Potent Anti-Cancer Properties of Phthalimide-Based Curcumin Derivatives on Prostate Tumor Cells

Metastatic castration-resistant prostate cancer is commonly treated with chemotherapy, whose effect is less than satisfactory. This raised the need for novel agents for the treatment of prostate cancer. In the present study, five phthalimide-based curcumin derivatives were synthesized and completely characterized to assess improved stability, pharmacodynamics, and radical scavenging ability. To investigate the potential application in anti-cancer therapy, the anti-proliferative activity of the synthesized molecules was determined on aggressive prostate tumor cells. We demonstrated that the K3F21 derivative has increased potency compared to curcumin, in terms of GI50, anti-proliferative and anti-migrating activities. K3F21 inhibits anchorage-dependent and -independent growth of prostate cancer cells by altering the expression of key genes controlling cell proliferation, such as Cylins D1, B1 and B2, and apoptosis, among which Puma, Noxa, and Bcl-2 family members. Finally, the anti-cancer activity of K3F21 was demonstrated by the analysis of cancer-associated PI3K/AKT, ERK, and p38 signaling pathways.


Introduction
Prostate cancer (PCa) is the second most commonly diagnosed cancer and leading cause of cancer death in western males [1]. The progression from normal prostate to prostatic intraepithelial neoplasia (PIN) and localized adenocarcinoma occurs over many years. Adenocarcinoma is a slow-growing tumor and progression to advanced and metastatic cancer is a relatively late process. Androgen deprivation therapy (ADT), the gold standard for the treatment of advanced PCa, has initial response rates of about 80-90%, but nearly all men eventually develop castration-resistant PCa (CRPC), which is currently incurable [2]. CRPC is commonly treated with different approaches: taxane chemotherapy (such as docetaxel and cabazitaxel), secondary hormonal therapeutic agents (abiraterone and enzalutamide), cellular immunotherapy (sipuleucel-T), or radiotherapy (radium-223). Despite this, inevitable progression of disease frequently occurs [3,4]. A considerable number of novel agents are currently under investigation worldwide to cure PCa and metastatic CRPC [5].
Curcumin (1,7-bis [4-hydroxy-3-methoxyphenyl]-1,6-heptadiene-3,5-Dione), a polyphenol obtained from the rhizome Curcuma longa L., and its analogs have shown anti-cancer properties by suppressing tumor initiation and progression [6,7], through the modulation of multiple signaling pathways and the inhibition of cell proliferation, invasion, metastasis, and angiogenesis [8]. pH-metric titrations were performed in order to assess K3F acidity and predict the most abundant species in physiological condition. On increasing pH, a titration trend was observed for all the compounds, and, as reported for K3F21, the plot of A vs. pH at fixed λ highlighted the presence of at least one equivalent point (Supplementary Figure S1). Overall protonation constants (βHL) were calculated from spectrophotometric data and optimized by means of HypSpec software [15] (Supplementary Table S1). K3F23, K3F24 and K3F33 behaved as weak monoprotic acids, with a pKa value close to 8-8.50 associated to the keto-enol moiety. As for the other two ligands, namely K3F21 and K3F31, the acid-base equilibria were more complex in view of the polyprotic nature of the compounds, and three logβLH values were calculated. Species distribution curves (Supplementary Figure S2) provided reliable prediction of neutral and charged forms as function of pH, and consequently helped to envisage cell penetration and distribution in physiological conditions. For monoprotic acids K3F23, K3F24, and K3F33, the prevailing species at pH 7 was the neutral one (HL). As for K3F21 and K3F31, they behaved as less weak acids, in fact at pH 7 58% of K3F21 and 65% of K3F31 were found in the mono-dissociated negatively charged form (H2L -), these percentages increased up to 74% and 86% respectively at pH 7.4. K3F31 resulted to be more dissociated than K3F21 at the same pH, hence more acid. This outcome could be attributed to the stabilizing effect of the intramolecular hydrogen bond between the methoxyl and hydroxyl groups on the aromatic ring in the indissociated form of K3F21, which shifts the equilibrium in favor of the neutral species (H3L).
The coupling of curcumin backbone with phtalimide moiety improved stability with respect to the lead molecule that degrades up to 40% within the first hour [23], in particular K3F33 was the most stable compound of the series, with a residual percentage close to 60% after 24 h (Figure 2A).
The phenolic derivatives K3F21 and K3F31 were tested for their radical scavenging ability by DPPH assay ( Figure 2B). Both K3F21 and K3F31 exerted radical scavenging activity against DPPH radical, however K3F21 was more effective as suggested by IC50 values (24 µM vs. 120 µM). The intramolecular hydrogen bond between the phenolic hydrogen and the methoxyl oxygen seems to be a conditio sine qua non for the formation of a stable radical species. pH-metric titrations were performed in order to assess K3F acidity and predict the most abundant species in physiological condition. On increasing pH, a titration trend was observed for all the compounds, and, as reported for K3F21, the plot of A vs. pH at fixed λ highlighted the presence of at least one equivalent point (Supplementary Figure S1). Overall protonation constants (β HL ) were calculated from spectrophotometric data and optimized by means of HypSpec software [15] ( Supplementary Table S1). K3F23, K3F24 and K3F33 behaved as weak monoprotic acids, with a pK a value close to 8-8.50 associated to the keto-enol moiety. As for the other two ligands, namely K3F21 and K3F31, the acid-base equilibria were more complex in view of the polyprotic nature of the compounds, and three logβ LH values were calculated. Species distribution curves (Supplementary Figure S2) provided reliable prediction of neutral and charged forms as function of pH, and consequently helped to envisage cell penetration and distribution in physiological conditions. For monoprotic acids K3F23, K3F24, and K3F33, the prevailing species at pH 7 was the neutral one (HL). As for K3F21 and K3F31, they behaved as less weak acids, in fact at pH 7 58% of K3F21 and 65% of K3F31 were found in the mono-dissociated negatively charged form (H 2 L − ), these percentages increased up to 74% and 86% respectively at pH 7.4. K3F31 resulted to be more dissociated than K3F21 at the same pH, hence more acid. This outcome could be attributed to the stabilizing effect of the intramolecular hydrogen bond between the methoxyl and hydroxyl groups on the aromatic ring in the indissociated form of K3F21, which shifts the equilibrium in favor of the neutral species (H 3 L).
The coupling of curcumin backbone with phtalimide moiety improved stability with respect to the lead molecule that degrades up to 40% within the first hour [23], in particular K3F33 was the most stable compound of the series, with a residual percentage close to 60% after 24 h (Figure 2A).
The phenolic derivatives K3F21 and K3F31 were tested for their radical scavenging ability by DPPH assay ( Figure 2B). Both K3F21 and K3F31 exerted radical scavenging activity against DPPH radical, however K3F21 was more effective as suggested by IC 50 values (24 µM vs. 120 µM). The intramolecular hydrogen bond between the phenolic hydrogen and the methoxyl oxygen seems to be a conditio sine qua non for the formation of a stable radical species.

