This review addresses the oncopharmacological properties of curcumin at the molecular level. First, the interactions between curcumin and its molecular targets are addressed on the basis of curcumin’s distinct chemical properties, which include H-bond donating and accepting capacity of the β-dicarbonyl moiety and the phenylic hydroxyl groups, H-bond accepting capacity of the methoxy ethers, multivalent metal and nonmetal cation binding properties, high partition coefficient, rotamerization around multiple C–C bonds, and the ability to act as a Michael acceptor. Next, the in vitro chemical stability of curcumin is elaborated in the context of its susceptibility to photochemical and chemical modification and degradation (e.g., alkaline hydrolysis). Specific modification and degradatory pathways are provided, which mainly entail radical-based intermediates, and the in vitro catabolites are identified. The implications of curcumin’s (photo)chemical instability are addressed in light of pharmaceutical curcumin preparations, the use of curcumin analogues, and implementation of nanoparticulate drug delivery systems. Furthermore, the pharmacokinetics of curcumin and its most important degradation products are detailed in light of curcumin’s poor bioavailability. Particular emphasis is placed on xenobiotic phase I and II metabolism as well as excretion of curcumin in the intestines (first pass), the liver (second pass), and other organs in addition to the pharmacokinetics of curcumin metabolites and their systemic clearance. Lastly, a summary is provided of the clinical pharmacodynamics of curcumin followed by a detailed account of curcumin’s direct molecular targets, whereby the phenotypical/biological changes induced in cancer cells upon completion of the curcumin-triggered signaling cascade(s) are addressed in the framework of the hallmarks of cancer. The direct molecular targets include the ErbB family of receptors, protein kinase C, enzymes involved in prostaglandin synthesis, vitamin D receptor, and DNA.
Carcinogenesis, cancer cell proliferation, and metastasis encompass a myriad of complex signaling pathways governed by multitudinous intracellular and extracellular (bio)molecules and ions that are collectively responsible for forging the cancer phenotype. Typically, malignant cellular transformation is driven by congenital or acquired genetic in- or transactivation of endogenous signaling pathways. These pathways are tightly regulated in healthy cells but become derailed (hypo- or hyperactivated) in neoplasms. In addition, after a neoplastic state has materialized, the cancer phenotype is constantly subjected to molecular changes in response to an inductive micro-environment, further altering the activity and involvement of different signaling pathways.
To fight cancer effectively, therapeutic strategies should take the dynamic nature of the cancer phenotype into account. The most ideal chemopreventive and chemotherapeutic agents therefore comprise those that affect multiple molecular targets, preferably in the different phenotypic manifestations of the cancer cell, with minimal toxicity for healthy cells. In view of the latter, the majority of synthetic chemotherapeutics (e.g., alkylating agents, topoisomerase inhibitors, and antimetabolites), plant alkaloids (e.g., vinca alkaloids and taxanes), as well as cytotoxic antibiotics (e.g., anthracyclines) targets rapidly dividing and metabolically hyperactive cancer cells at a limited number of molecular loci. Although these compounds are notably lethal to cancer cells, the inadvertent uptake by non-cancerous cells triggers their death also, accounting for the chemotherapy-induced sequelae that patients often experience.
The use of a multitargeted approach to cancer chemotherapy (Lee et al., 2011b) may therefore be instrumental in curtailing therapy-related side effects while preserving therapeutic efficacy. Such a multitargeted modality comprises either a lower dose cocktail of highly toxic chemotherapeutic agents [referred to as metronomic dosing (Hanahan et al., 2000)] or a single, mildly to nontoxic compound that targets multiple components of vital signaling pathways in predominantly cancer cells. Certain phytochemicals are particularly eligible for the single compound type modality inasmuch as they are pleiotropic modulators of manifold signal transduction pathways and exhibit relatively low toxicity in non-cancerous cells (Khan et al., 2008; Lee et al., 2011b). Unfortunately, the currently Food and Drug Administration (FDA)-approved chemotherapeutic phytochemicals such as paclitaxel, docetaxel, vinblastine, and vincristine have only a singular molecular target (tubulin), whereas other phytochemical candidates such as (–)-epigallocatechin gallate and delphinidin have multiple molecular targets (Ermakova et al., 2006; Li et al., 2007; He et al., 2008; Kang et al., 2008; Shim et al., 2008, 2010; Hwang et al., 2009; Ozbay and Nahta, 2011) and in vitro cytostatic/cytotoxic potency (Bin Hafeez et al., 2008; Hsieh and Wu, 2009; Philips et al., 2009; Qiao et al., 2009; Yun et al., 2009; Cvorovic et al., 2010; Das et al., 2010) but lack an advanced clinical development trajectory and regulatory approval status.
One phytochemical that fully meets the criteria of the single compound-type modality is curcumin [(1E,6E)-1,7-bis-(4-hydroxy-3-methoxyphenyl)-1,6-heptadiene-3,5-dione/diferuloylmethane], a polyphenolic compound derived from the rhizome of the Curcuma longa. Curcumin inhibits multiple vital pathways in cancer cells (Kunnumakkara et al., 2008) that affect all hallmarks of cancer (Hanahan and Weinberg, 2011), rendering curcumin a chemopreventive, oncostatic, and antimetastatic agent. Interestingly, curcumin appears to be toxic to cancer cells but cytoprotective to healthy cells (Dinkova-Kostova and Talalay, 2008; Bisht et al., 2010), owing in part to its strong antioxidative capacity (Priyadarsini et al., 2003). Numerous studies in humans have established that oral consumption of high doses of curcumin, up to 12,000 mg/day, is tolerable and safe (Shoba et al., 1998; Cheng et al., 2001; Sharma et al., 2004; Lao et al., 2006), and several clinical phase I and II trials have been conducted with curcumin for the treatment of multiple types of cancer (Cheng et al., 2001; Anand et al., 2008; Goel et al., 2008; Hatcher et al., 2008; Shehzad et al., 2010; Gupta et al., 2013). Moreover, curcumin has been shown to alleviate senescence (Sikora et al., 2010) and various forms of neurodegenerative (Jomova et al., 2010), infectious (Zhao et al., 2011), and autoimmune disease (Bright, 2007). The medical disciplines and research fields in which curcumin has been studied exemplify the breadth of medical and biotechnological potential of this phytochemical, with cancer clearly leading the way.
This review focuses on the oncopharmacological properties of curcumin in the context of the hallmarks of cancer. First, the relationship between curcumin structure and its ability to undergo pleiotropic intermolecular bonding and chemical stability and solubility are addressed to provide a backdrop against which information on the pharmacokinetics and pharmacodynamics of curcumin as well as its in vitro and in vivo metabolites is presented. Following a brief summary of the cancer hallmarks, the direct molecular targets of curcumin are identified, and a detailed account is given of the molecular pathways that are affected as a result of curcumin binding. Subsequently, the phenotypical/biological changes induced in cancer cells upon completion of the curcumin-triggered signaling cascade are addressed in the framework of the hallmarks of cancer. Whereas previous reviews have typically provided lists of (bio)molecules that are directly and indirectly affected by curcumin (Johnson and Mukhtar, 2007; Goel et al., 2008; Gupta et al., 2011; Zhou et al., 2011), this review focuses exclusively on direct curcumin-ligand binding as a starting point for explaining the downstream cellular effects pertinent to cancer biology and treatment. The molecular targets were selected on the criteria that (1) the curcumin-ligand interaction affects multiple cancer hallmarks, (2) curcumin binds selectively to multiple targets in a single pathway, and/or (3) curcumin binding imposes imminent lethality on the cell (e.g., by inducing interference with vital transcriptional activity and subsequent cell death). The molecular targets addressed in this review are therefore the ErbB family of receptors, protein kinase C, enzymes involved in prostaglandin synthesis, vitamin D receptor, and DNA.
Readers should note that, although it is our opinion that all sections presented in this review are imperative to fully appreciate curcumin as a pharmaceutical compound, we understand that not everyone may be equally interested in all parts. Accordingly, the sections were divided into chemical properties, pharmacokinetics, and pharmacodynamics to facilitate selective reading and written such that each section can be understood in a stand-alone manner.
II. Chemical Properties of Curcumin
Information on the structure, solubility, and stability of curcumin is imperative for the proper interpretation of its pharmacokinetic and pharmacodynamic properties. The majority of curcumin-related data have been generated using cell-based assays that, particularly in the case of this compound, are not necessarily representative of the in vivo situation.
A. Curcumin Structure: Implications on Intermolecular Interactions
The generic structures of the turmeric-derived curcuminoids are provided in Fig. 1A, which, in order of their relative abundance in the root, comprise curcumin, demethoxycurcumin, bis-demethoxycurcumin, and cyclocurcumin. Curcumin (R1 and R2 = OCH3) is an amphipathic molecule with polar central and flanking regions that are separated by a lipophilic methine segment (Fig. 1B). Curcumin possesses seven distinct chemical properties that facilitate intermolecular interactions and thus association with its biomolecular targets. These include (1) H-bond donating and accepting capacity of the β-dicarbonyl moiety, (2) H-bond accepting and donating capacity of the phenylic hydroxyl residues, (3) H-bond accepting capacity of the ether residue in the methoxy groups, (4) multivalent metal and nonmetal cation binding properties, (5) high partition coefficient (log P), (6) rotamerization around multiple C–C bonds, and (7) behavior as a Michael reaction acceptor. The interactions that are based on H-bonding with specific functional groups of curcumin are summarized in sections II.A.1 through II.A.3, and the hydrophobic interactions are addressed in section II.A.5. A full overview of the curcumin-target molecule interactions is provided in Supplemental Table 1, which summarizes additional molecular targets for which the interacting protein or nucleic acid residues were identified by docking studies or site-directed mutagenesis but not in relation to curcumin’s specific functional groups.
1. H-bond Donating and Accepting Capacity of the β-Dicarbonyl Moiety.
Curcumin’s β-dicarbonyl moiety deviates in some respects from the typical behavior of β-diketones in solution, including acetylacetone (H3C–CO–CH2–CO–CH3), the central structure of curcumin. β-Diketones other than curcumin generally prevail in the diketo tautomeric form in solutions with high relative permittivity (εr) or polarity [e.g., water and dimethylsulfoxide (DMSO)] and in the enol tautomeric form in solutions with low εr [e.g., (cyclo)hexane and carbon tetrachloride] (Emsley, 1984). Antithetically, the β-diketone of curcumin is proposed to exist predominantly as an enol (Fig. 1A) in both aqueous solution and in organic solvents with a lower εr than water, particularly in polar protic solvents (e.g., alcohols) and polar aprotic solvents (e.g., acetonitrile, DMSO) (Fig. 1B). The tendency of the keto↔enol equilibrium to shift toward the enol tautomer in these solvent systems has been confirmed experimentally by Fourier transform infrared spectroscopy (FT-IR) (Kolev et al., 2005), nuclear magnetic resonance spectroscopy (NMR) (Roughley and Whiting, 1973; Unterhalt, 1980; Gorman et al., 1994; Khopde et al., 2000), fluorescence spectroscopy (Khopde et al., 2000; Nardo et al., 2008), and absorption spectroscopy (Khopde et al., 2000; Nardo et al., 2008) as well as computationally (Kolev et al., 2005; Balasubramanian, 2006; Payton et al., 2007; Galano et al., 2009). In nonpolar solvents, the steady-state absorption spectra of curcumin in cyclohexane (Nardo et al., 2008), toluene (Chignell et al., 1994), benzene (Khopde et al., 2000), and others (Tønnesen et al., 1995) exhibit a concrete red band or shoulder, which has been ascribed to the diketo tautomer (Khopde et al., 2000; Nardo et al., 2008, 2009). Curcumin hence adopts both the enol and diketo tautomeric form in nonpolar environments (Balasubramanian, 2006), albeit the diketo tautomer is quantitatively exiguous relative to the enol tautomer (Kolev et al., 2005), and exclusively the enol tautomeric form in solvents with increasing εr (Fig. 1B). This is underscored by the fact that the red shoulder in the absorption spectrum of curcumin dissolved in water-dioxane, a slightly nonpolar but water-miscible solvent, disappears progressively at increasing water:dioxane ratios (Ortica and Rodgers, 2001).
Accordingly, the conformation of the β-diketone is highly dependent on the chemical environment that in turn dictates intermolecular bonding behavior. In nonpolar solvents, curcumin primarily exists as a closed cis-enol tautomer, where the proton is believed to be symmetrical (Nardo et al., 2008) in an electron-delocalized system (Emsley, 1984; Balasubramanian, 2006) (Fig. 1B). The enolic proton and carbonyl oxygen are therefore unavailable for H-bonding to other molecules. The intramolecular H-bonding is perturbed in weakly as well as strongly H-bonding solvents by the polar solvent molecules, as a result of which the enol tautomer adopts an open conformation (Emsley, 1984; Toullec, 1990; Khopde et al., 2000; Nardo et al., 2008), thereby abrogating the symmetry of the semiaromatic ring (Nardo et al., 2008, 2009) (Fig. 1B). In the open conformation, the valence electrons of the carbonyl and enolic oxygen act as H-bond acceptors and the enolic proton serves as an H-bond donor, possessing charges of −0.73, −0.70, and 0.50 in water (Fig. 1C), respectively (Balasubramanian, 2006). The interaction between curcumin’s β-carbonyls and molecules with increasing εr is reflected by the generally larger Stokes shifts (ΔS) with increasing solvent polarity (Fig. 1B). Accordingly, under physiological circumstances the open enol tautomer can interact with water molecules in, e.g., plasma and cytoplasm, or with polar amino acid residues in target or carrier proteins, either directly or indirectly via water molecules.
A summary of the H-bond-mediated interactions between curcumin’s β-diketone moiety and amino acid residues in different proteins is provided in Fig. 2 for the H-bond accepting carbonyls and in Fig. 3 for the H-bond donating enolic proton. Some of the proteins that curcumin interacts with through the β-diketone moiety that may bear relevance to cancer include DNA methyltransferase 1 (Yoo and Medina-Franco, 2011), aldose reductase 1 and 2 (Muthenna et al., 2009), lipoxygenase (Katsori et al., 2011), 20S proteasome (Milacic et al., 2008), tubulin (Li et al., 2012), cyclooxygenase (COX)-2 (Selvam et al., 2005) (sections III.E.1.e and III.E.7), glycogen synthase kinase-3 β (GSK-3β) (Bustanji et al., 2009) (section III.E.1.b.ii), and glyoxalase I (Liu et al., 2010).
2. H-bond Accepting and Donating Capacity of the Phenylic Hydroxyl Groups.
As the centrally positioned β-diketone moiety, the flanking hydroxyl groups impart H-bond accepting and donating capabilities on the molecule, expanding the number of possible interaction sites that account for curcumin’s pleiotropic binding behavior. The intermolecular interactions facilitated by the phenylic hydroxyl groups are summarized in Fig. 2 for the H-bond accepting phenylic oxygens and in Fig. 3 for the H-bond donating phenylic protons. In addition to most of the protein targets mentioned in the preceding section, histone deacetylase 8 (Bora-Tatar et al., 2009) and protein kinase C (PKC)δ (Majhi et al., 2010; Mamidi et al., 2012) (section III.E.6) have been identified as cancer-pertinent proteins with which curcumin associates as a result of phenylic hydroxyl group-mediated H-bonding.
3. H-bond Accepting Capacity of the Methoxy Groups.
The third functional group that is capable of mediating intermolecular interactions is the methoxy group, which can only act as an H-bond acceptor through the valence electrons of the ether. The intermolecular interactions facilitated by the methoxy ether are summarized in Fig. 2. No additional protein targets than those previously described in sections II.A.1 and II.A.2 were found to bind curcumin through the methoxy groups. Interestingly, docking studies with diketonic curcumin, i.e., the physiologically less abundant isoform (section II.A.1), revealed H-bonding with 14 amino acids of seven different proteins but no interactions for enolic curcumin (Figs. 2 and 3).
4. Multivalent Metal and Nonmetal Cation Binding.
Molecules containing a β-diketone are known to form stable complexes with multivalent metal cations (Skopenko et al., 2004). For example, acetylacetone, curcumin’s central moiety (section II.A.1), can complex metals at both oxygen atoms (in case of the diketo form), the ketone oxygen (in case of the keto-enol form), the olefin (Cα═Cβ), or the Cα, which has a partial negative charge (Fig. 1C) (Cotton and Wilkinson, 1980). Accordingly, curcumin forms chelates with metals such as boron (Mohri et al., 2003; Rao and Aggarwal, 2008), copper (Baum and Ng, 2004; Barik et al., 2007; Zebib et al., 2010; Addicoat et al., 2011), aluminum (Jiang et al., 2011), magnesium (Zebib et al., 2010), zinc (Zebib et al., 2010), lead (Daniel et al., 2004), cadmium (Daniel et al., 2004), and ferrous (Fe2+) and ferric (Fe3+) iron (Tønnesen and Greenhill, 1992; Borsari et al., 2002; Baum and Ng, 2004; Bernabe-Pineda et al., 2004b; Ak and Gulcin, 2008; Dairam et al., 2008), but also with metal oxides such as vanadyl (Thompson et al., 2004) and nonmetals such as selenium (Zebib et al., 2010). Although the majority of experimental data indicate that the metal cations bind the oxygens of the β-diketo moiety (Borsari et al., 2002; Zebib et al., 2010; Jiang et al., 2011), some have proposed that metal chelation may also be facilitated by the valence electrons of the methoxyphenyl oxygens (Ak and Gulcin, 2008; Jiang et al., 2011) based on the negative charge of these oxygens (Fig. 1C) and the metal chelating activity of β-diketone-lacking (poly)phenolics such as quercetin (Fiorucci et al., 2007) and catechol (Borsari et al., 2002).
Aside from its ability to chelate free metal cations, curcumin may use the metal atoms of metalloproteins as a means to interact with these proteins. Metal-mediated interactions between curcumin and its target molecules have been described for several different classes of metalloproteins, including glyoxalase I (Liu et al., 2010; Yuan et al., 2011), thioredoxin reductase (TrxR) (Singh and Misra, 2009), and HIV-1 integrase (Vajragupta et al., 2004). Molecular docking studies revealed that curcumin coordinated with the zinc atom in the catalytic site of glyoxalase I through a carbonyl oxygen (Liu et al., 2010; Yuan et al., 2011), whereby the binding was further stabilized by H-bonding between the phenylic hydroxyl group and a single lysine (Yuan et al., 2011) or multiple amino acid residues and hydrophobic interactions (Liu et al., 2010). With respect to TrxR, docking studies showed an important interaction between the selenium atom of a selenocysteine residue (Sec498) in the enzyme’s active site and the π-electrons of one of the methine bridge alkenes (Singh and Misra, 2009). Experimental studies have confirmed that curcumin in fact binds to the redox-active residues Cys496 and Sec497 in rat TrxR (Fang et al., 2005), which corresponds to the docking results with the human protein (Singh and Misra, 2009). The binding of curcumin in the active site is mediated by a Michael reaction (Fang et al., 2005), as addressed in section II.A.7, and is further strengthened by H-bonding and hydrophobic interactions (Singh and Misra, 2009). Finally, the catalytic pocket of HIV-1 integrase contains a magnesium atom to which both carbonyl oxygens of curcumin’s diketo moiety coordinated (Vajragupta et al., 2005). The bonding distances between the oxygens and the magnesium were considerably shorter (1.7 and 2.1 Å) than the H-bonding distances between curcumin and several amino acid residues (≥2.7 Å), indicating that the metal-diketone interaction is quite significant in the binding of curcumin to HIV-1 integrase.
