• Open Access

Incorporation of multicellular spheroids into 3-D polymeric scaffolds provides an improved tumor model for screening anticancer drugs


To whom correspondence should be addressed.
E-mail: benwu@seas.ucla.edu


Development of cancer therapeutics requires a thorough evaluation of drug efficacy in vitro before animal testing and subsequent clinical trials. Three-dimensional (3-D) in vitro models have therefore been investigated for drug screening. In this study, we have developed a novel in vitro model in which multicellular aggregates, or spheroids, were incorporated into 3-D porous scaffolds. Drug resistance assays showed that spheroid-seeded scaffolds have much higher drug resistance than monolayer cultures, spheroids on flat substrates, or scaffolds seeded with dispersed cells. Furthermore, spheroid-seeded scaffolds demonstrated higher lactate production leading to acidosis, and higher expression of angiogenic factors. These data suggest that the spheroid-seeded 3-D scaffolds might serve as a useful in vitro system for screening cancer therapeutics. (Cancer Sci 2010; 101: 2637–2643)

In the process of developing new cancer therapeutics, candidate compounds are thoroughly investigated in vitro before any in vivo or clinical trials are performed. However, many therapeutics that demonstrate efficacy in vitro fail to produce the same effect in vivo.(1) This might be attributed to inadequate or limited modeling of in vivo tumor behavior by the in vitro systems. While two-dimensional (2-D) cultures have provided a variety of high-throughput applications,(2,3) there still exists a significant need for a more realistic tumor model for effective screening of anticancer drugs.(4,5)

The use of spherical aggregates of cells, or spheroids, has been an integral part of cancer research since Holtfreter and Moscona(4,5) introduced them in 1944 and 1957, respectively. Spheroids possess more realistic, three-dimensional (3-D) cell–cell and cell–matrix interactions than 2-D cell cultures.(6,7) Many studies have suggested both correlative and causal effects on cellular drug response due to various characteristics of spheroids including necrotic cores induced by hypoxia,(8)gene expressions,(9) drug permeability(10) and inhibited apoptosis.(11) Therefore, tumor models using spheroids have provided several advantages for a multitude of cancer research fields.

A more recent approach for creating realistic tumor models is the use of biocompatible synthetic polymers such as poly(lactic-co-glycolide) and poly(ε-caprolactone) to engineer the desired 3-D environments for cell growth.(12–14) Culturing cancer cells in these 3-D structures has been a useful tool for various biomedical fields. A major advantage of creating 3-D matrices involves the ability to manipulate and incorporate various components such as microspheres containing cytokines,(15,16) protein content(17,18) and structural complexity(19) in the culture environment.

Despite the success of these models, they still present inherent limitations. Specifically, spheroids are entirely cell-based, so they do not offer as much flexibility for engineering the surrounding environment as 3-D structural scaffolds. They also do not represent the microenvironment as well as 3-D scaffold cultures.(20) On the other hand, culturing cells in 3-D matrices cannot capture the cell–cell interactions that arise from the aggregated morphology of tumors. Although some studies with 3-D scaffolds have presented cultures in which cells begin to cluster after a certain period of time, the aggregations are difficult to control or reproduce.(13)

In this study, we explore a novel 3-D in vitro system for evaluating the efficacy of cancer therapeutics, in which multicellular spheroids of reproducible dimensions and cell numbers are incorporated into 3-D polymeric biodegradable scaffolds. Seeding spheroids into the scaffolds harnesses the engineering prospects of using a polymeric scaffold, as well as captures the cell–cell interactions. To observe this effect, the cellular drug response of 3-D scaffolds with spheroids (which we will refer to as “SS” hereafter), 3-D scaffolds with dispersed cells, or “MS”, and 2-D monolayer cell cultures have been compared.