Phthalimide-Based Curcumin Derivatives Decrease Viability of Human Cancer Cells
To investigate the potential anti-tumor activity of the new synthesized molecules, we tested their effect on the proliferation of cancer cell lines, in particular PC3 and DU145, two of the most representative in vitro models of PCa, in which AR is not expressed. Dose-response assays were performed with K3F21, K3F23, K3F24, and K3F33 ( Figure 3A,B) and GI50 values were determined as the concentration that causes 50% growth inhibition following 48 h of treatment ( Figure 3C). The comparison between curcumin and K3Fderivatives highlighted that the addition of phthalimide significantly enhances the anti-proliferative activity of K3F21 and K3F23 in PC3 and DU145 cells ( Figure 3C). Similarly, we tested the activity of the compounds in another cancer cell line, the colon carcinoma HCT116 cells, in which the effect of curcumin has been largely investigated and characterized [24]. Also in these cells, K3F21 is the most active molecule compared to the other derivatives (Supplementary Figure S3A

Phthalimide-Based Curcumin Derivatives Decrease Viability of Human Cancer Cells
To investigate the potential anti-tumor activity of the new synthesized molecules, we tested their effect on the proliferation of cancer cell lines, in particular PC3 and DU145, two of the most representative in vitro models of PCa, in which AR is not expressed. Dose-response assays were performed with K3F21, K3F23, K3F24, and K3F33 ( Figure 3A,B) and GI50 values were determined as the concentration that causes 50% growth inhibition following 48 h of treatment ( Figure 3C). The comparison between curcumin and K3F-derivatives highlighted that the addition of phthalimide significantly enhances the anti-proliferative activity of K3F21 and K3F23 in PC3 and DU145 cells ( Figure 3C). Similarly, we tested the activity of the compounds in another cancer cell line, the colon carcinoma HCT116 cells, in which the effect of curcumin has been largely investigated and characterized [24]. Also in these cells, K3F21 is the most active molecule compared to the other derivatives (Supplementary Figure S3A  To determine whether K3F-derivatives could have cytotoxic rather than cytostatic activity, we analyzed the distribution of the cells within the different phases of the cell cycle following the administration of the molecules at GI50 doses for 48 h (Figure 4 and Supplementary Figure S4A Only the administration of curcumin and K3F33 significantly increases SubG1 events, indicative of cell death ( Figure 4A). As for DU145, curcumin mainly acts as a cytotoxic drug, as indicated by the significant increase of SubG1 events from 1.75% to 5% in control and treated cells, respectively. K3F24 and K3F33 showed the ability to arrest cell cycle progression, while K3F21 and K3F23 both arrested the G2/M progression and induced cell death ( Figure 4B). The semisynthetic taxane docetaxel (DTX), chemotherapeutic agent approved by FDA for its ability to prolong survival in patients with metastatic CRPC [25], was used as positive anti-proliferative compound. As expected, DTX administration at GI50 dose showed a high cytotoxic effect associated to a robust increase in SubG1 events (Supplementary Figure S4C,D).
K3F-derivatives were also tested in HCT116 cells: all the molecules induced a significant increase in SubG1 events (Supplementary Figure S3C).
Taking into consideration that all K3F-derivatives showed anti-proliferative effects, but K3F21 is the molecule with the lowest GI50 value in all the tested cell lines, we decided to further investigate its molecular activity as anti-cancer drug in PCa cells.