Inasmuch as approximately one-half of all proteins contain a metal (Thomson and Gray, 1998) and an estimated 25–33% of proteins require metals to function properly (Waldron and Robinson, 2009), the binding of curcumin to metalloproteins may constitute an important chemical and biological phenomenon, provided that the metals are accessible to curcumin (e.g., in the catalytic site of enzymes).
5. High Partition Coefficient (log P).
Despite the polarity of the flanking functional groups and the central dicarbonyl moiety, curcumin overall is rather lipophilic. This is evidenced by its poor solubility in aqueous solvent and good solubility in organic solvents (section II.B) as well as by its log P value, which is a measure of the extent to which a solute prefers the lipophilic phase (typically octanol) over the aqueous phase in a biphasic, immiscible solvent system. The log P of curcumin has been reported in the range of 2.3 (Jankun et al., 2006) to 2.6 (Fujisawa et al., 2004; Tomren et al., 2007). This relatively high degree of lipophilicity, which is attributable to the methine-rich segments that connect the polar regions (Balasubramanian, 2006), has several important implications on intermolecular interactions.
a. Biochemical implications of curcumin’s lipophilicity.
The first important implication is that curcumin’s methines can interact with hydrophobic amino acids in substrate binding sites of proteins. For example, molecular docking studies have confirmed hydrophobic interactions between curcumin and alanine and tyrosine in human immunoglobulin G (Liu et al., 2008). Such interactions were also ascertained between the phenylic rings of curcumin and valine in human NADPH-dependent aldo-keto reductase, using site-directed mutagenesis (Matsunaga et al., 2009). Furthermore, the heptanoid region of curcumin, docked to COX-1, was found to be surrounded by numerous hydrophobic amino acids, including leucine, isoleucine, alanine, glycine, and valine (Selvam et al., 2005). Additional specific examples are provided in Supplemental Table 1, altogether indicating that the electrostatic bonding between curcumin and polar residues, as detailed in sections II.A.1 to II.A.4, is stabilized by these hydrophobic interactions. As a result, a greater net bonding strength is achieved between curcumin and the molecular target than would have been the case in the absence of curcumin’s lipophilic features. Alternatively, the association with a target molecule could rely predominantly on hydrophobic interactions, as proposed for superficial hydrophobic clefts in albumin (Kragh-Hansen, 1981; Zsila et al., 2003).
Second, the relatively high partition coefficient of curcumin is indicative of the molecule’s capacity to interact with biomembranes (Jaruga et al., 1998), which is reflected by the intercalation of curcumin into model membranes (liposomes) composed of saturated (dipalmitoylphosphatidyl choline or dimyristoylphosphatidyl choline) (Barry et al., 2009; Perez-Lara et al., 2010), mono-unsaturated (dioleoylphosphatidyl choline) (Ingolfsson et al., 2007; Hung et al., 2008; Sun et al., 2008), or egg yolk phosphatidyl cholines (Kunwar et al., 2006; Karewicz et al., 2011). The association of curcumin with model membranes occurs at relatively high partition constants, i.e., in the range of 2.5 × 104 M−1 (Kunwar et al., 2006) to 4.3 × 104 M−1 (Karewicz et al., 2011) for egg yolk phosphatidylcholine liposomes and 2.4 × 104 M−1 for dioleoylphosphatidyl choline liposomes (Hung et al., 2008), indicating that curcumin can be taken up by cells by direct intercalation into the cell membrane.
b. Curcumin-biomembrane interactions.
The kinetics of these interactions (Sun et al., 2008) correspond to those of other amphipathic drugs and peptides (Sheetz and Singer, 1974; Lee et al., 2004) and entail a well-documented biphasic process (Banerjee et al., 1985; Huang and Wu, 1991; Heller et al., 1998). In the first phase, curcumin adsorbs to the outer membrane leaflet at the water-membrane interface in a low-energy binding state, which is associated with membrane thinning due to the lateral displacement of primarily polar phospholipid head groups (Sun et al., 2008). Within 60 seconds, a transition to a high-energy binding state occurs, whereby curcumin inserts deeper into the hydrophobic compartment of the membrane (Sun et al., 2008). No di- or multimerization of curcumin occurs during the partitioning process (Sun et al., 2008), even when saturation levels are reached at a curcumin:lipid molar ratio of ∼0.1 (Sun et al., 2008; Perez-Lara et al., 2010).
The exact intermolecular interactions and corollary molecular conformation of curcumin in either membrane binding state are, however, elusive. On the basis of data obtained with differential scanning calorimetry (DSC), X-ray diffraction, 2H NMR, and FT-IR spectroscopy it has been proposed that curcumin inserts into bilayers (Barry et al., 2009; Perez-Lara et al., 2010) with its main axis parallel to the bilayer normal (Perez-Lara et al., 2010). Theoretically, a membrane core-crossing conformation is possible, inasmuch as curcumin in planar conformation has a maximum length of roughly 22 C-atoms, whereas the acyl chain regions of dioleoylphosphatidyl choline- and dipalmitoylphosphatidyl choline-based bilayers have a length of 36 and 32 C-atoms, respectively. FT-IR analysis of the phospholipid PO2− and fatty acid ester C═O vibrational modes revealed that curcumin does not interact with these moieties (Perez-Lara et al., 2010), indicating that curcumin preferably localizes in the hydrocarbon-rich domain of the lipid bilayer, which is in agreement with the thermographic data obtained by DSC (Barry et al., 2009; Perez-Lara et al., 2010). Fluorescence quenching experiments established that liposome-encapsulated curcumin fluorescence is quenched more extensively by membrane-penetrating quenchers (acrylamide) than by less membrane-permeant quenchers (iodine) (Kunwar et al., 2006). Similar assays with brominated carboxylic acid derivatives (2- or 16-bromohexadecanoic acid or 11-bromoundecanoic acid), which more or less behave as component phospholipids in model membranes, corroborated that curcumin fluorescence is most intensely quenched by the bromine at C∆11 (Karewicz et al., 2011), i.e., in a very hydrophobic region of the bilayer (Bemporad et al., 2005). Accordingly, the lipophilic methines of curcumin most likely facilitate hydrophobic interactions with the acyl chains of phospholipids in cell and subcellular membranes, as was shown for erythrocytes (Jaruga et al., 1998), whereby the polar flanking regions may undergo H-bonding with water molecules embedded in the more distal portions of the membrane relative to the bilayer center (Bemporad et al., 2005). Under physiological conditions, the β-diketones, methoxy ethers, and hydroxyl groups may also be prone to electrostatic interactions with polar transmembrane protein residues, as suggested by Ingolfsson et al. (2007).
It should be noted that some discrepancies exist regarding the planar, membrane core-crossing conformation of curcumin. Curcumin’s electrons are delocalized over the entire molecule (Balasubramanian, 2006; Nardo et al., 2009), accounting for the lower transition energies of the π-electrons and therefore a considerably red-shifted absorption maximum (Kunwar et al., 2006; Ingolfsson et al., 2007; Hung et al., 2008; Karewicz et al., 2011) compared with feruloyl methane (Tønnesen et al., 1995), the exact half of the curcumin molecule. If indeed curcumin inserts into the membrane in a planar configuration parallel to the bilayer normal, then curcumin fluorescence should have been quenched by 16-bromohexadecanoic acid as well, i.e., in the most hydrophobic region of the membrane, in addition to the abovementioned quenching by 11-bromoundecanoic acid (Karewicz et al., 2011). In an electron-delocalized system such as that of curcumin, bromine-mediated quenching should be indiscriminate along the full length of the delocalized system, pleading against curcumin adopting a membrane core-crossing orientation. Moreover, such an orientation would place the highly polar central region of curcumin, where even the β-carbons and the Cα bear a slightly positive and negative charge, respectively (Balasubramanian, 2006), in the most hydrophobic part of the membrane (Bemporad et al., 2005), leading to thermodynamic instability. The adoption of the closed cis-enol conformation or diketo tautomerization, as has been described for curcumin in solvents (Khopde et al., 2000; Kolev et al., 2005; Balasubramanian, 2006; Nardo et al., 2008, 2009) that are chemically analogous to the bilayer core (Kragh-Hansen, 1981), or hydration of the β-diketone moiety by comigration of water molecules (Bemporad et al., 2005) might ameliorate the thermodynamic instability, but would not resolve it.
It is hence plausible that curcumin inserts deeper into the bilayer in the high-energy binding state but remains intertwined in the acyl chains of a single membrane leaflet, as illustrated in Barry et al. (2009), in a more perpendicular orientation relative to the bilayer normal. A perpendicular orientation would not be in conflict with earlier experimental data (Barry et al., 2009; Perez-Lara et al., 2010) inasmuch as the phase transition temperature of the lipid bilayer, including the pretransition of saturated phosphatidylcholines (DSC), would be affected in a fashion similar to the parallel orientation. The same applies to X-ray diffraction and vibrational (FT-IR spectroscopy) patterns as well as proton resonances (2H NMR) of the phospholipid methylenes in the presence of curcumin, which attest to interactions with curcumin but do not unequivocally reveal its orientation. Curcumin may also undergo rotamerization (section II.A.6), whereby for instance the polar moieties are oriented toward the more hydrophilic membrane region and the lipophilic segments toward the membrane core to achieve a thermodynamic optimum, by, e.g., adopting the a,s-trans,s-trans,a or a,s-cis,s-trans,a conformations in Fig. 4 or by adopting a nonplanar trans-diketo conformation as in Scheme 2 in Kolev et al. (2005).
In addition to the numerous H-bond donating and accepting sites (sections II.A.1 to II.A.3), one of the most favorable features of curcumin in regard to intermolecular interactions is its ability to undergo rotamerization at multiple locations in the methine bridge. As shown in Fig. 4, enolic curcumin contains five sites where the molecule can rotate about a C–C bond, whereas diketonic curcumin has six rotameric sites. With respect to enolic curcumin, rotation about the –Cα–Cβ– bond may be impaired due to resonance stabilization in the keto-enol moiety, depending on the chemical environment (section II.A.1 and Fig. 1B). However, in polar, strongly H-bonding solvents (or in cells), curcumin mainly comprises the open enol tautomer (Emsley, 1984; Khopde et al., 2000; Nardo et al., 2008) and could therefore undergo cis-to-trans isomerization around the Cα (Toullec, 1990). Density functional theory calculations on the structure of curcumin, where only four rotational axes were assumed (excluding the –Cα–Cβ– bond), yielded 24 possible rotamers of enolic curcumin, of which the nine most energetically favorable structures are presented in Fig. 4 (Kolev et al., 2005).
Curcumin’s rotameric capabilities considerably expand the versatility of intermolecular bonding inasmuch as both flanks can adopt the most suitable conformation to maximize the number of H-bonds between curcumin and its molecular target. This conformational flexibility is most evident from the docking studies that were used to construct Figs. 2 and 3, which revealed that curcumin tends to adopt an entirely different configuration for each molecular target. For instance, curcumin bends only slightly and retains most of its planar structure when interacting with the minor groove of duplexed oligonucleotides (Zsila et al., 2004; Koonammackal et al., 2011), whereas it becomes entirely nonplanar and often rotates around the longitudinal axis when interacting with, e.g., the C1B domain of PKCθ (Majhi et al., 2010; Das et al., 2011), PfATP6 (Ji and Shen, 2009), aldose reductases (Muthenna et al., 2009; Katsori et al., 2011), and beta amyloid peptides (Ngo and Li, 2012). As a result, curcumin is able to associate with a plethora of biomolecules due to its structural adaptability, which is "exploited" to maximize the number of interatomic bonds between curcumin and its target.
7. Michael Acceptor Capacity.
The final chemical attribute of curcumin that facilitates intermolecular interactions is the ability to act as a Michael acceptor (Dinkova-Kostova et al., 2001). A Michael reaction is an addition reaction whereby a nucleophile (i.e., molecules or ions with a lone pair of electrons such as thiols/thiolates and amines) is covalently attached to a compound containing an α,β-unsaturated carbonyl, which for β-diketones such as curcumin comprises the –Cδ=Cγ–(Cβ=O)– segment (Fig. 1C). The reaction has been proposed to proceed via a carbocation intermediate [–Cδ+–Cγ–(Cβ–O–)–], resulting in the binding of the nucleophilic atom (:N–R, where :N is the nucleophilic atom) to curcumin’s Cδ, yielding –(Cδ–N–R)=Cγ–(Cβ–OH)– (Fang et al., 2005).
Michael additions to curcumin have been described for molecular targets such as reduced glutathione (GSH) (Mathews and Rao, 1991; Awasthi et al., 2000), glutathione S-transferase (GST) (van Iersel et al., 1996, 1997), TrxR (Fang et al., 2005), interleukin (IL)-1 receptor-associated kinase (IRAK) (Jurrmann et al., 2005), the histone acetyltransferase E1A binding protein p300 (Marcu et al., 2006), calcium release-activated calcium channel protein 1 (Shin et al., 2012), and human ErbB-2 (HER2) (Jung et al., 2007), which is addressed separately in section III.E.1. The atoms through which the target molecules undergo a Michael addition include thiols (Mathews and Rao, 1991; Awasthi et al., 2000; Fang et al., 2005; Jurrmann et al., 2005) and selenols (Fang et al., 2005), both of which are mostly associated with a (seleno)cysteine residue.
The simplest and most illustrative example of a Michael reaction with curcumin is GSH, a tripeptide composed of glutamate-cysteine-glycine that is known for its different types of reactions involving nucleophiles, including its capacity to act as a Michael donor to α,β-unsaturated carbonyl compounds (Ketterer, 1988). Incubation of curcumin with increasing concentrations of GSH led to a proportional (Mathews and Rao, 1991) and time-dependent (Awasthi et al., 2000) decrease in absorbance at ∼430 nm, which is curcumin’s main absorption band in polar protic solvents (Tønnesen et al., 1995) such as buffered aqueous solutions (Fig. 1B), emanating from the electron delocalization over the entire conjugated system (Balasubramanian, 1990, 1991, 2006). A Michael reaction, which abrogates the conjugation in the methine segment, is therefore consistent with the observed decrease in absorbance at this wavelength. Although the experiments were performed at or above (near-)neutral pH [at which curcumin degradation is likely to occur (section II.C.1.a)] and curcumin was shown to cause GSH degradation (Awasthi et al., 2000), the formation of curcumin-GSH adducts was confirmed by high performance liquid chromatography (HPLC) and mass spectrometry (MS) (Awasthi et al., 2000). Moreover, curcumin is a very potent inhibitor of π-class GSTs (van Iersel et al., 1996) by binding to cysteine residues, mainly Cys47 in the active site of the P1-1 subunit of the GSTs, most likely via a Michael addition (van Iersel et al., 1997), altogether indicating that curcumin interferes in cellular glutathione redox metabolism through binding multiple targets via a Michael reaction.
Furthermore, curcumin has been shown to covalently bind to Cys496 and Sec497 in the catalytic pocket of TrxR in vitro, which resulted in curcumin concentration-dependent, irreversible inhibition of enzymatic activity (Fang et al., 2005). TrxR inhibition by curcumin also proceeded in a concentration-dependent manner in HeLa cells (Fang et al., 2005). The binding of curcumin to Cys496 and Sec497 was proposed to proceed via a nucleophilic attack by the thiol and selenol, respectively, on curcumin’s carbocation (Cδ+). A similar mechanism, i.e., alkylation of thiol residues, was observed in an IRAK-overexpressing murine T-cell line (EL-4IRAK) incubated with curcumin (Jurrmann et al., 2005). IRAK is recruited to the IL-1 receptor (IL-1R) upon IL-1 binding, which initiates the IL-1 signaling cascade (Bol et al., 2003). The recruitment of IRAK to IL-1R can be inhibited by numerous thiol-modifying compounds, including diamide, menadione, and phenylarsine oxide (Bol et al., 2003), as was also shown to be the case for curcumin (Jurrmann et al., 2005). In IL-1β-stimulated EL-4IRAK cells, the recruitment of IRAK to IL-1R was inhibited in a curcumin concentration-dependent manner without interfering with IL-1β binding to IL-1R. When the free thiol groups of IRAK were occupied by iodo-acetyl-[125I]iodotyrosine in IL-1-stimulated cells, curcumin blocked the extent of radiolabeling in a concentration-dependent manner, indicating thiol modification by curcumin. Inasmuch as the inhibition of IRAK thiols by, e.g., phenylarsine oxide, is reversible in the presence of thiol reductants such as dithiothreitol and dimercaptopropanol (Singh and Aggarwal, 1995; Friedrichs et al., 1998) but irreversible in case of curcumin, the thiol modification by curcumin does not entail a redox reaction but most likely a covalent Michael addition-mediated alkylation (Dinkova-Kostova and Talalay, 1999; Jurrmann et al., 2005).
The most compelling evidence for the Michael acceptor properties of curcumin come from studies in which a curcumin derivative that lacked the α,β-unsaturated carbonyl (i.e., tetrahydrocurcumin, Fig. 12) was employed as negative control or where site-directed mutagenesis was used to replace the target cysteine residues with a thiol-lacking amino acid. For example, radioactively labeled curcumin was shown to inhibit the acetyltransferase activity of p300 as a result of covalent binding to the protein. However, when enzyme activity assays were performed with radiolabeled tetrahydrocurcumin, no inhibition of p300 was observed (Marcu et al., 2006), confirming that the curcumin-p300 complex was the result of a Michael reaction. Similar results were obtained with ErbB-2 (section III.E.1). With respect to calcium release-activated calcium channel protein 1, which curcumin binds to and thereby inhibits Ca2+ flux through the Ca2+-release activated Ca2+ channel, replacement of the reactive Cys195 with a nonreactive serine reduced the inhibitory effect of curcumin. Moreover, tetrahydrocurcumin exhibited a less potent inhibitory effect on Ca2+-release activated Ca2+ channel, which was entirely abrogated as a result of the serine substitution (Shin et al., 2012).
These findings collectively attest to the fact that curcumin binds to target proteins containing Michael donating residues, i.e., mainly (seleno)cysteines. Inasmuch as cysteines play an important functional and catalytic role in the substrate binding site of enzymes (Saito, 1989; Carugo et al., 2003), the covalent binding of curcumin to target molecules via a Michael reaction likely constitutes an important contributory factor in curcumin’s pleiotropic binding behavior and corollary biological effects.
B. Curcumin Solubility
Because of its relatively high log P value (section II.A.5), curcumin is practically insoluble in aqueous medium. The solubility of curcumin is high in polar aprotic and polar protic solvents, as reflected by its order of solubility: acetone > 2-butanone > ethyl acetate > methanol > ethanol > 1,2-dichloroethane > 2-propanol > ether > benzene > hexane. DMSO is also a commonly used solvent, which dissolves curcumin up to a concentration of 11 mg/ml (versus 1 mg/ml for ethanol). Furthermore, curcumin is soluble in some nonpolar solvents with comparable εr values [benzene (2.3), toluene (2.38), diethyl ether (4.3), chloroform (4.81)] but does not dissolve well in alipathic or alicyclic organic solvents such as hexane and cyclohexane, respectively (Khopde et al., 2000).
