Activation of PKC supports the anticancer activity of tigilanol tiglate and related epoxytiglianes

The long-standing perception of Protein Kinase C (PKC) as a family of oncoproteins has increasingly been challenged by evidence that some PKC isoforms may act as tumor suppressors. To explore the hypothesis that activation, rather than inhibition, of these isoforms is critical for anticancer activity, we isolated and characterized a family of 16 novel phorboids closely-related to tigilanol tiglate (EBC-46), a PKC-activating epoxytigliane showing promising clinical safety and efficacy for intratumoral treatment of cancers. While alkyl branching features of the C12-ester influenced potency, the 6,7-epoxide structural motif and position was critical to PKC activation in vitro. A subset of the 6,7-epoxytiglianes were efficacious against established tumors in mice; which generally correlated with in vitro activation of PKC. Importantly, epoxytiglianes without evidence of PKC activation showed limited antitumor efficacy. Taken together, these findings provide a strong rationale to reassess the role of PKC isoforms in cancer, and suggest in some situations their activation can be a promising strategy for anticancer drug discovery.

www.nature.com/scientificreports/ PKC signaling and then induce proteolytic degradation of activated PKC isoforms even after short exposure times 13 . It is therefore extremely challenging to develop pharmacological agents capable of enhancing PKC signaling in a sustained way, while avoiding down-regulation or degradation 11,12 . Tigilanol tiglate (TT, also known as EBC- 46), is an epoxytigliane isolated from the seed of the native Australian rainforest plant Fontainea picrosperma (Euphorbiaceae) 14 . The presence of an α-oriented ring B epoxide, common in daphnane diterpenoids, is extremely rare in phorbol esters and complex to install by semi-synthesis from tiglianes 15 , explaining the paucity of data on the bioactivity of these natural products. TT was previously found to be an activator of Protein Kinase C (PKC) isoforms, and a single intratumoral dose of TT caused hemorrhagic necrosis and tumor ablation in melanoma, HNSCC and other mouse models of cancer 16,17 . TT was also found to induce a respiratory burst from human polymorphonuclear cells, and cause increased permeability of human umbilical vein endothelial cell monolayers. Furthermore, the anti-cancer mechanism of action of TT was found to be at least in part PKC-dependent, as the pan-PKC inhibitor bisindolylmaleimide-1 (BIS-1) partially inhibited efficacy 16 .
TT has recently been approved by the European Medicines Agency as a veterinary pharmaceutical for intratumoral treatment of non-metastatic, non-resectable mastocytomas in dogs. The drug is also under clinical evaluation for treatment of a range of other cutaneous and subcutaneous cancers in humans 18,19 , and companion animals 20 , and is showing a promising safety and efficacy profile. For example, in a Phase III veterinary trial for treatment of canine mastocytomas, complete and enduring response was recorded in 60 out of 80 patients 21 , while in a Phase I human safety/dose-escalation study, maximum tolerated dose was not reached and signs of efficacy were observed in 9 tumor types, including complete response in 4 patients 22 .
The clinical efficacy of a PKC-activating drug in treating a range of cancers provided a unique opportunity to gain further insights both into the complex role of PKC in cancer and into fundamental structure-activity relationships of epoxytiglianes relevant to their future development as new drug candidates. To address this potential, a detailed phytochemical investigation to obtain a library of epoxytiglianes and closely-related derivatives from F. picrosperma was undertaken. We present evidence of a link between activation of PKC isoforms and tumor regression after intratumoral administration.

Results
TT (EBC-46) and family member structure elucidation. Prior to this study only the flat structure of TT had been reported previously 14,16 , hence it was critical to determine the full stereochemical structure for stringent biological interpretation. In brief, a suitable crystalline derivative (1) (Fig. 1A) of TT was obtained using Appel conditions 23 , which confirmed all unassigned asymmetric centers and in turn the reported tigliane biosynthetic pathway 24 . Interestingly, x-ray crystallographic analysis of the tigliane system has only been sparsely reported [25][26][27] reinforcing precise elucidation challenges for this class of diterpene.
Structure-activity analysis and assessment of cell growth inhibition of epoxytiglianes. The sixteen highly novel tiglianes, and TT (EBC-46) as the benchmark, were broken into two groups for structureactivity analysis and assessment of inhibition of cancer cell growth.