K3F21 Administration Modulates the Transcription of Genes Involved in PCa Cell Growth
The anti-proliferative and pro-apoptotic activity of curcumin on PCa cells was previously reported [10,[26][27][28]. Among the mechanisms through which curcumin impacts on PCa proliferation, transcriptional modulation of genes controlling cell cycle progression and apoptosis has been described in vitro and in vivo (reviewed in [10]). We therefore decided to investigate the transcriptional effect of K3F21 administration on these processes in PC3 and DU145 cells, analyzing selected representative target genes. In both cell lines, we observed a general decrease, though not always significant, in the transcription of Cdc2, CyclinB1, CyclinB2, and CyclinD1 cell cycle genes in K3F21-treated cells compared to curcumin and control cells ( Figure 5). The transcription of genes encoding CDK inhibitors p21 and p27 similarly increased following curcumin and K3F21 administration in PC3 cells, while they raised only after K3F21 treatment in DU145 cells. These results are consistent with cell cycle analysis (Figure 4), showing that curcumin and K3F21 similarly arrested cell cycle progression in PC3 cells, while only K3F21 induced a G2/M block in DU145 cells. The analysis of EGFR, linked to the proliferation of PCa cells, highlighted that its expression is reduced by K3F21 only in DU145 cells. Finally, we analyzed few apoptotic genes, such as the pro-apoptotic Noxa, Puma, and Bad, and the anti-apoptotic Bcl-2 and Bcl-xl genes. While Noxa was induced by K3F21 as well as curcumin, Puma increased exclusively following K3F21 administration, in both PCa cells. As for anti-apoptotic genes, different transcriptional programs were activated in the two cell lines: Bcl-2, but not Bcl-xl, was weakly repressed by K3F21 in PC3, as opposed to DU145.

K3F21 Affects the Key Molecular Pathways Promoting PCa Progression
Multiple intracellular signaling pathways implicated in PCa progression are modulated by anti-cancer drugs, as corroborated by the analysis of whole cellular extracts from DTX-treated PCa cells (Supplementary Figure S5). By western blot analysis, we determined whether K3F21 is able to inhibit the activation of receptor tyrosine kinase signaling ( Figure 6A). Firstly, we studied the effect of K3F21 administration on PI3K/AKT pathway activation, significantly deregulated in PCa [29]. Both curcumin and K3F21 inhibited the phosphorylation of AKT in Ser473 and Thr308 in PC3 cells [30]. We were not able to detect AKT phosphorylation in DU145 cells, presumably due to the presence of the oncosuppressor PTEN that minimizes endogenous AKT activation [30]. Administration of chemotoxic drugs has been shown to induce ERK phosphorylation, and an inverse relation seems to link ERK and AKT activity in PCa [30][31][32]. Western blot analysis showed that both curcumin and K3F21 increased ERK1/2 phosphorylation in PC3 and, more evidently, in DU145 cells, likely as a consequence of reduced basal levels of PI3K/AKT activity [30].
Finally, p38 phosphorylation was investigated as important mechanism mediating cellular pro-inflammatory responses in PCa [10]. The administration of K3F21, but not curcumin, inhibited phospho-p38 expression in PC3, while no changes were observed in DU145 cells.
We also analyzed the effect of the molecules on the expression of γH2AX, as a marker of DNA damage commonly activated by chemoterapeutic drugs ( Figure 6B). While curcumin increased γH2AX in PC3 only, K3F21 triggered DNA damage in both PC3 and DU145 cells, indicating a genotoxic effect independent from the status of p53. Indeed, the oncosuppressor p53, involved in DNA damage response, is not expressed in PC3 and mutated in its activity in DU145 cells. Moreover, the evaluation of PARP-1 cleavage corroborated the activation of apoptotic cell death following K3F21 administration only in DU145 cells, as already observed by cell cycle analysis ( Figure 4).  Figure S5). By western blot analysis, we determined whether K3F21 is able to inhibit the activation of receptor tyrosine kinase signaling ( Figure 6A). Firstly, we studied the effect of K3F21 administration on PI3K/AKT pathway activation, significantly deregulated in PCa [29]. Both curcumin and K3F21 inhibited the phosphorylation of AKT in Ser473 and Thr308 in PC3 cells [30]. We were not able to detect AKT phosphorylation in DU145 cells, presumably due to the presence of the oncosuppressor PTEN that minimizes endogenous AKT activation [30]. Administration of chemotoxic drugs has been shown to induce ERK phosphorylation, and an inverse relation seems to link ERK and AKT activity in PCa [30][31][32]. Western blot analysis showed that both curcumin and K3F21 increased ERK1/2 phosphorylation in PC3 and, more evidently, in DU145 cells, likely as a consequence of reduced basal levels of PI3K/AKT activity [30].
Finally, p38 phosphorylation was investigated as important mechanism mediating cellular proinflammatory responses in PCa [10]. The administration of K3F21, but not curcumin, inhibited phospho-p38 expression in PC3, while no changes were observed in DU145 cells.
We also analyzed the effect of the molecules on the expression of γH2AX, as a marker of DNA damage commonly activated by chemoterapeutic drugs ( Figure 6B). While curcumin increased γH2AX in PC3 only, K3F21 triggered DNA damage in both PC3 and DU145 cells, indicating a genotoxic effect independent from the status of p53. Indeed, the oncosuppressor p53, involved in DNA damage response, is not expressed in PC3 and mutated in its activity in DU145 cells. Moreover, the evaluation of PARP-1 cleavage corroborated the activation of apoptotic cell death following K3F21 administration only in DU145 cells, as already observed by cell cycle analysis ( Figure 4).

