Curcuma longa L. belongs to the family Zingiberaceae, which is a perennial herb that measures up to 1 m height with a short stem. It is distributed throughout tropical and subtropical regions in the world, being widely cultivated in Asian countries, mainly in India and China (Kapoor, 1990). As a component of folklore medicine the use of C. longa has been documented both in Indian and Chinese cultures. The rhizomes are used as Ezhu according to the Chinese Pharmacopoeia (2010 edition; Zhao et al., 2010) and are a household remedy in Nepal (Eigner and Scholz, 1999). The long list of usages include antiseptic, analgesic, anti-inflammatory, antimalarial, and insect repellant (Chaudhri, 1950; Li et al., 1998; Niederau and Göpfert, 1999; 2001; Tawatsin et al., 2001; Duke, 2002). Traditional Indian medicine claims to use its powder against biliary disorders, anorexia, coryza, cough, diabetic wounds, hepatic disorders, rheumatism, and sinusitis (Ammon et al., 1992). In ancient Hindu medicine, C. longa is extensively used for the treatment of sprains and swelling caused by injury (Ammon and Wahl, 1991). The antioxidant activity of C. longa is well-known (Ammon and Wahl, 1991; Anand et al., 2007).

Extensive research during the past years revealed that curcumin has considerable potential against a wide variety of both malignant and non-malignant diseases. Curcumin exhibits activity against numerous inflammatory diseases, including pancreatitis (Gukovsky et al., 2003; Gülçubuk et al., 2005), arthritis (Joe et al., 1997; Liacini et al., 2002), inflammatory bowel disease (Holt et al., 2005), gastritis (Swarnakar et al., 2005), allergy (Baek et al., 2003; Ram et al., 2003), and fever (Lee et al., 2003; Shao et al., 2004), possibly through the downregulation of inflammatory markers. It is also active against autoimmune diseases, including scleroderma (Tourkina et al., 2004), psoriasis (Bosman, 1994), multiple sclerosis (Verbeek et al., 2005), and diabetes (Babu and Srinivasan, 1995, 1997; Sajithlal et al., 1998). Curcumin also exhibits a great potential against various types of cancers. Its mechanism of action involves, firstly, the suppression of tumor cell proliferation by down-regulation of anti-apoptotic gene products, activation of caspases, and induction of tumor suppressor genes (Jiang et al., 1996; Bush et al., 2001; Chan and Wu, 2004). Secondly, curcumin suppresses tumor invasion by down-regulation of matrix metalloproteinases and cell surface adhesion molecules (Lin et al., 1998; Fenton et al., 2002; Lee et al., 2006). Thirdly, curcumin inhibits angiogenic cytokines leading to suppression of angiogenesis (Shin et al., 2001; Leyon and Kuttan, 2003; Bobrovnikova-Marjon et al., 2004) and lastly the anti-inflammatory and cytotoxic effects of curcumin contribute to its antitumor activity (Srivastava, 1989; Fujiyama-Fujiwara et al., 1992; Ammon et al., 1993).

The activity of many medicinal plants results from the interaction action of several constituents, which may cooperatively act in an additive or synergistic manner. It has been repeatedly observed that extracts of medicinal plants reveal better activities than their isolated single compounds at comparable equivalent concentrations of the active components. This phenomenon is attributed to the absence of interacting substances present in crude extracts. For example, extracts were obtained from the fresh herb of Artemisa annua L. either by soaking the herb in water followed by wringing out the juice by hand or by pounding the fresh herb to a pulp followed by squeezing out the juice. The extracts were then analyzed for artemisinin concentration and tested against malaria parasites. It was found that the antiplasmodial IC₅₀-values were 6–18-fold lower than was expected in terms of their artemisinin content suggesting that the activity of the extracts could not be entirely accounted by their artemisinin content (Wright et al., 2010). Another example are the Cinchona alkaloids. There are almost 30 alkaloids described in the bark of Cinchona officinalis L. The most well-known of these are quinine, quinidine, and cinchonine and cinchonidine. However, quinine is not the most potent of the alkaloids: quinidine, dihydroquinidine, and cinchonine all have consistently lower 50% inhibitory concentrations (IC₅₀) in vitro. The combination of quinine with quinidine and cinchonine is 2–10 times more effective in vitro against quinine-resistant strains, and the mixture of alkaloids reveals more consistent effects than any of the alkaloids singly used (Druilhe et al., 1988; Karle and Bhattacharjee, 1999). In pharmacokinetic synergy, substances with little or no bioactivity may assist the main active principle to reach the disease target by several mechanisms, e.g., improving bioavailability, or decreasing metabolism and excretion. Comparable effects are not yet known for C. longa and needs to be elucidated. In the present investigation, we addressed this question. We have chosen ascorbic acid (vitamin C, ascorbate, C₆H₁₂O₆) as phytochemical constituent of C. longa to investigate the additive or synergistic effects of its combination with curcumin.