Group 1 contained an (S)-2-methylbutyrate at C13, allowing comparisons to focus on the length, branching and unsaturation of the C12 esters (Fig. 2C). The known growth inhibition of the K562 human myelogenous leukemia cell line at nM levels by the prototypic PKC activator PMA (Fig. 2B) 30 was used to compare antineoplastic potential, integrating possible differences in drug uptake and PKC activation in vitro. The latter property was confirmed with the pan-PKC inhibitor bisindolylmaleimide I (BIS-1; Fig. S3), in line with our previous findings 16 .

PKC isoform translocation and signaling induced by epoxytiglianes. Transient transfection of
HeLa cells with PKC isoforms tagged to EGFP and subsequent treatment with either 500, 50 or 5 nM PMA, or the epoxytiglianes and analogues, for 1 h was used to assess the PKC isoform translocation profile of the selected derivatives (Fig. 4A). As expected, efficient translocation to the plasma membrane of the classical isoforms − α, − βI, − βII and − γ, as well as the novel − δ, − θ, and − η subtypes was observed with treatment with PMA (Fig. 4B). Interestingly, differing patterns of translocation were observed between high and moderate concentrations of PMA for PKC isoforms − α and − γ, where a mixed perinuclear/plasma membrane pattern was seen at 500 nM ( Supplementary Fig. S4). No translocation was seen for the atypical PKC-ζ isoform lacking a C1 DAG-binding domain after treatment with PMA. Treatment with TT was consistent with previous studies 16 , with a high percentage of cells showing translocation of PKCβI and − βII isoforms, and less translocation seen   B l a n k www.nature.com/scientificreports/ for the − α, − γ, − δ, and − θ isoforms at both 500 and 50 nM (Fig. 4B). No translocation of the atypical PKC-ζ isoform was observed following TT treatment. In general, the 6,7-epoxytiglianes showed preference for translocation, and therefore activation, of PKC-βI and -βII isoforms. Additionally, the longer the carbon chain ester at C12, the more translocation was observed for PKCθ, with the exception of the branched chain case EBC-47. B-ring modifications, either type 1 (e.g., EBC-211) or type 2 (e.g., EBC-158) showed minimal translocation of any PKC isoforms in this assay ( Fig. 4B; full data shown in Supplementary Fig. S4), supporting the above data. Assessing translocation of PKC-ε was not possible due to the toxicity of high levels of expression of this protein to cells following transient transfection, as previously reported 16 . Importantly, the observed translocation data likely corresponded with PKC activation, as treatment with the active analogues at the same time points showed increased phosphorylation of ERK1/2 in both MM649 (Fig. 4C) and HeLa cells ( Supplementary Fig. S5), as well as increased cytokine release from peripheral blood mononuclear cells ( Supplementary Fig. S6).
Selected epoxytigliane analogues show anticancer efficacy in a xenograft melanoma model. Tumors established in BALB/c Foxn1 nu mice ("nude" mice, lacking T-cells) from intradermal injection of the human melanoma cell line MM649 were employed as a model for comparing the efficacies of TT and selected analogues administered by intratumoral injection. A dose response with TT showed that 27 nmol/ tumor site, lower than that used previously 16 , was a suitable suboptimal level to allow comparison with analogues of higher or lower potency in formed tumors (average 100 mm 3 ). Treatment with vehicle alone (40% propylene glycol in water) led to no decrease in tumor volume (Fig. 5B), with all tumors remaining over 100 mm 3 (Fig. 5A). Treatment with TT (27 nmol) led to rapid reduction in tumor volume (Fig. 5C), with a single tumor regrowing 20 days after initial treatment (n = 14 tumors). Loss of the C12 ester (i.e., EBC-186) led to total loss of antitumor potency (Fig. 5D). For straight chain C12 esters, the butanoyl (EBC-148; Fig. 5F) was equivalent in potency to TT, whereas the acetyl (EBC-147) and hexanoyl (EBC-83) moieties (Fig. 5E,G, respectively) were less potent than TT with relapse in all but one tumor in each case. The result for the hexanoyl was somewhat surprising compared to its significant potency in vitro (Fig. 2E). EBC-47 (Fig. 5H) had similar potency to TT, with only a single tumor recurrence (n = 10 tumors). As found for their bioactivities in vitro, both of the B-ring modified derivatives EBC-211 (Fig. 5I) and EBC-158 (Fig. 5J)   www.nature.com/scientificreports/ to correlate with in vivo tumor ablation efficacy. PMA was again used as a positive control, and showed reduction of protein levels of PKCβ following 24 h treatment of MM649 melanoma cells, likely from proteolytic