K3F21 Inhibits Metastatic Ability of PCa Cells to Migrate and Proliferate
Increased cell motility of cancer cells, among which PCa cells, strongly impacts on their metastatic potential. We therefore tested whether K3F21 could decrease the migration ability of PC3 and DU145 cells, characterized by high metastatic potential. In vitro wound assays were performed and cell migration of curcumin and K3F21-treated cells was calculated as percentage of wound recovery compared to control cells. As shown in Figure 7, while DMSO-treated cells filled about 100% of the wound area, K3F21 exerted inhibitory effects on the migratory capacity of PCa cells, similarly to curcumin. Interestingly, K3F21 showed more potent anti-migration activity compared to DTX, commonly used in anti-cancer clinical therapy.

K3F21 Inhibits Metastatic Ability of PCa Cells to Migrate and Proliferate
Increased cell motility of cancer cells, among which PCa cells, strongly impacts on their metastatic potential. We therefore tested whether K3F21 could decrease the migration ability of PC3 and DU145 cells, characterized by high metastatic potential. In vitro wound assays were performed and cell migration of curcumin and K3F21-treated cells was calculated as percentage of wound recovery compared to control cells. As shown in Figure 7, while DMSO-treated cells filled about 100% of the wound area, K3F21 exerted inhibitory effects on the migratory capacity of PCa cells, similarly to curcumin. Interestingly, K3F21 showed more potent anti-migration activity compared to DTX, commonly used in anti-cancer clinical therapy. Next, we performed a clonogenic assay, based on the ability of a single cell to grow into a colony, to determine the effectiveness of the anti-proliferative activity of K3F21 to inhibit cell growth of PCa cells also in this condition. Plated PC3 and DU145 cells were treated with different doses of curcumin and K3F21 and the number of colonies were counted following 7 or 10 days, respectively, and reported as percentage vs. DMSO control cells. K3F21 clearly inhibited cell growth when administered at lower concentrations than curcumin, in both cell lines ( Figure 8A,B). Interestingly, despite the GI50 value of K3F21 in traditional 2D monolayer PC3 culture was established as about 10 µM, colony formation was quite completely abolished already at 5 µM, at least in PC3 cells.
Finally, we investigated whether K3F21 was able to inhibit the proliferation of PCa cells when cultured in suspension within a semi-solid gel. The ability to grow in the absence of anchorage to the extracellular matrix (anchorage-independent condition) is one of the hallmarks of malignancy and is crucial in the tumor progression. The decrease in colony formation reached up to 95% following K3F21 treatment at 5 and 10 µM in PC3 and DU145 cells, respectively ( Figure 8C,D, Supplementary Figure S6). In PC3 cells, we observed Next, we performed a clonogenic assay, based on the ability of a single cell to grow into a colony, to determine the effectiveness of the anti-proliferative activity of K3F21 to inhibit cell growth of PCa cells also in this condition. Plated PC3 and DU145 cells were treated with different doses of curcumin and K3F21 and the number of colonies were counted following 7 or 10 days, respectively, and reported as percentage vs. DMSO control cells. K3F21 clearly inhibited cell growth when administered at lower concentrations than curcumin, in both cell lines ( Figure 8A,B). Interestingly, despite the GI50 value of K3F21 in traditional 2D monolayer PC3 culture was established as about 10 µM, colony formation was quite completely abolished already at 5 µM, at least in PC3 cells.
Finally, we investigated whether K3F21 was able to inhibit the proliferation of PCa cells when cultured in suspension within a semi-solid gel. The ability to grow in the absence of anchorage to the extracellular matrix (anchorage-independent condition) is one of the hallmarks of malignancy and is crucial in the tumor progression. The decrease in colony formation reached up to 95% following K3F21 treatment at 5 and 10 µM in PC3 and DU145 cells, respectively ( Figure 8C,D, Supplementary Figure  S6). In PC3 cells, we observed enhanced anti-tumor activity of K3F21 compared to curcumin: similar effects were induced by curcumin when administered at 10 and 20 µM compared to K3F21 5 µM. enhanced anti-tumor activity of K3F21 compared to curcumin: similar effects were induced by curcumin when administered at 10 and 20 µM compared to K3F21 5 µM.

