A dynamic in-vitro model of the intestinal epithelium for the investigation of P-glycoprotein activity

Multidrug resistance is still an obstacle for chemotherapeutic treatments. One of the proteins involved in this phenomenon is the P-glycoprotein, P-gp, which is known to be responsible for the efflux of therapeutic substances from the cell cytoplasm. To date, the identification of a drug that can efficiently inhibit P-gp activity remains a challenge, nevertheless some studies have identified natural compounds suitable for that purpose. Amongst them, curcumin has shown an inhibitory effect on the protein in in vitro studies using Caco-2 cells. Aiming at understanding the role of physiological flow on the modulation of membrane protein activity, we studied the uptake of a P-gp substrate under static and dynamic conditions. Caco-2 cells were cultured in bioreactors and in Transwells and the basolateral transport of Rhodamine-123 assessed in the two systems as a function of P-gp activity. Experiments were performed with and without pre-treatment of the cells with an extract of curcumin or an arylmethyloxy-phenyl derivative to evaluate the inhibitory effect of the natural substance with respect to a synthetic compound. The results indicated that the P-gp activity of the cells cultured in the bioreactors was intrinsically lower, and that the effect of both natural and synthetic inhibitors was up modulated by the presence of flow. Our study underlies the fact that the use of more sophisticated and physiologically relevant in vitro models can bring new insights on the therapeutic effects of natural substances such as curcumin. P-gp modulating effects of rhizomes that the Rh-123 uptake a inhibiting

substances, such as chemotherapeutic agents, from cell cytoplasm. In fact, P-gp has been the main pharmacological target in combating MDR for several years (3).
In the human gastrointestinal tract, P-gp is found in high concentrations on the apical surfaces of superficial columnar epithelial cells of the colon and distal small bowel. Its expression appears to increase when the tissues pass into the tumor state, as occurs in the colorectal epithelium (4). Many attempts have been made to identify effective P-gp inhibitors, known as chemosensitizers or MDR modulators, which sensitize resistant cells to the action of cytotoxic drugs. The approaches used up to now have not returned significant results and have failed in clinical trials (1). The first attempts to modulate P-gp were carried out by testing calcium-channel blockers and antisteroids (first-generation drugs), however these compounds interfered with several enzyme systems (5). Later, second-and third-generation drugs were developed, but have failed in clinical trials mainly due to unpredictable pharmacokinetic interactions (6).
Alternatively, researchers have tried to find a valid P-gp inhibitor from substances of natural origin.
Curcuma Longa has been one of the most intensively studied plant extracts and its action on the intestinal glycoprotein has been investigated in static in vitro models. Figure 1 illustrates the inhibition of P-gp pumping activity in the presence of curcumin.
In a study using intestinal cell monolayers as a model of the gastro-intestinal tract, the authors found that curcumin down regulated the expression and function of intestinal P-gp (11). Specifically, curcumin at a concentration of 30 µM increased Rhodamine-123 (Rh-123) (a substrate of P-gp) accumulation by two-fold compared to the negative control and reduced efflux by 30% in the apical site. Ampasavate et al. investigated P-gp modulating effects of extracts prepared from rhizomes of Curcuma Longa and Curcuma sp. ''Khamin-oi''. They showed that the extract of Curcuma Longa increased Rh-123 uptake in a dose dependent manner, inhibiting the activity of intestinal P-gp. (10).
Both studies were performed using Caco-2 cells, cultured in a Transwell® system, which differentiate to form a polarized epithelial cell monolayer providing a physical and biochemical barrier to the passage of ions and small molecules (12,13). This cell line is widely used across the pharmaceutical industry as an in vitro model of the human small intestinal mucosa to predict the absorption of orally administered drugs (14).
In order to understand how the presence of flow could affect the modulation of P-gp, we studied the inhibition of intestinal P-gp by culturing Caco-2 cells in bioreactors. The systems we used combine a two-compartment cell culture device with apical and basal media flow and have been described by Giusti (Giusti et al. 2014) and Cacopardo (16). Both studies demonstrate that they are capable of better mimicking physiological barriers as they recapitulate the dynamic environment of the intestine.
Here we compared the apical efflux of a P-gp substrate after preconditioning with a Curcumin extract and a synthetic compound (an arylmethyloxy-phenyl derivative or APD (17)) which proved to inhibit Pgp activity in a [ 3 H]-vinblastine transport inhibition test with IC 50 of 0.19 µΜ. The experiments were performed in Transwells (traditional static system) and using bioreactors. To the best of our knowledge this is the first application of such devices to the investigation of P-gp inhibition by curcumin.

