3D Fluid-Dynamic Ovarian Cancer Model Resembling Systemic Drug Administration for Efficacy Assay

Recently, 3D in vitro cancer models have become important alternatives to animal tests for establishing the efficacy of anticancer treatments. In this work, 3D SKOV-3 cell-laden alginate hydrogels were established as ovarian tumor models and cultured within a fluid-dynamic bioreactor (MIVO®) device able to mimic the capillary flow dynamics feeding the tumor. Cisplatin efficacy tests were performed within the device over time and compared with (i) the in vitro culture under static conditions and (ii) a xenograft mouse model with SKOV-3 cells, by monitoring and measuring cell proliferation or tumor regression, respectively, over time. After one week of treatment with 10 μM cisplatin, viability of cells within the 3D hydrogels cultured under static conditions remained above 80%. In contrast, the viability of cells within the 3D hydrogels cultured within dynamic MIVO® decreased by up to 50%, and very few proliferating Ki67-positive cells were observed through immunostaining. Analysis of drug diffusion, confirmed by computational analysis, explained that these results are due to different cisplatin diffusion mechanisms in the two culture conditions. Interestingly, the outcome of the drug efficacy test in the xenograft model was about 44% of tumor regression after 5 weeks, as predicted in a shorter time in the fluiddynamic in vitro tests carried out in the MIVO® device. These results indicate that the in vivo-like dynamic environment provided by the MIVO® device allows to better model the 3D tumor environment and predict in vivo drug efficacy than a static in vitro model. This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 International license (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution and reproduction in any medium, provided the original work is appropriately cited.


Introduction
Ovarian cancer is one of the main causes of death in female cancer patients and has one of the highest gynecological mortality rates (Sankaranarayanan and Ferlay, 2006). The poor survival rate is mainly due to chemoresistance to established drug protocols, as also happens in many other cancer cases (Lowe et al., 2013). In this context, the increasing prevalence of drugresistant cancers necessitates further research and treatment development. Currently, an anticancer drug candidate that enters Phase I trials will successfully proceed further with a probability of only 8%, highlighting the urgent need for new physiologically relevant in vitro tumor models better resembling the in vivo conditions to test novel drugs and therapies (Suggitt and Bibby, 2005). Based on current regulatory guidelines, the screening of new cancer drugs is carried out by using high-throughput assays, where in vitro toxicity and efficacy tests are performed on cells grown as monolayers over planar plastic surfaces; then, in the preclinical development, in vivo toxicological and ADME (adsorption, distribution, metabolism, excretion) studies are performed in animal models. However, it is now widely demonstrated that 2D cell cultures are oversimplified and poorly resemble the complex 3D tumor microenvironme nt (Abbott, 2003;Loessner et al., 2010;Marrella et al., 2019). On the other hand, animal models commonly fail to predict human safety and efficacy in clinical studies, besides being expensive and associated with ethical issues (Liu et al., 2013). Therefore, in the last years, novel 3D human in vitro culture systems have increasingly gained attention as potential compromises between traditional 2D cultures and in vivo models (Hoarau-Véchot et al., 2018). They aim to combine the advantages of the former (better control of the experimental conditions, relative ease of manipulation and analysis, species-specificity) and approach the latter by better representing in vivo physiology.
In this scenario, 3D tumor spheroids have been proposed as in vitro human cancer models (Raghavan et al., 2015;Herter et al., 2017). Spheroids are scaffold-free aggregations of cells suitable for prolonged in vitro culture and high-throughput drug testing. They have been shown to resemble many physiological aspects better than cells grown in monolayers (e.g. cells were incubated in a humidified, 5% CO2 atmosphere at 37° C. Medium was changed 2 days after the original plating and then twice a week. When culture dishes were nearly confluent, cells were detached with 1X trypsin (EuroClone) after two washes in D-PBS 1X and replated until the next confluence. After two passages, cells were used for the in vitro and in vivo experiments.
Alginate (Alg) powder (Manugel GMB, FMC Biopolymer) was dissolved in physiologic solution (0.9% NaCl solution) at 1% w/v and the solution was then filtered under sterile conditions. SKOV-3 were detached from plastic tissue culture flasks with 0.05% trypsin and resuspended in DMEM supplemented with 10% FBS and 1% penicillin/streptomycin. The SKOV-3 suspension then was mixed with the sterile Alg solution to obtain a final Alg concentration of 0.5% w/v. The SKOV-3/Alg suspension was dripped into a sterile 0.5M CaCl2 (Sigma Aldrich) gelling bath to form alginate spheres with a final concentration of cells of 1.3 × 10 6 cells/mL.
After washing the spheres with DI water to remove the excess of Ca, the OCM were gently moved into 6 well-plates and cultured with 4.5 ml DMEM supplemented with 10% FBS, 1% penicillin/streptomycin and CaCl2 (5 mM) in a humidified environment (5% CO2) at 37°C.