Ascorbic acid is a ketolactone and a water-soluble antioxidant. There are two chemical forms of ascorbic acid: the reduced form (ascorbic acid; AA) and the oxidized form (dehydroascorbic acid; DHA; Mamede et al., 2011). AA is well-known for its potent antioxidant properties, as it is able to scavenge free radicals and reactive oxygen species (ROS). Thus, it has been associated with decreased oxidative stress in vivo (Carr A. and Frei, 1999; Carr A. C. and Frei, 1999). High doses of AA can reduce inflammatory biomarkers such as C-reactive protein (CRP), tumor necrosis factor (TNF-α), interferon-γ (IFN-γ), and the interleukins IL-1, IL-2, IL-6, IL-8 (Gilliam and St Clair, 2011; Mikirova et al., 2012). AA is a cofactor in vivo for enzymes involved in the biosynthesis of collagen, carnitine, neurotransmitters, and neuropeptide hormones as well as enzymes involved in regulation of epigenetic or transcription factors (Rebouche, 1991; Du et al., 2012). Furthermore, it is a cofactor in the synthesis of the neurotransmitters norepinephrine, dopamine, and serotonin and neuropeptide hormones such as oxytocin (Harrison and May, 2009; May et al., 2012). Extensive studies have been carried out on AA in the treatment of cancer and vast literature exists on AA and cancer. In 1949, AA was first proposed to be used for cancer therapy (Klenner, 1949; Mc, 1952). The first comprehensive review on AA and cancer was published in 1979 and an updated review 25 years later (Cameron et al., 1979; González et al., 2005). AA may act as a prodrug causing the formation of AA radical and hydrogen peroxide in the extracellular space (Chen et al., 2007). In a clinical trial, AA exerted antitumor activity in patients with advanced cancer as a stand-alone therapy as well as in combination with other anticancer agents (Hoffer et al., 2008). The conjugation of AA with extracts of medical herbs stimulated apoptosis and disrupted the cell cycle in different cancer cell lines. Furthermore, AA was pro-oxidant generating hydrogen peroxide-dependent cytotoxicity toward various cancer cells without adversely affecting normal cells. AA together with sodium nitrite induced genotoxicity due to oxidative DNA damage. High concentrations of AA killed tumor cells in vitro with high efficiency and inhibited angiogenesis in mice bearing sarcoma (Chen et al., 2008; Kuroiwa et al., 2008; Verrax and Calderon, 2009; Yeom et al., 2009; Rozanova Torshina et al., 2010).

In this paper, we have chosen AA because it is widely distributed in many plants and it is also synthetically available. This allows us to investigate the cytotoxicity of the combination of curcumin and AA outside of the plant at exactly defined conditions to clarify their pharmacological effects in the combination. Our strategy was to investigate the combination by isobologram analysis and to investigate the genes which are related to the cytotoxicity induced by curcumin and AA in a broad spectrum of cancers. We postulate that the identification of genes that are specifically regulated by AA and curcumin could improve the understanding of the efficacy of C. longa in cancer treatment. We have systematically studied the microarray-based mRNA expression of genes, which influence the cellular response to curcumin and AA in the tumor cell line panel of the National Cancer Institute (NCI), USA to find possible mechanistic explanations for the interaction of these two compounds.