Methods
First, we performed a preliminary cytotoxicity test to determine a safe curcumin dose to be employed in the P-gp modulation study. Then, we set-up the culture of Caco-2 cells in the two systems for 21 days, with regular cell monitoring, including TEER and TEEI (Trans Epithelial Electrical Resistance and Impedance respectively). Finally, the P-gp activity in response to treatment with the two different compounds (Curcumin extract and the synthetic compound APD) was assessed through a transport study using a P-gp substrate: rhodamine-123 (Rh-123). Caco-2 cells (between passages 30 and 45) were seeded at a density of 25000 cells/cm 2 in bioreactors (through their tubing system) and PET 12-well-plate Transwell® (Corning, Italy). After cell adhesion, the apical chambers of the bioreactor were filled with 2 mL of medium and the basolateral ones with 1 mL of medium. In the Transwells, the apical and basolateral compartments were filled with 500 µL and 1 mL of medium, respectively. All the samples were incubated at 37 °C in a humidified atmosphere of 5% CO 2 for three weeks. In the static samples the medium was changed every three days.
After one week, the basal and apical circuits were filled up to a total volume of 10 mL of media respectively and kept in dynamic conditions (under flow) at a flow rate of 150 µL/min, which according to the manufacturer should provide an average shear stress of around 6 × 10 − 4 Pa on the membrane.
In this set-up, the medium was changed once a week: half the volume was removed and replaced by fresh medium.

Curcuma extract and APD compound preparation
The extract of Turmeric was taken from a common encapsulated supplement (Curcumina Santé, from Santé Naturels, Italy) containing 450 mg of pure curcumin, titled at 95%. It is regularly registered as a food supplement notified to the Italian Ministry of Health. A stock solution of curcumin (45 mg/mL) was diluted in DMSO and stored at − 20 °C. APD (kindly provided by the medicinal chemistry lab of Rapposelli Simona, Dept.of Pharmacy, UNIPI, Italy) was dissolved in Hank's Balanced Salt Solution (HBSS) and the stock solution (10 mM) was stored at -20 °C. This compound belongs to a new class of molecules that showed appreciable modulatory activity on P-gp and a high degree of selectivity (17)(18)(19). The authors studied P-gp inhibition activity by 3 combined biological assays among which there was the inhibition of P-gp mediated transport of vinblastine. This derivative was shown to compete with radiolabeled vinblastine (a well-known P-gp substrate) for the P-gp binding site, with an IC 50 value of 0.19 mM. (17).

Assessment of curcumin toxicity
The toxicity of curcumin was measured through a preliminary assessment of the metabolic activity of the cells, before and after incubation with the compound at different concentrations. Usually, TEER is measured applying a low-frequency (f < 5 kHz) current stimulus across the cellular barrier and recording the resulting voltage (16,20). In this study, it was monitored at 40 Hz with an integrated cellular impedance-meter (16,21), in the bioreactors, and at 12. Finally, transepithelial electric impedance (TEEI) spectra were acquired in the bioreactors within a frequency range of 40-10000 Hz, providing a more precise indication regarding the formation of a tight monolayer.

Assessment of P-gp Activity
On the twentieth day of culture, all the samples were rinsed with PBS and then incubated with a curcumin solution (50 µg/mL) that was added to the apical compartments through the mixing chambers in the bioreactors (10 mL) and directly in the apical compartment in the Transwells

Data analysis
The analysis of the effect of curcumin or APD on P-gp activity is based on Rh-123 efflux from the basolateral side to the apical side of planar cell cultures. Mathematically, P-gp activity is expressed as a function of Rh-123 mass (µg) on both apical and basal sides at 120 min (t1 apical and t1 basal , respectively), as well as Rh-123 initially administered to the experimental set-up (t0 basal ). All these parameters were normalized with respect to the blanks (i.e., measurements without cells) and combined into the equation as follows: To appreciate how much the treatment with the substances (curcumin vs APD) inhibited the activity of the protein, the following equation was used: Statistics based on two-way ANOVA analysis was performed using the Tukey's test for multiple comparisons and setting statistical significance at p < 0.05, with GraphPad Prism (GraphPad Software, San Diego, CA, USA). Error bars in results section represent standard deviations (n = 6).