Ovarian cancer model viability and proliferation
SKOV-3 viability within OCM was evaluated qualitatively through a live/dead assay (Sigma Aldrich). Briefly, after 24 h, OCM were washed with PBS and incubated in 2 mM calcein-AM and in 4 mM EthD-1 in PBS for 15 min at 37°C in a dark environment to detect live and dead cells, respectively. OCM were washed three times in PBS and then observed by means of fluorescence microscopy (Nikon H550L).
SKOV-3 proliferation within tumor models was quantitatively assessed by Alamar Blue Assay (Thermo Fisher Scientific). SKOV-3 cultured in monolayers were used as 2D control. In particular, 15 × 10 3 cells were cultured over a glass slide placed in a 6-well plate with the same volume of medium as the 3D models (4.5 ml). For 3D proliferation rate analysis, after 3 hours (T0), 2 (T2), 4 (T4), 7 days (T7) of culture, the OCM were placed in 96 well-plates containing 0.2 ml of 1% v/v of Alamar Blue solution, as indicated by the manufacturer. Samples were incubated at 37°C for 4 h in the dark. The supernatants containing Alamar blue were collected and absorbance readings assessed spectrophotometrically. A calibration curve was derived by seeding known number of cells in 96-well plates to find the correspondence between the number of cells and the output of the absorbance readings. The proliferation rate was calculated as the ratio between the number of cells detected at each time point respect the number of cells after 3 hours of culture (T0). Then, OCM were washed with physiologic solution and placed in 6-well plates containing the original culture medium. (N=3 biological replicates; n= 2 technical replicates)

In vitro drug efficacy tests
In vitro drug efficacy tests were performed by using a compartmental fluidic device (commercialized as MIVO ® by React4life S.r.l., IT). The system design is shown schematically in Figure 1.
24-well Transwell inserts (Corning) containing an OCM were placed and cultured within the bioreactor, forming two fluidically independent chambers: the tissue culture chamber, which was filled with culture medium (0.3 ml), and the circulatory chamber, connected to a closed loop fluidic circuit containing 4.2 ml medium where cisplatin (purchased from Sigma Aldrich srl.) was allowed to circulate at a rate of 0.3 cm/s, simulating the capillary flow rate. Four hours after their formation, OCM were placed into the bioreactor chamber and cisplatin was added into the bioreactor circuit connected with the receiver chamber ( Fig. 1).
3D hydrogels cultured within 6-well plates with 4.5 ml medium with or without cisplatin were used as static controls. Cell viability of cisplatin-treated SKOV-3 was assessed quantitatively by Alamar Blue Assay at different time points. Briefly, the samples were placed in 96-well plates and incubated with fresh medium containing 0.2 ml of 1% v/v Alamar Blue solution, at 37°C for 4 h in the dark. Cell viability was derived as % of live cells normalized to the untreated controls. Student's paired t-test between each dynamic condition and the respective static one was performed for each time point and statistical significance was set at *P < 0.05, (N=3 biological replicates; n= 2 technical replicates).