Plants are perhaps earth's most accomplished chemists. They produce thousands of specialized secondary metabolites, and during the evolution of life plants developed multi-targeted chemicals to fulfill diverse tasks. C. longa contains some 30 different phytochemicals. In the present study, we analyzed the cytotoxicity of a combination of two of these phytochemicals of C. longa, curcumin and AA. If applied alone, curcumin and AA were cytotoxic toward cell lines of different tumor types with curcumin exhibiting stronger cytotoxicity than AA. Our results are in line with other reports on the inhibitory activity of curcumin (Bimonte et al., 2016; Guzzarlamudi et al., 2016; Kasi et al., 2016; Ye et al., 2016; Yu et al., 2016; Zeng et al., 2016) and AA (Chen et al., 2015; Fukui et al., 2015; Jacobs et al., 2015; Sunil Kumar et al., 2015; Venturelli et al., 2015). The anticancer effects of curcumin in vitro and in vivo are primarily due to the activation of apoptotic pathways in cancer cells as well as the inhibition of mechanisms related to the tumor microenvironments such as inflammation, angiogenesis, invasion, and metastasis. In particular, curcumin targets numerous therapeutically important cancer signaling pathways such as p53, Ras, PI3K, AKT, Wnt, β-catenin, mTOR, and so on.

AA also reveals anticancer activity in vitro and in vivo, however at higher concentration than curcumin. A number of suggestions have been put forth on the potential mechanisms, by which AA causes death of cancer cells. The most common is that AA is a precursor for hydrogen peroxide (H₂O₂) generation, which is considered to be preferentially cytotoxic to cancer cells (Parrow et al., 2013). Exposure to AA concentrations of up to 5 mM for 1 h resulted in decreased survival of cancer cells and cell death was dependent on H₂O₂ production mediated by extracellular AA oxidation (Chen et al., 2005). Furthermore, intercellular metals contribute to the production of H₂O₂, with AA losing an electron to form a radical molecule. The free electron is donated to a transition metal. This reduced metal is then available to react with molecular oxygen resulting in the generation of H₂O₂. In the presence of AA, H₂O₂ reacts with another transition metal ion such as ferrous ion to generate a hydroxyl radical (Kehrer, 2000). The tumor suppressor p53 may also play a role for this activity. P53-positive cell lines were more sensitive to both AA and H₂O₂ treatment than p53-deficient ones (Kim et al., 2012).

Thus, having two compounds from one plant with diverse mechanisms of action, we were interested to evaluate, whether their combination would reveal synergistic, antagonistic, or additive interactions. Understanding drug-drug interactions always represents a critical issue in the drug development process, since clinically relevant changes in exposure of co-administered drugs can lead to reduced efficacy or, conversely, adverse drug reactions, depending on the therapeutic window of the drugs. The latter becomes especially important with anti-cancer medications, since they are typically administered at or close to the maximally tolerated dose (Waters, 2015). In this study, we applied isobologram analyses to evaluate the nature of interaction of curcumin and AA. Isobologram analysis is considered as gold standard to provide evidence for drug interactions. The combination of curcumin and AA exhibited additive effects in leukemia and colon cancer cell lines and supra-additive effects in the glioblastoma cell lines. Our hypothesis is that both natural products may work together by different molecular pathways to achieve their overall cytotoxicity. Unlike chemically synthesized drugs, natural products might be active at lower doses and over longer periods of incubation, which could further support the appearance of additive effects.

We further analyzed molecular determinants of sensitivity and resistance of cancer tumor cell lines toward curcumin and AA. We correlated the IC₅₀-values expressed on induction by curcumin and AA of 60 tumor cell lines by COMPARE analysis of microarray-based transcriptome-wide mRNA expression levels of these cell lines (Scherf et al., 2000). We identified genes from diverse functional groups, which were associated with response of the tumor cells toward curcumin and AA. Under curcumin treatment, these groups of genes included DNA repair, mRNA metabolism, signal transduction, angiogenesis, proliferation, apoptosis etc. While for AA treatment, the COMPARE analysis provided genes that are involved in signal transduction, transcription factors, and apoptosis. Although the exact function of these genes for cellular responsiveness to curcumin or AA treatment is still unknown, we have some clues of explanation. On treatment with AA the following genes were downregulated: SETD4, TAF9, PNN, CDK7, TFAM, and PPWD1. SETD4 is a methyltransferase, which is involved in carcinogenesis. Its down-regulation suppressed cellular proliferation and delayed the G1/S cell cycle transition without affecting apoptosis. Furthermore, its knockdown decreased cyclin D1 (Faria et al., 2013). The TATA-binding protein associated factor 9 (TAF9) interacts with oncogenic GLI family members to form GLI-TAF9 binding, which is important for carcinogenesis activity and malignant growth (Yoon et al., 2015). PNN is a nuclear and cell adhesion-related protein participating in the regulation of gene expression and thereby, positively promoting cell-cell adhesion, and negatively affecting cell migration and cell proliferation (Shi et al., 2001). Cyclin-dependent kinase 7 (CDK7), which promotes transcription during the cell cycle, is critical for the survival of cancer cells. The inhibition of CDK7 suppressed proliferation and induced apoptotic cell death (Wang et al., 2015). Mitochondrial transcription factor A (TFAM), a member of the high mobility group (HMG) box protein family, is required for mitochondrial DNA replication and transcription. HMG proteins are often overexpressed in cancer cells and are involved in apoptosis regulation (Krynetskaia et al., 2009; Vander Heiden et al., 2009). TFAM may play a significant role in tumorigenesis (Guo et al., 2011). PPWD1 has a well-characterized peptide domain (WD40 domain for PPWD1), and this domain has also a critical role in carcinogenesis. The WD40 domain mediates signal transduction and transcriptional regulation during cell cycle and apoptosis. PPWD1 may serve as target for drug development (Davis et al., 2008; Jeon et al., 2014). From the above discussion, AA mechanism revolves around the genes affecting cell proliferation and cell cycle events; it can be assumed that these genes contribute to sensitivity of the tumor cells to AA.