Assessment of toxicity of curcumin
The cytotoxicity of curcumin on Caco-2 cells was assessed by incubating the cells with different concentrations of curcumin. The cells were grown in a 24-well-plate for 7 days and their metabolic activity was determined by performing the Alamar blue assay, before and after 48h incubation with the different curcumin extracts. The results ( Figure 4A) showed that the metabolic activity of the cells was not substantially affected by curcumin solutions at the concentration range 45 -11.25 µM. Only for the highest concentration tested, 90 µM, did the cell metabolic activity decrease down to roughly 70%. A 50 µM curcumin solution was within the non-toxic range, so it was used for the P-gp activity modulation experiments.

Cell monitoring and morphological analysis
After the seeding in the bioreactors and in Transwells, Caco-2 cells were regularly monitored over a 21 day differentiation period, using an optical microscope and by measuring the TEER and TEEI.
The cells progressively formed a compact monolayer, exhibiting features of differentiated cells after 10-12 days, as shown by the images in Figure 5.
The formation of a compact monolayer was also verified by the TEER and TEEI analysis. As shown in Figure 4B, TEER increased during the culture reaching a plateau at day 14 (~ 800 Ω.cm 2 in the Transwells and ~ 1000 Ω.cm 2 in the bioreactors), which was maintained until day 21. Moreover, TEEI trends ( Figure 4C), measured in the bioreactors, suggest the presence of an integral cell layer at all the time points (16).
After the P-gp modulation experiment, the cells were fixed and the actin fibers and the tight junction protein occludin were stained. The monolayer was maintained intact during the transport assay and the cells preserved their tight junctions, as pointed out by the presence of occludin ( Figure 6A).

Assessment of the activity of P-gp
We measured P-gp activity as a function of Rh-123 efflux from the basal side to the apical side of cells cultured in the Transwell and in the bioreactors. Additionally, we assessed the modulation of curcumin and APD (i.e., positive control) by their inhibition effect.
Statistical analysis showed that in the absence of inhibitors P-gp activity observed in the bioreactors is lower than that measured in static conditions ( Figure 4D). In the presence of natural and synthetic inhibitors, P-gp activity is further reduced in the bioreactors with respect to the Transwells ( Figure   4E). Moreover, in both the static and dynamic conditions the inhibition of P-gp activity by curcumin is higher than that observed in the positive controls demonstrating that the natural substance is able to more markedly affect the Rh-123 efflux than the selective P-gp modulator APD.

Discussion
In the preliminary cytotoxicity test, we observed a decrease of Caco-2 metabolic activity (< 20%) at the maximum curcumin concentration tested (90 µM). Therefore, a non-toxic concentration of 50 µM curcumin was used for the P-gp modulation experiment. Similar concentrations have been employed in other reports in the literature (10,24,25).

