A tissue-engineered gastric cancer model for mechanistic study of anti-tumor drugs

The scaffold was electrospun as previously described [20]. Briefly, polydioxanone (PDO) in the form of surgical sutures (Advanced Inventory Management, Mokena, IL), porcine gelatin A (Sigma-Aldrich, St. Louis, MO) and bovine elastin (Elastin Products Co. Inc., Owensville, MO) were dissolved in 1,1,1,3,3,3-hexafluoro-2-propanol (Sigma-Aldrich, St. Louis, MO) to achieve a total concentration of 10% (w/v) (weight ratio = 8:1:2). The solution was loaded into a syringe capped with a 27-gauge needle mounted on a motorized pump (PHD 2000, Harvard Apparatus, Holliston, MA). A total of 1 mL of the blend solution was electrospun onto a grounded sheet collector at a feeding rate of 2 mL h⁻¹ in a 1 kV cm⁻¹ electric field. The electrospun scaffold was then desiccated in vacuum for at least 24 h. The morphology of the electrospun scaffold was studied with scanning electron microscope. The degradation of scaffolding materials was characterized by incubating the electrospun scaffold in tumor cell media at 37 °C and 5% CO₂ for up to six weeks. At the end of each week, the sample was retrieved, desiccated in vacuum for 24 h and weighed on a digital balance. The weight loss was calculated as ${\rm Weight \; loss}(\%) =\vspace*{3pt} \frac{{{\rm Weight \; before \; incubation} - {\rm Weight \; after \;incubation}}}{{{\rm Weight \; before \; incubation}}}{\rm \times 100 \% }{\rm .}$

Gastric tumors were resected from human patients at Shanghai Xinhua hospital. Tumor masses were immediately disassociated using sterile scissors and blades. To obtain a single-cell suspension, disassociated tissues were incubated with ultrapure collagenase IV (Worthington Biochemicals, Freehold, NJ) in medium 199 (200 units of collagenase per mL) at 37 °C for 4 h followed by a filtration through a 40 µm nylon mesh and a wash with HBSS/20% fetal bovine serum. The collected suspension was thoroughly rinsed with HBSS. CD24⁺ and CD44⁺ cells were sorted out by fluorescent activated cell sorting (FACS) (BD LSRII, BD Biosciences, San Jose, CA), using anti-human CD24 and anti-human CD44 antibodies (Pharmingen, Franklin Lakes, NJ). The sorting was repeated until the purification reached at least 90% in respective populations. Freshly sorted out CD24⁺ and CD44⁺ cells were used in the apoptosis assay in the chemotherapeutic effect study [26]. Written consent was obtained from patients prior to the resection of the pancreatic tumor specimens.

Desiccated scaffolds (D = 6 mm) were sterilized by in 70% ethanol followed by an extensive rinse in sterile phosphate buffered saline. The whole population of tumor cells were seeded on the scaffold (1000 cells/scaffold) and cultured for up to seven days for an in vitro tumorigenesis. 3-(4,5-dimethylthiazol-2-yl)-2,5- diphenyltetrazolium bromide (Promega, Madison, WI) and Caspase-Glo® 3/7 assay kit (Promega) were used to measure the viability and apoptosis, respectively, per manufacturer's protocols. Tumor cells cultured in tissue culture plates (TCP) were used as control.

Freshly resected gastric tumors from human patients were completely minced (∼1 mm × 1 mm) and cultured on sterilized scaffold (∼2 mm × 2 mm) at 37 °C and 5% CO₂ overnight. Then the tumor/scaffold construct was surgically implanted subcutaneously into a BALB/c nu/nu female nude mouse (4–6 weeks old) (Shanghai Laboratory Animal Co., Shanghai, China). One edge of the scaffold was sutured to the skin to temporarily immobilize the construct. In the control group, each piece of the tumor tissue was implanted s.b. without the scaffold as in the traditional xenograft model. The animal study was approved by the Research Ethics and Compliance Committee at Shanghai Xinhua Hospital. The tumor site was photographed bi-weekly after the surgery and the tumor volume calculated as previously described: tumor volume = [L × W²]/2 [27]. By the end of week 12 post-surgery, all mice were euthanized and tumor masses retrieved and weighed on a digital scale prior to subsequent biological analyses.

The phenotype of cells in retrieved gastric tumor mass was determined by immunohistochemistry staining. Retrieved gastric tumors from mice were incubated in 10% formalin at 4 °C for at least 48 h, then embedded in paraffin and sliced into 4 µm thick serial sections on charged glass slides. For immunohistochemistry staining, sample were deparaffinized in xylene and rehydrated in ethanol gradients. Antigen retrieval was carried out by heating slides in sodium citrate buffer (10 mM sodium citrate, 0.05% Tween 20, pH 6.0) for 15 min at 100 °C. Samples were blocked in 5% BSA buffer at 4 °C overnight. Thereafter, samples were incubated at 4 °C for 24 h with primary antibodies, including carbohydrate antigen 19–9 (CA199) (Abcam, Cambridge, MA), carcinoembryonic antigen (CEA) (Dako, Carpinteria, CA) and P53 (BD Pharmingen, San Jose, CA), which had been diluted in 1% BSA to recommended dilutions per manufacturers' protocols. At the end of the incubation, samples were extensively rinsed in PBS and then incubated at room temperature for 1 h with rabbit anti-human secondary antibodies conjugated with horseradish peroxidase (Abcam). All samples were counterstained with hematoxylin (Sigma Aldrich) before clearing and mounting. To measure the percentage of CD24⁺ and CD44⁺ cells, tumor masses were prepared as described above to obtain single-cell suspension and analyzed by flow cytometry using respective antibodies.