The poor aqueous solubility of curcumin bears several important implications for in vitro and in vivo research. First, curcumin should always be dissolved in solvents that are miscible in water, which include acetone, butanone (to an extent), methanol, ethanol, 1,2-dichloroethane (up to 8.7 g/l w/w), 2-propanol, and DMSO. Of these solvents, the least toxic should preferably be used in experiments, which is best judged by their lethal 50% dose values, given here for the oral administration route in rats, unless noted otherwise (obtained from the respective material safety data sheet): acetone, 9.8 g/kg; 2-butanone, 2.7 g/kg; methanol, 5.6 g/kg; ethanol, 7.1 g/kg; 1,2-dichloroethane, 0.41 g/kg (mouse); isopropanol, 5.0 g/kg; DMSO, 18.0 g/kg. Accordingly, DMSO constitutes the most suitable solvent for in vitro and in vivo studies, although proper controls (solvent alone) should always be employed in all assays.
Second, some researchers report dissolving curcumin in slightly basic water or aqueous buffer, and several suppliers of curcumin even advocate raising the pH to improve solubility in aqueous medium. Making the conditions more alkaline, however, does not yield a tremendous increase in solubility relative to the 1.1 mM solubility at pH = 7.3 (Tønnesen, 1989b). More importantly, those who conduct experiments with curcumin should note that curcumin becomes very susceptible to degradation under alkaline conditions, i.e., a pH > 6.5 [Fig. 6B and "Solvolysis: (alkaline) hydrolysis," section II.C.1.b.i], due in part to the formation of the phenylate anion (Fig. 8 and "Chemical degradation and modification of curcumin," section II.C.1.b.i). As described in the latter section, the phenylate anion can give rise to curcumin radicals that in turn mediate degradation of the molecule, react with other curcumin radicals to form dimeric catabolites, or react with biomolecules in the cells, which may lead to different experimental outcomes. Inasmuch as protonated curcumin (i.e., at acidic pH) is even less soluble in water, preference should be given to DMSO as solvent system.
C. Curcumin Stability In Vitro
A drug must remain stable through all formulation stages and in the body after administration until a pharmacological effect has been conveyed. The stability of a drug, characterized by its chemical, physical, microbiological, therapeutic, and toxicological stability (O'Donnell and Bokser, 2005), is critical for drug safety and efficacy and can be affected by multiple factors, including oxidative/nitrosative degradation and/or modification, solvolysis, and aggregation. For curcumin, there are two major in vitro stability issues that complicate its use as a pharmaceutical, namely oxidative degradation and modification and solvolysis (Fig. 5). The oxidative degradation and modification can be categorized into photochemical processes (i.e., those induced by light absorption) and chemical processes (i.e., those induced in the absence of light).
Chemical modification and/or degradation change curcumin’s chemical structure and properties and thus affect its intermolecular bonding behavior (section II.A). This in turn may drastically affect curcumin’s pharmacokinetic and pharmacodynamic properties (section III), depending on whether the site interacting with the molecular target (section II.A) is chemically altered. Consequently, curcumin may entirely lose its anticancer attributes as a result of the modification/degradatory processes described in section II.C.1. Additionally, there are intracellular and in vivo stability concerns. These are alluded to in section II.C.2.a.
1. In Vitro Oxidative Degradation and Modification of Curcumin.
The main mode of photochemical and chemical degradation and modification of curcumin is through oxidation, which is primarily mediated by reactive oxygen species (ROS). It is well-established that curcumin is a potent antioxidant that interacts with different types of physiologically produced oxygen-centered radicals, including hydroxyl radical (∙OH) (Tønnesen, 1989a; Kunchandy and Rao, 1990; Tønnesen and Greenhill, 1992; Reddy and Lokesh, 1994; Ruby et al., 1995; Das and Das, 2002; Vajragupta et al., 2004; Biswas et al., 2005; Agnihotri and Mishra, 2011), superoxide anion (O2•–) (Kunchandy and Rao, 1990; Reddy and Lokesh, 1994; Ruby et al., 1995; Priyadarsini, 1997; Das and Das, 2002; Vajragupta et al., 2004; Biswas et al., 2005; Ak and Gulcin, 2008; Dairam et al., 2008), and peroxyl radicals (Priyadarsini, 1997; Masuda et al., 2001; Deng et al., 2006). Curcumin also reacts with physiologically produced nitrogen-centered radicals (nitric oxide, ∙NO, and nitrogen dioxide radicals) (Unnikrishnan and Rao, 1995; Sreejayan and Rao, 1997; Onoda and Inano, 2000), sulfur-centered radicals (oxidized glutathione) (Khopde et al., 1999), and oxidants such as hydrogen peroxide (H2O2) (Tønnesen and Greenhill, 1992; Iwunze, 2004; Ak and Gulcin, 2008; Griesser et al., 2011) as well as singlet oxygen (1O2) (Tønnesen et al., 1986; Chignell et al., 1994; Subramanian et al., 1994; Das and Das, 2002). Curcumin is further capable of reacting with nonphysiological radicals such as azide radicals (Gorman et al., 1994; Priyadarsini, 1997; Khopde et al., 1999; Priyadarsini et al., 2003), 2,2-diphenyl-1-picrylhydrazyl (DPPH, a stable nitrogen-centered radical) (Venkatesan and Rao, 2000; Priyadarsini et al., 2003; Fujisawa et al., 2004; Ak and Gulcin, 2008; Feng and Liu, 2009), 2,2'-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid) radical cations (ABTS•+) (Venkatesan and Rao, 2000; Ak and Gulcin, 2008; Feng and Liu, 2009), dimethyl-4-phenylenediamine dihydrochloride radical cations (Ak and Gulcin, 2008), halocarbonperoxyl radicals (Priyadarsini, 1997; Khopde et al., 1999), dibromine radical anions (Khopde et al., 1999), galvinoxyl radicals (Feng and Liu, 2009), and Triton-X 100 radicals (Priyadarsini, 1997). Lastly, it has been shown to interact with nonphysiological peroxides such as tert-butyl hydroperoxide (Sugiyama et al., 1996).
The reactivity of radicals toward curcumin is chiefly dictated by the bond dissociation energies (Fujisawa et al., 2002) of curcumin’s functional groups, being approximately 80 kcal∙mol−1 for the phenylic O−H, > 90 kcal∙mol−1 for the central C−H (Cα, Fig. 1C), and 98 kcal∙mol−1 for the enolic O−H (Sun et al., 2002; Wright, 2002). Accordingly, radicals predominantly attack the phenylic hydroxyl group (Barclay et al., 2000; Sun et al., 2002; Wright, 2002; Priyadarsini et al., 2003; Chen et al., 2006c; Feng and Liu, 2009; Agnihotri and Mishra, 2011), either by hydrogen atom transfer (HAT; H3R + R′• → H2R• + HR′, where H3R represents fully protonated curcumin) or by single electron transfer (SET) from curcumin to the radical (H3R + R′• → H3R•+ + R′–) (Barclay et al., 2000; Sun et al., 2002; Priyadarsini et al., 2003; Litwinienko and Ingold, 2004; Galano et al., 2009; Agnihotri and Mishra, 2011). HAT can also occur at the methines, with the Cα as the most frequently proposed location (Sugiyama et al., 1996; Jovanovic et al., 1999; Sun et al., 2002; Wright, 2002; Priyadarsini et al., 2003; Agnihotri and Mishra, 2011). Additionally, the enol moiety is subject to HAT (Sugiyama et al., 1996; Ak and Gulcin, 2008; Agnihotri and Mishra, 2011) or to ionization of the enol proton (pKa1 = 7.5) and subsequent electron transfer from the radical via sequential proton loss electron transfer (SPLET; H3R → H2R– + H+, H2R– + R′• → H2R• + R′–, H2R• → HR•– + H+) (Litwinienko and Ingold, 2004; Galano et al., 2009), but in all instances the site of attack depends strongly on the reactivity and thus type of the radical (i.e., the bond dissociation energy of the parent compound and the redox potential), the ionization potential of abstractable substrate atoms, and the solvent (Zheng et al., 1997; Wright, 2002; Litwinienko and Ingold, 2004; Feng and Liu, 2009; Galano et al., 2009; Agnihotri and Mishra, 2011). Consequently, ∙OH can oxidize curcumin at almost every O−H and C−H bond (Wright, 2002; Agnihotri and Mishra, 2011), whereas less reactive radicals such as O2•– will mainly attack phenylic O−H bonds (Sawyer and Valentine, 1981; Jurasek and Argyropoulos, 2006), as has been reported for other (poly)phenolic compounds (Paya et al., 1992; Yen and Hsieh, 1997; Mochizuki et al., 2002; Cetinkaya et al., 2012).
Curcumin derivatives have been employed to demonstrate the importance of the type of radical in relation to the HAT/SET/SPLET site on curcumin. For certain ROS (e.g., tert-butyl hydroperoxide and ABTS•+), the phenylic groups and/or the alkenes in the methine segment are not required for an antioxidant effect, as was shown for dimethoxytetrahydrocurcumin (keto-enolic curcumin with both methoxyphenyls substituted by ortho-dimethoxybenzenes and a hepta-4-ene bridge) (Sugiyama et al., 1996) and for benzylated curcumin derivatives, where the phenylic hydroxyl groups are protected by benzyl groups (Feng and Liu, 2009). Lipid peroxidation by tert-butyl hydroperoxide was inhibited by 36% by dimethoxytetrahydrocurcumin compared with ∼71% by curcumin and ∼83% by tetrahydrocurcumin (keto-enolic curcumin with hepta-4-ene bridge) (Sugiyama et al., 1996). Similarly, the radical scavenging capacity of the benzyl-protected curcumin was equal to curcumin and tetrahydrocurcumin for ABTS•+ but nonexistent for galvinoxyl or DPPH radicals (Feng and Liu, 2009). For the latter two radical species, the phenylic groups were of paramount importance (Sreejayan and Rao, 1996; Feng and Liu, 2009). The order of antioxidative capacity of curcumin toward DPPH is tetrahydrocurcumin > hexahydrocurcumin (keto-enolic curcumin with heptane bridge) ≈ octahydrocurcumin (di-enolic curcumin heptane bridge) > curcumin > > demethoxycurcumin (R1 = H; Fig. 1A) > >> bisdemethoxycurcumin (R1 = R2 = H; Fig. 1A) (Somparn et al., 2007; Feng and Liu, 2009). With respect to galvinoxyl radicals, the order was curcumin > > octahydrocurcumin > tetrahydrocurcumin (Feng and Liu, 2009).
The HAT/SET/SPLET reactions between curcumin and radicals cause curcumin to become a radical itself, which subsequently may undergo radical delocalization over the conjugated segments (Masuda et al., 1999; Priyadarsini et al., 2003; Litwinienko and Ingold, 2004; Chen et al., 2006c; Griesser et al., 2011), intramolecular rearrangement of bonds (Tønnesen and Greenhill, 1992; Griesser et al., 2011; Gordon and Schneider, 2012), degradation (Tønnesen et al., 1986; Tønnesen and Greenhill, 1992; Chignell et al., 1994; Masuda et al., 1999; Agnihotri and Mishra, 2011), dimerization through radical coupling between curcumin radicals (Masuda et al., 1999, 2002; Fujisawa et al., 2004), or radical coupling with other molecular radicals to yield heterodimeric complexes (Masuda et al., 2001) or possibly multimeric complexes (in case of double deprotonation at the phenylic oxygens) (Priyadarsini et al., 2003). In the latter case, noncurcumin radicals may comprise, for example, lipid (cholesterol) radicals (Iuliano, 2011) co-encapsulated in liposomal formulations of curcumin (Thangapazham et al., 2008a; Chen et al., 2009a; Shi et al., 2012) or radicals of antioxidants (Suntres, 2011) added to the curcumin formulation (Oetari et al., 1996), both formed by curcumin-mediated type I or II photochemical reactions ("Mechanisms of photo-induced reactive oxygen species production by curcumin and photochemical degradation/modification of dissolved curcumin," section II.C.1.a.ii), Haber and Weiss (1932) or Fenton (1894) reactions ("Chemical degradation and modification of curcumin," section II.C.1.b.i), and/or oxidation reactions in the absence of transition metals ("Chemical degradation and modification of curcumin," section II.C.1.b.i).
With respect to degradation, cleavage of the molecule occurs at different locations in the methine bridge (Tønnesen et al., 1986; Tønnesen and Greenhill, 1992; Chignell et al., 1994; Masuda et al., 1999; Agnihotri and Mishra, 2011), yielding two nonidentical, monophenylic catabolites. The degradation of curcumin does not require alkenes [which are involved in the formation of cyclic ether (Griesser et al., 2011) and dioxetane (Masuda et al., 1999) intermediates in curcumin radicals] or the phenylic groups per se, as catabolites were found after ROS-induced degradation of dimethoxy tetrahydrocurcumin in O2-saturated acetonitrile that were structurally analogous to the catabolites found after curcumin degradation (i.e., 3,4-dimethoxybenzoic acid, 3′,4′-dimethoxyacetophenone, and 3-(3,4-dimethoxyphenyl)-propionic acid) (Sugiyama et al., 1996).
a. Photochemical degradation and modification of curcumin.
Curcumin is a chromophore that absorbs strongly in the visible wavelength range, making it susceptible to photo(oxidative) degradation and modification in daylight and artificial lighting. Photo-induced degradation of curcumin occurs irrespectively of the chemical environment, i.e., in solid state as well as in different organic solvents (Tønnesen et al., 1986; Khurana and Ho, 1988; Ansari et al., 2005), even in the absence of UV light and oxygen (Tønnesen et al., 1986). However, the composition, degradation kinetics, and relative abundance of the degradation products differ depending on the physical state of the compound and the conditions.
i. Photochemical Degradation of Solid State (Crystalline) Curcumin.
In regard to the crystalline form of curcumin, the type of solvent used for extracting and purifying curcumin appears to have an effect on the degree of degradation. For example, 120-hour sunlight exposure of crystalline curcumin obtained by ethanol extraction yielded three unidentified compounds with a cumulative relative concentration of 33% in addition to vanillin (34%), ferulic aldehyde (0.5%), ferulic acid (0.5%), and vanillic acid (0.5%) (Khurana and Ho, 1988) (Fig. 6A). For sunlight-exposed, methanol-extracted curcumin in solid state, the degradation products were identical but the concentrations of vanillin, ferulic aldehyde, ferulic acid, and vanillic acid were mostly lower, namely 2.0%, 0.2%, 0.1%, and 1.5%, respectively (Khurana and Ho, 1988). The degradation of crystalline curcumin abides by second order kinetics, at least when exposed to light in the 400- to 750-nm wavelength range for 4 hours (Tønnesen et al., 1986).
ii. Photochemical Degradation and Modification of Dissolved Curcumin.
The photochemical degradation of solubilized curcumin yields similar end products as light-irradiated crystalline curcumin, namely vanillin, vanillic acid, 4-vinylguaiacol, ferulic acid, and ferulic aldehyde (Fig. 6A), when dissolved in isopropanol and irradiated for 4 hours at 400-510 nm (Tønnesen et al., 1986). However, the degradation kinetics are solvent dependent. The degradation rate constants (kdeg) of 40 µM curcumin dissolved in different organic solvents and exposed for 4 hours to visible light (400–750 nm) conform to first-order kinetics and constitute kdeg = 1.4×10−2 h−1 (methanol), kdeg = 1.4×10−1 h−1 (ethyl acetate), kdeg = 2.9×10−1 h−1 (chloroform), and kdeg = 5.2×10−1 h−1 (acetonitrile), corresponding to half-lives of 50.2 hours, 5.1 hours, 2.4 hours, and 1.3 hours, respectively (Tønnesen et al., 1986). Visible light imposes a broader degree of degradation than UV light inasmuch as irradiation of curcumin in methanol with 254-nm light produced three unspecified degradation products, whereas irradiation with daylight produced five unspecified degradation products (Ansari et al., 2005), indicating that some of the degradative reactions require the first excited or triplet state induced through the violet absorption band ("Mechanisms of photo-induced reactive oxygen species production by curcumin and photochemical degradation/modification of dissolved curcumin," section II.C.1.a.ii).
Exposure of curcumin to visible light also causes solvent-dependent structural modifications in addition to degradation. Chromatographic separation and subsequent analysis by MS and NMR revealed that curcumin subjected to irradiation with 400- to 750-nm light for 4 hours was associated with double deprotonation and cyclization at one of the two o-methoxyphenyl residues (Tønnesen et al., 1986), yielding 7-hydroxy-1-[(2E)-3-(4-hydroxy-3-methoxyphenyl)prop-2-enoyl]-6-methoxynaphthalen-2(1H)-one (Fig. 6A). The internal cyclization occurred in isopropanol, methanol, and chloroform but not in acetonitrile and ethyl acetate (Tønnesen et al., 1986). Moreover, the photo-induced degradation products of curcumin can further react with solvent molecules under the influence of visible light irradiation, as has been shown for ferulic aldehyde in isopropanol, yielding the condensation product 4-[(1E)-3-(propan-2-yloxy)prop-1-en-1-yl]guaiacol (Fig. 6A) (Tønnesen et al., 1986).
Mechanisms of photo-induced reactive oxygen species production by curcumin and photochemical degradation/modification of dissolved curcumin.
Visible light-induced curcumin fragmentation or modification of its chemical structure is caused mostly by oxidative processes through ROS, which in this case are produced by triplet state curcumin (3Cur) itself upon irradiation. The production of ROS by curcumin is mediated via type I (electron transfer to molecular oxygen (O2): 3Cur + O2 → Cur•+ + O2•–) (Tønnesen et al., 1986; Chignell et al., 1994; Dahl et al., 1994; Shen et al., 2005) and type II (energy transfer to O2: 3Cur + O2 → Cur + 1O2) (Chignell et al., 1994; Dahl et al., 1994; Shen et al., 2005) photochemical reactions. However, the extent of O2•– and 1O2 generation by photochemical means is probably limited (Priyadarsini, 2009) insofar as singlet curcumin predominantly undergoes nonradiative decay to the ground state via intramolecular charge transfer (ICT) (Khopde et al., 2000; Zsila et al., 2003; Shen and Ji, 2007) and subsequent excited state intramolecular proton transfer (ESIPT) (Khopde et al., 2000; Nardo et al., 2008; Priyadarsini, 2009). Intersystem crossing to the triplet state, from which the type I and II reactions proceed, constitutes a minor pathway (Priyadarsini, 2009). Nevertheless, these photochemical reactions produce sufficient ROS to cause substantial curcumin modification and degradation, as addressed in the previous section.