On the other hand, genes potentially responsible for responsiveness on curcumin treatment were identified and curcumin downregulated, i.e., AK2, PDK1, NR2F2, DFFB, MTHFD2, and ALDH3B1. Adenylate kinases (AKs) represent enzymes that catalyze reversible high-energy phosphoryl transfer reactions between adenine nucleotides in the intermembrane space. During periods of metabolic stress, AK2 increases the amount of available adenosine monophosphate and therefore activates downstream ATP-sensing mechanisms—such as AMP-activated protein kinase (AMPK)—to regulate the cellular metabolism. Inhibition of AK2 expression significantly inhibited the proliferation of cancer cells (Dzeja and Terzic, 2009). PDK1 plays a key role in several cancer types. Alterations of PDK1 are critical for oncogenic PI3K signaling. PDK1 has an essential role in regulating cell migration, especially in the context of PTEN deficiency. Downregulation of PDK1 levels inhibits migration and metastasis. PDK1 inhibitors may be useful to prevent cancer progression and abnormal tissue dissemination (Raimondi and Falasca, 2011). The nuclear receptor subfamily 2, group F, member 2 (NR2F2) is a master regulator of angiogenesis and acts as oncogene in prostate and other human cancers. NR2F2 is robustly expressed in the stroma of healthy ovary with little or no expression in epithelia lining the ovarian surface, clefts, or crypts. The pattern of NR2F2 expression was severely disrupted in ovarian cancers, in which decreased levels of stromal expression and ectopic epithelial expression were exhibited. Targeting NR2F2 expression in ovarian cancer cell lines enhanced apoptosis and increased proliferation (Hawkins et al., 2013). DFFB contributes to both chromosomal condensation and DNA degradation during apoptosis, decreased DFFB expression favors DNA damage, which in turn may contribute to both tumorigenesis and better response to DNA damaging chemotherapy (McDonald et al., 2005). MTHFD2 mRNA and protein expression is markedly elevated in many cancers and correlated with poor survival in breast cancer. MTHFD2 is integral to mitochondrial one-carbon metabolism, a metabolic system recently implicated in rapid cancer cell proliferation. Synthesis of one-carbon units carried by the tetrahydrofolate (THF) cofactor is important for proliferating cells, required for nucleotide synthesis and methylation reactions. MTHFD2 is a bifunctional enzyme, catalyzing the NAD⁺ dependent CH2-THF dehydrogenase and CH⁺-THF cyclohydrolase reactions within the mitochondria. Within the mitochondrial folate pathway, MTHFD2 is of special interest, because MTHFD2 was one of the most consistently overexpressed mRNAs genome-wide across 19 different tumor types. The MTHFD2 protein is specifically expressed in transformed cells, but not the stroma surrounding the tumor tissues. MTHFD2 by RNAi impairs proliferation in a variety of cancer cell lines, independent of the tissue of origin, and decreases invasion and migration in breast cancer cell lines. MTHFD2 is broadly required for cancer cell proliferation and viability (Lehtinen et al., 2013; Nilsson et al., 2014). ALDH3B1 is a metabolically active enzyme with distinct specificity for various aldehyde substrates, particularly medium-, and long-chain aliphatic aldehydes. These substrates include many products that are formed during LPO, such as hexanal, 4-hydroxy-2-nonenal (4-HNE), octanal, and trans-2-nonenal. ALDH3B1 plays an important physiological role against cellular oxidative stress by detoxifying aldehydes derived from oxidative processes, such as ethanol metabolism and LPO (Marchitti et al., 2010). Our pharmacogenomics data shows curcumin suppressing cell proliferation by downregulation of anti-apoptotic genes and cell surface adhesion molecules. Curcumin is also seen to regulate cellular metabolism and inhibition of angiogenic cytokines. We can then suggest that the downregulated genes affect the sensitivity of the tumor cells to curcumin.

