The human embryonal carcinoma cell line NT2/D1 is a single-cell clone derived in multiple steps from a parental line designated as Tera-2 
. NT2/D1 cells are considered pluripotent and are in that respect similar to embryonic stem cells 
. In contrast to stem cells, these cells are malignant and they contain karyotype abnormalities including polyploidy and chromosomal translocations 
. However, similar rearrangements are also to varying degree found in most other cancer-derived cell lines. In addition, proteomic comparison of the cell membrane composition between embryonic stem cells and NT2/D1 cells revealed unique features, but also extensive similarities 
. For comparative purposes with respect to the effects of curcuminoids, NT2/D1 cells should be considered a cancer cell line with some features in common with normal embryonic cells.
The effect of curcumin on cell division has been the subject of extensive studies with numerous cell lines and primary cells 
. Curcumin induces growth arrest at various phases of the cell cycle, including G1/S, S, and G2/M. This is accomplished by modulating signal transduction pathways, including those regulated by e.g. NFκB 
, Akt 
, and NrF2 
. These in turn affect the expression of gene products such as among others COX-2, cyclins, Bcl-2, and p53, thereby ultimately reducing or abolishing cell division (for reviews and additional references: 
). The results presented here are qualitatively consistent with those observations. Curcuminoids diminish NT2/D1 cell division in a dose dependent manner upon repeated exposure to curcuminoids at 1–5 µM initial concentrations (). Reports from other studies on the inhibition of cell division or cell survival following curcumin exposure vary over a wide curcumin concentration range. In cases where IC50
values for cell proliferation or viablility were determined for a variety of cancer cell lines, these ranged from 2.5–170 µM after curcumin exposure times of 24–120 h. Comparative values for primary cells ranged from 2.8 to >1000 µM (). However, in most reports curcumin was added as a single initial dose and not replenished before the time of analysis. It is therefore difficult to compare those studies with the results reported here, where curcuminoids were regularly replenished every 24 h. Indeed, comparable IC50
values for NT2/D1 proliferation would be about 2.3 µM after 96 h or 3.8 µM after 48 h of exposure. These values would likely be higher after a single application of curcuminoids, since these would be depleted after 24 h (). However, it is clear that different cell lines vary in their sensitivity to curcumin. For example, HeLa cells can be repeatedly exposed to more than 10 µM curcuminoid concentrations without apparent cell death under the same conditions as those applied here 
(). The reasons for these different responses remain to be determined, but they may be related to the configuration of cellular sites available for curcuminoid binding and their role in the regulation of cell division. Although the effects of curcuminoids on normal cells were not examined in this study, it may be noted that a number primary cells from different tissues are sensitive to the effects of curcumin, albeit at generally higher concentrations than is typically the case for cancer-derived cell lines (). Furthermore, curcumin inhibits the proliferation of human retinal endothelial cells at concentrations of 1–30 µM 
. Curcumin also inhibits cell division in mouse oocytes and early embryonic development at concentrations of <50 µM 
. In a different study, significant apoptosis and inhibition of cell division was reported at the mouse blastocyst stage at a 24 µM curcumin concentration 
. Survival of zebra fish embryos and larvae was also impaired after a 24 h exposure to curcumin with respective LD50
-values of 7.5 µM and 5 µM 
. These results suggest that the effects of curcumin on normal cells may not be as benign as generally presumed.
Curcumin effect on cell proliferation.
At higher initial curcuminoid concentrations (>8 µM), extensive NT2/D1 cell death occurs within two days of exposure (). At 46–49 µM curcuminoid concentrations, cell death is largely due to apoptosis via the activation of caspase 3/7 by both extrinsic (caspase 8) and intrinsic pathways (caspase 9). In this respect, these results are consistent with those reported for other cell lines at 20–100 µM curcumin concentrations 
and for mouse embryos at 6–24 µM concentrations 
. However, in NT2/D1 cells there is also a generalized activation of caspases in response to higher levels of curcuminoids (). This includes caspases 1,4 and 5, which are activated by extracellular inflammatory stimuli 
. Caspase 2 is considered to be another initiator caspase whose function remains undetermined, whereas caspase 10 is functionally related to caspase 8. Finally, caspase 6 is an effector caspase with a different substrate specificity from caspases 3/7 
. Although at higher curcuminoid concentrations there is a universal activation of caspases in NT2/D1 cells, the dose-dependent cellular response to curcuminoids suggests that caspase activation may be similarly differentially regulated. In particular since caspases also have a function in cellular proliferation and differentiation 
. Although caspase-dependent apoptosis is a prominent feature at high curcuminoid concentrations, alternative mechanisms of cell death such as autophagy 
or other caspase-independent pathways 
have been described.
