tak1), CKI7 (for CK1), and SAR405 (for VPS34). As shown in Supplementary Fig. 3, there was no significant difference in the level of p62 phosphorylation at S351 in cells pretreated with Tak1, CK1, VPS34, and other inhibitors. However, mTORC1 inhibitors were the exception, suggesting that mTORC1 might be involved in curcumin-mediated augmentation of p62 phosphorylation at S351. Therefore, we investigated whether mTORC1 was activated in neuronal cells in the presence of curcumin, by measuring the phosphorylation levels of mTORC1 and ULK1, a protein downstream of mTORC1. Surprisingly, the levels of mTORC1 phosphorylation at S2448 and ULK1 phosphorylation at S757 were largely unchanged by curcumin ( Supplementary Fig. 4), thereby indicating that mTORC1 might not be activated by curcumin in neuronal cells.
To determine whether PKCδ was involved in curcumin-induced phosphorylation of p62 at S351, we first investigated the activation of PKCδ using a T505 phospho-specific PKCδ antibody. As shown in Fig. 5A,B, PKCδ phosphorylation levels were increased in cells treated with curcumin, and peaked after 12 h of curcumin treatment, thereby indicating that PKCδ was activated in cells in the presence of curcumin. Notably, the amount of PKCδ which co-immunoprecipitated with p62 was also increased in cells treated with curcumin (Fig. 5C). Additionally, the level of p62 phosphorylation at S349 was enhanced in cells transfected with a PKCδ expression plasmid (Fig. 5D), but was diminished in cells with a PKCδ-specific shRNA expression plasmid (Fig. 6B). To confirm this, we pretreated neuronal cells with PKCδ-specific siRNA, and examined the level of p62 phosphorylation at S351 following curcumin treatment. As expected, p62 phosphorylation at S351 was abrogated in www.nature.com/scientificreports/ cells pretreated with PKCδ-specific siRNA (Fig. 6A). Taken together, these results strongly suggest that PKCδ is directly involved in the phosphorylation of p62 at S351 in the presence of curcumin.
VPS34 is not involved in p62 phosphorylation at S351. A previous study has suggested that VPS34 induces PKCδ-dependent phosphorylation of p62 at S351 24 . Therefore, we examined whether VPS34 was also involved in curcumin-induced p62 phosphorylation at S351 via PKCδ. To accomplish this, we determined the protein level of VPS34 in neuronal cells following curcumin treatment. The expression level of VPS34 in neuronal cells increased in the presence of curcumin (Fig. 7A). Unexpectedly, however, curcumin-mediated phosphorylation of p62 at S351 was not inhibited, but increased by transfection with VPS34-specific siRNA (Fig. 7B). Considering our results shown in Figs. 5 and 6, it could be expected that the increase of phosphorylated p62 at S351 in the cells treated with VPS34-specific siRNA might be resulted from an increased activity of PKCδ. To www.nature.com/scientificreports/ confirm it, we examined phosphorylation levels of PKCδ at T505 (active form) in the cells treated with VPS34specific siRNA. As a result, any significant change in the phosphorylation levels of PKCδ in the cells treated with VPS34-specific siRNA was not observed compared to those in cells treated with scramble siRNA as a control, suggesting that PKCδ may be not involved in the increase of p62 phosphorylaton level by VPS34 knockdown (Fig. 7B). Additionally, p62 phosphorylation at S351 was induced in cells transfected with PKCδ, but the phosphorylation level was abated by co-expression of VPS34 (Fig. 7C). Altogether, these results demonstrate that curcumin induces PKCδ-mediated p62 phosphorylation at S351 but in a VPS34-independent manner.

Discussion
Nrf2 can be targeted pharmacologically in a variety of diseases, including neurodegenerative, vascular, and metabolic diseases underlined by oxidative stress and inflammation 28,29 . In the last two decades, many organic compounds have been developed as Nrf2 activators or inhibitors, some of which are under clinical trials for their therapeutic effect on different chronic diseases 16 . Among them is the natural compound curcumin, an electrophilic Nrf2 inducer and a ROS scavenger [30][31][32] . However, the exact mechanism by which curcumin activates Nrf2 remains to be explored. In the present study, we provide the first evidence that curcumin activates Nrf2 via PKCδ-mediated p62 phosphorylation at S351, which would inhibit the ability of Keap1 to trap Nrf2, resulting in Nrf2-stabilization and activation. Given that Nrf2 also induces the expression of p62 through transcriptional activation 25 , we speculate that Nrf2 activation by curcumin would be further enhanced by this positive feedback loop (Fig. 8).