K3F21 Enhances Anti-Tumor Activity of Docetaxel in DU145 Cells
The combination of DTX with other anti-proliferative molecules has been proposed as a promising strategy for enhancing the sensitivity to chemotherapy in CRPC. Curcumin co-administration to DTX was shown to increase the anti-proliferative activity of the two molecules with respect to single treatments. In particular, their combination modulates the expression and activity of proteins commonly hyper-activated in PCa [33]. We therefore decided to investigate the effect of K3F21 co-administration with DTX on PC3 and DU145 cells (Figure 9). Treatment with DTX resulted in a dose-dependent decrease in cell viability of PCa cells. In DU145 cells, co-administration of K3F21 and DTX 20nM for 48 h induced a significant decrease in cell viability compared to independent treatments with DTX or K3F21. The determination of the Combination Index (CI) [34] revealed a slight synergism (CI = 0.8) between the drugs with a DTX drug reduction index (DRI) of 2.9, hinting that concurrent administration with K3F21 could allow an effective reduction of DTX dose. Differently, none of the evaluated K3F21/DTX combinations affected cell growth when compared simultaneously to both DTX and K3F21 single administration in PC3 cells.

K3F21 Enhances Anti-Tumor Activity of Docetaxel in DU145 Cells
The combination of DTX with other anti-proliferative molecules has been proposed as a promising strategy for enhancing the sensitivity to chemotherapy in CRPC. Curcumin co-administration to DTX was shown to increase the anti-proliferative activity of the two molecules with respect to single treatments. In particular, their combination modulates the expression and activity of proteins commonly hyper-activated in PCa [33]. We therefore decided to investigate the effect of K3F21 co-administration with DTX on PC3 and DU145 cells (Figure 9). Treatment with DTX resulted in a dose-dependent decrease in cell viability of PCa cells. In DU145 cells, co-administration of K3F21 and DTX 20nM for 48 h induced a significant decrease in cell viability compared to independent treatments with DTX or K3F21. The determination of the Combination Index (CI) [34] revealed a slight synergism (CI = 0.8) between the drugs with a DTX drug reduction index (DRI) of 2.9, hinting that concurrent administration with K3F21 could allow an effective reduction of DTX dose. Differently, none of the evaluated K3F21/DTX combinations affected cell growth when compared simultaneously to both DTX and K3F21 single administration in PC3 cells.

Discussion
In preclinical models, thalidomide and its derivatives have shown growth inhibition and enhanced susceptibility to apoptosis of tumor cells [35]. However, in terms of the achieved response rates, i.e., the percentage of patients whose cancer shrinks or disappears after treatment, thalidomide showed mild activity towards CRPC. Phase II clinical trial showed that about 30-40% of patients with metastatic CRPC had a decline in PSA of ≥40%, associated with an improvement of clinical symptoms, following thalidomide treatment [36,37].
As regards curcumin, several studies have shown its potential therapeutic effects on PCa cells survival, both in vitro and in vivo [16,26,27,38]. In particular, the activity of curcumin has been well described on PC3 cells, accepted in vitro model of highly aggressive prostate cancer [16,39]. Nevertheless, the growth suppressive activity and bioavailability of curcumin in humans showed limitations for its use as an effective therapeutic agent in cancer. Therefore, the conjugation of phtalimide with curcumin and its derivatives could represent a successful strategy to develop new multi-target derivatives to defeat PCa. In addition, as previously observed for other β-diketo substituted curcuminoids [22], the presence of a bulky substituent in α position to the keto-enol moiety shifts the equilibrium towards the diketo form, improving stability and potentially bioavailability. K3F33, K3F23, and K3F24 show an acid-base behavior similar to curcumin, with a typical pKa value ranging from 8 to 8.5 that can be attributed to the dissociation of enol proton. Differently, as