The photogeneration of O2•– by curcumin has been studied by electron paramagnetic resonance (EPR) using 5,5-dimethyl-1-pyrroline-N-oxide (DMPO) as spin trap in combination with concomitant resonant light (420 nm) irradiation (Chignell et al., 1994). Curcumin was found to generate O2•– in different types of solvents, including benzene, toluene, acetone, acetonitrile, DMSO, and ethanol, which could not be inhibited by the 1O2 quencher strychnine (Chignell et al., 1994). Furthermore, irradiation of curcumin-encapsulating sodium dodecyl sulfate micelles with a 100-W tungsten-halogen lamp (400-nm cutoff filter) for 3 hours concurred with complete consumption of dissolved oxygen, of which ∼5% was recovered after the addition of catalase (Chignell et al., 1994), an enzyme responsible for disproportionating H2O2 to H2O and O2. The H2O2 was most likely formed by the spontaneous disproportionation of O2•– (Sawyer and Valentine, 1981), which had formed as a result of curcumin irradiation. These experimental data confirm that O2 undergoes a one-electron reduction by accepting a triplet state electron from curcumin, yielding a curcumin radical in the process (Shen et al., 2005).
The site of the type I process, and hence the location where the curcumin radical forms, appears to be the diketo moiety following excited state ICT from the methoxyphenyl group (Khopde et al., 2000; Zsila et al., 2003; Shen and Ji, 2007). Accordingly, type I reactions in curcumin produce two radical species, O2•– and a ketonic O-centered curcumin radical (Khopde et al., 2000; Zsila et al., 2003; Shen et al., 2005), which is most likely stabilized by the dicarbonyl moiety (O=C−C=C−O• ↔ O=C−C•−C=O ↔ •O−C=C−C=O). In turn, photogenerated O2•– can be scavenged by curcumin, producing a phenylic O-centered curcumin radical in the process. Both types of O-centered curcumin radicals are catalysts for the degradation and modification of curcumin.
The scavenging of O2•– by curcumin has been demonstrated in several studies. By use of an in vitro xanthine-xanthine oxidase system to generate O2•–, curcumin was shown to inhibit the O2•–-mediated reduction of ferricytochrome c by 25–40% at a concentration of 20–80 μM in a concentration-dependent manner (Reddy and Lokesh, 1994; Das and Das, 2002) and of nitroblue tetrazolium (NBT) by 19% at a curcumin concentration of 270 μM [this system also contained catalase and superoxide dismutase (SOD)] (Vajragupta et al., 2004). The latter results were confirmed by EPR in the same O2•–-generating system employing DMPO as O2•– trap, showing that 10 mM curcumin inhibited 30% of DMPO-OOH adduct formation at a 250-mM DMPO final concentration (Vajragupta et al., 2004). Similarly, in an O2•–-generating system based on the photosensitization of riboflavin, curcumin inhibited the O2•–-mediated reduction of NBT by 50% at a concentration of 17 μM (Ruby et al., 1995) and by 43% at a concentration of 41 μM (Ak and Gulcin, 2008). Although the quantitative data are not in complete agreement because of slight differences in the assay method and possibly differences in the employed light source and irradiation parameters, the trend is the same, namely a curcumin-mediated decrease in the reduction of NBT. The findings hence collectively attest to the fact that curcumin is capable of reacting with O2•–. Reactions between O2•– and antioxidants can proceed via SET (Buettner, 1993; Bielksi et al., 1985), but in the case of curcumin, O2•– is more likely to disproportionate via HAT from the phenylic hydroxyl group to yield a curcumin phenoxyl radical (Jovanovic et al., 1999; Masuda et al., 1999; Barclay et al., 2000; Sun et al., 2002; Wright, 2002; Priyadarsini et al., 2003; Litwinienko and Ingold, 2004; Chen et al., 2006c; Agnihotri and Mishra, 2011) and a hydroperoxyl anion (O2•– + ROH → HO2– + RO•) (Nakanishi et al., 2003; Lee et al., 2006; Jung and Ham, 2007), given the relatively low bond dissociation energy of O2•– (Jurasek and Argyropoulos, 2006). However, the disproportionation of O2•– in aqueous medium (2 O2•– + 2 H2O → O2 + H2O2 + 2 OH–) occurs at a rate constant of < 3×10−1 L∙mol−1∙s−1 (Sawyer and Valentine, 1981). The oxidation of curcumin by this route is therefore associated with a marginal phenoxyl radical yield, as O2•– reacts with comparable antioxidants such as ascorbic acid and α-tocopherol at considerably lower rate constants (k = 10−4-10−5 L∙mol−1∙s−1) (Bielksi et al., 1985; Buettner, 1993). The oxidation of the phenylic hydroxyl group is largely outcompeted by the spontaneous disproportionation process.
Once a phenylic O-centered curcumin radical has been formed, however, the radical can translocate along the conjugated system to the methine carbons (Sugiyama et al., 1996; Priyadarsini et al., 2003; Fujisawa et al., 2004; Litwinienko and Ingold, 2004; Chen et al., 2006c; Griesser et al., 2011) to produce C-centered radicals. The carbonyl O-centered radical formed by triplet state electron transfer to O2 exhibits a similar pattern, albeit the radical is only confined to the Cα and not the other methine carbons. The reactions that proceed from the C-centered curcumin radicals ultimately yield cyclized end products (Masuda et al., 2002; Fujisawa et al., 2004; Griesser et al., 2011) and degradatory catabolites (Tønnesen et al., 1986; Tønnesen and Greenhill, 1992; Chignell et al., 1994; Masuda et al., 1999; Agnihotri and Mishra, 2011), as pointed out previously (section II.C.1) and detailed in the section on chemical degradation ("Chemical degradation and modification of curcumin," section II.C.1.b.i).
Additionally, curcumin may be oxidized by H2O2 formed from the spontaneous disproportionation of O2•–. H2O2 apparently oxidizes curcumin at the phenylic hydroxyl group, as evidenced by the decrease in the rate of curcumin oxidation from 4.4 ± 0.4 to 3.9 ± 1.0 µM∙min−1 (11%) for 4′-methoxycurcumin (see Fig. 1C for numbering) and to 0.07 ± 0.01 µM∙min−1 (98%) for 4′,4′′-dimethoxycurcumin in Tris-HCl buffered solution (pH = 8.0) upon addition of H2O2 (Griesser et al., 2011). The incidental formation of ∙OH from H2O2 is precluded in these experiments, as ∙OH would have easily oxidized 4′,4′′-dimethoxycurcumin. H2O2-mediated curcumin oxidation produces a phenoxyl radical, the fate of which has been clarified above. Corroboratively, oxidation of curcumin by H2O2 was reported to yield ferulic acid (Tønnesen and Greenhill, 1992), one of the identified catabolites of irradiated curcumin (Fig. 6A) produced by C−C bond cleavage.
H2O2 could also facilitate the production of ∙OH through Haber-Weiss/Fenton chemistry (Fenton, 1894; Haber and Weiss, 1932) or in the absence of redox-sensitive transition metals (Blanksby et al., 2007). The indirect formation of more reactive radicals such as ∙OH from O2•– (via H2O2) is not unlikely, provided that p-anisyl and veratryl derivatives of curcumin, which lack both phenylic hydroxyl groups (Babu and Rajasekharan, 1994), were able to scavenge O2•– in an isolated O2•–-generating system (photosensitization of riboflavin) (Anto et al., 1996b). This could have only occurred via enolic O−H or C−H bonds that have higher bond dissociation energies than O2•–, and thus by more reactive ROS than O2•–. The scavenging of ∙OH by curcumin and its subsequent degradation have been proven in numerous studies (Tønnesen, 1989a; Kunchandy and Rao, 1990; Tønnesen and Greenhill, 1992; Reddy and Lokesh, 1994; Ruby et al., 1995; Das and Das, 2002), the specifics of which are further discussed in the section on chemical degradation ("Chemical degradation and modification of curcumin," section II.C.1.b.i), where this process is more prominent and experimentally better supported.
Lastly, it has been reported that the water-mediated disproportionation of O2•– as well as electron transfer reactions between O2•– and redox-sensitive transition metals yield 1O2 (Khan, 1981, 1989; Corey et al., 1987a,b). 1O2 production also occurs in the Haber-Weiss reaction (Khan and Kasha, 1994). Consequently, the curcumin-mediated photoproduction of the relatively innocuous O2•– could leverage into the generation of the highly toxic 1O2 via multiple pathways, which will have profound implications on curcumin degradation and modification ("Mechanisms of photo-induced reactive oxygen species production by curcumin and photochemical degradation/modification of dissolved curcumin," section II.C.1.a.ii).
The abovementioned set of reactions, including additional O2•–-mediated reactions as described in Sawyer et al. (1981) and Winterbourn and Kettle (2003), in all probability constitute the starting conditions from which curcumin is chemically degraded (Tønnesen et al., 1986; Ansari et al., 2005) or modified via pathways as proposed by, e.g., Tønnesen and Greenhill (1992), Masuda et al. (1999, 2002), Fujisawa et al. (2004), Agnihotri and Mishra (2011), and Griesser et al. (2011). These reactions also account for the formation of C-centered radicals as observed by Chignell et al. (1994) during irradiation of dissolved curcumin.
As opposed to O2•–, type II reaction-derived 1O2 constitutes a highly reactive ROS that attacks numerous biomolecules (e.g., fatty acids, amino acids, nucleic acids, steroids, and endogenous pigments) as well as olefins and (hetero-)aromatics, both of which are present in curcumin. The reactions typically proceed via oxygenation (H-abstraction and O-addition) or cycloaddition (forming 1,2-dioxetane or an endoperoxide).
In case of curcumin, 1O2 formation as a result of irradiation with resonant visible light has been demonstrated directly by 1O2 phosphorescence (Chignell et al., 1994), steady-state luminescence spectroscopy (Nardo et al., 2008), and EPR spin trapping (Chignell et al., 1994) and indirectly by absorption spectroscopy of curcumin in the presence of 1O2 quenchers (Tønnesen et al., 1986; Chignell et al., 1994) as well as by density functional theory calculations (Shen et al., 2005). The 1O2 yields after irradiation of curcumin with resonant visible light are strongly solvent dependent (Chignell et al., 1994), which indicates that the energy transfer from triplet state curcumin to O2 proceeds from the β-dicarbonyl moiety. In nonpolar (toluene and benzene) and polar aprotic (acetonitrile) solvents, the quantum yield of 1O2 formation is 0.11–0.12 (Chignell et al., 1994; Gorman et al., 1994), whereas in polar protic solvents such as ethanol and iso-propanol as well as in surfactant micelles (sodium dodecyl sulfate and Triton X-100) the quantum yields are approximately 10-fold lower (Chignell et al., 1994) and fourfold lower (in methanol versus acetonitrile, Hanne Hjort Tønnesen, personal communication). This may stem from the shift in curcumin’s keto-enol equilibrium toward the enol form in solvents with increasing proticity (section II.A.1). Ketones have long-lived triplet states (required for 1O2 generation) in aprotic solvents (Wagner, 1976), and curcumin partially adopts a diketo conformation in aprotic solvents (Fig. 1B). The longevity of the keto triplet state is perturbed by inter- and intramolecular H-bonding in alcohols and water (Fig. 1B), accounting for the marginal 1O2 generation by curcumin in these solvents and the more extensive 1O2 generation in nonpolar and aprotic solvents (Chignell et al., 1994). On the other hand, deactivation of curcumin’s first excited state may proceed via mechanisms other than nonradiative decay through intermolecular H-bonding with solvent molecules and may entail ESIPT that involves reketonization, as detailed in Nardo et al. (2008). ESIPT-mediated reketonization would interfere with intersystem crossing of the excited state electron to the triplet state, subsequent energy transfer to O2, and corollary production of 1O2.
On top of producing 1O2, curcumin has been shown to scavenge 1O2 in two different in vitro 1O2-generating systems. The first system employed peroxidized 3,3′-(1,4-naphthylene)-dipropionate (NDP-O2), which, when incubated at 37°C, yields NDP, O2, and 1O2. The addition of NDP-O2 to plasmid DNA at 37°C caused 1O2-mediated single strand breaks in the DNA (measured by agarose gel electrophoresis and densitometry), a process that was inhibited by 48% in the presence of 100 μM curcumin and by 31% in the presence of 10 μM curcumin (Subramanian et al., 1994). Moreover, the extent of protection was in the order of curcumin > demethoxycurcumin > bis-demethoxycurcumin (Subramanian et al., 1994), suggesting that the methoxy groups play a role in 1O2 scavenging. In the second system, 1O2 was generated by photosensitization of Rose Bengal and quantified by EPR spectroscopy of 2,2,6,6-tetramethyl piperidine (TEMP)-O2 adducts (TEMPO). TEMPO formation was inhibited by 85% at a 3.12 μM curcumin concentration and by 50% at 2.75 μM curcumin concentration (Das and Das, 2002). Both studies hence provide evidence that curcumin is a 1O2 scavenger.
Curcumin-mediated 1O2 photoproduction and subsequent scavenging is accompanied by curcumin degradation that concurs with a reduction in 1O2 phosphorescence (Chignell et al., 1994), underscoring the notion that, just as singlet state curcumin ("Chemical degradation and modification of curcumin," section II.C.1.b.i), triplet state curcumin is autodestructive. Some or all of the resulting degradation products (vanillin, vanillic acid, ferulic acid, ferulic aldehyde, 4-vinylguaiacol, and 7-hydroxy-1-[(2E)-3-(4-hydroxy-3-methoxyphenyl)prop-2-enoyl]-6-methoxynaphthalen-2(1H)-one, Fig. 6A) (Tønnesen et al., 1986) are photosensitizers themselves, albeit weaker than curcumin, and propagate the degradation process with continued light exposure (Chignell et al., 1994). Unfortunately, the precise 1O2-mediated degradation mechanism is presently elusive, although it is known that certain hindered phenyls such as 2,6-di-tert-butylphenol, which is somewhat similar to the methoxyphenyl residues of curcumin, are susceptible to H-abstraction by 1O2 to yield substituted phenoxyl radicals and corollary dimers (Matsuura et al., 1972). Corroboratively, the breakdown of curcumin in isopropanol by visible light (400–750 nm) could be inhibited by the addition of 1O2 quenchers β-carotene and 1,4-diazabicyclo(2,2,2)octane and induced by photoactivation of methylene blue (an 1O2 generator) at wavelengths not absorbed by curcumin (500–750 nm) (Tønnesen et al., 1986). Methylene blue-facilitated degradation of curcumin could in turn be inhibited by β-carotene (Tønnesen et al., 1986), altogether confirming that 1O2 is responsible for the catalytic effects. Similarly, photo-induced curcumin degradation in toluene could be prevented by the addition of strychnine (Chignell et al., 1994).
The autodestructiveness of light-exposed curcumin does not necessarily require an aerated environment and can manifest itself under deoxygenated conditions, i.e., in the absence of O2•– or 1O2. When curcumin was embedded in an oxygen-impermeable carbohydrate matrix and irradiated with broadband (240–600 nm) or filtered (400–510 nm) light, extensive photobleaching was observed, but chromatographic analysis did not reveal any of the common degradation products (Tønnesen et al., 1986). A comparable effect was evidenced for curcumin in deoxygenated toluene, where photobleaching occurred that could not have been caused by ROS (Chignell et al., 1994). However, the exact mechanism underlying these O2-independent processes has remained at large, mainly because the fundamentals of nonoxidative photobleaching are elusive.
b. Chemical degradation.
In addition to photo-induced degradation, curcumin undergoes rapid and considerable degradation in alkaline buffered aqueous solutions, cell culture medium (RPMI 1640), and human plasma (Tønnesen and Karlsen, 1985a,b; Wang et al., 1997; Griesser et al., 2011; Gordon and Schneider, 2012). The catalytic end products of curcumin in phosphate-buffered solutions equilibrated at 31.5°C (pH = 8.5) (Tønnesen and Karlsen, 1985a) or 37°C (pH = 7.2) (Wang et al., 1997) were vanillin, ferulic acid, and feruloyl methane (Fig. 6B). Incubation at 31.5°C for 5 minutes or 28 hours at pH = 8.5 yielded predominantly feruloyl methane, the further degradation of which resulted in the formation of vanillin and acetone as byproduct (Tønnesen and Karlsen, 1985a). Incubation of curcumin in phosphate-buffered solution, RPMI 1640 medium, or human plasma at 37°C for 1 hour at pH = 7.2 produced vanillin as the major degradation product as well as (2Z,5E)-2-hydroxy-6-(4-hydroxy-3-methoxyphenyl)-4-oxohexa-2,5-dienal (Fig. 6B) (Wang et al., 1997). At longer incubation times the (2Z,5E)-2-hydroxy-6-(4-hydroxy-3-methoxyphenyl)-4-oxohexa-2,5-dienal was further degraded to vanillin (Wang et al., 1997), implying the cogeneration of (2Z)-2-hydroxy-4-oxopent-2-enal as catabolic byproduct, albeit this was not confirmed experimentally. Furthermore, the suspension of curcumin in Tris-HCl buffer (pH = 8) at room temperature for 30 minutes or phosphate buffer (pH = 7.5) for 2 hours yielded vanillin, ferulic acid, and feruloylmethane as minor degradation products and 6-hydroxy-1-(4-hydroxy-3-methoxyphenoxy)-3-(4-hydroxy-3-methoxyphenyl)-1,3,3a,6a-tetrahydro-4H-cyclopenta[c]furan-4-one as the major degradation product (Fig. 6B) (Griesser et al., 2011; Gordon and Schneider, 2012), indicating that internal cyclization can also occur in the absence of type I or type II photochemical reactions.
With respect to buffered aqueous solutions, the rate and rate order of the decomposition are pH and temperature dependent, respectively. The degradation rate constants and curcumin half-lives in phosphate-buffered solutions at 31.5°C (pH = 8.5) (Tønnesen and Karlsen, 1985a) or 37°C (pH = 7.2) (Wang et al., 1997) are plotted as a function of pH in Fig. 6B. It is evident that at both temperatures curcumin is relatively stable in a pH range of 1–6, i.e., acidities at which curcumin is expected to be (almost) fully protonated in the enol form (Tønnesen and Karlsen, 1985b) (Fig. 1C). At pH > 6, degradation rate constants accelerate rapidly by more than two orders of magnitude and reach a maximum around pH = 8. The degradation kinetics at 31.5°C are of a second order (Tønnesen and Karlsen, 1985b), whereas at body temperature the kinetics entail a first-order process (Wang et al., 1997). Curcumin degrades much faster at 37°C (50% in 9.4 minutes at pH = 7.2 and in 1.1 minute at pH = 8.0) (Wang et al., 1997) than at 31.5°C (50% in 4.6 hours at pH = 7.3 and in 2.1 minutes at pH = 8.0) (Tønnesen and Karlsen, 1985b).
i. Mechanisms of Chemical Degradation.
Solvolysis: (alkaline) hydrolysis
Solvolysis encompasses the degradation of a compound through a nucleophilic substitution or elimination reaction by solvent molecules. Here, only hydrolysis will be considered, a process that yields a protonated and a hydroxylated fragment of the parent molecule. The same principally applies to alkaline hydrolysis, where, in the case of β-dicarbonyl compounds, the reaction advances according to RCOCH2COR′ + OH– → RCOCH3 + R′CO2– (Rahil and Pratt, 1977).