In vivo xenograft model
A total of twelve 5-week-old female nude mice (Mice Hsd: Athymic Nude Foxn 1 nu female) were purchased from Envigo RMS srl, San Pietro al Natisone, Italy. The mice had a body weight of 21.1/ 20.1-24.3 g (median/interquartile range) at the beginning of the experiment.
Animals were delivered to the animal facility 10 days prior to the beginning of the study for acclimatization. The mice were housed in Sealsafe Plus GM500 plastic cages (Tecniplast Spa, Buguggiate, Italy) with a light dark cycle of 12h/12h at a temperature of 21 ± 2°C (dawn: 6:30-7:00 am) and a relative humidity of 60 ± 20%. Food (pellets, 10 mm, 2018 Envigo RMS Srl, San Pietro al Natisone, Italy) and sterilized water were provided ad libitum. Enrichment was supplied by Mouse house (Tecniplast Spa, Buguggiate, Italy). The animals in each group were divided into 3 animals per cage. Cages were clearly labelled with an ID card indicating study number, group, gender and treatment schedule. All animals were subjected to the same environmental conditions. The study was carried out according to the guidelines enforced in Italy and in compliance with the Guide for the Care and Use of Laboratory Animals, 8 th Edition, 2011.
SKOV-3 derived tumors were established via subcutaneous injection of 1 x 10 6 cells into the right flank of mice. The cells were resuspended in phosphate buffered saline (PBS) for s.c. injection. Tumor size was monitored over time. After 10 days, when tumor volumes had increased to 50 mm 3 , mice were randomized into two treatment groups, i.e., a control group (sham treated, N=6) and an experimental group (treated, N=6). The mice were administered PBS (control group) or cisplatin (6 mg/kg) intravenously once every seven days for 3 weeks. Tumor growth was quantified three times a week using a digital caliper. The tumor volume was calculated as follows: 0.5 x length x width 2 .
The results are expressed as tumor growth inhibition (%TGI), which was calculated as the percentage of reduction of tumor volume compared to the control: For comparison with the in vitro cell viability data, the reciprocal trendline %TGI values (100-%TGI) were used.
The effect size f and total sample size were calculated using the G*Power software. The input data for the analysis were extrapolated from the published literature (Faul et al., 2007).
The in vivo experiments were authorized by the Ministry of Health for in vivo studies and by the Body for the Protection of Animals (OPBA).

Immunostaining
Hydrogels were fixed with 4% para-formaldehyde in PBS (PFA; pH 7.4) for 1 h and incubated for 1 h in blocking buffer (0.5% Triton X-100, BSA 2% w/v, CaCl2 5mM in physiologic saline solution). Subsequently, hydrogels were incubated with primary antibody for 1 h at room temperature. Cell proliferation was detected by staining cells with a rabbit anti-Ki67 antibody (Abcam, USA; 1:400 dilution in blocking buffer); while apoptotic cells were stained by using rabbit Anti-Cleaved Caspase-3 antibody (Abcam, USA; 1:100 dilution in blocking buffer). Samples were then washed three times with PBS and incubated for 1 h with Alexa Fluor 488-conjugated goat antirabbit secondary antibody (Abcam, USA; 1:200 dilution in blocking buffer) for anti-ki67 and with Alexa Fluor 555-conjugated goat antirabbit secondary antibody (Abcam, USA; 1:200 dilution in blocking buffer) for anti-Caspase-3 . Nuclei were counter-labeled with DAPI (Sigma-Aldrich). Imaging was performed with a fluorescence microscope (Nikon H550L). Images obtained by fluorescence microscopy were analyzed using ImageJ software.

Drug diffusion within the 3D tumor hydrogel
Drug diffusion into the 3D tumor hydrogel was determined via HPLC. Alginate hydrogels were cultured either under static conditions or within the MIVO ® for 7 days with cisplatin 10 µM. The hydrogels were then left in the incubator for 3 days in 5 mM CaCl2 physiologic solution to allow the cisplatin to diffuse from the hydrogel to the solution. Then the samples were analyzed by HPLC. The HPLC system consisted of a pump, column compartment and RS variable wavelength detector (all UltiMate 3000, Thermo Fisher Scientific). The injection valve fitted with a 20 µL sample loop and an Accucore 150-C18 (Dimensions = 150 x 3 cm and particle size 2.6µm) was purchased from Thermo Fisher Scientific. The mobile phase consisted of methanol-water (80:20, v/v). The UV detector was adjusted at 254 nm. The flow rate was set at 0.2 mL/min (isocratic flux) and the column temperature at 40°C (Kaushik et al., 2010;Tezcan et al., 2013) (N=3 biological replicates).