Despite structural evidence of apoptosis and caspase activation, nucleosomal DNA fragmentation could not be demonstrated in NT2/D1 cells exposed to curcuminoids. In contrast, apoptosis induction by camptothecin did show concomitant DNA fragmentation indicating that this result was not due to methodological reasons. This phenomenon may indeed be cell-type specific since variable induction of DNA fragmentation was observed in leukemia and fibroblastic cell lines 
. Lack of DNA fragmentation was also observed in Jurkat cells and and primary T-cells. This was interpreted as the direct inhibition of the CAD endonuclease by the binding of curcumin to the active site 
. However, CAD activation by caspase 3 occurs in the nucleus 
and significant nuclear localization of curcumin has generally not been observed by histological methods 
. Nevertheless, at least one study has reported some degree of nuclear localization both by fluorescence microscopy and differential centrifugation 
. It is therefore conceivable that a sufficient amount of curcuminoids are present in the nucleus to achieve CAD inhibition, or that under apoptotic conditions there is a different distribution of curcuminoids within cells due to altered exposure of subcellular compartments.
At intermediate curcuminoid concentrations (6–7 µM) a limited degree of cell death is observed and the remaining NT2/D1 cells gradually develop a phenotype that is consistent with senescence (). Cellular senescence is often induced by external stress signals such as hypoxia or by the action of cytokines 
. These in turn induce the production of reactive oxygen species either at the cytoplasmic level or in the mitochondrial transport chain. In particular in human cells, the activation of the p16INK4A/rb signaling pathway leading to the irreversible cessation of cell division has been implicated in the induction of senescence 
. The observation that only a subpopulation of cells comprising about 40–50% of the original number () developed into a senescent phenotype (), suggests that this may not be a random process. Since NT2/D1 cells are not homogeneous in terms of chromosome distribution, it is conceivable that some of the cells express biomarkers that make them more susceptible to cell death upon exposure to intermediate curcuminoid concentrations. In addition, the initial change in morphology in the remaining cells () may indicate a form of cellular differentiation, which is consistent with the notion that this cell line can be induced to differentiate into various ectodermal cell types 
. This concept may be extended to other cancer cell lines. For example, at intermediate curcuminoid concentrations a majority CCF-STTG1 cells showed robust differentiation into a phenotype resembling astrocytes before progressing to senescence. In contrast, only a minority of HeLa cells exhibited similar signs of differentiation and senescence (not shown). These variable effects among dedifferentiated cancer cell lines may again reflect the distribution of cells within the total population with biomarker profiles that designate them either for cell death or differentiation at intermediate curcuminoid concentrations.
There is a concentration-dependent gradient of biological effects of curcuminoids on NT2/D1 cells ranging from a decrease in the rate of cell division to cell death by apoptosis. Since most of these events can be accounted for by the interaction of extracellular signals on cell membrane receptors, it is not surprising that curcuminoids bind preferentially to membrane-associated fractions (). Although it has not been proven that the cell membrane is the primary target for curcuminoid binding, it is reasonable to assume that it is the initial target, in particular, in view of the rapid dissociation of non-metabolized curcuminoids from the cells following their removal from the medium (). Data based on fluorescence microscopy seem to suggest that curcumin also accumulates intracellularly 
. Although fluorescence microscopic resolution does not appear to be sufficient to clearly differentiate between cell membrane and intracellular distribution, a study using confocal microscopy to determine curcumin localization concluded that curcumin almost exclusively partitioned into membrane structures 
. A different study employing differential centrifugation also showed preferential binding to membrane fractions 
. However, Kunwar et al. 
used 0.6% Nonidet P-40 to disrupt cells. Even at such a concentration, this detergent substantially increases the solubility of curcumin, which is likely to dislodge membrane bound curcumin during extraction and fractionation. This would explain the significantly larger proportion of curcumin recovered from the cytoplasmic fraction in that study compared to the very low levels of cytoplasmic curcuminoids reported here (). Another study employed sonication as a method of cell disruption 
. Cells were then centrifuged at 9,000×g and the supernatant was further centrifuged at 100,000×g. This resulted in total amounts of curcumin 2–3 times higher in the 100,000×g membrane pellet than in the cytosolic supernatant, although the amount bound to the initial 9,000×g pellet was not reported.
