Tumor growth is slower in spheroid-plug model
First, we compared the kinetics of tumor growth in classical and spheroid-plug model using noninvasive 3D-USG. Size of B16 tumors injected by classical method increased logarithmically (14.37 ± 4.51-fold increase at day 14) (Fig. 2a). In contrast, in spheroid-plug model, B16 tumor plugs did not significantly change their size (1.07 ± 0.12-fold increase) (Fig. 2a). Consistently, LLC tumors injected by spheroid-plug model were also smaller than tumors injected classically (1.82 ± 0.20- vs. 2.69 ± 0.39-fold increase at day 14, p < 0.01); however, the difference between the models was much less pronounced than in B16 tumors (Fig. 2a).
In addition to 3D-USG technique, we compared kinetics of tumor growth by in vivo detection of bioluminescence (IVIS). Two-way ANOVA indicates that model applied for implantation of B16 cells influences the tumor growth (p < 0.01), with significantly faster increase in luminescence in case of classical method (Fig. 2b). However, at day 14, the IVIS signal intensity from B16 tumors in spheroid-plug model was much higher than on the day 1 and similar to that in classical model (2.04 ± 0.46 × 107 vs. 3.99 ± 0.23 × 107 at day 14, NS) (Fig. 2b). Thus, at day 14, the difference between models in tumor volumes was notably higher (15.16-fold, Fig. 2a) than the difference in IVIS signal (1.96-fold, Fig. 2b). Accordingly, in the case of LLC tumors, we did not notice differences in IVIS signal (Fig. 2b) between models (4.26 ± 0.53 × 106 vs. 3.79 ± 0.86 × 106), despite small but statistically significant difference in LLC tumor volumes (Fig. 2a).
We concluded that spheroid-plug tumors grew slower than tumors formed in classical model. However, measurements of tumor volumes by 3D-USG were not paralleled by detection of tumor cell-derived luminescence detected by IVIS.
Tumors in spheroid model are less necrotic
In the next step, we investigated possible reasons for the discrepancy between tumor volumes and IVIS signal intensities. Because only viable cells are able to express luciferase and produce detectable bioluminescence signal, the inconsistency between tumor volumes and IVIS measurements may indicate the presence of necrotic, not viable areas within tumor. Indeed, histological analysis using hematoxylin and eosin staining confirmed that B16 and LLC tumors from spheroid-plug model had smaller area of necrotic tissue than tumors from classical method (B16 10.46 ± 2.91 vs. 21.49 ± 3.13 %, p < 0.05; LLC 17.95 ± 5.06 vs. 34.26 ± 3.06 %, p < 0.05) (Fig. 3a).
Given that variable necrosis content results in discrepancy between tumor volumes and IVIS signal, we proposed that a ratio of tumor volume (measured, e.g., using 3D-USG) to the IVIS signal intensity from the tumor (Fig. 3b) may be used to evaluate necrosis content within tumors in quantitative and noninvasive manner. We demonstrated that this parameter, named here the necrosis factor (NF), correlates significantly with necrosis content in all analyzed tumors estimated by histology (r = 0.48, p < 0.0068, Spearman rang correlation) (Fig. 3b). The calculated NF showed significant, consistent with histological analysis, differences between models: its value was lower in spheroid-plug model, both in B16 (p < 0.01) and LLC tumors (p < 0.05) (Fig. 3c).
Altogether, the spheroid-plug tumors were less necrotic. Our analysis showed also that in vivo imaging could be used to calculate necrosis factor as noninvasive indicator of tumor necrosis.
Spheroid-plug model ensures stable vascularization of tumors
Necrotic areas of tumors are often the consequence of inadequate and nonfunctional tumor vasculature [15, 16]. We monitored development of vasculature within tumors in vivo (days 1, 7, and 14 after tumor injection) using 3D-USG with perfusion analysis. Performed measurements indicated different kinetics of tumor angiogenesis between classical and spheroid-plug model in B16 tumors. While at day 7 we observed a higher perfusion in classical model (17.84 ± 2.31 vs. 9.42 ± 1.74 %, p < 0.01) (Fig. 4a), at day 14, it decreased and was lower than in spheroid model (5.45 ± 1.82 vs. 14.33 ± 2.35 %, p < 0.01) (Fig. 4a). The observed pattern — slower development of vasculature at day 7 (14.71 ± 1.99 vs. 10.93 ± 1.82 %) but higher vascularization at day 14 (4.77 ± 0.73 vs. 7.90 ± 0.84 %) in spheroid-plug model compared to classical model — was also visible in LLC tumors; however, the differences did not reach statistical significance (Fig. 4b).