Computational fluid-dynamic (CFD) simulations
Fluid dynamic and mass transport simulations were performed both in static and dynamic conditions to simulate the concentration of cisplatin within the cell-laden hydrogels over time.
The 3D domain, the related size and dimensions were calculated based on the real dimensions of the microfluidic circuit used during the test. As shown in Figure S1 1 , Domain 1 represents the circulatory chamber of the bioreactor and the fluidic pattern of the circuit, Domain 2 is the tissue culture chamber, which is represented by a well filled with medium, and Domain 3 represents the alginate-based hydrogel sphere. The obtained geometry, the fluid dynamics within the circuit, and the mass transport of cisplatin through the entire system was modeled using Comsol Multiphysics 5.3a. For the numerical solution of the physics involved in this system, the Laminar Fluid Flow module and the Transport of Diluted Species module were used.
The first physical phenomenon involved in this system was represented by the fluid dynamics from the circulatory chamber of the bioreactor (i.e. Domain 1) to the tissue culture chamber (Domain 2) where the cell-laden hydrogel is cultured (Domain 3). The fluid was supposed to be laminar, incompressible, and not turbulent. The velocity and pressure field profiles were calculated according to Navier-Stokes and the continuity equation (Equation 1): where , and are the velocity, the density (10 3 / 3 ) and the viscosity (10 −3 • ) of the fluid, and is the pressure, respectively. The flow rate was set to = 3 / to generate velocity resembling the capillary blood flow. An iterative geometric multigrid (GMRES) algorithm was used to solve the equations. Discretization was chosen P2+P1 for velocity and pressure field, respectively. In the outlet, the pressure was set equal to zero with no backflow. A no-slip condition was fixed on the boundary of the geometry. As the initial value, the velocity was set equal to zero in the entire system.
The general mass transport equation was used to describe cisplatin mass transport through the system: where is the cisplatin concentration in the system, is the diffusivity of cisplatin, is the velocity field, and is the reaction term. The diffusivity of cisplatin in the medium (approximated as water) at 37°C ( − ), included in the equation for domains 1 and 2, is set equal to 1.034 * 10 −5 ²/ , as reported in the literature (Modok et al., 2007;Panczyk et al., 2013). On the contrary, the diffusion constant of cisplatin within the alginate hydrogels ( ℎ − ) was calculated by using the empirical Ogston model, as shown in equation 3. The model takes into account several parameters including the polymer volume fraction , the solute (cisplatin) radius , the alginate fiber radius and the diffusion value of cisplatin in water − . In particular, the value was calculated using the following expression: where the polymer volume fraction was approximated to 0.01 while the radius of cisplatin and alginate fiber were 4 • 10 −10 and 8 • 10 −10 , respectively (Amsden, 1998;Modok et al., 2007). Moreover, it should be considered that the diffusivity value of cisplatin in the alginate hydrogels is affected by the presence of dead cells, which decrease the mass transport of cisplatin within the polymer.
Firstly, the Michelis-Menten parameters were calculated by considering that 50.000 molecules of cisplatin are enough to kill one cell (Amsden, 1998), as reported in the following reaction (equation 4): Based on the best fitting of the experimental data ( Fig. S3 1 ), the obtained values were = 1.66 • 10 −12 [ * −3 * −1 ] and = 6.64 • 10 −3 [ * −3 ]. In this case, the Ogston model considers an added corrective term to describe the mass transport in the system. Therefore, the diffusion of cisplatin within the hydrogel could be described by the following equation: where − is the diffusivity of cisplatin in water, is the ratio calculated according to the Ogston model (equal to 0.86 in (3)), and the exponential term defines the decrease of the diffusion during the time due to the presence of dead cells.
Parameters and (Tab. 1) depend on the velocity field (either in static or dynamic conditions) and were calculated considering the amount of cisplatin within the alginate hydrogels experimentally measured through HPLC analysis after 7 days of culture.

Tab. 1: k and b values for the cisplatin diffusivity in static and dynamic conditions
Static The initial concentration of cisplatin ( 0 ) in the bioreactor was set at 100 µM or 10 µM.
Danckwerts conditions were selected in the inlet, while in correspondence of the boundary surfaces no-flux condition was considered. For the outlet, the diffusion term was considered equal to zero. A direct backward differentiation formula (BDF) algorithm was required for the transient study. Linear discretization was chosen for the concentration field. The reaction term was defined according to the Michaelis-Menten kinetics: where is the maximum consumption rate and is the concentration of cisplatin when the rate is equal to /2. The reaction term takes into account the deactivation of cisplatin molecules bound to the DNA of cancer cells and no longer available during the experimental tests. This term was considered as a consumption rate of cisplatin.
The average concentration profile of cisplatin within the hydrogels was calculated as follows:

3D cancer cell viability and proliferation
Firstly, ovarian cancer cell viability and growth within the OCM were investigated and visualized by calcein-AM staining. Four h after OCM generation (day 0), most of the cells were alive, indicating the suitability of the procedure for cellular embedment within the alginate hydrogels. Cell viability was evaluated also after 7 days of culture. A higher cell density was observed at day 7, proving the cells' ability to proliferate within the hydrogels ( Fig. 2A).
To quantify this result, the cell proliferation rate was measured by Alamar Blue assay after 2, 4 and 7 days of culture. Figure 2B shows that cancer cells cultured within the OCM proliferate more slowly than cells grown in 2D monolayers, in agreement with other reports (Chitcholtan et al., 2013).