Accordingly, ferulic acid and feruloyl methane are justifiable byproducts of hydrolytic degradation of curcumin (Fig. 7A). Hydrolysis is facilitated by OH– ions, and the rate at which curcumin degrades increases with decreasing pOH (Fig. 6B and 7A). This premise notwithstanding, radical-induced strand cleavage at the Cα also constitutes a possible route to ferulic acid and feruloyl methane formation, as explained in the next section.
On the other hand, the formation of the other catabolites cannot be explained by hydrolytic processes, even under basic conditions. In the absence of substitutable N- or S-atoms, alkynes, specific ketone-producing reagents, ester or amide groups, or resonant light, oxo-additions to a terminal methyl group [in case of vanillin and (2Z,5E)-2-hydroxy-6-(4-hydroxy-3-methoxyphenyl)-4-oxohexa-2,5-dienal] or to an alkene chain [in case of 6-hydroxy-1-(4-hydroxy-3-methoxyphenoxy)-3-(4-hydroxy-3-methoxyphenyl)-1,3,3a,6a-tetrahydro-4H-cyclopenta[c]furan-4-one] typically cannot occur in aqueous solutions (Larson and Weber, 1994) by other processes than radical-mediated reactions.
Chemical degradation and modification of curcumin.
On the basis of the previous paragraph, it is evident that ROS-mediated redox reactions not initiated by photochemical means also result in curcumin degradation and modification. The chemical degradation of curcumin in sodium phosphate/citrate buffer (pH = 8.0) coincides with the depletion of dissolved O2 at approximately equal rates, namely 4.3 µM∙min−1 and 4.5 µM∙min−1, respectively (Griesser et al., 2011). One of the recently identified catabolites of curcumin, 6-hydroxy-1-(4-hydroxy-3-methoxyphenoxy)-3-(4-hydroxy-3-methoxyphenyl)-1,3,3a,6a-tetrahydro-4H-cyclopenta[c]furan-4-one (Fig. 7B) (Griesser et al., 2011; Gordon and Schneider, 2012), contains 8 O-atoms instead of the 6 O-atoms of the parent compound, altogether corroborating the role of O2 and the potential involvement of ROS in the reactions that lead to the chemical degradation and modification of curcumin.
With the exception of feruloyl methane and ferulic acid, the cyclized, dimerized, and monomethoxyphenylic catabolites of curcumin have one distinct feature in common, namely that they are formed from a curcumin (phenoxyl, alkoxy, or C-centered) radical (Masuda et al., 1999, 2002; Fujisawa et al., 2004; Agnihotri and Mishra, 2011; Griesser et al., 2011). Curcumin "radicalization" as a precursory condition for subsequent degradation/modification is unequivocal and uncontested, but the mechanism(s) underlying the formation of a curcumin radical in the absence of ROS-generating compounds (e.g., enzymes such as xanthine oxidase), noncurcumin radicals (e.g., DPPH), and self-sensitized type I or type II photochemical reactions is partly elusive.
Both O2•– and ∙OH have been implicated as the ROS generated by curcumin in physiological buffers. In potassium phosphate buffer (pH = 8.0), 50 μM curcumin was shown to generate O2•– at a rate of 0.1 μM∙min−1 over the course of 20 minutes, measured spectrophotometrically by the O2•–-mediated reduction of cytochrome c (Griesser et al., 2011). The kinetics of curcumin degradation and O2 depletion were superimposable, and the reduction of cytochrome c was inhibited by SOD, confirming the production of specifically O2•–. The authors proposed that a phenoxyl radical is formed by oxidation of the phenylate anion by O2, whereby O2 is reduced to O2•–, forming a curcumin phenylate anion radical in the process. This mechanism most probably applies to curcumin, given the fact that at pH = 8 the relative abundance of the phenylate anion is ∼25% (pKa2 = 8.5, Fig. 1C) and that the formation of a phenoxyl radical in eugenol (half-curcumin minus the Cα and carbonyl) starts at pH = 9 (Fujisawa et al., 2002) (i.e., close to pKa2 of curcumin) in a pH-dependent pattern that is very similar to the pattern of pH-dependent degradation of curcumin (Fig. 6B). Nevertheless, this mechanism does not account for all aspects of the degradation kinetics. In the absence of additional reducing agents, redox reactions between redox couples (curcumin and ROS) proceed at stoichiometric parity (Schafer and Buettner, 2001). It is therefore peculiar that, at a ∼15 µM (∼25%) concentration of the phenylate anion, 45 µM (75%) of curcumin was degraded, especially when only 1 μM O2•– was produced during these 10-minute lasting reactions (Griesser et al., 2011). Hydrolysis as the predominant degradation mechanism is unlikely here because the amount of ferulic acid produced was marginal relative to the internally cyclized catabolite, 6-hydroxy-1-(4-hydroxy-3-methoxyphenoxy)-3-(4-hydroxy-3-methoxyphenyl)-1,3,3a,6a-tetrahydro-4H-cyclopenta[c]furan-4-one, a well-characterized curcumin radical end product (Griesser et al., 2011; Gordon and Schneider, 2012).
The arguments presented in the previous paragraph provide compelling evidence for the possibility that other mechanisms/ROS govern the chemical degradation of curcumin. Several studies have been published on the catalytic properties of curcumin in the Haber-Weiss/Fenton reactions. The Haber-Weiss reaction starts with the reduction of ferric iron to ferrous iron by O2•– (Fe3+ + O2•– → Fe2+ + O2) followed by the oxidation of Fe2+ by H2O2 to yield ∙OH (Fe2+ + H2O2 → Fe3+ + ∙OH + OH–), also referred to as the Fenton reaction. When the Haber-Weiss reaction was carried out with hypoxanthine/xanthine oxidase as a generator of O2•–, FeCl3 as a source of Fe3+, and hyaluronic acid as a substrate for ∙OH, the oxidative degradation of hyaluronic acid (as evidenced by a decrease in viscosity) was exacerbated by the addition of curcumin (Tønnesen, 1989a). In similar experiments, the reduction in viscosity of curcumin-lacking hyaluronic acid mixtures was 68% in the presence of Fe3+ and 7% in the absence of Fe3+ after 20 minutes (Tønnesen, 1989b), which showed that hyaluronic acid was in fact degraded as a result of the iron-catalyzed Haber-Weiss reaction. A 66% decrease in viscosity was found when curcumin was added to mixtures not containing Fe3+ (Tønnesen, 1989b), indicating that curcumin is capable of generating ∙OH even when the first step of the Haber-Weiss reaction is circumvented. It should be noted that no H2O2 was added to these reaction mixtures, as the authors relied on the production of H2O2 from the spontaneous disproportionation of O2•– (Marklund, 1976; Tønnesen, 1989b). The inhibition of hyaluronic acid degradation by mannitol, a specific ∙OH scavenger (Shen et al., 1997), substantiates the formation of H2O2 by spontaneous disproportionation of O2•– in these experiments.
Subsequent studies, which focused on the Fenton component only, revealed that curcumin in aqueous solution exhibits a Jekyll and Hyde-type of ROS personality that is comparable to ascorbic acid (Miller and Aust, 1989), namely one with both antioxidative and pro-oxidative properties (Kunchandy and Rao, 1990). At higher concentrations, curcumin behaves as an antioxidant, whereas at lower concentrations it acts as a pro-oxidant. The turning point was reported to lie between a curcumin concentration of 0.61 and 2.70 µM and depends on the type of substrate used to measure ∙OH production/scavenging (Kunchandy and Rao, 1990). It was further demonstrated that curcumin, just as ascorbic acid (Miller and Aust, 1989), is capable of reducing transition metals such as Fe3+ in a concentration-dependent manner (Kunchandy and Rao, 1989; Tønnesen and Greenhill, 1992). In fact, curcumin was more effective in reducing Fe3+ than ascorbic acid at equimolar concentrations (Reddy and Lokesh, 1994). However, the reduction of Fe3+ (100 μM) by curcumin was not directly proportional to the consequent ∙OH-mediated degradation of deoxyribose (used as substrate for ∙OH in the Fenton reaction) in the presence of H2O2, which was directly proportional in case of ascorbic acid (Reddy and Lokesh, 1994). The extent of deoxyribose degradation occurred in the order 10 > 20 > 5 = 50 = 100 > > 0 μM curcumin (Reddy and Lokesh, 1994), indicating that beyond a threshold ∙OH:curcumin ratio the Fenton reaction becomes debilitated, possibly due to curcumin oxidative degradation/modification and corollary inability to chelate and reduce transition metals. The reduction of Fe3+ occurs at the keto-enol moiety, i.e., curcumin’s metal chelating center (section II.A.4), inasmuch as the redox reaction is unaffected in curcumin derivatives lacking both phenylic hydroxyl and methoxy groups (Tønnesen and Greenhill, 1992). This, however, does not preclude the possibility that SET reactions to transition metals do not occur from the phenylic hydroxyl group, as has been demonstrated for ferulic acid (Mellican et al., 2003). In both instances, reduction of Fe3+ concurs with the generation of O-centered curcumin radicals.
Based on the above, there are several possible pathways that lead to radical-mediated degradation and modification of curcumin in aqueous solution, which have been summarized in Fig. 8. Essentially, these pathways can be categorized into metal-free oxidation reactions and transition metal-catalyzed reactions, both yielding a curcumin radical. The metal-free oxidation reactions require curcumin phenylate anions for SET to molecular oxygen (Nakanishi et al., 2003; Lee et al., 2006; Jung and Ham, 2007) and are therefore both pH and O2 dependent. At pH = pKa2 = 8.5, ∼50% of the curcumin molecules comprise the phenylate anion tautomer, which translates to ∼5% at pH = 7.5 and ∼0.5% at pH = 6.5. The fact that curcumin degradation rapidly accelerates from pH = 6.5 is hence in compliance with the increasing presence of the phenylate anion and the metal-free oxidation of this curcumin tautomer by codissolved O2 in neutral-to-alkaline buffers. The O2•– that is formed in the process can in turn oxidize phenylic residues by HAT ("Mechanisms of photo-induced reactive oxygen species production by curcumin and photochemical degradation/modification of dissolved curcumin," section II.C.1.a.ii) or spontaneously disproportionate into H2O2, which is also capable of oxidizing the phenylic hydroxyl group ("Mechanisms of photo-induced reactive oxygen species production by curcumin and photochemical degradation/modification of dissolved curcumin," section II.C.1.a.ii) and possibly form ∙OH in the absence of redox-active metals (Blanksby et al., 2007). The resulting curcumin phenylate anion radicals can subsequently undergo radical delocalization, giving rise to radical-radical coupling-, cyclization-, and C−C bond cleavage reactions (Fig. 8). The latter reactions further deplete dissolved O2 during the formation of 1,2-dioxetane adducts as reaction intermediates.
The metal-free oxidation reactions most likely proceed concomitantly with transition metal-catalyzed reactions, because redox-active metals such iron and copper are often present in trace amounts in aqueous buffers (Powell and Wapnir, 1994), including phosphate-buffered saline (Huang et al., 2004), cell culture medium (Huang et al., 2004), and possibly in (purified) curcumin extracts. There are two main transition metal-facilitated pathways that result in curcumin degradation or modification, of which the most ubiquitous pathway entails a redox reaction between curcumin and the metal cation, producing a carbonyl O-centered curcumin radical (Fig. 8) that is resonance stabilized by the diketo moiety. The Cα radicals may undergo internal cyclization (Fig. 7B) or radical-radical coupling with other curcumin radicals (Fig. 7C) or peroxidation by O2 in aerated solvents. The latter progresses to either C−C bond cleavage or ketonization under alkaline conditions to yield O2•–, which in turn can react with curcumin as described in the previous paragraph, oxidize ferric iron to its ferrous state, or disproportionate into H2O2. The second pathway requires the presence of (curcumin-reduced) Fe2+, whereby H2O2 facilitates the generation of the most reactive ROS, ∙OH, via the Fenton reaction. ∙OH subsequently abstracts a proton from any curcumin atom to produce a curcumin radical from which modified catabolites and degradation products are formed (Fig. 8). The Fenton reaction causes oxidation of ferrous iron into its ferric state, which can be reduced again by curcumin to propagate the redox cycle (Cur + Fe3+ → Cur• + Fe2+ → Fe2+ + H2O2 → Fe3+ + Cur → repetition). Evidently, the Fenton reaction is highly exacerbated at neutral and basic pH, where O2•– and its disproportionated derivative H2O2 are formed more abundantly due to the tautomerization of curcumin into a phenylate anion.
c. Uncharacterized curcumin degradation products (in vitro).
As addressed in section II.C.1, curcumin undergoes degradation/modification upon exposure to (alkaline) aqueous solutions, light, and oxygen (particularly when dissolved), yielding numerous catabolites (Figs. 6–8), of which some have not been characterized. The in vitro studies in which the curcumin degradation products were examined mainly employed thin layer chromatography (Tønnesen and Karlsen, 1985a; Tønnesen et al., 1986; Tønnesen and Greenhill, 1992; Masuda et al., 1999, 2001, 2002; Ansari et al., 2005) or HPLC (Tønnesen and Karlsen, 1985a,b; Tønnesen et al., 1986; Khurana and Ho, 1988; Tønnesen and Greenhill, 1992; Wang et al., 1997; Masuda et al., 1999, 2001, 2002; Griesser et al., 2011; Gordon and Schneider, 2012) to isolate the curcumin catabolites and subsequently MS (Tønnesen and Karlsen, 1985a; Tønnesen et al., 1986; Masuda et al., 1999, 2001, 2002; Griesser et al., 2011; Gordon and Schneider, 2012), gas chromatography (Tønnesen et al., 1986), gas chromotography-MS (Wang et al., 1997), or NMR (Tønnesen et al., 1986; Masuda et al., 1999, 2001, 2002; Griesser et al., 2011) to characterize the catabolites. All HPLC isolations were based on reversed phase chromatography (Tønnesen et al., 1986; Khurana and Ho, 1988; Tønnesen and Greenhill, 1992; Wang et al., 1997; Masuda et al., 1999, 2001, 2002; Griesser et al., 2011; Gordon and Schneider, 2012) with detection of the eluates by fluorescence spectroscopy (λex = 420 nm, λem = 470 nm) (Tønnesen and Karlsen, 1985a,b) or absorption spectroscopy at predominantly single wavelengths in the UV (Khurana and Ho, 1988), such as 240 nm (Masuda et al., 2002), 280 nm (Tønnesen and Karlsen, 1985a; Tønnesen et al., 1986), 300 nm (Masuda et al., 1999, 2002), 350 nm (Tønnesen et al., 1986; Tønnesen and Greenhill, 1992), and 360 nm (Masuda et al., 1999), as well as the visible spectrum, including 420 nm (Tønnesen et al., 1986; Masuda et al., 1999, 2001) and 430 nm (Ansari et al., 2005). Variable wavelength UV/visible spectrum or diode array spectrometers were employed only in a limited number of studies to detect the eluted compounds (Wang et al., 1997; Griesser et al., 2011; Gordon and Schneider, 2012).
In the majority of these curcumin modification/degradation studies, the chromatograms revealed peaks of which the compounds were not further characterized. For example, in the study by Khurana and Ho (1988), three compounds with a relative abundance of 33% were isolated but not characterized. These compounds were not vanillin, ferulic aldehyde, ferulic acid, or vanillic acid. Similarly, the chromatograms in Tønnesen et al. (1986) display several peaks after photochemical degradation of curcumin that were not further characterized, as do chromatograms acquired after curcumin degradation in buffered solutions (Tønnesen and Karlsen, 1985a; Wang et al., 1997; Griesser et al., 2011; Gordon and Schneider, 2012) and after oxidative modification (Masuda et al., 1999, 2002). These observations demonstrate that the list of curcumin catabolites presented in the literature and summarized in section II.C.1 is incomplete.
In addition, the isolation, and with it the subsequent characterization of curcumin modification/degradation products by liquid chromatography, depends chiefly on the stationary phase to capture the catabolites, the mobile phase with which the catabolites are eluted, and the spectroscopic detection system used to identify and collect the eluted catabolites. For any given combination of these chromatographic parameters, a specific set of molecules with specific chemical properties can be isolated and characterized. Consequently, more hydrophilic catabolites [e.g., those that contain nucleophilic groups due to oxidation/(di)oxo-addition] may not be captured by the reversed phase column material or may coelute early with other catabolites of similarly low log P value, as a result of which they will not be adequately separated. Such effects are nicely exemplified in Ansari et al. (2005), where curcumin that was degraded photochemically, oxidatively (by H2O2), or under acidic and basic conditions yielded peaks on the silica gel thin layer chromatography with a fractionation range that was lower and higher than that of curcumin. These peak positions indicate that, in light of a chloroform:methanol (92.5:7.5) mobile phase, degradation products were formed that were respectively more hydrophobic and more hydrophilic than curcumin. Depending on the HPLC conditions, these catabolites may fall outside of the properly separated range of eluents and hence go undetected.
The use of a single wavelength detection system further restricts the spectrum of compounds that can be detected, because curcumin has an absorption maximum at >400 nm (Tønnesen et al., 1995), whereas vanillin, vanillic acid, ferulic acid, feruloyl methane absorb at 327–340 nm (Tønnesen et al., 1995; Gordon and Schneider, 2012), the methoxyphenyl as well as tetrahydrocurcumin absorb at ∼276 nm (Tønnesen et al., 1995), and the methine bridge containing only the β-dicarbonyl moiety absorbs at 272 nm (Tønnesen et al., 1995). On top of that, photochemically modified catabolites such as 7-hydroxy-1-[(2E)-3-(4-hydroxy-3-methoxyphenyl)prop-2-enoyl]-6-methoxynaphthalen-2(1H)-one (Fig. 6A) (Tønnesen et al., 1986), chemically modified catabolites such as (2Z,5E)-2-hydroxy-6-(4-hydroxy-3-methoxyphenyl)-4-oxohexa-2,5-dienal (Fig. 6B) (Wang et al., 1997) and 6-hydroxy-1-(4-hydroxy-3-methoxyphenoxy)-3-(4-hydroxy-3-methoxyphenyl)-1,3,3a,6a-tetrahydro-4H-cyclopenta[c]furan-4-one (Fig. 7B) (Griesser et al., 2011; Gordon and Schneider, 2012), and the dimerized derivatives (Fig. 7C) (Fujisawa et al., 2004; Masuda et al., 1999, 2002) exhibit different absorption maxima than curcumin and its monophenylic degradation products (Gordon and Schneider, 2012), the position of which depends on whether the conjugated system has been perturbed. These catabolites may therefore go undetected when a single wavelength detection system is used during HPLC. The chemically modified catabolites may also exhibit different chromatographic properties, altogether adding to the fact that not all curcumin modification/degradation products have been characterized.
2. Implications of the (Photo)Chemical Instability of Curcumin.
a. Implications on pharmaceutical preparations of curcumin.