The cellular sites that are targets for curcuminoid binding remain largely unknown. However, they are likely to be numerous with varying individual binding affinities. Consequently, the overall KD
values for specific cellular binding of curcuminoids determined here are not to be considered as representing the affinity for a single binding site. Instead, they represent the sum of numerous binding sites with similar curcuminoid binding affinities. Protein components embedded in the membrane such as cellular receptors, a wide variety of which have been identified in NT2/D1 cells 
, seem to be likely targets. In particular, since curcumin has been shown to interact with proteins, including albumin and numerous mediators of cellular transduction (for references see: 
). However, direct binding of curcuminoids to the exposed polar moieties of membrane lipids is also likely. This clearly has implications in the context of membrane compartmentalization into lipid rafts 
. The binding of curcuminoids to such ordered lipid-protein assemblies could affect both signal transduction events and provide an avenue for internalization 
. Indeed, lipid raft and receptor mediated endocytosis has been described for the cellular uptake of resveratrol, a biochemically and functionally related polyphenol 
The overall KD
for total specific binding of soluble curcuminoids to NT2/D1 cells was here determined to be in the 6–10 µM range using variable amounts of curcuminoid-saturated serum (). These KD
values are based on cellular binding versus the concentration of total soluble curcuminoids, which are primarily present as complexes with serum components under saturating conditions. Although this value is in the same range as the concentrations necessary to induce biological effects, actual cellular binding under the employed experimental conditions of variable curcuminoid and constant serum concentration would be considerably lower due to competition for cellular binding by free serum (). This also has therapeutic implications, since in vivo
the lower degree of cellular curcuminoid binding in the presence of excess serum would further limit the cellular response at an already low level of bioavailability. The situation is further complicated by the observation that there are at least two different binding sites for curcumin on serum albumin 
, which is a primary curcuminoid-binding component in serum 
. Indeed, the stoichiometric ratio of curcuminoids/albumin at binding saturation suggests the presence of at least three potential albumin binding sites 
. It is likely that the albumin-bound curcuminoids are released sequentially for binding to cellular receptors. It is also plausible that the serum- or albumin-bound curcuminoids occupy different binding sites, depending on whether they are SOLID- or DMSO-solubilized. Upon mixing such saturated preparations, intermolecular rearrangements could occur leading to a redistribution of curcuminoid binding sites within the serum that would have a similar competitive effect as adding curcuminoid-free serum. This would then result in an increase in the overall apparent KD
for cellular binding as was observed for albumin and serum solutions (). Consequently, differences in apparent KD
s for cellular binding resulting from using different soluble curcuminoid preparations would not represent variable affinities for cellular receptors, but rather varying affinities for curcuminoid-binding to serum components.
A comparison of the cell lines NT2/D1 (embryonal carcinoma), HeLa (cervical carcinoma), and CCF-STTG1 (astrocytoma), shows that these cells differ in their rate of cell division, their affinities for curcuminoids, and their response to curcuminoids (). The cellular responses to curcuminoids in the form of cell death and senescence are related to the cellular affinities for curcuminoids. The reasons for these differences in cellular affinities for curcuminoids between cell lines are currently unclear. However, they are likely associated with the lipid and protein organization within the cell membranes of the different cell lines. As a result, the cellular affinity and response to endogenous or serum-derived growth factors may also vary, resulting in different rates of cell division.
Although differential effects of individual curcuminoids on a variety of biological responses have been reported (review: 
), it did not matter in these experiments whether the curcuminoids solubilized in FCS contained either predominantly CUR (DMSO-solubilized) or BDMC (SOLID-solubilized). The effects on cell division, apoptosis, induction of senescence, or activation of caspases were indistinguishable for the two preparations within the limits of the established experimental parameters. This could be because the two preparations contained a complement of all curcuminoids, albeit at different relative abundance, and as such the experimental design did not allow for a sufficient resolution to detect differential effects. In addition, the lack of difference may be explained by the different properties of curcuminoids in solution and in terms of binding and metabolism. Although BDMC and DMC have higher apparent binding affinities for cellular receptors, they are also metabolized at a relatively faster rate, resulting in a faster depletion from cells and media ( and ). Conversely, over a 24 h period CUR is more readily subject to chemical decomposition than BDMC or DMC 
. Overall, the sum of these effects may neutralize any differential biological effects of the individual curcuminoids in these preparations.