Materials and Methods

Incorporation of spheroids into scaffolds.  U251 cells were cultured as described in the Data S1. The combination of spheroids consisting of 2000 cells and scaffolds with 500–1000 μm diameter pores (Data S1) were used in the present study. After 48 h of culture, spheroids were seeded onto scaffolds placed on sterile filter paper (Whatman, Florham Park, NJ, USA) to absorb excess culture medium through capillary action. One hundred spheroids were pipetted onto each scaffold using specialty pipette tips with large orifices (Fisher Scientific, Pittsburgh, PA, USA). The seeded scaffolds were moved onto poly-HEMA-coated wells of 12-well non-tissue culture treated plates (Fisher Scientific) (500 μL of 5 mg/mL polyHEMA in anhydrous ethanol in each well dried for 48 h at 37°C), and 50 μL of culture medium was added to each scaffold. The spheroids were left to attach for 4 h in humidified incubators, and 2 mL of culture medium was subsequently added to each well. After 24 h of culture, the plates were placed on an orbital shaker. Successful incorporation of spheroids into scaffolds was confirmed with microscopic imaging (Data S1).

Drug resistance assay.  For the 2-D systems, cells were seeded at four different densities (10 k/cm2, 20 k/cm2, 40 k/cm2 and 60 k/cm2) and cultured for 1 day. For the 3-D systems, 200 k monolayer-cultured cells or 100 spheroids (2000 cells each) were seeded onto the scaffolds and were allowed to grow for 2 days with standard culture conditions on an orbital shaker. Samples were subsequently incubated in solutions of various concentrations of doxorubicin (Sigma-Aldrich, St Louis, MO, USA) and irinotecan (Sigma-Aldrich). The cells were left to grow for 2 more days, and the bicinchoninic acid (BCA) assay (Thermo Scientific, Rockford, IL, USA) was performed following the manufacturer’s protocol. The assay was used to quantify cell viability at each drug concentration relative to the wells without drugs. Some modifications to the standard BCA assay include using cell scrapers (Fisher Scientific) to reduce experimental error when retrieving the cell lysates and cutting scaffold samples into fine pieces prior to assaying. The Infinite 200 Microplate Reader (Tecan Systems, San Jose, CA, USA) was used to measure optical densities. Sigmoidal drug-response curves for the different systems were then generated, and each IC50 value, which is the drug concentration at which half the cells are killed, was interpolated. The amount of doxorubicin sequestered by the scaffolds was also quantified colorimetrically (Data S1).

Visualization of drug penetration.  Spheroid-seeded scaffolds were incubated in culture medium with doxorubicin (1.8 mL of 100 μM, 104 μg total) for 3 h. To identify dead cells, the samples were then incubated in propidium iodide (Invitrogen, Carlsbad, CA, USA) for 20 min. After rinsing with PBS, the samples were subsequently frozen in Tissue-Tek OCT compound (Sakura Finetek, Torrance, CA, USA) at −80°C. Frozen samples were cryosectioned (20 μm thick) onto superfrost plus microslides. Doxorubicin was visualized at 480 nm/560 nm ex/em, dead cells were identified at 530 nm/620 nm ex/em using propidium iodide, and all cell nuclei were visualized with DAPI staining at 360 nm/460 nm ex/em. Appropriate controls were also visualized (Data S1). The intercellular spaces inside the spheroids in scaffolds were visualized using fluoroscein isothiocyanate (FITC)-conjugated dextran molecules (Data S1).

Lactate production assay.  To assess the glycolytic activity of different systems, the amount of lactate produced was analyzed. The culture medium was first aspirated, and the cell cultures were incubated in Hanks’ Balanced Salt Solution (Mediatech, Manassas, VA, USA) for 2 h at 37°C. The lactate in each sample was quantified using the Lactate Assay kit (BioVision, Mountain View, CA, USA) following the manufacturer’s protocol. To assess the degree of acidosis, the intracellular pH of spheroids in the scaffolds was measured (Data S1).

VEGF and bFGF expression assay.  To quantify vascular endothelial growth factor (VEGF) and basic fibroblast growth factor (bFGF) production, aliquots of culture medium of cells were analyzed with Quantikine Human VEGF and Quantikine Human FGF basic enzyme-linked immunosorbent assay (ELISA) kits (R&D systems, Minneapolis, MN, USA), respectively, following the manufacturer’s protocol.

Statistical analysis.  To assess the statistical difference between two groups (fraction of aggregated cells in scaffolds, amount of drug sequestered, intracellular pH), two-tailed, unpaired t-test assuming equal variance (F-tests showed P > 0.05 for most pairs) was selected because no predictions about the changes in the means were made, and a Gaussian distribution for each population was assumed. As for the statistical difference among three or more groups (lactate production, angiogenic factor production), a one-way analysis of variance (anova) was selected. For pair-wise comparisons within each experimental group, Tukey’s post test was used.