degradation as previously reported 37 . In addition, levels of phosphorylated MARCKS, a direct PKC substrate, were also significantly reduced following 24 h of treatment with PMA (Fig. 6A). As per previous studies, 1 or 24 h exposure, and subsequent washing out of PMA, along with further incubation for 24 h clearly demonstrated extended decreased levels of PKCβ and phosphorylated MARCKS. Notably, this pattern was replicated by EBC-83, an epoxytigliane demonstrating modest efficacy in vivo. However, exposure of other 6,7-epoxytiglianes to MM649 melanoma cells (Fig. 6A) did not lead to short term loss of PKC isoforms. While some reduction in phosphorylated protein level of direct downstream targets was observed with extended treatment of the most potent epoxytiglianes (i.e., TT or EBC-47 at 24 h treatment and 24 h recovery), detectable levels of PKC isoforms were maintained for those exhibiting the highest efficacy against MM649 tumors in vivo. No reduction in protein level of the atypical isoform PKCι was observed following short or prolonged treatment with any epoxytigliane or analogue, suggesting the effect is mediated by the PKC C1 domain. The ongoing presence of PKC isoforms following treatment with active epoxytiglianes was not unique to MM649 cells; treatment of SK-MEL-28 melanoma (Fig. 6B) or B16-F0 murine melanoma cells (Fig. 6C) with TT showed maintenance of PKC isoform protein level compared to PMA even with 24 h exposure and 24 h recovery from treatment.

Discussion
The hallmark of most tigliane diterpenoids from F. picrosperma is the presence of multiple oxidation on the B-ring and a tigliate group on the ring C13-hydroxyl, with differences limited to the acyl decoration of the 12-hydroxyl and the glicidyl/triol nature of the trioxygenated ring B fragment. The configuration of the trioxygenated fragment on the B-ring was critical for bioactivity, since the 5-hydroxy-6,7-epoxy motif outperformed its regioisomeric (5,6-epoxy-7-hydroxy)-and the ring opened (triol) examples, the latter being almost inactive. Within the linear C12 alkyl esters, growth inhibition increased with chain length, as found in the phorbol esters [31][32][33] , but this relationship did not hold for alkenyl esters, where restricted conformational flexibility might interfere with membrane insertion/transport. The (S)-2-methylbutanoyl is a notable motif for both C12 and C13 esters. This residue had a significant influence on bioactivity, presumably related to an increasing hydrophobicity and more limited conformational freedom close to the polycyclic skeleton, an increased stability to esterase attack in vivo 38 , or a combination of both effects. The test cell line K562 expresses both classical and novel PKC isoforms, similar in expression pattern to MM649 (Supplementary Fig. S3C). However, these cell lines vary greatly in their sensitivities to growth inhibition by epoxytiglianes, implicating potential differences in downstream signaling or activation of alternative C1 domain proteins in addition to PKC activation. Nevertheless, a dissection between growth inhibition in vitro and antitumor potency was observed, the latter peaking with efficacy of TT. In vitro activity data on a variety of cell types could be rationalized in terms of ester hydrophobicity and 6,7-epoxide-dependent PKC activation, but antitumor action selectively emphasized the 6,7-epoxide structural element, being optimal for TT and EBC-148, but declining with higher or lower levels of ester hydrophobicity, the latter attributed to less PKC activation. However, the effect on tumor growth and cell viability in vivo may not be a direct effect of the compounds on the tumor cells, given the apparent insensitivity to the compounds in vitro. These results likely indicate that in vivo efficacy may act through other mechanisms than those seen in vitro. There are many instances where sensitivity to anti-cancer drugs differ between in vitro and in vivo or 3D situations [39][40][41] . This apparent contradiction between sensitivities to this class of natural products in vitro versus in vivo has been previously observed for TT 16 . We previously identified MM649 cells to be relatively insensitive to TT in vitro. In contrast, treatment in vivo with 50 nmol TT lead to 90% ablation of MM649 tumors. Subsequent experiments showed a rapid loss of cell viability when tumors were treated in vivo, potentially due to the effect of TT on intratumoral vasculature 16 . Moreover, the in vivo efficacy of TT was shown to be, at least in part, attributable to activation of PKC. Treatment with BIS-1 prior to TT significantly reduced the efficacy of treatment, demonstrating the role of PKC 16 .