In vitro drug efficacy test
The viability of SKOV-3 OCM cultured within MIVO ® and treated with cisplatin (10 M or 100 M) was measured over time. In particular, while in static conditions the drug was directly added to the medium surrounding the OCM, in dynamic conditions (i.e. MIVO ® ) it was injected into the fluidic circuit beneath the membrane from which it could reach the tumor tissue cultured in the upper chamber of the bioreactor by diffusion, resembling extravasation of the drug.
After two days of culture in the presence of 100 M cisplatin, cell viability in static and in dynamic conditions was significantly reduced and was even lower after 4 and 7 days with no significant differences observed between static and dynamic conditions (Fig. 3).
Viability was weakly but not significantly reduced after two days culture in the presence of 10 M cisplatin under static conditions. Interestingly, significantly decreased cell viability to 67.81% ±0.62 and 50.44% ±0.25 after 4 and 7 days respectively at 10 µM cisplatin was observed only under dynamic conditions within MIVO ® . OCM cultured under static conditions remained above 80% cell viability for the overall time of observation (Fig. 3).

In vivo drug efficacy test
The in vivo efficacy of cisplatin against ovarian cancer SKOV-3 cells was evaluated in a xenograft model. Treatment of six nude mice with 6 mg/kg cisplatin versus six control mice started 10 days after tumor induction, defined as Day 0. The experiment was stopped on Day 35 when the tumor volume in the control group reached 2000 ± 270 mm 3 and was more than double compared to the treated group (see Fig. S2 1 for tumor volume data). Figure 4A shows the inhibitory effect of cisplatin on tumor growth over time. The %TGI was 7.3%, 31.1% and 56.9% on Days 12, 21 and 35, respectively, as shown in Fig. 4A.
Interestingly, comparing the endpoints of the in vitro (static treatment), in MIVO ® (dynamic treatment) and in vivo (xenograft model), an excellent overlap of drug efficacy data is observed only between MIVO ® and in vivo data, although on a different time scale (Fig. 4B); comparable results are obtained after 2,4 and 7 days of treatment in MIVO ® and after 2, 11 and 25 days of treatment in mice. Due to the differences in drug distribution and metabolism (ADME profiles) between these two different experimental conditions, the time needs to be rescaled to obtain comparable results (~3 times faster in MIVO®). On the contrary, static in vitro results do not resemble data obtained in vivo.

CFD simulations and mass transport within MIVO ®
The concentration fields of 10 µM cisplatin within the OCM cultured in static and dynamic conditions are shown in Figure 6 based on the geometry and the model set-up defined in Materials and Methods.
After 12 h, the amount of cisplatin diffused within the alginate spheres seemed to be higher in the static than in the dynamic condition, i.e., values of 0.0053 mol/m 3 and 0.00707 mol/m 3 were detected for the dynamic and the static conditions, respectively ( Fig. 5A-B). Interestingly, after 4 days of culture the trend was opposite, with higher amounts of cisplatin in dynamic conditions than in static ( Fig. 5C-D). The same results were also obtained after 7 days. Specifically, during the dynamic culture, the alginate hydrogel was completely filled with cisplatin, while in static conditions concentration gradients through the sphere were still evident ( Fig. 5E-F).
Simulations were performed at the two concentrations tested in vitro (10 M and 100 M). No main differences between the amount of cisplatin simulated in the model and that experimentally measured were detected (Fig. 6), proving the reliability of the mathematical model.
The average concentration profiles of cisplatin within the hydrogels showed a significant difference among the different cases tested over 7 days, as shown in Figure 7.
The diffusion of the drug within the hydrogels in static conditions was faster at short time scales at both concentrations of 100 M and 10 M than in dynamic conditions. It is likely that when the hydrogel was placed in the well plate dipped in the cisplatin solution, the concentration gradients of the drug were high enough to allow a fast diffusion through the hydrogel. It should be noted that in static conditions, the corresponding convection term is missing.
On the contrary, in dynamic conditions the cisplatin concentration within the hydrogels increased slowly and reached the same concentration of the static condition after 18 hours both at 100 and 10 M. Under these conditions, the drug needed to move from domain 1 to domain 2, finally reaching domain 3 (Fig. S1 1 ). However, the concentration of cisplatin within the hydrogels became higher than in static conditions after one day of culture. After 4 days of culture in dynamic conditions, the cisplatin concentration reached a plateau, and a clear difference of cisplatin concentration between the two culture conditions was detected at both tested concentrations (i.e.100 and 10 M), as shown in Figure 7.
In summary, in static conditions, the absence of fluid motion generates a greater cisplatin resistance within the hydrogel due to the accumulation of dead cells, which limit the mass transport within the polymer. In contrast, in dynamic conditions, the flow allows a continuous removal of dead cells, thus leading to less hydrogel resistance to the cisplatin diffusion. days. While dynamic condition shows a complete drug distribution within the 3D volume after 7 days of culture, under static condition a strong gradient is still present.  For each drug concentration (10 or 100 µM) the dynamic condition shows a slower diffusion over time that reaches an higher plateau value.