Spheroids remain intact in scaffolds after seeding.  A novel in vitro tumor model is presented based on merging two widely used 3-D systems: multicellular aggregates and 3-D scaffolds (Fig. 1a). Using a previously reported method based on centrifugation,(21) spheroids of reproducible dimensions (∼300 μm) using the U251 glioma cell line were generated (Fig. 1b). We were also able to reproduce the results of a previous study showing that spheroid compaction increased throughout the period of culture as the diameter steadily decreased from 350 μm to 300 μm (Fig. 1c). This can be attributed to both cell death and stabilized cell–cell interactions, since it has been shown that U251 cells do not proliferate as much as other cell lines upon forming spheroids.(22) The majority of cells within the spheroids remained viable up to 4 days (Fig. 1d). After culturing the spheroids for 2 days to stabilize compaction, they were seeded into collagen-coated poly(lactic-co-glycolic acid) (PLGA) scaffolds (Fig. 1e). Images of cryosectioned scaffold samples with nuclear staining showed that the spheroids were indeed incorporated into and attached within the pores (0.8–1 mm in diameter) of the collagen-coated scaffold. Over time, cells detached from the spheroid, spreading and proliferating on the scaffold in a similar manner to their monolayer counterpart. Intact and spherical aggregations were still visible after 2 days. However, by the fourth day of culture most of the spheroids in the scaffold had dissociated (data not shown). This suggested that scaffold modification may be optimized to balance maintenance of spheroid morphology with dissociation of spheroids due to cells adhering to the scaffolds. Regardless, spheroids retained their spherical, aggregated morphology in collagen-coated scaffolds for up to 2 days of culture, providing a window of time during which the system can be analyzed.

Figure 1.

 Establishing spheroid-seeded scaffolds. (a) Previously established in vitro tumor models include monolayer cultures, spheroids and scaffolds with cells. A novel tumor model is developed by incorporating spheroids, or 3-D cellular aggregates, into 3-D scaffolds. (b) Microscopic image of a 2-day-old spheroid of 2000 cells confirmed the spherical morphology. (c) Spheroid diameter gradually decreased over the duration of the culture (n = 10). Mean diameters (±SEM) are plotted as a function of time. (d) The viability of cells within spheroids cultured for 4 days was observed using a live stain (calcein-AM). The control represents a dead spheroid that had been incubated in 70% ethanol for 30 min. (e) Images of poly(lactic-co-glycolic acid) (PLGA) scaffolds showed that spheroids were successfully incorporated into the pores of the scaffold. Scale bar, 100 μm.

Spheroid-seeded scaffolds have increased resistance to anticancer therapeutics.  In order to identify the effect of incorporating multicellular spheroids into the scaffolds on the system’s overall drug resistance, the drug response of four systems was compared: (i) cells cultured as 2-D monolayers; (ii) spheroids only; (iii) dispersed cells in scaffolds (MS); and (iv) spheroids in scaffolds (SS) (Fig. 2). Overall, all 3-D systems showed higher drug resistance when compared with 2-D. Among the 3-D systems, SS exhibited the highest drug resistance for both drugs (Table 1).

Figure 2.

 Resistance to anticancer drugs. Four different systems, 2-D at 20 000 cells/cm2 (▪), spheroids in 2-D (•; only for doxorubicin), 3-D scaffolds with monolayer-cultured cells (MS, bsl00066), and 3-D scaffolds with spheroid-cultured cells (SS, bsl00072) were treated with (a) doxorubicin and (b) irinotecan and cell death was quantified. Mean percentage of cell deaths (±SEM) plotted as a function of drug concentration (n = 4).