Discussion
In preclinical models, thalidomide and its derivatives have shown growth inhibition and enhanced susceptibility to apoptosis of tumor cells [35]. However, in terms of the achieved response rates, i.e., the percentage of patients whose cancer shrinks or disappears after treatment, thalidomide showed mild activity towards CRPC. Phase II clinical trial showed that about 30-40% of patients with metastatic CRPC had a decline in PSA of ≥40%, associated with an improvement of clinical symptoms, following thalidomide treatment [36,37].
As regards curcumin, several studies have shown its potential therapeutic effects on PCa cells survival, both in vitro and in vivo [16,26,27,38]. In particular, the activity of curcumin has been well described on PC3 cells, accepted in vitro model of highly aggressive prostate cancer [16,39]. Nevertheless, the growth suppressive activity and bioavailability of curcumin in humans showed limitations for its use as an effective therapeutic agent in cancer. Therefore, the conjugation of phtalimide with curcumin and its derivatives could represent a successful strategy to develop new multi-target derivatives to defeat PCa. In addition, as previously observed for other β-diketo substituted curcuminoids [22], the presence of a bulky substituent in α position to the keto-enol moiety shifts the equilibrium towards the diketo form, improving stability and potentially bioavailability. K3F33, K3F23, and K3F24 show an acid-base behavior similar to curcumin, with a typical pK a value ranging from 8 to 8.5 that can be attributed to the dissociation of enol proton. Differently, as suggested for other curcuminoids [40], the low value of the first pK a for K3F21 and K3F31 could be reasonably attributed to the phenol group, which acidity is decreased by the conjugated keto-enol group in para position. For these compounds, the second pK a is due to the keto-enol moiety, and the last one to the second phenolic group.
This study reports on the synthesis, chemical and biological characterization of the phthalimide-based curcuminoids as new potential therapeutic molecules in PCa. Chemical characterization and in vitro data on tumor cell lines identified K3F21 as the most attracting molecule for further cellular and molecular characterization of its anti-proliferative activity. In fact, K3F21 is a weak acid that is indissociated in physiological condition, hence present in its neutral form. The high lipophilicity (data not shown) together with its low molecular mass allow to predict high bioavailability. The presence of phtalimide moiety stabilizes the diketo tautomer, improving on one hand pharmacokinetics with respect to curcumin and maintaining similar radical scavenging activity on the other. K3F21 shows higher efficacy in terms of growth inhibitory activity (GI50) compared to curcumin, not only in PCa cells but also in human colon adenocarcinoma cells. Although K3F21 has a similar GI50 concentration on PC3 and DU145 cells, it exerts different anti-proliferative effects in the two cell lines. While its administration to PC3 significantly increases the G2/M population, without inducing cell death, DU145 treated cells shows both G2/M cell cycle arrest and apoptosis (Figure 4). The analysis by western blot of cleaved-PARP1 suggests the activation of a caspase-mediated cell death ( Figure 6).
PC3 and DU145 cells are characterized by non-luminal-like phenotypes and are representative of aggressive disease, with similarity in overall gene expression [41]. Indeed, the analysis of single gene transcription, performed by RT-qPCR, highlighted the activation of quite similar transcriptional programs by curcumin and K3F21 in the two cell lines ( Figure 5). Despite this, we should consider that PC3 and DU145 cells differ in the status of the p53 gene, which encodes for a transcription factor involved in the response to cellular stresses, DNA damage, and cytotoxic drugs. During PCa progression, epithelial cells acquire different mutations in key genes controlling cell proliferation and cell survival, among which p53 [42]. In particular, mutations that give rise to mutant gain of function (GOF) or dominant negative (DN) p53 allow the tumor to survive anticancer therapies [43]. The administration of DNA damaging agents to mutant p53 cells may potentially further increase cancer progression: mutant p53 cells usually escape cell cycle arrest following DNA damage, which eventually triggers the acquisition of more mutations allowing cancer cell survival. While PC3 cells are p53-null, therefore they do not express p53 at the protein level, DU145 cells express mutant DN p53 (P223L/V274F) [44,45]. K3F21 efficiently inhibits cell proliferation in DU145 as well as in PC3 cells when cultured in normal 2D conditions. Differently, curcumin and K3F21 activities on cell migration and colony formation in anchorage-independent condition, key futures of cancer cells, highlighted the increased sensitivity of PC3 cells vs. DU145 cells towards curcumin and K3F21 (Figure 8). It is interesting to note that in 3D culturing conditions, that better mimic the physiological in vivo tumor growth condition, K3F21 is more active than in normal 2D monolayer culture conditions. Finally, we should also consider the status of the PTEN/PI3K/AKT axis in the examined cells. PTEN is one of the anti-apoptotic genes frequently mutated in PCa: while PC3 are PTEN-negative, DU145 cells possess the functional protein. This difference plays a great role in the activation of the PI3K/AKT and ERK pathways: PTEN expression in DU145 cells minimizes aberrant activation of PI3K/AKT, which is sufficient for upregulation of ERK phosphorylation. ERK activation seems to be necessary for drug-induced death and its inactivation has been identified as a hallmark of advanced PCa [31,32,46]. Consistent with these published data, our western blot analysis highlighted that K3F21-induced inhibition of AKT phosphorylation is associated with increased ERK activation in both PC3 and DU145 cells ( Figure 6).
Treatment of PCa cells with DTX inhibits their proliferation and activates apoptosis. Since curcumin has been recently shown to increase the efficacy of DTX treatment in PCa [33], we decided to investigate the effect of K3F21 and DTX combination. In DU145 only, we were able to quantitatively determine synergism with CI<1 between the two molecules. Moreover, the favorable DTX dose-reduction index (DRI) suggested the promising potentiality of K3F21 as a therapeutic adjunct with DTX in the treatment of PCa. Further studies will help to elucidate why this effect is observed only in DU145 and not in PC3 cells.
Overall, our results indicate that K3F21 exhibits similar or more potent activity than curcumin in the inhibition of prostate cancer cell growth and migration, as well as in the modulation of key molecular pathways involved in tumor progression and survival. These results open the opportunity to develop K3F21 molecule as a new therapeutic drug in aggressive PCa tumors.