Caco-2 cells grown in static Transwells and in bioreactors formed tight and well differentiated
monolayers. It is worth noticing that the expansion of cells in the monolayer did not occur on a single plane, but gave rise to domes, as shown in Fig. 5B and 5C, resembling the natural arrangement of cells in the intestinal environment (26). Both in the Transwells and in the bioreactors, it was possible to observe the progressive formation of the monolayer, which, as illustrated in Fig. 4B, came to confluence in about seven days (TEER values higher than 600 Ω.cm 2 ) reaching a maximum at day 14 and 21 (~ 800 Ω.cm 2 in the Transwells and ~ 1000 Ω.cm 2 in the bioreactors). The integrity of the monolayer in the bioreactors was confirmed by the TEEI measurements (Fig. 4C), which follow a typical RC circuit trend. Indeed, the cell monolayer can be represented by an equivalent electric circuit composed of the resistive paracellular pathway and cell capacity (transcellular pathway) in parallel. At low frequency (f < 5 kHz), the capacitor is fully charged and current cannot flow across it, thus ions can only flow paracellulary i.e. through the gaps between cells. At higher frequency, the capacitor becomes progressively more conductive and, in the presence of a compact cell layer with well-established tight junctions, the current is able to flow across the cells (transcellular current).
Moreover, the higher TEER values in the bioreactors suggest that the fluidic conditions improve the barrier's integrity and tightness, as reported in (16).
The presence of tight junctions was also confirmed by occluding staining, after 21 days in culture and after the passage test. Occludin is a protein that protrudes on the outer face of the membrane and mediates cell to cell adhesion. It has also been linked with the regulation of intermembrane and paracellular diffusion of small molecules (27). Furthermore, as demonstrated in Fig. 6, the P-gp activity assessment experiment did not interfere with monolayer integrity and tightness.
The effect of curcumin on P-gp was evaluated through the efflux of Rhodamine-123 from the basal side to the apical side of the cells. Since P-gp is a membrane glycoprotein, expressed especially on the apical side of cells, the presence of Rhodamine-123 on the apical side can be correlated with its activity (28)(4).
Data collected to date on APD shows that it inhibits the activity of P-gp at even lower doses than that used in this study (100 µM) ). Interestingly, we observed that the inhibition of Pgp by APD was surpassed by the inhibitory effect of curcumin (Fig. 4E) in both static and dynamic conditions. As curcumin may also interact with other ABC-transporters expressed in Caco-2 cells (such as MRP2 and BCRP/ABCG2 (29,30)) for which APD does not show any activity  These differences could be due to mechanotransduction effects arising from shear stress, which are known to modulate P-gp activity (31). For instance, a study by Garcia-Polite et al, demonstrated that the protein activity had a peak when cells were subjected to a shear stress of 1 Pa and that it decreased for a shear stress of 4 Pa (32). Interestingly, there is a contrasting hypothesis that P-gp was down regulated in arterioles of rat brain, in response to a higher shear stress (33,34). The shear stress provided by the flow in the present study is below the mentioned ranges and suggests that these cells are highly sensitive to even low levels of mechanical stimuli. It should however be noted that the analysis of the effects of increased shear stress are very difficult to decouple from those due to increased nutrient supply (35).
Remarkably, although the P-gp activity of the cells in the bioreactors was intrinsically lower, the activity inhibition efficiency, by both the ADP and curcumin, was higher in the bioreactors. This suggests that the presence of a flow not only influences P-gp by itself but also affects the action of external modulators. These contrasting events could be due to the effect of flow on the binding between substrate (Rh-123) and the protein and/or on the inhibition mechanism by both the Arylmethyloxy-phenyl derivative and curcumin.
We demonstrated that the culturing the cells and performing the efflux test under more physiological conditions has an impact on P-gp activity. At this point, further research is necessary to identify the precise mechanism by which P-gp modulation by curcumin was different in the dynamic system. It should be noted that the pharmacokinetic properties of curcumin in humans are not completely elucidated. Studies performed on mice suggest that curcumin is promptly metabolized by conjugation and reduction, resulting in poor systemic bioavailability (36). The proposed dynamic model constitutes a more faithful representation of the microenvironment of the human intestinal epithelium in vitro, enabling better translation between in vivo and in vitro studies.

Conclusions
We show that the presence of a physiological microenvironment can affect how P-gp is modulated in vitro. Our results indicate that the inhibition of the protein by the action of curcumin is higher using a dynamic system compared with traditional static models. The potential therapeutic application of natural substances is of interest both in drug development and nutraceuticals, and it is crucial to perform in vitro studies using conditions which replicate the in vivo environment as faithfully as possible. As such, the present study is the first evidence that the use of more sophisticated tools can bring new insights on the therapeutic effects of natural substances such as curcumin.       Caco-2 monolayer cultured in the bioreactor after the P-gp activity assay. Fluorescence staining of the nuclei, actin microfilaments (A) and occludin (B).