PKCδ is ubiquitously expressed and involved in numerous signal transduction pathways related to cell proliferation and differentiation, inflammation, and apoptosis 33,34 . Various stimuli lead to PKCδ activation through phosphorylation or proteolytic cleavage to an active fragment 35,36 . It has been shown that PKCδ requires phosphorylation at S643, S662, and T505 for its full activation 37 , and of these three sites, phosphorylation at S662 and T505  www.nature.com/scientificreports/ is mediated by PKCζ or mTOR and PDK1, respectively [37][38][39] . We found that curcumin increased the level of PDK1 phosphorylation at S241 with a similar time point as when PKCδ is phosphorylated at T505 ( Supplementary  Fig. 5A). Therefore, it seems that curcumin-activated PDK1 may participate in the activation of PKCδ, leading to the phosphorylation of p62 at S351. This notion is consistent with that of a previous study which suggests that curcumin activates Nrf2 in human monocytes via PKCδ 40 . Indeed, our data showed that curcumin induced the activation of PKCδ in different cells. Additionally, PKCδ is known to directly phosphorylate Nrf2 at S40, which is required for the dissociation of Nrf2 from Keap1 41 . Therefore, it is expected that curcumin-mediated PKCδ activation not only phosphorylates p62 at S351, but also phosphorylates Nrf2 at S40, thus cooperatively enhancing the dissociation of Nrf2 from Keap1. In addition to PKCδ, several kinases such as mTORC1, CK1, and Tak1 have been shown to participate in the phosphorylation of p62 at S351 under normal cellular conditions 21,23,24 . We found that, in neuronal cells treated with curcumin, the level of p62 phosphorylation at S351 was not changed by pretreatment with (5Z)-7-Oxozeanol, a Tak1 inhibitor, and CKI7, a CK1 inhibitor. In contrast, the level of p62 phosphorylation at S351 was attenuated in the presence of mTORC1 inhibitors such as PP242 or Torin1 ( Supplementary Fig. 3), suggesting that mTORC1 might be involved in p62 phosphorylation. However, accumulating evidence suggests that curcumin suppresses mTORC1 activation 42,43 . In this study, there was no significant change in the phosphorylation levels of ULK1 at S757, a substrate of mTORC1 and of mTORC1 at S2448, in cells treated with curcumin, compared with untreated control cells (Supplementary Fig. 4). Thus, we postulate that phosphorylation of p62 at S351 occurs continuously through the basal activity of mTORC1, rather than via induction by curcumin. However, given that PKCδ directly phosphorylates p62 at S351 (Figs. 5, 6), it is possible that the phosphorylation of p62 at S351 may be marginally inhibited by mTORC1 inhibitors because the phosphorylation of PKCδ at S662 by mTORC1 is necessary for its full activation as previously described 37 .
VPS34, a class III lipid kinase generating PI(3)P, is essential for protein sorting to the vacuole via the endolysosomal pathway 44,45 . Interestingly, a previous study suggested that VPS34 promotes the association of PKCδ with p62, thus enhancing the phosphorylation of p62 at S351 24 . Here, we observed increased expression of VPS34 protein following curcumin treatment (Fig. 7A); however, pretreatment of neuronal cells with SAR405, a VPS34specific inhibitor, did not affect curcumin-induced phosphorylation of p62 at S351 (Supplementary Fig. 3). Additionally, ectopic expression of VPS34 and knockdown of VPS34 using a VPS34-specific siRNA reduced and augmented the level of p62 phosphorylation, respectively (Fig. 7B,C). Intriguingly, VPS34 likely plays a negative regulator in curcumin-induced p62 phosphorylation at S351 according to our results (Fig. 7). Our www.nature.com/scientificreports/ findings suggest that PKCδ-mediated phosphorylation of p62 at S351 is not dependent on VPS34, in contrast to the previous report 24 . p62 undergoes extensive post-translational modifications including phosphorylation, and interacts with various signaling molecules, thereby allowing fine-tuned regulation of its function 17 . The phosphorylation of p62 at S351 increases its binding affinity with Keap1, thereby activating Nrf2 via a non-canonical mechanism 21,22 . Importantly, the p62 S351 is phosphorylated in selective autophagy conditions such as impaired proteostasis, mitochondrial depolarization, and bacterial infection 21,46,47 . In a previous study, we observed increased phosphorylation of p62 at S351 in addition to Nrf2 activation in TFEB-expressing cells 48 . Thus, the Keap1-Nrf2 pathway and selective autophagy, the major stress response pathways, are coupled to each other via the phosphorylation of p62 17 . We also observed that the addition of curcumin to neuronal cells markedly induced p62 phosphorylation at S403 in its ubiquitin-associated (UBA) domain, which is known to increase the affinity between UBA and polyubiquitin chain 49 (data not shown). Consequently, p62 associated with Keap1 via the P-STGE motif is expected to be translocated to the ubiquitinated cargos in the presence of curcumin. Then, Nrf2 free from Keap1 is accumulated in the cells, thus resulting in Nrf2-activation. To further prove that PKCδ is directly involved in p62 phosphorylation at S351, we pretreated neuronal cells with various PKCδ chemical inhibitors including rottlerin, bisindolylmaleimide I, and Gӧ6983 etc., before curcumin treatment. Unexpectedly, the phosphorylation level of p62 at S351 was highly increased with only treatment of PKCδ inhibitors to the cells. Given that rottlerin is well known as a strong uncoupler 50 , the effect was comprehensive, but not in cases of other inhibitors. PKC inhibitors have a wide spectrum of inhibitory activity for PKCs; therefore, it seems that the inhibition of PKC signaling raises some stress to cells, and then resulting in p62 phosphorylation at S351. The precise mechanism remains to be elusive.