Table 1.   IC50 of systems assayed for drug resistance (μM)
Culture systemDoxorubicin (95% CI)Irinotecan (95% CI)
2-D (10 000/cm2)0.05 (0.04–0.07)20.7 (12.2–35.3)
2-D (20 000/cm2)0.09 (0.06–0.12)15.4 (9.53–25.0)
2-D (40 000/cm2)0.22 (0.14–0.35)16.2 (13.3–19.7)
2-D (60 000/cm2)0.19 (0.10–0.34)16.1 (12.7–20.4)
Spheroids in 2-D3.46 (1.93–6.21)N/A
3-D MS1.96 (1.21–3.18)253 (155–413)
3-D SS (collagen-coated)20.2 (11.5–35.7)798 (442–1440)

Because the apparent increase in drug resistance could be due to other effects resulting in reduced drug availability, we explored two factors. First, since cell density can influence drug resistance by affecting the drug-to-cell ratio within a sample and since we did not keep the cell density constant for all systems, drug resistance at various cell densities were compared in 2-D for both drugs (Fig. 3a,b). For doxorubicin, a sixfold increase in cell density only resulted in approximately a fourfold increase in IC50, and for irinotecan, the IC50 values were unaffected by an increase in cell density (Table 1).

Figure 3.

 Effects of reduced drug-to-cell ratio and polymer drug sequestering on drug resistance. To assess the significance of drug-to-cell ratio on the observed drug resistance, four different cell densities (10 k/cm2, 20 k/cm2, 40 k/cm2 and 60 k/cm2) were subjected to varying concentrations of either doxorubicin or irinotecan, and cell viability was quantified. Mean percentage of cell deaths (±SEM, n = 3–4) were plotted as a function of concentration of (a) doxorubicin or (b) irinotecan. The amount of doxorubicin sequestered onto poly(lactic-co-glycolic acid) (PLGA) scaffolds was also quantified. Bar graph (c) represents μg of doxorubicin sequestered (mean + SEM, n = 3). ***P < 0.001.

Second, when comparing 2-D and 3-D systems involving scaffolds, the presence of hydrophobic polymers may sequester hydrophobic drug molecules, thereby also contributing to the observed drug resistance. Because doxorubicin possesses an intrinsic orange-red color, the amount of doxorubicin sequestered onto the scaffold was quantified colorimetrically after solubilizing the scaffolds. When compared with the control, scaffolds incubated in 100 μM drug solution (104 μg total) for 2 days seemed to sequester only approximately 9.5 μg (9% of the total drug) in the solution (Fig. 3c).

Drugs readily penetrate the spheroids inside scaffolds.  Since the aggregated morphology may limit drug penetration and reduce drug exposure to cells at the interior of the spheroids, we explored a third factor related to these diffusional limitations. Specifically, the presence of doxorubicin in the cells cultured in SS was evaluated. Although penetration of doxorubicin into spheroids has been studied previously,(23) to our knowledge, drug penetration into spheroids within 3-D scaffolds has not yet been evaluated. After incubating scaffolds seeded with spheroids in culture medium containing doxorubicin, the presence of doxorubicin within the spheroids was visualized using its intrinsic green fluorescence. Cell death was also assessed using propidium iodide. After 3 h of incubation, doxorubicin was found throughout each spheroid (Fig. 4a). Furthermore, cells with doxorubicin did not emit red fluorescence, showing that the cells were still alive within the 3 h of incubation, albeit the drug uptake.

Figure 4.

 Assessing drug penetration. (a) Visualization of doxorubicin (dox) penetration into spheroids within the scaffold was based on doxorubicin’s intrinsic fluorescence (white: negative bright field; blue (middle left panel): DAPI; green (right panel): doxorubicin; red (bottom left panel): propidium iodide [PI]; scale bar, 100 μm). Controls include SS incubated in culture medium without doxorubicin and blank collagen-coated scaffolds incubated in propidium iodide only (not shown). (b) The presence of intercellular spaces within the spheroids were visualized by incubating SS in PBS containing fluorescein isothiocyanate-conjugated 150 kD dextran (scale bar, 50 μm).

We also hypothesized that the rapid drug penetration might be possible due to the intercellular space within the spheroids. To explore this idea, FITC-conjugated 150 kD dextran molecules were used as membrane-impermeant markers for intercellular spaces. Intense fluorescence was only detected in the spaces between the cells of the spheroids (Fig. 4b), confirming the presence of intercellular regions within the spheroids.