Previously, we reported the mRNA expression profile induced by curcumin in the NCI cell line panel, which was merged from four different microarray platforms (Novartis, Stanford, Chiron, and Genelogic; Sertel et al., 2012). In the present investigation, we focused only on the Novartis microarray platform for the comparison of curcumin and vitamin C and to reduce the degree of complexity. If we compared the top ranked genes in the previous analysis with those of the present investigation, we found some genes in common (MTHFD2, AK2, NFATC21P, BTAF1, RBBP4), although the majority of genes were different in both analyses. This result points to an observation that was frequently made by many investigators: different microarray platforms deliver different results. Nevertheless, several biological functional groups were found to be common in our previous and the present paper, e.g. cell cycle, DNA damage response, cell migration, inflammation, signaling pathways, and apoptosis-regulating genes. To our opinion, microarray data are reliable, if they are used for the generation of testable hypotheses. In this respect, both of our microarray analyses were useful. Microarray data represents the starting point for the elucidation of modes of action of cytotoxic compounds rather than completed end results.

Cluster analyses were applied in the present investigation under the assumption that responsiveness of cancer cells might be predicted by using gene expression patterns and that appropriate gene expression profiles might be sufficient to predict whether a cancer cell line is sensitive or resistant to a cytotoxic compound (Sertel et al., 2010). Curcumin revealed two clusters with predominantly sensitive and three with predominantly resistant cell lines. For AA cluster analysis revealed two clusters containing mainly resistant and two clusters containing mainly sensitive cell lines in a comparable fashion to curcumin. The prediction of sensitivity or resistance to cytotoxic drugs by mRNA expression profiles is interesting in the context of individualized or precision medicine, because it may open the possibility to determine prior to treatment, whether or not a tumor will respond to specific drugs. Our data demonstrate that this may not only be feasible to established anticancer drugs, but also to investigative natural products such as curcumin or AA.

The fact that the mRNA expression profiles induced by curcumin and AA related to different gene expression patterns may be related to different modes of actions of both compounds. Medicinal herbs generally contain mixtures of active compounds, which may interact in an additive or synergistic manner. Synergistic interactions may need common mechanisms e.g., a common specific pathway that they inhibit. From an evolutionary point of view, synergistic interactions need co-evolutionary selection pressures to evolve. Hence, it can be speculated that synergisms are less likely to occur than additive effects. Therefore, additive drug interactions can be more frequently found in medicinal herbs. Compounds with different modes of action can efficiently and sufficiently fulfill the requirements for plants to survive under specific evolutionary selection pressure. This may explain that we found additive rather than synergistic interactions in isobologram analyses between curcumin and AA in the panel of cell lines tested. This observation is in accordance with previous data with several cytotoxic compounds from Artemisia annua L., which also showed additive rather than synergistic interactions (Efferth et al., 2011).

In summary, we have identified some genes which were downregulated by AA and curcumin. The genes may be responsible for cellular responsiveness of the varied cancer cells to AA and curcumin treatments. The cellular activities tackled by the downregulated genes include inhibition of cell proliferation and cell cycle activities, reduction in cellular metabolism, downregulation of anti-apoptotic gene products and inhibition of angiogenic cytokines. The two natural products seem to induce cytotoxicity by different mechanisms and this may lead to achieve tumor eradication in vivo. The varied cellular functionalities represented by the genes downregulated may support additive effects observed in the isobologram analyses.

Title selection and research design: TE, EO. Laboratory experiments, results generation, data analysis, and interpretation: EO, OK. Manuscript writing and submission: EO. Manuscript writing: HG. Advising in research design, results interpretation, and manuscript writing: TE.