General Procedures and Chemicals
Elemental analysis was performed on Thermo ScientificTM FLASH 2000 organic elemental analyzer. The percentage of C, H, and N were experimentally calculated. Liquid chromatography/mass spectrometry (LC/MS) was performed on Agilent 6300 Ion Trap LC/MS System equipped with an electrospray ionization (ESI) interface. The compounds were separated using Agilent Zorbax SB C18 30 × 2.1 mm, 3.5 µm. Samples were prepared in MeOH and diluted to 10 ppm in MilliQ water; blank was MilliQ water. Eluent phase: pump A H 2 O (formic acid 1%), pump B ACN (formic acid 1%), gradient: 10% of B for 1 min, 10-100% of B for 5 min, then 100% of B for 4 minutes, flux 0.3 mL/min, injection volume 10 µL. The ion spectra were obtained in positive mode, using a scan range between 100 and 1500 m/z. High-purity nitrogen was used as nebulizer and drying gas (flow rate 10 L/min, 350 • C). The nebulizer gas pressure was 32 psi and the capillary voltage was 3.5 kV. All reagent grade chemicals were purchased from Sigma-Aldrich (St. Louis, MO, USA) and used without further purification unless otherwise specified. NMR spectra were recorded on a Bruker FT-NMR AVANCE III HD 600 MHz spectrometer with 5 mm CryoProbe BBO H&F at 298 K. Nominal frequencies were 600.13 MHz for 1 H and 150.9 MHz for 13 C. For each sample,~5 mg were weighed and diluted up to 0.6 mL with the proper deuterated solvent into 5 mm NMR tube. 90 • pulse was calibrated for each sample and standard NMR parameters were used to achieve quantitative results (relaxation delay 10 s). Proton and carbon chemical shifts are given in parts per million (ppm) vs. external TMS, and were determined by reference to the solvent residual signals. Typical 2D homo-and hetero-nuclear techniques were used for assignment, i.e., 1 H, 1 H-COSY, 1 H, 13 C-HSQC, 1 H, 13 C-HMBC. The purity of all final compounds was determined to be at least 95% pure by a combination of LC-MS, NMR, and combustion analysis.

Acid-Base Behavior and Stability in Physiological Conditions
UV-vis spectrophotometric measurements were performed using Jasco V-570 spectrophotometer at 25.0 ± 0.1 • C in the 200-600 nm spectral range employing 1 cm quartz cells. Owing to the poor water solubility of the compounds, a methanol mother solution (5 × 10 −3 M) was diluted in water in order to give 50 µM solutions used for pH-metric titrations. The pH value was varied by adding small amounts of concentrated NaOH or HCl, 25 points in the pH range 3-10. A constant ionic strength of 0.1 M (NaNO 3 ) was maintained in all the experiments. Each titration was performed at least three times. The overall protonation constants (logβ LH ) were evaluated from spectrophotometric data using the software HypSpec [48].
Kinetic stability studies were performed at 37 • C in darkness, the change in absorbance in the 200−600 nm range over an overall period of 8 h was estimated for all the K3F samples by UV-vis spectroscopy. 100 µM solutions of the ligands were prepared in 0.1 M Tris-HCl buffer (pH 7.4) with constant ionic strength (0.1 M NaNO 3 ). Spectra were recorded every 30 min. All profiles were linearized by hyperbolic function (Equation (1)), which represents an empirical model that well describes drug decomposition or release [49].
t/f % = at + b (1) where f % is the fraction of residual compound at time t (min) expressed as percentage referred to starting concentration at time zero.

Antioxidant Activity (DPPH Assay)
The antioxidant activity of the compounds was evaluated in terms of hydrogen donating or radical scavenging ability, using the stable DPPH radical (1,1-diphenyl-2-picrylhydrazyl radical). A variable amount (15,30,45,75,105, and 150 µL) of a methanol solution (1.2 mM) of each compound was placed in a cuvette, and 3 mL of a 60 µM methanol solution of DPPH was added. Absorbance measurements were initiated immediately. The decrease in absorbance at 517 nm was monitored every minute up to 5 , then every 5 up to 30 and every 30 until reaction reaches completion and absorbance stabilizes attaining a plateau after 120 . Methanol was used to zero the spectrophotometer. The absorbance of the DPPH radical without K3F compound, i.e., the control, was measured daily, and concentration was calculated applying Equation (2) [50] [DPPH · ] = A − 1.006 10970 The percentage of inhibition (%In) of the DPPH radical by each sample was calculated according to the formula where A 0 represents the absorbance of the control (DPPH radical) at time 0, while A t refers to the absorbance of the mixture DPPH/antioxidant at time t (120 min). Values of absorbance were corrected taking into account volume dilution and all determinations were performed in triplicate.
Synthesized curcumin, K3F compounds and docetaxel (MedChem Express, Monmouth Junction, NJ, USA) were dissolved in DMSO (Sigma-Aldrich, St. Louis, MO, USA) and added to cell medium at the concentrations described in the text for 48 h. DMSO was used as control. Docetaxel IC50 values were calculated through MTT assay and estimated as 17 nM and 30 nM in DU145 and PC3, respectively.