Increased drug resistance may be due to increased glycolysis and angiogenic factors.  Many studies have suggested the role of various cytokines in determining the drug response of tumor cells. For instance, because spheroid morphology creates regions of hypoxia, metabolic dependence on glycolysis might be increased. In fact, high glycolytic activity is yet another distinguishing feature of tumor cells, and a previous study demonstrated that inhibiting glycolysis in cancer cells leads to increased drug sensitivity.(24) To evaluate the glycolytic activity of the four systems, their lactate production was quantified. The SS and 2-D cell cultures exhibited the highest and lowest lactate produced per cell, respectively (Fig. 5a). With enhanced output of lactate into the surrounding environment, cells inside the spheroids would be subjected to acidic conditions. By using an intracellular pH probe, the degree of acidosis was evaluated. Indeed, the pH values inside the spheroids were lower than 7.4. Furthermore, cells in the interior of the spheroids had pH values that are significantly lower than those of the cells seeded in the 50-μm-wide perimeter of the spheroids (Fig. 5b).

Figure 5.

 Hypoxia-associated characteristics of spheroid-seeded scaffolds. (a) The amount of lactate produced by each tumor model was quantified using a lactate assay kit. All data (mean + SEM, n = 4–8) are normalized to 2-D values. ***P < 0.001. (b) Degree of acidosis in SS was also evaluated by measuring the intracellular pH. Exterior represents the 50-μm-wide perimeter of spheroids within the scaffolds, while interior is defined as the inner core of the spheroids. Mean pH values (±SEM) from the two regions are plotted (n = 4). *P < 0.05. (c) The relative production of angiogenic factors, VEGF and bFGF were also quantified. All data (mean + SEM, n = 3–4) are normalized to their respective 2-D values. *P < 0.05. **P < 0.01. ***P < 0.001 (all compared with 2-D).

As a result of both hypoxia and acidosis, angiogenic factors might be induced.(25,26) Moreover, angiogenic factors such as VEGF and bFGF have been shown to be correlated with drug resistance.(27,28) Thus, expression of the cytokines by the different systems was quantified using ELISA assay kits and compared. The 2-D systems had significantly lower expression of both angiogenic factors than the 3-D systems (Fig. 5c). All 3-D culture systems produced similar cytokine expression levels.


Utility of spheroid-seeded scaffolds as a drug screening tool.  It has been previously shown that 3-D systems lead to an increased drug resistance when compared with 2-D systems. More specifically, 3-D systems utilizing spheroids only or polymeric scaffolds only have been demonstrated as useful tumor models to mimic the in vivo tumor behavior.(13,20,22) Our study has shown that combining the two techniques by seeding 3-D scaffolds with spheroids instead of dispersed, monolayer-cultured cells increases the drug resistance significantly. For our studies using doxorubicin and irinotecan, our 2-D IC50 values, 0.05–0.19 μM for doxorubicin and 16–20 μM for irinotecan, corresponded well with previous studies.(29,30) Also, cells in SS exhibited approximately 10-fold and threefold, respectively, higher drug resistance than cells in MS. Since both doxorubicin and irinotecan have been used in vivo, it was possible to compare the dose response of our in vitro system with that of the in vivo studies to assess the system’s ability to screen drugs more accurately. One study has reported the concentration of liposomal doxorubicin around a glioma tumor to be approximately 7 μM for an effective intravenous injection dose.(31) Interestingly, this compared fairly well with our IC50 of 20 μM. Another study using irinotecan and rats with implanted xenograft glioma tumor models showed that delivery of 3 mg/mL (4800 μM) of irinotecan slightly improved survival time.(32) Although the IC50 value of 798 μM is still well below the effective in vivo concentration, the chemoresistance of SS was closer to the in vivo data than that of MS. Moreover, in this analysis, two drugs of very different physiochemical properties were used. While doxorubicin and irinotecan both have very similar molecular weights (543 and 586 g/mol, respectively), their respective partition coefficients (−0.5 and 3.2; octanol-water), predicted water solubilities (1.2 and 0.1 mg/mL), and half-lives (55 and 6–12 h) are considerably different.(33) Thus, because the drug response of our system for both doxorubicin and irinotecan was the closest to in vivo data among all in vitro systems observed, it suggests that the system may be a useful in vitro screening tool for evaluating the efficacy of drugs of various properties. It should be emphasized that these comparisons do not suggest that our current system can replace any in vivo models.