Previous studies have outlined the role of neutrophils in antitumor efficacy of PKC agonists 42,43 . In keeping, intralesional injection of active epoxytiglianes into tumors in BALB/c Foxn1 nu mice induces an initial swelling that recedes after days, potentially suggesting influx of components of the innate immune system. Our previous work with TT showed a minor role for neutrophils in antitumor efficacy with the use of blocking antibodies 16 . However, recent results have potentially highlighted the recruitment of neutrophils following treatment as an important factor in TT efficacy 17 . Therefore, the efficacy of the active epoxytiglianes may be due to the effects of the class on the innate immune system, in particular recruitment and activation of neutrophils, and warrants further investigation.
The TT analogues with modest activity in K562 growth inhibition, translocation of PKC and subsequent downstream phosphorylation of signaling pathways, resulted only in an initial tumor volume reduction (i.e., EBC-147, Fig. 5E). Whereas family members with low bioactivity against PKC and downstream signaling, showed no reduction of tumor volume following intralesional treatment (i.e. EBC-211, Fig. 5I and EBC-158, Fig. 5J). The 6,7-epoxytiglianes showing strong K562 growth inhibition, translocation of PKC and phosphorylation of signaling pathways could ablate established tumors upon intralesional injection (i.e., TT, Fig. 5C; EBC-47, Fig. 5H). Epoxytiglianes with acetyl (EBC-147; Fig. 5E) and hexanoyl (EBC-83; Fig. 5G) moieties showed reduced in vivo efficacy although they exhibited PKC activation. The loss of antitumor potency by EBC-83, the n-hexanoate analogue of TT, was surprising in the light of its higher hydrophobicity and PKC activation capacity compared to TT. While the reason for these discrepancies are not immediately clear, it appears that EBC-147 does not elicit as strong PKC activation or downstream signaling as EBC-148 or TT (Fig. 1, Figs. S5 and S6). It is also possible that tumor ablation in the melanoma xenograft model may be less dependent on PKC activation and more reliant on alternate targets, potentially including but not limited to those proteins containing a C1 domain for this  Figure 6. Limited PKC isoform degradation in cells following selected epoxytigliane treatment. (A) Western blot analysis of MM649 melanoma cells treated with 500 nM epoxytigliane analogues for the indicated times before harvesting. W-cells exposed to the epoxytiglianes for the indicated times before removal, washing and normal media for an additional 24 h prior to harvesting. (B,C) Western blot analysis of (B) SK-MEL-28 human melanoma or (C) B16-F0 murine melanoma cells treated with 500 nM TT (EBC-46) or PMA for the indicated times prior to harvesting. W-cells exposed to the epoxytiglianes for the indicated times before removal, washing and normal media for an additional 24 h prior to harvesting. www.nature.com/scientificreports/ group of natural epoxytiglianes and analogues. Further investigation of novel targets of the epoxytiglianes, as well as activation of alternate downstream signaling pathways is therefore needed. The role of PKC activation in the antitumor activity of other phorboids is unclear. The 12-deoxyphorbol ester prostratin has been reported to reduce tumor volume in mouse models of pancreatic cancer 44 , but is only a weak modulator of PKC, while the activity of ingenol-3-angelate (marketed as ingenol mebutate or Picato used for treatment of actinic keratosis 45,46 and basal cell carcinoma 47 ) may also involve additional targets 48 .