Immunostaining
Immunostaining of OCM collected at different time points was carried out to determine the expression of Ki67 to identify proliferating cells, and of Caspase-3 to mark the cells undergoing apoptosis in response to treatment with 10 M and 100 M cisplatin ( Fig. 8 and Fig. S4 1 respectively).
Ki67 positive cells were detected in untreated OCM cultured in static (Fig. 8, Fig. S4) or dynamic conditions (Fig.  S5A 1 ) after 2 days of culture and numbers increased further after 7 days.
Interestingly, cell-laden hydrogels treated with 10 M cisplatin under dynamic conditions displayed an overall positive staining for proliferating cells comparable to untreated hydrogels after 2 days, however the Ki67 staining was drastically reduced after 7 days (Fig. 8A). This behavior was even more evident for the high dose of the drug (i.e. 100 M) (Fig. S4 1 ). On the other side, when the tumor tissue was treated under static conditions, there was less Ki67 staining than in untreated hydrogels or under dynamic conditions after 2 days and Ki67 staining was still well evident after 7 days of 10M ALTEX preprint published August 3, 2020 doi:10.14573/altex.2003131

Fig. 8: Fluorescence images showing immunostaining of Ki67 (green) as index of proliferation and Caspase-3 (red) as marker of apoptosis of SKOV-3 cultured within alginate hydrogels treated with 10 µM cisplatin in static or dynamic conditions (MIVO). The untreated controls were cultured in static conditions.
Cells were stained after 2 or 7 days and counter-labeled with DAPI (blue). Scale bar is 500 µm. (N=3 biological replicates; n= 2 technical replicates). drug treatment, especially in the inner part of the hydrogel (Fig. 8A). When the treatment was carried out at a higher drug concentration, Ki67 staining was reduced compared to the 10M dose (Fig. S4 1 ). These data are in line with the cell viability reduction shown in Figure 3.
The expression of caspase-3 was weak in the untreated control after 2 days and almost undetectable after 7 days in static and dynamic conditions (Fig. 8, Fig. S4 and Fig. S5A 1 ). Positive caspase staining was observed in both static and dynamic conditions after 2 days of 10 M drug treatment (Fig. 8B); this was much more evident for samples treated with 100 M (Fig.  S4 1 ); interestingly, after 7 days of treatment within the MIVO ® device, cell-laden hydrogels displayed a homogeneous spatial distribution of apoptotic cells, while samples treated in static conditions displayed caspase-positive staining mainly in the outer ALTEX preprint published August 3, 2020 doi:10.14573/altex.2003131 rim of the OCM, indicating an accumulation of dead cells that could form a physical barrier hindering cisplatin diffusion (Fig.  8B). Again, this was much more evident for samples treated with 100 M (Fig. S4 1 ).