We have further determined that the polymer itself in the system can play a role, albeit small, in raising the apparent chemoresistance by sequestering the drugs. Although not the focus of this study, polymer matrices could have more deterring effects on the drug screening process depending on its size and chemical properties. Moreover, studies have shown that some cell lines readily form spheroids while others do not,(34) and that for some drugs, culturing as spheroids instead of a monolayer has no observable effects on cellular drug response.(35,36) Thus, the dimensions and properties of the 3-D scaffold as well as the cell line used for spheroid generation would both be important considerations when screening drugs with different properties. We have generated and characterized the morphology of another glioma cell line, LN-229, to be similar to U251 in terms of spheroid production (Won Jin Ho, Edward A. Pham, Christopher W. Ng, Daniel T. Kamei, Benjamin M. Wu, unpublished data). This may be another candidate cell line to be incorporated into 3-D scaffolds for further studies.

One clear limitation of the current spheroid-seeded scaffold system is that the spheroids dissociate in the scaffolds by the fourth day of culture. This restricts its utility in any long-term studies. However, we have observed in additional experiments that differential modifications of the scaffold surfaces such as conjugating Arg-Gly-Asp (RGD) peptide, which is a prevalent extracellular peptide involved in cellular adhesion, can control the spheroid dissociation rate (data not shown). Since cell spreading is dependent on the density of RGD peptides on a surface,(37) the amount of RGD-conjugation could be altered in order to optimize spheroid adhesion to the scaffolds. Efforts to promote cell–cell adhesion within the spheroid, as well as attempts to prevent cell adhesion to the 3-D scaffold surface, may maintain the aggregations of spheroids for longer periods.

Understanding the basis of increased drug resistance.  Previous studies have explained the higher drug resistance of 3-D systems based on a variety of mechanisms. One possible component of chemoresistance is drug penetration.(38) Cellular aggregation as well as the presence of tortuous, 3-D scaffold structures may limit the drug exposure to some cells. However, we found that drug uptake in SS was efficient and that doxorubicin was found inside the spheroids within 3 h of exposure. This was consistent with spheroid permeability data previously obtained for various drugs including doxorubicin.(23,39) In fact, one previous study has shown that even at relatively low concentrations (∼10 μM), doxorubicin rapidly traverses more than 120 μm from the periphery upon exposure.(23) Thus, coupled with the observation of intercellular spaces throughout the spheroids using 150 kD FITC-dextran, the presence of doxorubicin throughout the 300 μm diameter spheroids was not surprising. Others have shown that despite efficient drug penetration, spheroids still exhibited higher drug resistance than cells in a monolayer.(40) Therefore, our study confirms the idea presented by many others that increased drug resistance in 3-D systems cannot merely be explained by the effects on drug transport.

Another distinguishing characteristic of real-life tumors is high glycolytic activity. Although oxygen readily diffuses, it is a required metabolite that is rapidly consumed by cells. Thus, oxygen tension in a given cellular construct is a function of several factors including varying metabolic activities in the surrounding environment and the spatial arrangement of cells. Since 3-D systems are more susceptible to forming regions of hypoxia, cells compensate for the need of nutrients by increasing their dependence on the less efficient anaerobic pathways following glycolysis instead of oxidative phosphorylation. As expected, the highest level of lactate production was seen in the spheroid-seeded scaffolds. Importantly, as confirmed by our measurements of the intracellular pH, increased production of lactate leads to acidosis. Consequently, acidosis can trigger a variety of mechanisms leading to higher drug resistance. A previous study has shown that inhibiting glycolysis results in reduced drug resistance.(24) Further studies exploring the relevance of this mechanism to the observed behavior of SS, especially on a molecular level, would be valuable.

Another widely studied mechanism linked with hypoxia and tumor development is angiogenesis. Cancer cells respond to a lack of oxygen by stimulating neovascularization and inducing the expression of angiogenic factors such as VEGF(41) and bFGF.(42) Many earlier studies showed that inhibiting the activities of these growth factors leads to increased chemosensitivity, increased expression of proteins that participate in the induction of apoptosis, and thus reduced cell survival.(27,28,43,44) Although our results did not show a definitive role of angiogenic factors in the enhanced drug resistance of SS, there was a significantly higher production of VEGF in SS, which may partly be due to the role of acidosis in up-regulating VEGF.(25) These results are also suggestive of the usefulness of SS in screening VEGF-targeted therapies.

Disclosure Statement

The authors have no conflict of interest.


confidence interval


3-D scaffolds with dispersed cells


3-D scaffolds with spheroids