The (photo)chemical instability (Fig. 5) makes curcumin a rather "onerous" pharmaceutical and brings forth several critical implications for the posology and formulation process. The first implication is that oral curcumin formulations should be administered at relatively high doses to compensate for the inevitable loss of pharmacological potency due to (photo)chemical degradation (section II.C.1) in aqueous environments. Although curcumin is initially dissolved in solvents in which it is stable (e.g., methanol, ethanol, DMSO), it will ultimately come in contact with an aqueous environment (buffer, culture medium, blood, cytosol) that is, in most instances, equilibrated/sustained at 37°C and pH ≈ 7.4 as well as replete with enzymes that mediate its modification/degradation. Additionally, poor uptake from the gut (section III.A), uptake by nontarget tissue (section III.A), and enzyme-mediated biotransformation in healthy and cancer cells (section III.C) further reduce curcumin bioavailability after oral intake. Although these issues can to an extent be counteracted by administering higher dosages, such regimens are not always well-tolerated by patients. Clinical trials have reported patient discontent and withdrawal of patients from the high dosage cohorts solely due to the discomfort they experienced when ingesting a large amount of curcumin pills (Cheng et al., 2001; Irving et al., 2013).
A second implication is that, in case of oral delivery, curcumin formulations should preferably contain crystalline curcumin or curcumin suspended in a stabilizing, physiologically compatible solvent (e.g., oils) or emulsifier. Pills prepared in an oxygen-free environment are most ideal because such formulations are minimally susceptible to oxidative degradation/modification and can be sufficiently protected from light, although it has been shown that pharmaceutical curcumin preparations in tablet form may contain numerous catabolites (Ansari et al., 2005). Nevertheless, dissolution of crystalline or otherwise chemically stable curcumin formulations in the stomach and transport through the gastrointestinal tract is not expected to have detrimental effects on curcumin stability due to the dark and acidic (Fig. 6B) environment.
Lastly, formulations designed for oral administration are associated with poor systemic bioavailability as a result of in vitro and in vivo degradation/modification and unfavorable pharmacokinetics (section III.A). Curcumin formulated in pills is therefore highly suitable for the treatment of gastrointestinal malignancies (Sharma et al., 2001; Garcea et al., 2005; Villegas et al., 2008) and for long-term chemopreventive therapy (Anto et al., 1996a; Perkins et al., 2002; Lao et al., 2006; Villegas et al., 2008), particularly in the gastrointestinal tract (Huang et al., 1994; Rao et al., 1995), but ill-suited for the treatment of cancers outside of the gastrointestinal system. For these reasons, curcumin for oral chemotherapy of nongastrointestinal cancers should at least be coadministered with uptake-enhancing agents such as piperine (Shoba et al., 1998; Singh et al., 2013) and turmerones (Yue et al., 2012) or in nanoparticle-encapsulated form (Shaikh et al., 2009) when systemic delivery by a (targeted) drug delivery system is not possible (section II.C.2.a.i). For the treatment of non-gastrointestinal cancers, curcumin is best administered via the intravenous route (Bisht and Maitra, 2009) to achieve maximal systemic concentrations and delivery to the tumor. Inasmuch as curcumin is poorly soluble in water (section II.B), either a delivery system is required to solubilize and stabilize curcumin in plasma or curcumin must be chemically modified, e.g., with more hydrophilic functional groups.
i. Curcumin Drug Delivery Systems.
Numerous nanoparticulate drug delivery systems have been developed for curcumin, including lipid-based (liposomes, micelles, solid lipid particles, nanodisks) and polymer-based vehicles, nano-emulsions/gels/suspensions, and molecular complexes such as curcumin-plasma proteins, -amino acid, and -water-soluble polymers (Ghosh et al., 2011; Sun et al., 2012a). In addition to the equal or improved pharmacodynamic efficacy compared with unencapsulated curcumin (Li et al., 2005b; Ma et al., 2008; Thangapazham et al., 2008a; Song et al., 2011), potential benefits of curcumin encapsulation include protection from degradation/modification in buffered solutions, cell culture medium, and blood and plasma, as has been described for lipid-based nanocarrier systems (Barry et al., 2009; Chen et al., 2009a; Karewicz et al., 2011), and reduced uptake by organs (Song et al., 2011), altogether leading to improved bioavailability (Maiti et al., 2007; Marczylo et al., 2007; Gupta and Dixit, 2011). However, the level of protection is strongly dependent on the type and composition of the drug vehicle, particularly in the case of liposomes (Chen et al., 2009a), and there seems to be no protection from photochemical degradation. Whereas curcumin encapsulation in surfactant (SDS and Triton X-100) micelles relayed no protective effect on the photochemical degradation rate compared with free curcumin (Chignell et al., 1994), complexation of curcumin with different cyclodextrins even accelerated its photochemical decomposition (Tønnesen and Karlsen, 1985a,b; Tomren et al., 2007).
As the in vitro and in vivo chemical stability of curcumin generally improves upon encapsulation, there are several steps during the nanocarrier preparation process where curcumin may be susceptible to (photo)chemical degradation. This is best exemplified for liposomal formulations, the most commonly used and typically nontoxic (Wang et al., 2008a) nanocarriers for curcumin, although the principles will mostly apply to other drug delivery systems as well. First, photochemical degradation of curcumin can occur in crystalline and solubilized forms of curcumin (section II.C.1.a), as a result of which the curcumin formulations should always be kept out of direct light at any stage of the formulation process, i.e., by using amber glassware or foil-covered containers. Where possible, (biocompatible) 1O2 quenchers (e.g., β-carotene) and transitional metal chelators (Welch et al., 2002) should be added to the formulation at nontoxic concentrations to deter curcumin (per)oxidation and Fenton reactions (Fig. 8). Second, liposomes can be prepared by at least six different methods (Lasch et al., 2003), all of which include a stage in which lipids are mixed with an aqueous phase to facilitate vesicle self-assembly. To deter chemical degradation of curcumin during this stage, the aqueous phase should be pH-adjusted (pH < 6.5) and iso-osmolar (0.292 osmol/kg) to prevent formation of the phenolate tautomer ("Chemical degradation and modification of curcumin," section II.C.1.b.i) and osmolar gradient-driven destabilization of the liposomes under physiological conditions (Drummond et al., 1999), respectively. If these prerequisites are not met, the carriers may ultimately comprise a mixture of curcumin and curcumin catabolites, the implications of which are addressed in section II.C.2.
ii. Synthetic Curcumin Analogues.
Similarly, a variety of curcumin analogues have been synthesized for the purpose of increasing water solubility, deterring degradation/modification (Ferrari et al., 2009; Liang et al., 2009; Wan et al., 2010; Abdel Aziz et al., 2012; Gagliardi et al., 2012; Kulkarni et al., 2012), and improving therapeutic and chemopreventive efficacy (Ishida et al., 2002; Ohtsu et al., 2002; Venkateswarlu et al., 2005; Tamvakopoulos et al., 2007a; Fuchs et al., 2009; Liang et al., 2009). A schematic overview of synthetic curcumin analogues that are both more stable in aqueous media at neutral pH and pharmacologically equally or more effective than curcumin in different cancer cell lines is provided in Fig. 9.
For the curcumin analogues, different preparatory prerequisites apply than for the curcumin nanocarriers, where the encapsulated curcumin is largely shielded from the elements that mediate its breakdown. The prerequisites are mainly structure-driven and comprise three main components. First, stability of synthetic curcumin analogues is achieved by eliminating the potential oxidation sites, i.e., the phenylic and enolic hydroxyl groups, which precludes the formation of radicals and consequent chemical degradation/modification as described in "Chemical degradation and modification of curcumin" in section II.C.1.b.i and Fig. 8. Corroboratively, substitution of the phenylic hydroxyl groups by methoxy groups (i.e., dimethoxycurcumin) conferred considerable stability in cultured cells for 48 hours and in mice for up to 4 hours (Tamvakopoulos et al., 2007a).
Second, attention should be given to the fact that, in cells and in vivo, curcumin undergoes enzymatic cleavage and modification by proteins it directly associates with (section III.C). To eliminate enzymatic cleavage/modification of curcumin at the β-diketo moiety, for instance, numerous monocarbonyl derivatives have been synthesized (e.g., Fig. 9) (Robinson et al., 2005; Fuchs et al., 2009; Liang et al., 2009). In vivo testing in rats evinced that the circulation half time of some of the monocarbonyl analogues was longer than that of curcumin after oral administration, suggesting an improved systemic stability (Liang et al., 2009). It should be noted, however, that these data were not corrected for other important pharmacokinetic parameters such as uptake, biotransformation, and excretion and may therefore not reflect chemical stability per se.
Third, inasmuch as intermolecular bonding between curcumin and target molecules may be perturbed by the elimination of the enolic and phenylic hydroxyl groups (section II.A), synthetic curcumin analogues should be substituted with other functional groups or atoms capable of electrostatic interactions to maximally retain the pleiotropic binding behavior of the parent compound. Accordingly, it has been proposed (Liang et al., 2009) that the benzene should be substituted with electron withdrawing groups (e.g., halogens, carbonyls) because the electronegativity of these groups correlates positively with cytotoxicity (Fig. 9), and/or a weak electron donating groups (e.g., alkyls, alcohols, amines) for maximum antitumor effects (Fuchs et al., 2009; Liang et al., 2009). Highly suitable electron withdrawing groups are the halogens (X = F, Cl, Br, see Fig. 9) because of their halogen bonding capacity (X∙∙∙:O) (Politzer et al., 2007) and because carbohalides are not readily oxidized by O2 or ROS [except for ∙OH (Merga et al., 1996)] under physiological conditions (Yang et al., 2004a) to yield radicals or enzymatically degraded in eukaryotic organisms [humans lack the necessary enzymes (Werlen et al., 1996)]. As shown in Fig. 9 and reported by (Liang et al., 2009), the halogenated curcumin analogues can be very toxic to cancer cells, although several halogenated derivates were inactive (Ishida et al., 2002). Proper weak electron-donating groups include methoxy and alkyl peroxy (R−O−O−CH3) groups, which were shown to be more cytotoxic than curcumin in various cancer cell lines (Ishida et al., 2002; Ohori et al., 2006; Mosley et al., 2007; Cen et al., 2009; Fuchs et al., 2009; Liang et al., 2009).
On a final note, whether the curcumin analogues are toxic to cells should at all times be investigated experimentally in addition to in silico and preferably in a multitude of cancer cell lines as well as in noncancer cells (control), as was done by, for instance, Fuchs et al. (2009), Liang et al. (2009), and Katsori et al. (2011). The extent of cytotoxicity is not always predictable (Robinson et al., 2005) and may vary in both healthy cells and cancer cell lines (Fuchs et al., 2009; Liang et al., 2009). Moreover, given that curcumin can inhibit cancer cells at the level of metastasis, proliferation, and viability, assays with curcumin analogues should be performed at all these levels to determine the full pharmacodynamic spectrum of the analogue. Some curcumin analogues can strongly inhibit cancer cell proliferation without imposing any lethality at that concentration [such as compound 1b in (Katsori et al., 2011)], rendering them unsuitable for cancer treatment. Similarly, structural isomers may exert entirely different pharmacodynamic effects, as for example evidenced by the curcumin analogue 1d (highly toxic at <10 µM) versus its isomer 1e (nontoxic at concentrations up to 100 µM) in Katsori et al. (2011).
III. Anti-Cancer Properties of Curcumin
A. Curcumin Pharmacokinetics and Pharmacodynamics
In section II.A the chemical basis was provided for the binding behavior of curcumin that, because of its distinct chemical properties, is capable of interacting with a plethora of molecules and thereby detrimentally affect numerous vital pathways in cancer cells (Johnson and Mukhtar, 2007; Anand et al., 2008; Gupta et al., 2011), causing their demise. In fact, curcumin is able to induce lethal effects in virtually every cancer type, at least in vitro. In support of this, a literature study was performed to determine the in vitro tumor killing capacity of curcumin per cancer type. The results, extrapolated from 137 published papers, have been summarized for the most common cancers in Fig. 10 and specified per cancer cell line for the reproductive system (Supplemental Fig. 1), digestive system (Supplemental Fig. 2), lymphatic and immune system (Supplemental Fig. 3), nervous system (Supplemental Fig. 4), as well as the pulmonary, urinary, and skeletal system and skin (Supplemental Fig. 5), whereby the in vitro 50% lethal or inhibitory curcumin concentrations (LC50 or IC50, respectively) were plotted as a function of incubation time and cancer cell line. The complete data set containing additional information is provided in Supplemental Table 1. The actual LC50 values are expected to be lower than reported, inasmuch as these values are hardly ever corrected for the considerable extent of curcumin degradation in aqueous media ["Solvolysis: (alkaline) hydrolysis," section II.C.1.b.i]. Readers should also note that the data were not stratified for the method with which the LC50 value was determined (typically an MTT or water-soluble tetrazolium-1 assay) or the experimental procedures and conditions (e.g., medium, solvent for curcumin, degree of cell monolayer confluence, culture plate configuration, culture conditions, etc.). Although this makes valid interstudy comparison of the data difficult, some general observations and implications are worth highlighting in regard to curcumin pharmacodynamics and pharmacokinetics.
First, the LC50 values range mostly from 1 to 100 µM, with a mean ± SD LC50 value of 21 ± 17 µM when all data are clustered (n = 309). The relatively high LC50 values are pharmacologically problematic in light of curcumin’s instability (section II.C), although the LC50 values may be skewed inasmuch as neither curcumin degradation nor the pharmacodynamics of the degradation products were accounted for (see also section III.C.5). Moreover, curcumin exhibits poor systemic uptake and profuse conjugation and modification (section III.C.2), altogether accounting for the submicromolar-to-low nanomolar curcumin levels found in blood after oral intake (Fig. 11 and Supplemental Table 2). Curcumin is ineffectively transported across the intestinal mucosa into the circulation (Holder et al., 1978; Wahlstrom and Blennow, 1978; Ravindranath and Chandrasekhara, 1980; Cheng et al., 2001; Sharma et al., 2001, 2004; Garcea et al., 2004, 2005; Yang et al., 2007a; Vareed et al., 2008; Villegas et al., 2008; Suresh and Srinivasan, 2010; Wahlang et al., 2011; Berginc et al., 2012). Furthermore, the curcumin molecules that bypass transport hurdles and escape biotransformation in the intestinal mucosa (Ireson et al., 2002; Hoehle et al., 2007; Wahlang et al., 2011; Berginc et al., 2012; Dempe et al., 2012), i.e., by definition the first pass effect for orally administered curcumin, and manage to reach the circulation instantaneously become susceptible to chemical modification in blood (Pan et al., 1999); uptake and biotransformation by the liver (second pass effect), kidneys, and other organs (Holder et al., 1978; Wahlstrom and Blennow, 1978; Pan et al., 1999; Asai and Miyazawa, 2000; Garcea et al., 2004; Hoehle et al., 2006; Tamvakopoulos et al., 2007b; Vareed et al., 2008; Marczylo et al., 2009), and excretion via the biliary or urinary system, albeit the latter occurs to a limited extent (Holder et al., 1978; Ravindranath and Chandrasekhara, 1980; Sharma et al., 2004; Marczylo et al., 2009; Suresh and Srinivasan, 2010). These phenomena are further elaborated in the context of biological curcumin metabolites in sections III.C.1 through III.C.3.
Accordingly, in human studies it has been shown that, at a single oral dose of 500 to 8000 mg, no curcumin was detected in serum 1, 2, and 4 hours after administration (Lao et al., 2006). In only one of the three subjects that received a 10,000-mg oral dose, serum curcumin levels reached a concentration of 66, 91, and 121 nM (baseline corrected) after 1, 2, and 4 hours, respectively. Similarly, one of the three subjects in the 12,000 mg dosage group exhibited serum curcumin levels of 81, 156, and 139 nM after 1, 2, and 4 hours, respectively. In two of the three subjects in both high-dosage groups, no curcumin was detected in serum (Lao et al., 2006). Other pharmacokinetic studies (Cheng et al., 2001; Sharma et al., 2004) in humans revealed that plasma curcumin concentrations peak at 1–2 hours after oral administration and reach a maximum of 1.77 ± 1.87 μM after an oral dose of 8000 mg (Cheng et al., 2001). Thus, plasma levels of curcumin after oral administration are so low relative to the reported LC50 values and are sustained for such a short period of time that they are not expected to be effective against non-gastrointestinal cancers, notwithstanding the tumoricidal efficacy of some of the curcumin degradation products and biological metabolites (sections III.B.2 and III.C.4).
For cancers outside the gastrointestinal system, the clinical data, albeit somewhat restricted because of small patient populations, generally reflect the low systemic bioavailability of orally administered curcumin (Fig. 11). In a phase I dose-escalation study (oral administration of curcumin at 1000–12,000 mg/day for 3 months), histologic improvements in neoplastic malignancies were found in 1 of 4 patients (25%) with uterine cervical intraepithelial neoplasia, in 1 in 2 (50%) patients with recently resected urinary bladder cancer, and in 2 of 6 patients (33%) with squamous cell carcinoma in situ (Cheng et al., 2001). Poorer outcomes were obtained in a phase II trial with pancreatic adenocarcinoma patients, in which only 2 of 21 pancreatic adenocarcinoma patients (10%) that were given a daily oral dose of curcumin (8000 mg) showed a response (Dhillon et al., 2008). In the first patient, curcumin treatment caused a gradual decrease in cancer antigen 125 (also known as carbohydrate antigen 125 or mucin 16—a cancer biomarker) levels over the course of 1 year, which was accompanied by stabilization of lesion size and a reduction in the PET-CT standardized uptake value. The second patient exhibited a 73% reduction in tumor size that lasted 1 month, after which the lesions that had regressed remained small, whereas new lesions had increased in size.
The marginal systemic bioavailability gives rise to a second important consideration, namely that the cancer types targeted by curcumin chemotherapy should be scaled according to their combined degree of curcumin susceptibility and accessibility so as to roughly gauge the potential therapeutic efficacy and, with it, utility. For example, cancers of the lymphatic and immune system (Supplemental Fig. 3) appear to be more susceptible to curcumin than for instance cancers of the reproductive system (Supplemental Fig. 1), as evidenced by the generally lower LC50 values (notwithstanding the fact that cell lines may phenotypically differ from native cancer cells and thus respond differently to curcumin). On top of the favorable degree of susceptibility, cancers of the lymphatic and immune system, and particularly the lymphoid leukemias and other hematologic neoplasms, are more accessible for curcumin, insofar as these cancer cells are either blood-borne or arise in highly vascularized tissue (bone marrow). Consequently, malignant B-cells, T-cells, and natural killer cells are more likely to come in direct contact with gut-derived, circulating curcumin before excessive uptake/biotransformation/chemical modification has occurred. This principle of high-susceptibility–high-accessibility may also apply to lymphomas and certain bone cancers, which also consistently exhibit relatively low LC50 values (Supplemental Figs. 3 and 5, respectively), given that lymph nodes (Osogoe and Courtice, 1968; Webster et al., 2006; Martinez-Corral et al., 2012) and some bone tumors (Verstraete et al., 1996; Wenger and Wold, 2000) are well perfused. By extension, the same principle holds for cancer cells undergoing metastasis, which occurs mainly via the circulation (van Zijl et al., 2011).