Discussion
The ability to rapidly and efficiently screen drugs with a more accurate preclinical tumor model is of great importance in drug development, because currently used assays still have severe limitations and poor predictivity. Considering the limitations of animal experiments to predict human response and the high need for novel drugs, it is necessary to include highly reproducible human systemic tumor models in preclinical analyses to validate more accurately the efficiency of drug candidates.
Recently, regulatory authorities have voiced a common desire to standardize the preclinical tests of drugs by replacing/reducing animal experimentation during early product development with specific in vitro systems, summarized in Table 2 (Daniel et al., 2018). This can be achieved only by strictly linking academic research outcomes with key industry entities to foster the implementation and optimization of relevant in vitro models on a large scale. The production of scientific data supporting the high reliability of alternative animal models is the first step to support industries in reducing time and costs of preclinical research. In our previous work, a fluidic device was adopted to culture a breast cancer tumor model in vitro, resembling some crucial steps of tumor growth: cell migration within the 3D hydrogel, cell evasion from the hydrogel and their intravasation into the fluid circuit (Cavo et al., 2018). Likewise, oral administration and the subsequent intestinal passage of other kinds of molecules and drugs can be modeled with the MIVO ® chamber, as already reported (Marrella et al., 2020). Starting from these promising results, we have here combined the use of a 3D cell-laden hydrogel as an ovarian tumor model of clinically relevant size with the MIVO ® fluidic device to resemble the human circulation and drug extravasation to reach the tumor mass. To test this technological approach, a drug efficacy assay was carried out as a proof-of-principle in parallel to a xenograft model, which represents the current gold standard. Moreover, the 3D tumor models were also cultured under static conditions, resembling the traditional assays using organoids/spheroids.
In order to further decrease the use of animals in pre-clinical research, serum-free medium can be a valuable option to perform in vitro cell culture. Along this line, also the use of recombinant antibodies can be desirable, respect to animalderived ones. In this work, the bovine-derived serum and animal-derived antibodies were used to validate this novel technological approach by comparing these results with data present in the literature and internal to our laboratory practice. In the next future we will move towards the evaluation of the use of alternative animal-free strategies for cellular assays, accordingly with the 3R principles.
Here the ovarian tumor was selected since it has the highest mortality rate of all the gynecological cancers worldwide (Siegel et al., 2012). However, this approach can be extended to many other solid tumors. SKOV-3 were embedded within 3D hydrogels or injected into mice, as previously reported (Cavo et al., 2018;Marrella et al., 2019).
The tumor model is composed of a cell-laden hydrogel. Among polymers, alginate was selected for its well-known advantages like inertness, chemical stability and no intrinsic bioactivity (Khurana and Godugu, 2018), to better focus on the drug penetration mechanisms and compare the in vitro model with the scaffold-free xenograft model. Further experiments could investigate increasing the level of complexity of the tissue model by combining different biopolymers and by incorporating multiple cell types associated with the ovarian tumor niche (e.g. fibroblasts, myofibroblasts, pericytes, vascular or lymphatic endothelial cells, and undifferentiated mesenchymal stem cells) to better model tumor microenvironment, which ALTEX preprint published August 3, 2020 doi: 10.14573/altex.2003131 11 is in fact a complex, heterogeneous and multi-cellular environment involving dynamic interactions between malignant cells and their surrounding stroma, including both cellular and acellular components.
Cisplatin was selected since it represents, together with its analogs, the first-line chemotherapeutic agent for the treatment of human ovarian cancer, exerting its cytotoxicity by forming DNA-links, which trigger cell apoptosis (Tiwari et al., 2005). The drug concentration tested in vivo (6 mg/kg) was adopted because it represents the "maximum tolerated dose", i.e. the most efficient dose without toxic effect (Aston et al., 2017). Two concentrations were employed in vitro to investigate a concentration-dependent cytotoxic effect and also to enable a comparison with in vivo data. For the higher drug concentration tested (100 M), we observed that most of the cells died after a few days of culture both in the static and the dynamic conditions indicating excessive toxicity. Differently, with the lower drug concentration (10 M), the decrease in tumor cell proliferation was greater in the dynamic conditions than in the static ones. Moreover, this drug concentration, which is commonly used to perform in vitro drug efficacy tests (Gao et al., 2015;Tang et al., 2015), enabled to observe within one week the cytotoxic effects of the drug over time comparable with those observed in vivo within three weeks.
It is an important aspect to determine at what concentration and for what duration a drug should be administered in vitro to allow an in vitro-in vivo comparison so that in the future the use of animal models in the preclinical phase can be reduced.