In vivo antitumor efficacy of epoxytiglianes is consistent with a role for PKC activation, as evidenced by isoform translocation, activation of downstream signaling and increased cytokine production. Remarkably, and in sharp contrast with the archetypal phorbol ester PMA, treatment with the most efficacious epoxytiglianes was associated with retention of detectable intracellular levels of the PKCβ isoforms. The molecular details of this biased activation are unknown, but tumor ablation in the melanoma xenograft model appears to be dependent on PKC activation in addition to longevity of isoform stability and downstream signaling for this group of natural epoxytiglianes and analogues. In addition, activation of PKCθ mostly corresponded with in vivo efficacy. Interestingly, both PKCθ and PKCβ are known to activate NF-κB signalling 49,50 . Importantly, the epoxytiglianes without evidence of PKC activation showed limited antitumor efficacy. These combined factors are suggestive of a proposition that the potent antitumor activity observed with several of the epoxytiglianes arises from stimulation of PKC signaling. This hypothesis is supported by recent findings suggesting particular PKC isoforms may act as tumor suppressors, and not as oncoproteins as long thought 5,11,12 and that in certain contexts, PKC may be considered a tumor suppressor 5,51 . The role of PKC isoforms in cancer is extremely complex and often contradictory, likely due to the effects being cell lineage specific. For example, PKCα inhibits cell motility in breast cancer 52 , but promotes migration of colon cancer 53 . Further, over-expression of PKCβ promotes breast cancer cell growth 54 , while other studies suggest PKCβ expression reduces breast cancer tumor formation 55 . Complete understanding of the specific roles of different PKC isoforms in cancer formation and progression remains an outstanding goal. In any case, our results suggest a reassessment of the role of PKC in cancer may be necessary, refocusing drug discovery efforts on activation rather than inhibition of a specific set of isoforms of this family of enzymes.
PKC translocation assays. HeLa  Mitogen activated protein kinase (MAPK) pathway activation. Quantitative detection of mitogen activated protein kinase (MAPK) pathway activation in response to compound addition was performed using AlphaLISA SureFire Ultra p-ERK1/2 (T202/Y204) and total ERK1/2 kits as per the manufacturer's instructions (Perkin-Elmer). Here, 2.5 × 10 4 HeLa or MM649 were seeded into a 96-well plate (Corning #3595) in media (100 μl) and incubated overnight at 37 °C in 5% CO 2 . After 24 h, cells were washed × 1 with PBS and treated with 500, 50 and 5 nM of compound, prediluted in media (100 µl). Cells were incubated for 1 h with compound as above, after which media was aspirated, cells washed once with PBS and lysed using 50 μl of kit derived lysis buffer. 10 μl of sample was incubated with 5 μl of donor and acceptor mix for 2 h prior to measurement using an Enspire alphascreen reader (Perkin-Elmer). Mean fluorescence intensity (MFI) values from phospho-ERK experiments were normalized to MFI values from total ERK (p-ERK F /ERK F ) for each compound.
Analysis of cytokine release from PBMCs. Human peripheral blood mononuclear cells (PBMCs) were isolated from heparinised blood (acquired from 3 different donors) by Ficol-Paque sedimentation. Briefly, whole blood was diluted 3:1 in prewarmed RPMI (no FCS) and layered on top of Ficol-Paque. Samples were centrifuged at 400×g, giving a red blood cell pellet and PBMC layer. The PBMC layer was harvested, washed × 3 with RPMI-1640 (no FCS) and finally resuspended in complete media. PBMCs were seeded at a density of 1.5 × 10 5 cells per well in media and stimulated with compound at three concentrations (500, 50, 5 nM) in duplicate for 24 h. Media samples were taken from each of the required wells and frozen at − 80 °C until use. Each media sample was assayed for the presence of soluble IL-1β, IL-6, IL-8, IL-10, IL-12p70 and TNF using a CBA Human Inflammatory Cytokine Detection Kit according to the manufacturers' instructions (Becton Dickinson). Mean fluorescence intensity values from each sample were compared against a standard curve to determine cytokine concentrations in cell culture supernatants (pg/ml ± S.D.) using FCAP Array analysis software (v3.0; http://www. bdbio scien ces.com/cba).

Xenograft tumor studies.
Xenograft tumor studies were performed as previously described. Briefly, MM649 tumors (2 × 10 6 cells in initial injection, 2 sites per BALB/c Foxn1 nu mouse) were grown to approximately 75-100 mm 3 , when mice were treated with either vehicle (40% propylene glycol in water, 50 µl), or 25 nmol of compound in vehicle (50 µl), via a single intra-tumoral injection. Tumor volume was measured twice weekly, recorded using digital calipers and expressed as mm 3 according to the formula A × b × b × 0.5 (A = length, b = breadth).
Human ethics statement. This

Data availability
Crystallographic data including structure factors in CIF format have been deposited with the Cambridge Crystallographic Data Centre (CCDC 1889589).