Even more suitable pharmacodynamic targets are malignancies of the gastrointestinal tract (Supplemental Fig. 2), where the major pharmacokinetic obstacles principally do not apply. The most prominent example is gastric cancer, a major contributor to cancer-related deaths worldwide (Jemal et al., 2011). Neoplasms typically form in the inner lining of the stomach (stage 0) and then progress to either the second and third layers (stage 1A) or the second layer and proximal lymph nodes (stage 1B). Consequently, orally ingested curcumin comes in direct contact with the venue of potential oncogenesis or malignantly transformed tissue at very high concentrations (because 100% of the orally administered curcumin is deposited directly into the stomach in intact form, rendering LC50 considerations obsolete) and under protective (dark, acidic) conditions. In case of gastric cancer, curcumin not only prevents malignant transformation of mucosal epithelial cells but also removes a cause of carcinogenesis. A persistent infection by Helicobacter pylori, the bacterial species that causes stomach ulcers, constitutes a significant risk factor for developing gastric cancer (Suganuma et al., 2012; Zabaleta, 2012) either due to chronic inflammation and/or the release of virulence factors (Hatakeyama and Higashi, 2005). Curcumin hence acts through its anti-inflammatory (Foryst-Ludwig et al., 2004; Sintara et al., 2010; Kundu et al., 2011) and bactericidal properties (Mahady et al., 2002; De et al., 2009) while favorably modulating numerous infection-induced, carcinogenesis-related pathways in mucosal epithelial cells (Foryst-Ludwig et al., 2004; Sintara et al., 2010; Kundu et al., 2011) and deterring procarcinogenic signaling in H. pylori (Zaidi et al., 2009). It should be noted that results obtained in vitro and in animal models are not necessarily translatable to the clinical setting with respect to H. pylori infection and gastric cancer. For instance, only 1 of 17 chronic gastritis patients (6%) with confirmed H. pylori infection who received a turmeric tablet (700 mg of curcumin and curcumin derivatives) three times a day for 4 weeks experienced complete eradication of the infection (Koosirirat et al., 2010). Similarly, a 7-day combinatorial treatment with orally administered curcumin (30 mg), lactoferrin (100 mg), N-acetylcysteine (600 mg), and pantoprazole (20 mg) was able to cure the H. pylori infection in only 3 of 25 patients (12%), although a significant decrease in the severity of dyspeptic symptoms was observed (Di Mario et al., 2007). A phase I study in patients with intestinal metaplasia of the gastric mucosa, a condition mainly instigated by a chronic H. pylori infection, found that only 1 of 6 patients (17%) taking curcumin at an oral dose of 1000 mg per day for 3 months exhibited histologic improvement in the lesion (Cheng et al., 2001).
Another example where curcumin’s pharmacokinetic hurdles play a minor role is colorectal cancer. Although colon cancer cell lines exhibit some of the highest LC50 values of all studied cancer cell lines (Supplemental Fig. 2), there is limited concern about the therapeutic efficacy of curcumin. A study in rats found that approximately 40% of per gavage administered curcumin remains in the gastrointestinal tract in chemically intact form over a period of 5 days and that the highest concentration of curcumin shifts along the gastrointestinal tract in time, i.e., stomach (15 minutes, 53 ± 5% of administered curcumin) → small intestine (30 minutes, 59 ± 11%) → cecum (3–7 hours, 43–51%) → large intestine (24 hours, 5 ± 3%) (Ravindranath and Chandrasekhara, 1980). These data suggest that both the stomach and the large intestine absorb a large fraction of the curcumin, enabling anticancer protection in cells of the mucosal lining of both gastrointestinal organs. In line with these findings, mice fed a curcumin-enriched diet exhibited curcumin levels that ranged from 39 ± 9 to 240 ± 69 nmol/g mucosal tissue in the small intestine (0.1 and 0.5% dietary curcumin, respectively) and from 15 ± 9 to 715 ± 448 nmol/g mucosal tissue in the colon (0.1 and 0.5% dietary curcumin, respectively) (Perkins et al., 2002). The therapeutic efficacy of the higher concentrations of orally administered curcumin was corroborated in this animal model of intestinal cancer (Perkins et al., 2002) but also in clinical trials (Sharma et al., 2001; Cruz-Correa et al., 2006; Carroll et al., 2011).
Clinical studies on the therapeutic efficacy of curcumin in relation to colorectal malignancies and premalignancies are most abundant and show the greatest promise. A recent phase IIa trial that included patients with colorectal neoplasias revealed that a 1-month-long, 4000 mg daily oral dose of curcumin reduced the number of aberrant crypt foci in 17 of 20 patients (85%) (Carroll et al., 2011). Another clinical study on familial adenomatous polyposis demonstrated that oral administration of 480 mg of curcumin three times a day for an average of 6 months led to a 60.4 and 50.9% decrease in the number and size of the polyps, respectively, in all patients (n = 5) when coadministered with 20 mg of quercetin (Cruz-Correa et al., 2006). Even in patients with advanced colorectal cancer refractory to standard chemotherapy, oral curcumin administration (ranging from 36 to 144 mg per day for 3–4 months) resulted in stabilization of the disease in 5 of 15 patients (33%) and yielded a 44% reduction in the plasma tumor marker carcinoembryonic antigen in one patient (Sharma et al., 2001). These effects have been attributed to curcumin-induced acceleration of apoptosis in the patients’ colorectal cancer cells, characterized by elevated DNA fragmentation, increased p53 (involved in cancer cell proliferation) and Bax (pro-apoptotic) expression, and reduced Bcl-2 (anti-apoptotic) levels (He et al., 2011).
Regardless of curcumin’s general anticancer effects, several peculiar pharmacodynamic and pharmacokinetic features of curcumin have surfaced in the clinical trial reports. First, a number of trials that entailed dose escalation monitoring reported positive biological effects for some of the lower curcumin concentrations but no effect for the highest concentration in the same patient cohort (Cheng et al., 2001; Sharma et al., 2001; Dhillon et al., 2008). For example, in the previously mentioned phase II trial with advanced pancreatic cancer patients, the peak plasma level of curcumin was 7 nM at 6 hours postadministration in the first patient that had remained stable for 1 year but almost sixfold higher (40 nM) in the second patient that had exhibited the tumor regression (Dhillon et al., 2008). Second, curcumin concentrations in tissue (Carroll et al., 2011) and fecal (Sharma et al., 2001) samples were occasionally higher for the lower oral curcumin dosages than for the highest oral dose given in the same patient cohort. Although plasma levels were found to be proportional to the administered dose in one clinical study (Cheng et al., 2001), a striking incongruence between oral dose and plasma concentrations was observed in a rat study (Ravindranath and Chandrasekhara, 1982), where the 10-mg and 400-mg dosages of radioactively labeled curcumin yielded similar blood concentrations at various time points after administration, both of which were higher than the blood concentrations after an 80-mg dose. Third, blood curcumin concentrations exhibit a notable variance in patients subjected to the same dosing regimen (Supplemental Table 2), indicating lack of dose linearity. For instance, vastly distinct mean blood concentrations were found in patients 2 hours after oral administration of curcumin at 2000 mg, namely 0 nM (Shoba et al., 1998), 7 nM (Anand et al., 2007), and 402 nM (Antony et al., 2008), and in patients who had been given 8000 mg, namely 60 nM (Dhillon et al., 2008) versus 4615 nM (Cheng et al., 2001). Despite the possibility that this may be attributable to the dissimilar curcumin extraction and quantification processes, the differences also arose when the same analytical procedures were implemented. Kanai et al. (2011) observed an almost 10-fold difference in blood curcumin levels (149 versus 1118 nM) in 2 patients 4 hours after an oral dose of 8000 mg. The exact reasons for the lacking dose linearity notwithstanding, it is incontrovertible that the pharmacodynamics and pharmacokinetics of orally administered curcumin are very complex and apparently difficult to control, which is underscored by the large standard deviations associated with the corrected blood curcumin concentrations in both the human and rat studies (Fig. 11, bottom panels, respectively). Moreover, the overall clinical evidence on the in vivo anticancer properties of curcumin is currently too limited to draw any definitive conclusions as to its therapeutic efficacy.
B. Pharmacokinetics and Pharmacodynamics of Curcumin Catabolites Generated In Vitro
Given the ease, propensity, and extent of curcumin metabolism in vitro and in vivo, the question arises whether curcumin alone exerts the pharmacological effects or whether it does so in conjunction with its metabolites, as has been propagated by some (Shen and Ji, 2009, 2012). Ample evidence points to the latter, namely that some of the curcumin metabolites contribute to the documented oncostatic effects of the parent compound, and may even do so to a sizeable degree. This is not very surprising given that some of the curcumin catabolites generated in vitro retain the methoxyphenyl residue (Figs. 6 and 7) that facilitates intermolecular interactions with target molecules through the H-bond donating and accepting properties (sections II.A.2 and II.A.3), Michael acceptor sites (section II.A.7), and strong antioxidant capacity (section II.C.1). The same applies to curcumin metabolites produced in cells and living organisms, which is addressed in section III.C. Naturally, the full extent of the oncostatic effects at the account of the curcumin metabolites cannot be accurately gauged insofar as not all metabolites have been characterized (section II.C.1.b) or evaluated in pharmacodynamic context. But, despite the previously stated "onerousness" of curcumin as a pharmaceutical, the consequences of its instability and biotransformation seem to be considerably less detrimental for oncostatic pharmacodynamics than for many other pharmacological compounds, which lose biological activity after degradation/modification.
With respect to the characterized curcumin catabolites in nonbiological samples and in in vitro test systems, vanillin and ferulic acid, which are naturally occurring phenolics themselves, have most extensively been studied for their pharmacokinetic behavior and antimutagenic, anticarcinogenic, tumoricidal, and antimetastatic properties.
1. Pharmacokinetics of Curcumin Catabolites (Generated In Vitro).
In vitro curcumin catabolism generally involves cleavage of the molecule in the methine bridge to a stable, substituted phenyl containing a terminal ketone (vanillin and (2Z,5E)-2-hydroxy-6-(4-hydroxy-3-methoxyphenyl)-4-oxohexa-2,5-dienal), methyl ketone (feruloyl methane), or acid (vanillic and ferulic acid) (Fig. 6). These catabolites, and particularly the acids that are deprotonated at physiological pH (Brown et al., 1955), have different physicochemical properties than curcumin (e.g., lower log P, covalent/H-bonding sites) and are therefore expected to exhibit slightly different pharmacokinetics than the parent compound. The extent to which the catabolites form in oral formulations, especially those for human use, is marginal relative to the degree to which the curcumin metabolites form in vivo (section III.C), making them biologically less relevant in regard to pharmacokinetics. Accordingly, the pharmacokinetics of the in vitro catabolites will only be briefly addressed, mainly in regard to vanillin and ferulic acid. However, it should be noted that the formation of the (photo)chemical degradation products is highly relevant in the context of in vitro research, where curcumin is added to the culture medium or physiological buffer (often at 37°C and neutral pH) and thus becomes susceptible to degradation with corollary effects on pharmacodynamics (sections III.B.2).
a. Pharmacokinetics of vanillin.
Studies in rats have shown that orally administered vanillin (100 mg/kg) is absorbed in the gastrointestinal tract and reaches a ∼3 μM peak plasma concentration 4 hours after administration (Beaudry et al., 2010), suggesting poor uptake, rapid clearance, and/or excessive metabolism as has been described for curcumin (section III.A and Fig. 11). Intravenously administered vanillin is almost completely cleared or metabolized within 2 hours (Beaudry et al., 2010). Correspondingly, analysis of urine retrieved from rats 24 hours after an oral dose of vanillin (100 mg/kg) revealed that vanillin is extensively cleared via the renal system in conjugated form, mainly as vanillin glucuronide and sulfate (Kirwin and Galvin, 1993). After 48 hours, 94% of the administered vanillin had accumulated in urine as vanillin (7%), vanillic acid (47%), vanillyl alcohol (19%), vanilloylglycine (19%), catechol (8%), 4-methylcatechol (2%), guaiacol (0.5%), and 4-methylguaiacol (0.6%) (Kirwin and Galvin, 1993), indicating that vanillin does not only undergo phase II xenobiotic metabolism but also extensive chemical modification, although it is not clear whether this occurs in the kidneys or before renal uptake. It is certain that the liver is able to metabolize vanillin, as experiments with liver slices demonstrated that vanillin is rapidly oxidized to vanillic acid by aldehyde oxidase and subsequently O-demethylated to protocatechuic acid and that small amounts of vanillyl alcohol are produced as well by a different reaction mechanism (Panoutsopoulos and Beedham, 2005).
The vanillin metabolites may exhibit different pharmacokinetics and pharmacodynamics than the parent compound, which may have consequences for cancer cells in terms of cytostatic effects. For instance, vanillic acid and protocatechuic acid are inhibitors of phenolsulfotransferases (Yeh and Yen, 2003), which are responsible for sulfating xenobiotics during phase II metabolism (Brix et al., 1998; Coughtrie et al., 1998), including curcumin (section III.C.2.b), vanillin, and ferulic acid. Their formation may therefore benefit the disposition of curcumin and some of its metabolites and improve the cytostatic potency. With respect to pharmacodynamics, vanillic acid has been shown to exhibit a range of biological effects in melanocytes on transcription factors, receptors, and enzymes that were absent for vanillin (Chou et al., 2010).
b. Pharmacokinetics of ferulic acid.
Ferulic acid is very effectively absorbed in both the stomach (Zhao et al., 2003a, 2004) and the intestines (Spencer et al., 1999), as a result of which ferulic acid exhibits high bioavailability after ingestion (Adam et al., 2002; Rondini et al., 2002; Zhao et al., 2003b). Studies in rats showed that, when stomachs containing intragastrically deposited ferulic acid were incubated ex vivo versus in situ for 25 minutes, respectively 80 ± 6% and 26 ± 11% of the ferulic acid was retrieved from the gastric content and 7 ± 1% and 4 ± 2% from the gastric mucosa (Zhao et al., 2004), indicating that more than one-half of the ferulic acid enters the circulation by means of gastric absorption. Similar experiments in rats revealed that 56 ± 2% of the ferulic acid perfused through the small intestine in situ is absorbed (Adam et al., 2002). The finding that no ferulic acid was detected in the ileum and cecum of rats that had been fed dietary ferulic acid for 9 days suggests that virtually all the ferulic acid is absorbed in the stomach and the duodenal and jejunal segments (Zhao et al., 2003a). Intestinal perfusion experiments with ferulic acid in bicarbonate buffer confirmed uptake of ferulic acid by the jejunum and minimal uptake by the ileum (∼10% of the jejunal uptake), with the maximum absorption rate occurring between 60 and 70 minutes of perfusion (Spencer et al., 1999).
Biotransformation of ferulic acid does not appear to occur in the stomach, including the gastric mucosa (Zhao et al., 2004), or in the intestinal lumen (Adam et al., 2002). However, extensive sulfation, glucuronidation, and sulfoglucuronidation takes place after absorption of ferulic acid from the stomach and intestines, as these conjugates have been detected in plasma, urine, and bile after intragastric deposition (Zhao et al., 2004), transenteral perfusion (Adam et al., 2002), and per gavage or dietary intake (Rondini et al., 2002; Zhao et al., 2003b). Experiments with isolated rat jejunums perfused with ferulic acid demonstrated that approximately 80% of the perfused ferulic acid passes through the intestine in unconjugated form, whereas approximately 20% is glucuronidated (Spencer et al., 1999), indicating that first pass metabolism occurs in enterocytes. The plasma concentrations of ferulic acid and its metabolites peak at 15 minutes after a single, per gavage dose and the compounds are cleared from the blood within 2 (Zhao et al., 2003b) to 4 hours (Rondini et al., 2002). The metabolites are present in blood within 5 minutes postadministration (ferulic acid sulfoglucuronide > unconjugated form > > sulfate = glucuronide) (Zhao et al., 2003b), altogether reflecting rapid metabolism and systemic clearance.
In the in situ gastric deposition experiments (Zhao et al., 2004), ferulic acid was the predominant species in portal vein plasma (approximately 50% of total ferulic acid) but was found at very low concentrations in arterial plasma, bile, and urine. This indicates that ingested ferulic acid is highly susceptible to the second pass effect by the liver. Correspondingly, the largest fraction of total ferulic acid, including its metabolites, was found in bile 25 minutes after intragastric deposition (Zhao et al., 2004). Conjugation in the liver occurs in the order of sulfoglucuronidation > glucuronidation > sulfation = no conjugation (Zhao et al., 2004), although analysis of bile extracted from rats that had been subjected to in situ intestinal perfusion with ferulic acid showed that biliary ferulic acid, which comprised 5–7% of the perfused dose, was only present as glucuronidated and sulfated conjugates (Adam et al., 2002). The degree of sulfation may be influenced by the fact that, as vanillic acid and protocatechuic acid (section III.B.1.a), ferulic acid is an inhibitor of phenolsulfotransferases, albeit less potent (Yeh and Yen, 2003).
Ferulic acid and its conjugates were also retrieved in urine 25 minutes after intragastric deposition in the order of sulfoglucuronidated > glucoronidated ≈ sulfated ≈ unconjugated, although the total ferulic acid content was approximately one-half of that in the liver (Zhao et al., 2004). In rats that were given dietary ferulic acid, the maximum total ferulic acid content in the bladder was reached within 1 hour and the cumulative excretion of free ferulic acid and its metabolites (glucuronidate-, sulfate-, and sulfoglucuronodate adducts) plateaued at 1.5 hours after ingestion (Rondini et al., 2002). Because approximately 70% of per gavage administered total ferulic acid (single dose) was retrieved in urine 5 hours after dosing (Zhao et al., 2003b), it is probable that a fraction of the hepatic metabolites is basolaterally excreted or re-enters the systemic circulation after biliary excretion to ultimately undergo elimination via the renal system.
The pharmacokinetics of ferulic acid in humans differ in some respects from that in rats. First, the cumulative urinary excretion of ferulic acid and its conjugates (at least ferulic acid glucuronide) is considerably slower and reaches a plateau after approximately 9 hours (Bourne and Rice-Evans, 1998). Second, the total fraction of free ferulic acid excreted over a 24-hour period comprises only 4–5% (0.9–3.0 mg in 1.2–3.0 liters of urine, i.e., 2–13 μM) of the ingested ferulic acid, whereas the fraction of total ferulic acid, including its metabolites, is less than 10% (Bourne and Rice-Evans, 1998). It should be noted that these data were collected in healthy volunteers who ingested tomatoes as a source of ferulic acid. In another study, where healthy volunteers drank coffee as a source of ferulic acid, the peak concentration of ferulic acid in urine was reached within 2 hours (Rechner et al., 2001). The bioavailability and pharmacokinetics of ferulic acid are impacted by the dietary source of the ferulic acid (Adam et al., 2002), which may explain the temporal differences in urinary peak concentration between tomato-derived and coffee-derived ferulic acid.
2. Pharmacodynamic Properties of Curcumin Catabolites (Generated In Vitro).
Based on the pharmacokinetic properties it is evident that vanillin is more likely to exert an adjuvant chemopreventive and oncostatic effect in cancers of the gastrointestinal tract, given its poor uptake and systemic bioavailability. In contrast, ferulic acid is completely taken up from the gut, but undergoes biotransformation and systemic clearance via the urinary system within several hours after oral intake. If, however, ferulic acid is capable of accumulating in the tumor during the time its blood levels are still considerable, it may conspire with curcumin in killing cancer cells.
a. Pharmacodynamic properties of vanillin.