Nrf2 activity is tightly controlled at several steps via its ubiquitination and proteasomal degradation. To date, the following three types of E3 ubiquitin ligases have been reported to participate in Nrf2 degradation by the proteasome, independently of Keap1: β-TrCP (β-transducin repeat-containing protein) 51 , DCAF11 (DDB1 and Cullin4 associated factor 11, also referred to as WDR23) 52 , and Hrd1 (HMG-CoA reductase degradation 1 homolog, also referred to as synoviolin) 53 . β-TrCP interacts with the Neh6 domain phosphorylated by GSK-3β, www.nature.com/scientificreports/ while DCAF11 binds to the DIDLID sequence of the Neh2 domain in Nrf2 52,54 . Some research groups have suggested that the PI3K-AKT signaling pathway is involved in curcumin-induced Nrf2 activation 6,31 . Since the phosphorylation of AKT and GSK-3β was increased in neuronal cells treated with curcumin ( Supplementary  Fig. 5A), we speculate that β-TrCP-mediated Nrf2 degradation may be inhibited by the decreased activity of GSK-3β. However, cells pretreated with AKT inhibitor IV, surprisingly exhibited higher levels of Nrf2 protein and p62 phosphorylation at S351 (Supplementary Fig. 5B). Also, the increased level of Nrf2 in the presence of curcumin was not ameliorated by transfection of constitutively active GSK-3β ( Supplementary Fig. 6). Thus, it seems that the PI3K-AKT/GSK-3β signaling pathway is, if any, partially involved in Nrf2 activation by curcumin. Interestingly, in the late-stage following curcumin treatment, the level of β-TrCP protein dramatically increased, whereas that of Nrf2 decreased ( Supplementary Fig. 8), thereby indicating that β-TrCP played a critical role in the termination of curcumin-induced Nrf2 activation by promoting Nrf2 degradation. Although curcumin has been known to be passed through the blood brain barrier 4,5 , there are two imitations: poor solubility in water and low permeability 55 . In fact, about 96% of oral curcumin is excreted in the feces and urine after 72 h in rats 56 . Interestingly, by a single oral administration of 5 mg curcumin in mice, approximately 5 ng of curcumin in one gram brain tissue was detected after 8 h, corresponding to 13.5 nM 57 . We observed that 1 μM curcumin was able to significantly increase Nrf2 protein level and ARE promoter activity (above 60%) ( Supplementary Fig. 7). Here, we used 10 μM curcumin to investigate whether it activates Nrf2, enhancing the expression of its down-stream genes. Even though the permeability could be different according to species, our study supports that taking 5 g of curcumin via the oral route could lead to 13.5 μM curcumin in brains. Recently, to increase the transport efficiency of curcumin to the brain, a variety of formulations have been developed 58 , which is expected to help to increase its possibility in successful prevention and therapeutic use for brain diseases.
Abundant preclinical and clinical evidence indicates that curcumin has potential as a therapy for various chronic diseases including neurodegenerative and inflammatory diseases 2 . Unlike other drugs, it can impact a diverse range of targets and signaling pathways 2,12 . To date, several studies on the activation of Nrf2 by curcumin have been reported, but the exact mechanism of activation remains controversial. Here, we have proposed a novel, non-canonical mechanism whereby curcumin activates Nrf2 via p62 phosphorylation at S351 through PKCδ activation. Thus, our results provide a scientific clue to understand the biological activity of curcumin in cells and to develop this natural compound as a therapeutic drug for chronic diseases, including neurodegenerative diseases.
Cell culture. Cells were cultured as described in the previous studies 48,59    Immunohistochemical staining. Immunohistochemical staining was performed as described in the previous study 48 . Mouse cortical cells were fixed on coverslips with 4% paraformaldehyde for 30 min. After cell permeabilization, the cells were incubated with a blocking solution (2% normal goat serum, 0.1% Triton-X-100 in PBS) for 1 h. After briefly washing with PBS, the cells were probed with rat anti-Nrf2 (1:250) antibody diluted in the blocking solution at 4 °C overnight. Then, the cells were incubated with donkey anti-rat Alexa488 conjugated antibody (1:500) for 1 h at room temperature. The coverslips were mounted on glass slides with the ProLong Gold Antifade Reagent (P36935, Invitrogen) and analyzed by the Olympus FV1000 laser scanning confocal microscope.
Quantitative real-time PCR. For cDNA synthesis, total RNA was extracted from cells using the RNeasy Mini Kit (Qiagen) according to the manufacturer's instructions. cDNA was synthesized from 2 μg of RNA using RT-PCR EcoDry Premix (Clontech). Quantitative real-time PCR (qRT-PCR) was performed using the SYBR Green Real-Time PCR protocol (4344463, Invitrogen) on the QuantStudio 6 Flex Real-Time PCR System (Applied Biosystems) as previously described in the study 48