Docetaxel, cisplatin and fluorouracil (DCF) and cisplatin and fluorouracil (CF) were both obtained from Shanghai Xinhua Hospital and diluted in sterile saline to indicated concentrations. The scaffold tumor model was established as described above. Implanted tumor cells were allowed to grow for four weeks before the drug administration. Immediately before the drug administration, the mice were randomly divided into three groups: DCF group (n = 7) that would receive docetaxel, cisplatin and fluorouracil, CF group (n = 6) that would receive cisplatin and fluorouracil, and NaCl group (n = 6) that would receive sodium chloride. In the DCF group, each mouse received by intravenous infusion docetaxel (2.14 mg kg⁻¹) and cisplatin (2.14 mg kg⁻¹) on day 1 followed by fluorouracil (21.4 mg kg⁻¹ d⁻¹) for three consecutive days. In the CF group, each mouse received by intravenous infusion cisplatin (2.86 mg kg⁻¹) on day 1 followed by fluorouracil (28.6 mg kg⁻¹ d⁻¹) for three consecutive days. Dose of each drug was calculated based on previous research [28, 29]. The drug administration in both DCF and CF groups were repeated every two weeks for a total of four times by the end of the 8-week study. Mice in the NaCl group received equal volume of sodium chloride following the same drug administration schedule. All mice were euthanized after the 8-week-long drug administration. The tumor volume, weight, CD24⁺ and CD44⁺ cells were measured as described above.

CD24⁺ and CD44⁺ cells sorted out by FACS were seeded on sterilized scaffolds in a 96-well plate (5000 cells/well) and cultured for up to 48 h at 37 °C and 5% CO₂. After seeding, samples were randomly divided into DCF, CF and NaCl groups (n = 3). In the DCF group, the regimen of docetaxel (200 µM), cisplatin (200 µM) and fluorouracil (2 mM) was added to the cell media at 8 h after seeding. Similarly, in the CF group, the regimen of cisplatin (300 µM) and fluorouracil (3 mM) was added at the same time point. In the control group, an equal volume of sterile NaCl was added to the cell media. The Caspase-Glo® 3/7 assay kit was used to measure the apoptosis immediately before the drug administration and after a 40 h drug treatment.

ImageJ was used to process and analyze all images. Student t-test and ANOVA with Tukey test were used where applicable (α = 0.05) in Prism (GraphPad Software, La Jolla, CA). All flow cytometric data was analyzed and graphically presented using FlowJo (Tree Star Inc., Ashland, OR). The median fluorescent intensity (MFI) of each population in each sample was measured by FlowJo and calculated using the formula below. An MFI% greater than 100% indicates the presence of interested cell population while an MFI% equal to 100% the lack of interested cell population. The MFI% increases along with the increase of the population size:

$$\begin{equation*} {\rm MFI}\% = \left( {\frac{{{\rm MFI}( {{\rm sample}} )}}{{{\rm MFI}( {{\rm isotype}} )}}} \right) \times 100\% . \end{equation*}$$

The electrospun scaffold featured randomly distributed non-woven micro/nano-fibers with a highly porous structure, which mirrors the morphology of native extracellular matrix (figure 1(A)). It should be noted that each individual fiber was a composite of PDO, gelatin and elastin. We used PDO as the backbone of this electrospun scaffold to confer a structural stability to the scaffold while it would degrade in vivo without unleashing cytotoxic materials. To enhance cell adhesion and proliferation, we incorporated gelatin and elastin into the scaffold, two dominant proteins found in native extracellular matrix. This electrospun composite scaffold was extensively characterized by physical and chemical analyses in a previous research and showed promising capabilities to promote tissue regeneration [20]. Like most tissue-engineered scaffolds, we expect the scaffold to gradually degrade over the course of native tissue growth. Therefore, we measured the weight loss of this scaffold incubated in cell culture media to draw its degradation profile. The result showed that the weight loss reached 31.77% ± 7.29% within the first week and gradually climbed to 72.03% ± 12.45% by week 6, evidencing the degradability of scaffolding materials (figure 1(B)). The fact that the weight loss rate was high in the first week but much depressed thereafter could be attributed to the rapid disintegration of gelatin and elastin in the scaffold with a slow yet steady degradation of PDO. Scaffold samples became too fragile to manipulate for measurement after being incubated for over six weeks.