Vanillin possesses potent antimutagenic properties, exemplified by its ability to reduce the extent of DNA lesions induced by several chemical mutagens (Ohta et al., 1986), UV light (Takahashi et al., 1990), as well as spontaneous mutations (Shaughnessy et al., 2006) in different Esherichia coli strains, mitomycin C-induced DNA lesions in (hybrid) Chinese hamster ovary cells (Sasaki et al., 1987; Gustafson et al., 2000), mouse bone marrow cells (Inouye et al., 1988), and somatic cells of Drosophila melanogaster (Santos et al., 1999), as well as X-ray-, UV light-, or H2O2-induced mutations in bone marrow-derived cells (Sasaki et al., 1990; Maurya et al., 2007) and hamster fibroblasts (V79 cells) (Imanishi et al., 1990; Tamai et al., 1992) and hybrid ovary cells (Gustafson et al., 2000). In human mismatch repair-deficient (hMLH1−) HCT116 colon cancer cells, vanillin decreased the number of spontaneous mutations in a concentration-dependent manner (19–73% at concentrations of 0.5–-2.5 mM) (King et al., 2007). Vanillin also improved the efficacy of DNA damage repair mechanisms under conditions of oxidative stress (Maurya et al., 2007) and in spontaneously mutating colon cancer (HCT116) cells (King et al., 2007). Ironically, vanillin (2.5 mM) itself appears to inflict a specific type of DNA damage, in consequence to which genes related to DNA damage, oxidative stress and stress responses, cell growth, and apoptosis are activated that in turn trigger DNA repair (by homologous recombination) of mutations not induced by vanillin without causing substantial cell death (King et al., 2007). Such a mechanism was also reported by others (Santos et al., 1999; Shaughnessy et al., 2006).
Microarray analysis of vanillin-treated human hepatocellular carcinoma (HepG2) cells provided insight into the comprehensive biochemical pathway-modulating properties of this catabolite in relation to its oncostatic and antiproliferative properties (Cheng et al., 2007a). At the LC50 (25 mM), 213 genes were downregulated and 347 genes were upregulated by vanillin. Transcriptional effects were also observed at sublethal concentrations; at 1 and 5 mM concentration, vanillin caused the downregulation of 26 and 119 genes and the upregulation of 28 and 84 genes, respectively. At a 5 mM concentration, 47 gene ontology classes were affected, mostly related to the cell cycle and apoptosis but also to tumor progression. These effects appeared to be centered around the Fos gene pathway, which codes for the protein c-Fos, a proto-oncogene that complexes with activator protein-1 (AP-1) to induce transformation and progression of cancer (Milde-Langosch, 2005). Proteomic follow up revealed that vanillin significantly suppressed the activity of AP-1 as a result of decreased levels of several nonphosphorylated mitogen-activated protein kinases (MAPKs), of which the phosphorylated extracellular signal-regulated protein kinases 1 and 2 (ERK-1/2) were also decreased, indicating that AP-1 activity in HepG2 cells was suppressed by vanillin through the ERK signaling pathway. Despite the fact that the vanillin concentrations used in this study were exceptionally high, high doses of vanillin were not toxic in healthy rats when administered per gavage or intraperitoneally up to 300 mg/kg (Ho et al., 2011). Vanillin is also capable of inducing apoptosis in human colorectal adenocarcinoma (HT-29) cells at a ∼2.4-fold lower LC50 (2.6 mM) than in noncancerous fibroblasts (LC50 = 6.6 mM in NIH/3T3 cells) because of its concentration-dependent inhibition of cell cycle checkpoints (G0/G1 arrest at IC50 = 1.3 mM and G2/M arrest at IC50 = 6.2 mM) (Ho et al., 2009).
Lastly, vanillin exhibits antimetastatic properties. Vanillin, but not vanillic acid, suppressed the metastasis of 4T1 mammary adenocarcinoma cells in mouse lungs and inhibited the invasion and migration of cancer cells as well as the enzymatic activity of cancer cell-secreted matrix metalloproteinase 9 (MMP-9) in vitro (Lirdprapamongkol et al., 2005). Inhibition of cell migration and angiogenesis by vanillin was also observed in hepatocyte growth factor-stimulated human lung adenocarcinoma (A549) cells (Lirdprapamongkol et al., 2009). These effects were attributed to vanillin-mediated inhibition of the α, β, δ (class IA), and γ (class IB) phosphoinositide 3-kinase (PI3K) isoforms and Akt (Lirdprapamongkol et al., 2009), and thereby the PI3K/Akt signaling pathway, which in general is constitutively activated in cancers to facilitate proliferation and to reduce apoptosis (Hennessy et al., 2005; De Luca et al., 2012).
b. Pharmacodynamic properties of ferulic acid.
As vanillin, ferulic acid possesses antimutagenic, anticarcinogenic, and antiproliferative properties. Ferulic acid was shown to inhibit mutations induced by several types of mutagens in Salmonella typhymurium (Wood et al., 1982; Yamada and Tomita, 1996) and Eisenia gracilis (Krizkova et al., 2000) as well as benzo[a]pyrene-induced mutagenesis in Chinese hamster fibroblasts (V79 cells) (Wood et al., 1982) at IC50 concentrations in the micromolar range. A ∼55% reduction in benzo[a]pyrene-induced nuclear aberrations was also observed in intestinal cells of mice after dietary ferulic acid treatment (Wargovich et al., 1985).
In regard to carcinogenesis, the induction of tongue squamous cell papillomas and carcinomas by 4-nitroquinoline-1-oxide (Tanaka et al., 1993; Mori et al., 1999) and colonic neoplasms by azoxymethane (Mori et al., 1999, 2000; Kawabata et al., 2000) was inhibited by diet-fed ferulic acid in rats. The ferulic acid-mediated reduction in colon adenomas and adenocarcinomas concurred with increases in GST and quinine reductase in the liver and colon of colon tumor-bearing rats in a ferulic acid dose-dependent manner (Kawabata et al., 2000). Similarly, oral administration of ferulic acid resulted in the normalization of cytochrome P450 and b5 (Alias et al., 2009) and of SOD, catalase, glutathione peroxidase, GST, and GSH to control levels in the liver and skin of skin tumor-bearing mice (Alias et al., 2009) and in the plasma and mammary tissue of mammary tumor-bearing rats (Baskaran et al., 2010). These data suggest that in vivo, both phase I and II detoxifying enzymes may aid in the chemopreventive action of ferulic acid.
Ferulic acid further exhibited anticarcinogenic effects in benzo[a]pyrene-induced pulmonary adenoma development in mice (Lesca, 1983) and ameliorated the 7,12-dimethylbenz[a]anthracene-induced development of skin cancers in mice (Kaul and Khanduja, 1998; Alias et al., 2009), mammary adenocarcinomas in rats (Baskaran et al., 2010), and carcinomas in the buccal pouch of Syrian golden hamsters (Balakrishnan et al., 2008). At the administered dosages, ferulic acid was not carcinogenic (Kaul and Khanduja, 1998; Balakrishnan et al., 2008; Alias et al., 2009; Baskaran et al., 2010). Moreover, in human breast cancer (T47D) cells, ferulic acid inhibited proliferation at an IC50 of 2.3 nM (Kampa et al., 2004), although inhibition of cell proliferation and apoptosis could not be reproduced in other human breast cancer cell lines at ferulic acid concentrations of up to 75 μM (Serafim et al., 2011). At concentrations of 150–1500 μM, however, ferulic acid was able to reduce cell viability in cultured rat hepatoma (HTC) cells (Maistro et al., 2011).
Several possible mechanisms have been elucidated that may in part explain the cytostatic effects of ferulic acid. Administration of ferulic acid to rats with 7,12-dimethylbenz[a]anthracene-induced mammary adenocarcinomas was associated with significantly downregulated expression of mutated p53, which is abundant in breast cancers (Chen et al., 2004; Tennis et al., 2006; Girardini et al., 2011) but also other cancers (Olivier et al., 2010; Freed-Pastor and Prives, 2012), and anti-apoptotic bcl-2 (Baskaran et al., 2010). In HepG2 cells, ferulic acid induced apoptosis in a dose-dependent manner by activating NADPH oxidase, which in turn produced exuberant amounts of ROS that ultimately led to cell death (Lee, 2005). Although ferulic acid failed to trigger apoptosis in estrogen-sensitive (MCF-7) and estrogen-insensitive (MDA-MB-231 and HS578T) human breast cancer cell lines, it was show to induce swelling and mitochondrial permeability transition in isolated rat mitochondria (Serafim et al., 2011), which is a precursor event for apoptosis (Petit et al., 1996; Hirsch et al., 1997).
c. Antioxidant and anti-inflammatory properties of curcumin catabolites (generated in vitro).
The anticancer effects of curcumin catabolites are to an extent related to the intricate relationship between oxidative stress, inflammation, and mutagenesis/carcinogenesis as well as metastasis. Oxidative modification of DNA and lipids during oxidative stress is associated with replication defects, transcriptional deregulation, genomic instability, and aberrant modulation of signaling pathways, i.e., factors that directly induce carcinogenesis (Cerutti, 1985, 1989; Cerutti and Trump, 1991; Trush and Kensler, 1991; Cerda and Weitzman, 1997; Marnett, 1999; Bartsch and Nair, 2006; Federico et al., 2007). Carcinogenesis is also promoted indirectly through inflammation. Oxidative stress triggers inflammation (Rolo et al., 2012; van Golen et al., 2012a,b) by, for example, ROS-mediated modulation of redox-sensitive transcription factors and release of damage-associated molecular patterns, which activate nuclear factor κ-light-chain-enhancer of activated B cells (NF-κB) and other signaling pathways that in turn lead to the release of various pro-inflammatory chemokines and cytokines (Karin and Greten, 2005; Reuter et al., 2010). When pervasive for longer periods of time, these inflammatory mediators can trigger carcinogenesis (Balkwill and Mantovani, 2001; Karin and Greten, 2005; Bartsch and Nair, 2006; Grivennikov and Karin, 2010; Grivennikov et al., 2010), which may be further exacerbated by the augmented levels of ROS and reactive nitrogen species produced during inflammation by leukocytes (Rolo et al., 2012; van Golen et al., 2012a,b). With respect to the link between oxidative stress and metastasis, elevated ROS production becomes instrumental once cells have adopted a cancer phenotype, because ROS intermediates are required for the cytoskeletal rearrangements that underlie cell motility (Pani et al., 2010).
Several of the characterized curcumin catabolites are antioxidants, including vanillin (Sawa et al., 1999; Maurya et al., 2007; Shyamala et al., 2007; Chou et al., 2010; Makni et al., 2011; Tai et al., 2011; Galano et al., 2012), vanillic acid (Sawa et al., 1999; Shyamala et al., 2007; Chou et al., 2010; Kumar et al., 2011; Prince et al., 2011; Stanely Mainzen Prince et al., 2011; Galano et al., 2012), ferulic acid (Scott et al., 1993; Kaul and Khanduja, 1998; Kanski et al., 2002; Kikuzaki et al., 2002; Hirata et al., 2005; Srinivasan et al., 2007; Jung et al., 2009), and 4-vinylguaiacol (Tressl et al., 1976). These catabolites may therefore elicit anticancer effects in various phases of cancer biology. Ferulic acid has been shown to specifically deter lipid peroxidation in lipid-rich egg yolk homogenates (Islam et al., 2009), rat kidneys (Jung et al., 2009), plasma (Balakrishnan et al., 2008), and isolated liver microsomal membranes, especially in synergy with endogenous antioxidants such as α-tocopherol and β-carotene (Trombino et al., 2004). Peroxidized lipids and their reactive aldehyde derivatives such as 4-hydroxy-2-nonenal and malondialdehyde are known to be mutagenic and carcinogenic (Tudek et al., 2010) due to their DNA adduct-forming propensity (Voulgaridou et al., 2011). Ferulic acid may hence prevent mutagenesis/carcinogenesis by neutralizing the detrimental effects of peroxidized lipids and reactive aldehydes in addition to its general ROS scavenging activities. Accordingly, ferulic acid prevented DNA breakage induced by H2O2 in HT-29 cells (Ferguson et al., 2005). Feruloyl methane (Fig. 6B), (2Z,5E)-2-hydroxy-6-(4-hydroxy-3-methoxyphenyl)-4-oxohexa-2,5-dienal (Fig. 6B), the cyclized curcumin derivatives (Figs. 6, A and B, and 7B), and the dimerized catabolites (Fig. 7C) likely possess antioxidant properties as well given their phenylic hydroxyl group(s).
On top of the antioxidant properties, vanillin and ferulic acid can modulate the activity of numerous redox systems and help sustain normophysiological levels of endogenous antioxidants, which are generally compromised during carcinogenesis (Schwartz et al., 1993; Pappalardo et al., 1996; St Clair et al., 2005). In noncancerous hepatocytes, vanillin attenuated the CCl4-induced reduction in the activity of the antioxidant enzymes catalase and SOD as well as GSH (Makni et al., 2011). In tumor-bearing animals, ferulic acid was shown to augment the activity/concentration of the molecular antioxidants α-tocopherol, ascorbic acid, glutathione, and the redox-sensitive enzymes GST, glutathione reductase, glutathione peroxidase, catalase, and SOD to near-normophysiological levels of control animals (Dean et al., 1995; Han et al., 2001; Balakrishnan et al., 2008; Alias et al., 2009; Baskaran et al., 2010). The vast majority of curcumin catabolites is therefore able to ameliorate oxidative stress directly and via endogenous molecular and enzymatic systems, such as the phase II detoxification enzymes, and thereby hamper the potential development and progression of cancer.
Additionally, some curcumin catabolites may indirectly bestow their anticarcinogenic effects by impacting the inflammation component in the oxidative stress/inflammation/carcinogenesis axis—an effect that has been proven for several classes of anti-inflammatory drugs (Wang et al., 2003; Greene et al., 2011; Thun et al., 2012). Vanillin has been shown to inhibit lipopolysaccharide-induced COX-2 gene expression and NF-κB activation in RAW 264.7 murine macrophages (Murakami et al., 2007) and to considerably reduce the expression of the pro-inflammatory cytokines tumor necrosis factor-α (TNF-α), IL-1β, and IL-6 in rat livers after CCl4 treatment (Makni et al., 2011). Vanillin also enhances pro-apoptotic signaling in cancer cells induced by TNF-related apoptosis-inducing ligand (TRAIL), a protein excreted by activated neutrophils (Simons et al., 2008), monocytes (Halaas et al., 2004), and macrophages (Halaas et al., 2000), via inhibition of NF-κB activation (Lirdprapamongkol et al., 2010). Similarly, vanillic acid inhibited TNF-α and IL-6 expression, suppressed the activation of NF-κB and caspase-1, and reduced the expression levels of COX-2 and prostaglandin (PG)E2 in lipopolysaccharide-stimulated mouse peritoneal macrophages (Kim et al., 2011). Ferulic acid was able to decrease the extent of tetradecanoylphorbol acetate-induced inflammation (edema) in mouse ears (Fernandez et al., 1998) and inhibit prostanoid production in IL-1β-stimulated human colon fibroblasts (Russell et al., 2008) as well as COX-2 production in HT-29 cells (Ferguson et al., 2005).
C. Pharmacokinetics and Pharmacodynamic Implications of Curcumin Metabolites Generated in Biological Systems
The chemical modification of curcumin in vivo is significantly different from the in vitro degradation and modification mechanisms because the metabolic processes are mainly enzyme driven. Consequently, the metabolites also differ and mainly comprise end products of reductase-mediated methine bridge hydrogenation, sulfation, glucuronidation, and (Michael reaction-mediated) protein complexation, as illustrated in Fig. 12 (Pan et al., 1999; Fang et al., 2005; Anand et al., 2007; Jung et al., 2007; Marczylo et al., 2007; Dhillon et al., 2008; Marczylo et al., 2009). The disposition of curcumin and its biological metabolites is summarized in Table 1.
1. Pre-enterocytic Pharmacokinetics.
The metabolism of orally ingested curcumin essentially starts in the epithelial cells of the intestinal mucosa, a highly regulated barrier composed of mucus-covered enterocytes that the curcumin molecules must pass through to enter the systemic circulation. However, before intracellular metabolism occurs, several pharmacokinetic hurdles must be overcome. The mucus—a low pH-sustaining aqueous layer replete with glycoproteins and lipids (Legen and Kristl, 2003)—is known to avidly retain curcumin (Berginc et al., 2012), as a result of which the pre-epithelial curcumin concentration gradient is lowered and transmucosal passage of curcumin is impaired. Secondly, intestinal mucus contains μ- and π-class GSTs that catalyze the conjugation of GSH to electrophilic compounds (Samiec et al., 2000), including curcumin (Awasthi et al., 2000; Usta et al., 2007), resulting in pre-enterocytic biotransformation and reduced systemic bioavailability of the native compound. Lastly, the curcumin molecules that are taken up by enterocytes are further subject to apical efflux from the cells (Usta et al., 2007; Berginc et al., 2012) (section III.C.2.f), i.e., back into the intestinal lumen. This also applies to curcumin that has been reduced or conjugated in the enterocytes (sections III.C.2.a through III.C.2.d).
The pre-enterocytic pharmacokinetic hurdles are partly responsible for the very low concentration of curcumin and its metabolites at the basolateral end (i.e., the circulation), as evidenced by the vast difference between the orally administered curcumin dose and the peak curcumin levels retrieved in plasma (Supplemental Table 2) and the very high levels in the intestines and feces. Studies in rats showed that 35–40% (Ravindranath and Chandrasekhara, 1982), 75 ± 8% (Wahlstrom and Blennow, 1978), and 89 ± 3% (Holder et al., 1978) of the orally administered single dose of curcumin was present in feces 3 days after administration. It should be noted that a portion of the retrieved fecal curcumin had undergone basolateral uptake and subsequent hepatic clearance (the second pass effect) and biliary transport back into the intestines inasmuch as the fraction of fecal curcumin was higher 12 days after administration than 3 days after administration (Ravindranath and Chandrasekhara, 1982). Consequently, intestinal retention of curcumin as a result of the first pass effect and enteral re-entry of curcumin and its metabolites due to the second pass effect account for the high levels of curcumin and its derivatives in the gut and poor systemic bioavailability after oral intake.
2. Xenobiotic Metabolism in Enterocytes.
The curcumin molecules that are taken up by enterocytes subsequently undergo phase I (modification) and phase II (conjugation) metabolism (Suzuki and Sugiyama, 2000) by several cytosolic and microsomal enzymes, most notably (oxido)reductases (section III.C.2.a), sulfotransferases (SULTs) (section III.C.2.b), glucuronosyltransferases (section III.C.2.c), and GSTs (section III.C.2.d), which further contributes to the low systemic curcumin levels. The poor bioavailability is also exacerbated by the fact that curcumin acts as a substrate for proteins that are present in enterocytes but not responsible for detoxification metabolism (section III.C.2.e). These proteins can bind and/or chemically modify curcumin. Aside from the specific examples addressed in section III.C.2.e, enzymes that may also catabolize curcumin include aldo-keto reductases (which can modify/cleave β-diketones) (Rosemond et al., 2004; Grogan, 2005), non-phase I metabolism (oxido)reductases that hydrogenate alkenes to alkanes (