The MIVO ® platform allows to drastically reduce the overall experimental time, since there is no need to wait for the tumor tissue to grow in mice after cancer cell injection. Moreover, the platform can cut out two highly time-consuming stages of pre-clinical trials: the complex bureaucratic procedures for obtaining ethical clearance, which can take 3-6 months to be completed, and the mice's quarantine stage, stabilization and acclimation, which requires around 1 month on average, reaching a total of 7 months to prepare a single experiment. The MIVO ® -based approach allows the reduction of these times to a couple of weeks of experimental tests, thus potentially reducing the release time of new medicines, animal use and expenditure.
Interestingly, previous studies reported a higher cancer cell resistance to chemotherapy of 3D tumor models when compared to traditional 2D assays (Talukdar and Kundu, 2012;Stock et al., 2016;Lhuissier et al., 2017;Curtin et al., 2018). This is confirmed by our data shown in Figure S6 1 , where SKOV-3 cultured in 2D conditions were treated with the same cisplatin concentration used in the 3D tumor models (i.e. 10M). Already after 2 days of culture, the cells' viability was reduced to about 40% in 2D in comparison to 80% found for the 3D culture (Fig. 2). These results confirm that monolayerbased assays overestimate the drug-cell interactions and the drug diffusion due to a lack of ECM.
For the first time to our knowledge, in this work we report a drug-induced in vitro tumor regression curve that is comparable to that measured in a xenograft model by modeling in vitro (i) 3D tumor tissue perfusion under a capillary circulation, and (ii) the systemic drug transport mechanisms. In contrast, the 3D tumor tissue treated under static conditions displayed resistance to the cytotoxic agent over time, in agreement with the literature. The combination of computational fluid dynamic simulation, drug diffusion measurements and immune-staining analysis allowed to highlight the reasons for these results, and in particular the poor reliability of static 3D tumor hydrogel culture for drug efficacy assays. In particular, when 3D tumor hydrogels were cultured within MIVO ® and the drug agent was flowing within the circuit, the reduction of cell viability began later than in static conditions. This is in line with the computational modeling results ( Fig. 6 and 7) showing that the cisplatin diffusion within the tumor hydrogel is initially faster when the drug agent is added around the 3D hydrogel (in static), while the drug extravasation under flow takes more time. Interestingly, cancer cell viability decreased over time at the lower cisplatin dose only under fluid-dynamic conditions in MIVO ® , displaying a continued cytotoxic effect of the drug comparable to that observed in vivo and in contrast with the alternative in vitro static approach. We hypothesize this is due to different mechanisms of drug penetration within the OCM over time, and to a poor and unreliable drug exposure under static conditions; in particular, when the cancer cell culture is carried out under static conditions, the drug initially enters the 3D hydrogel and may produce its cytotoxic effect mainly in the external layer of cells, which then create a kind of physical barrier for the drug, limiting its continuous diffusion towards the inner core of the polymeric matrix. On the contrary, when the OCM are cultured within MIVO ® resembling the capillary blood flow, the fluid circulation improves the mass transport and waste removal, thus supporting continuous drug diffusion within the hydrogel. These hypotheses were proven by measuring the cisplatin diffusion within the hydrogels in static and dynamic conditions, respectively (Fig. 7).
As a further control, the 3D tumor hydrogels were cultured for 7 days in dynamic versus static conditions without cisplatin (Fig. S5B 1 ). The proliferation rate of the cells was comparable, with a trend to a higher proliferation rate in dynamic conditions in comparison to static conditions. These data consolidate the results obtained, showing that the decrease of cell viability after 7 days observed in hydrogels cultured within MIVO ® treated with cisplatin is due to mechanisms of drug penetration within the hydrogel provided by the fluidic stimulation, not to a reduced cells proliferation under dynamic conditions. These data are consistent with previous work, which reported that the exposure of ovarian cancer cells to low magnitude fluid shear stress (as in MIVO ® chamber) can induce ovarian cancer progression (Ip et al., 2016) . This is supported by immunostaining against proliferating cells (i.e. Ki67) and apoptotic cells (i.e. caspase-3) in both tissue culture conditions ( Fig. 8 and S4 1 ). As expected, after 7 days of drug treatment, proliferating cells were observed under static conditions in the core volume of the OCM, while very few cells positive for Ki67 were found within the 3D hydrogels cultured in MIVO ® . Moreover, an external barrier of apoptotic cells was stained in static conditions, while cell-laden hydrogels within MIVO ® displayed a homogeneous spatial staining of caspase-3. These results were observed at both drug concentrations and were more evident at the higher concentration.
The experimental measurements of cisplatin concentration within the hydrogels through HPLC confirmed the different drug diffusion kinetics within the polymeric matrix: after 7 days of drug treatment, the drug nominal concentration (i.e. 10 M) was found within the hydrogels that had been cultured in dynamic conditions, while a lower concentration (7 M) was measured in the hydrogels that had been cultured under static conditions. These results confirm the key role of the fluid-dynamic environment in resembling the physiological 3D tumor tissue mass and drug transport through a fluid dynamic stream of the blood circulation in vitro. Under these conditions, obtained by