Cell Viability Assay
The inhibition of cell proliferation was measured by colorimetric MTT assay [51] and the concentration at which cellular growth is inhibited by 50% (GI50) was determined. Briefly, cells were seeded into 96-well plates at a density of 2000 cells/well. The following day, cells were treated with the indicated compounds and analyzed for cell viability after 48 h. 5 mg/mL thiazolyl blue tetrazolium bromide (Sigma-Aldrich, St. Louis, MO, USA) was added to each well at a final concentration of 0.5 mg/mL and the plates were incubated at 37 • C for 2 h. Medium containing unconverted MTT was removed and 100 µL MTT solvent (4 mM HCl, 0.1% NP-40 in isopropyl alcohol) was added to each well. The plate was gently rotated on an orbital shaker in the dark for 15 min, before the measurements of absorbance at 570 nm. The viability of untreated cells was arbitrarily set at 100%.
The ability of K3F21 and docetaxel to act synergistically with regard to growth inhibition was determined using CompuSyn software (ComboSyn, Inc., freely available at http://www.combosyn. com) [34]. Combination index (CI) and drug reduction index (DRI) were calculated for each concentration of drugs mixture, as indicated in the text. CI <1, =1 or >1 indicates synergism, additive effect, or antagonism, respectively. DRI >1 or <1 indicates favorable or not favorable dose-reduction.

Anchorage-Dependent and -Independent Colony Formation Assay
For anchorage-dependent colony formation assay, 1000 PC3 or 500 DU145 cells were plated in 6-well plates. Following overnight incubation, the cells were treated with curcumin or K3F21 (2, 5, 10, 20 µM) or with DMSO, as control. After 7-10 days, the plates were washed, fixed and stained with 0.5% crystal violet solution in 20% Methanol. Visible colonies with radius ≥3 pixels were counted with automated colony counter (OpenCFU 3.9.0 software, freely available at http: //opencfu.sourceforge.net/) and plotted against curcumin concentrations. Colony numbers were normalized as a percentage of colonies formed following DMSO treatment.
For anchorage-independent colony formation assay, 0.6% low melt agarose gel with 10% FBS in appropriate complete cell growth medium was prepared and added to the wells of a twelve-well culture dish as a base agar. 1200 cells per well were plated in 0.25% (PC3) or 0.2% (DU145) agarose gel in appropriate complete cell growth medium supplemented with varying concentrations of curcumin and K3F21 (5, 10, 20 µM) or with DMSO, as control. PC3 and DU145 cells were allowed to grow at 37 • C for three weeks (PC3) or five weeks (DU145). The effect of the drugs on anchorage-independent growth was determined by colony growth. Colonies were stained with 0.07% crystal violet and formed colonies were imaged. The numbers of colonies with radius ≥ 7 pixels were counted with automated colony counter (OpenCFU 3.9.0) and average number of colonies was plotted against drug concentrations. Colony numbers were normalized as a percentage of colonies formed following DMSO treatment.
4.9. Cell Migration Assay 4 × 10 5 prostate cancer cells were plated in a 12-well plate. Approximately 24 h later, when the cells were 100% confluent, the monolayer was scratched using a 200 µL pipette tip. Medium and non-adherent cells were removed and new medium containing 1% FBS and GI50 concentrations of Curcumin, K3F21 and DTX was added. Cells were observed under the microscope and the inhibition of migration was assessed when the wound in the control was closed, after 16 h and 40 h in DU145 and PC3, respectively. Three biological replicates were performed. Three random pictures were taken for each treated well per experiment and the areas of the wounds measured using ImageJ software. The percentage of wound recovery was then calculated using the formula: (final area-initial area)/initial area × 100%.

RNA Extraction and RT-qPCR
RNA was extracted from cells using RNeasy kit (Qiagen, Venlo, The Netherlands) and reversed transcribed with a Moloney murine leukemia virus reverse transcriptase (Promega, Madison, WI, USA). Quantitative real-time PCRs were performed using SsoAdvanced Universal SYBR Green Supermix (Bio-Rad, Hercules, CA, USA) on a Bio-Rad CFX Connect™ real-time PCR detection system. Primer sequences are available upon request. The housekeeping gene Rpl21 was used as a loading control. The relative fold change enrichments of real-time PCR samples were calculated with the formula 2 −(∆∆Ct) , where −(∆∆Ct) = −[(Ct target − Ct Rpl21)treated − (Ct target − Ct Rpl21)DMSO].