Curcuminoids bound to cell membrane receptors could act on signal transduction pathways by either blocking the access of growth factors present in the serum or by inducing cellular responses themselves. However, simply reducing or eliminating the serum in the media does not produce responses similar to those induced by curcuminoid exposure. Indeed, reducing serum concentrations to 1% merely resulted in a moderate decrease in the rate of cell division. Completely eliminating the serum induced extensive cell death within 24 hr while the induction of senescence could not be demonstrated at any serum concentration (not shown). Furthermore, exposing NT2/D1 cells to high concentrations of curcuminoids for shorter time periods of 0.5–4 h resulted in cellular responses similar to those obtained with chronic exposure at lower concentrations (). These observations suggest that curcuminoid-binding to cells activates signal transduction pathways that lead to dose- or time-dependent reductions in cell division, induction of senescence, or apoptosis. Indeed, curcumin has been identified as an activating ligand for the vitamin D receptor (VDR) 
, which is also highly expressed in NT2/D1 cells 
. The intracellular VDR could potentially be accessed by curcuminoids via the same cell membrane receptor- or caveolae-mediated mechanisms that are also used to transport steroid hormones 
. That curcuminoid binding per se
is responsible for these effects is further exemplified by the observations that presumably inactive metabolites such as those described here (), and conjugated products such as curcumin sulfate (not shown) or glucuronidate 
show no significant cellular binding. Similarly, upon incubating cell cultures with hexahydrocurcumin, the metabolite was found in the cytosolic but not in the membrane fraction 
, indicating a general loss of membrane binding upon curcumin reduction. It is indeed likely that the poor solubility of curcumin in water is a crucial property that forces it to be shuttled between amphophilic binding sites in serum components and the cellular binding sites, a property also shared by steroids and related compounds.
The primary metabolic reduction products of curcumin in vivo
, in tissue slices, and in cell extracts are tetrahydrocurcumin and hexahydrocurcumin with smaller amounts of dihydrocurcumin and octahydrocurcumin, a conversion that takes place primarily in the liver and intestine 
. Hexahydrocurcumin has also been identified in cultured Ishikawa and HepG2 cells 
. In this study, hexahydrocurcuminoids were identidied as the major metabolites in both HeLa and CCF-STTG1 cells, while the latter also generated significant amounts of octahydrocurcuminoids (). However, none of those reduction products were identified in NT2/D1 cells. Instead, novel curcuminoid reduction products with the same molecular weight as tetrahydrocurcuminoids but with different spectral and chromatographic elution properties were detected (). The preliminary structure assignment of these compounds shows different positions of the remaining double bonds compared to tetrahydrocurcumin. The same conversion products were initially also found in the astrocytoma cell line CCF-STTG1 and in HeLa cells upon shorter incubation times (<5 h) and are thus not unique to NT2/D1 cells (not shown). This shows that at least in culture, an alternative reduction pathway is available to inactivate curcuminoids. Whether this also exists in vivo
remains to be established. The enzymes responsible for the reductive conversion have not been identified, but at least in vitro
it can be accomplished with alcohol dehydrogenase 
. In any case, most reductases are cytoplasmic enzymes and consequently it would require the curcuminoids to be internalized after binding to the cell membrane. The resulting curcuminoid reduction products would then loose their membrane binding properties upon which they are expelled into the medium. That this is an efficient process is suggested by their lack of detection in cells.
In conclusion, this study has described the cellular binding and metabolic fate of curcuminoids. While focusing primarily on NT2/D1 cells as a model system, selected binding parameters and cellular effects were also compared to HeLa and CCF-STTG1 cells. For this purpose, commercial grade curcumin was first solubilized in FCS at saturating concentrations in the 1 mM range. As a result, maximal curcuminoid concentrations of about 50 µM can be achieved in culture media containing 5% total FCS. This method of solubilization is different from that typically employed in other studies, where curcumin is dissolved in an organic solvent and added at appropriate dilutions to the final culture medium. Directly solubilizing curcuminoids in serum offers several advantages over adding curcuminoids dissolved in organic solvents. These include the lack of initial precipitation, lower overall concentrations of organic solvents, and greater precision in achieving small increments of curcuminoid concentrations in the media. The lack of precipitation is particularly relevant to studies where curcuminoid concentrations in excess of 100 µM are employed with media containing 10% FCS. Since the vast majority of soluble curcuminoids are complexed with FCS, any excess beyond the saturation point will remain in the media as particulate matter with unknown consequences in terms of binding, metabolism and biological effects. Furthermore, commercial grade curcumin can be solubilized in serum by two procedures resulting in similar total concentrations but profoundly different final compositions. Direct mixing of such solutions at different ratios thus easily generates preparations with a range of curcuminoid profiles. This has allowed for the titration of their biological effects at curcuminoid concentration increments of about 1 µM in the media. Future projects will address the modulation of specific signal transduction pathways using the same range of curcuminoid concentrations that result in reduced rates of cell division, induction of senescence, and cell death. Based on such data, it should be possible to correlate specific biological responses with the incremental binding of curcuminoids to cells and specific signal transduction events. Given a sufficient resolution, this could help clarify the sometimes conflicting data that has emanated from the large body of literature encompassing curcumin research.