Zoom In Zoom Out Reset image size

To support the tumorigenesis the scaffold must be able to support tumor cells of all types without producing a cytotoxic effect. To that end we cultured gastric tumor cells on the scaffold and measured the viability as well as apoptosis of the population. The population of tumor cells on the scaffold featured a higher viability than that on TCP on day 7 but no difference was observed on day 1, suggesting that the scaffold provided a more favorable environment than the TCP for cells to adhere and proliferate (figure 2(A)). A significant apoptosis was found in neither population on either day 1 or day 7, evidencing a lack of cytotoxic effects from scaffolding materials.

Zoom In Zoom Out Reset image size

The in vivo study demonstrated that this electrospun composite scaffold enhanced tumorigenesis through week 12 post-surgery (figures 3(A) and (B)). The tumor volume in the scaffold group started to out-grow its peers in the traditional tumor group on week 8 post-surgery and sustained this advantage until week 12. By week 12 after the surgery, the tumor volume and weight in the scaffold model group was 43.27% and 75.58% greater than that the traditional model group, respectively. This increased tumorigenesis in the scaffold model substantiated our hypothesis that the use of an electrospun scaffold would facilitate the tumor growth.

Zoom In Zoom Out Reset image size

Previous research demonstrated that instructive information conveyed from tissue-engineered scaffolds could dictate cellular events like stem cell differentiation [12]. To detect if scaffolding materials led to unwanted phenotypic change, we probed the expression of gastric tumor markers, CA199, CEA and P53 (figure 4). All three markers were highly expressed in cells in the gastric tumor tissues in the scaffold model group as those the traditional model group on week 12 post-surgery. This result evidenced that neither the chemistry of scaffolding materials nor the physics of the scaffold caused aberrant phenotypic changes during the tumorigenesis, thus offering cancer scientists a reliable platform to evaluate chemotherapeutic drugs. The enhanced expression of P53 in the scaffold group can be attributed to the use of the scaffold, a variance among individuals or both.

Zoom In Zoom Out Reset image size

In gastric cancer, CD24⁺ and CD44⁺ cells in the tumor mass were believed to the most potent cancer stem cells (CSC) that drives the tumorigenesis in vivo [30, 31]. Therefore, we investigated the influence from the scaffold on the proliferation of CD24⁺ and CD44⁺ cells. The flow cytometric analysis showed that the scaffold model saw an increased percentage of both CD24⁺ and CD44⁺ populations by the end of the 12-week study (figure 5). The histogram of CD24⁺ and CD44⁺ cells in the scaffold model showed a wider shift than those in the traditional model group, indicating a larger population in the scaffold model. A quantitative analysis of the MFI, a widely used benchmark in flow cytometric analysis, showed that CD24⁺ and CD44⁺ populations in the scaffold model saw a 42% and 313% increase, respectively, compared to those in the traditional model. This increased proliferation of CSC in the tumor in the scaffold model could have greatly contributed to the accelerated tumorigenesis.

Zoom In Zoom Out Reset image size

After confirming the reliability and superiority of our scaffold model, we took advantage of it to comparatively investigate chemotherapeutic effects of DCF and CF regimens on gastric tumors. Both DCF and CF unequivocally arrested the growth of gastric tumor while the tumor underwent a steady and fast growth in the NaCl group (figure 6(A)). A significant difference between chemotherapeutic groups and NaCl group emerged at week 6 and was further pronounced by week 8. At the end of the 8-week treatment, tumor volumes in the DCF and CF groups were only 21.88% and 28.58% by average of that in the NaCl group, respectively. Correspondingly, tumor weights in DCF and CF groups were 26.03% and 32.25% of that from the NaCl group (figure 6(B)). The traditional tumor model showed a delayed onset of the chemotherapeutic effect of both DCF and CF as evidenced by the emergence of tumor volume difference as late as week 8 and a large variance within treatment groups, indicating a poor reliability (figure S1 (available from stacks.iop.org/BMM/8/045003/mmedia)). These results confirmed that both DCF and CF unleashed a proven chemotherapeutic effect on gastric tumor to arrest its growth and that the scaffold model was a robust and reliable platform for drug studies. These results prompted us to hypothesize that the arrest of tumor growth could be attributed to the chemotherapeutic effect on CD24⁺ and CD44⁺ CSC in the tumor masses. The flow cytometric analysis confirmed our hypothesis. Both DCF and CF diminished the CD24⁺ and CD44⁺ CSC populations after an 8-week treatment (figure 7(A)). The MFI analysis revealed that the DCF possessed a stronger ability than CF to deplete the CD44⁺ population whereas both showed similar competency in depleting CD24+ cells (figure 7(B)). To verify this finding, we measured the apoptosis level of CD24⁺ and CD44⁺ CSC due to DCF and CF treatment. The measurement agreed with the in vivo finding by showing that DCF induced a greater apoptosis of CD44⁺ cells than CF when no difference was observed prior to the drug treatment (figures 7(C) and (D)). However, the apoptosis levels of CD24⁺ cells were similar between DCF and CF treatments.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size