Furthermore, we noticed the difference in vessel localization within tumors (Fig. 4c, d). At day 14, in classical model, vessels were visible only near the tumor borders, whereas in spheroid-plug model, vessels penetrated the tumor core, as shown by visualization of 3D-USG analysis (Fig. 4c, d).
Infiltration of cells with progenitor phenotype is increased in spheroid-plug model
Infiltrating host cells may drive tumor angiogenesis [17, 18]. As spheroid-plug tumors were more vascularized, we checked if increased vascularization correlates with presence of other host cells. We took advantage of GFP-expressing tumor cells, what enabled us to distinguish GFP+ tumor cells from GFP− infiltrating cells by using flow cytometry (Fig. 5a).
First, we checked the content of endothelial cells. The total number of endothelial cells with CD45−Sca-1+CD31+c-Kit− phenotype (Fig. 5a) was significantly higher in the case of spheroid-plug model than in classical model, both in B16 (1.52 ± 0.55 vs. 0.039 ± 0.009 %, p < 0.01) and LLC (0.13 ± 0.061 vs. 0.083 ± 0.015 %, p < 0.05) tumors (Fig. 5b).
Next, among infiltrating cells, we distinguished population with non-hematopoietic progenitor cell phenotype CD45−Sca-1+CD31−cKit+ (Fig. 5a). This population was significantly increased in spheroid-plug model compared to the classical model in case of B16 tumors (4.92 ± 1.48 vs. 0.71 ± 0.36 %, p < 0.01), and similar tendency was observed in LLC tumors (0.067 ± 0.03 vs. 0.012 ± 0.004 %, p = 0.16) (Fig. 5c).
Apart from endothelial and mesenchymal cells [19–21], various hematopoietic progenitors may also regulate angiogenesis [22, 20, 21]. Indeed, we found that cells with hematopoietic progenitor phenotype CD45+Sca-1+c-Kit+ infiltrated tumors (Fig. 5d). These cells were more abundant in spheroid-plug model than in classical model in B16 tumors (10.16 ± 5.13 vs. 1.51 ± 0.71 %, p < 0.01) (Fig. 5e). However, in the case of LLC, there were no differences in the content of hematopoietic progenitors between tested models (Fig. 5e).
Taken together, we observed that in B16 tumors, the augmented angiogenesis in spheroid-plug model correlated not only with higher number of endothelial cells but also with increased infiltration of cells expressing progenitor markers (c-Kit and Sca-1). Such relationship seems to be tumor type specific, as it was not observed in tumors formed by LLC cells.
Spheroid-plug model increases the number of cells with cancer stem cell phenotype
In the next step, we compared heterogeneity of tumor cells between spheroid-plug and classical models. Given that in vitro 3D spheroid culture facilitates the acquisition of cancer stem cell properties [23–25], we wondered if there were more tumor cells with cancer stem cell phenotype in tested spheroid-plug model. To verify this hypothesis, we analyzed expression of several markers on GFP+ tumor cells that were previously thought to characterize cancer stem cells: c-Kit , Sca-1 , CD133 [28–30], CD49f , and CXCR4 [30, 32, 33].
The obtained results revealed that most of the LLC and B16 tumor cells expressed CD49f, what suggests that this antigen alone did not select unique cancer stem cells in investigated models (Fig. 6a). Nevertheless, we used CD49f together with CXCR4 to analyze heterogeneity of tumor cells (Fig. 6a). We detected a minor population expressing CXCR4 among LLC and B16 cells, both within CD49f+ and CD49f− subsets (Fig. 6a). The number of CXCR4+CD49f− tumor cells was higher in spheroid-plug model than in classical model either in B16 or LLC tumors (Fig. 6b), while CXCR4+CD49f+ population was more abundant in spheroid-plug model only in B16 tumors (Fig. 6c). Expression of c-Kit, Sca-1, or CD133 did not distinguish separate populations in B16 tumors. Differently, in LLC tumors, we found c-Kit+CD133− and c-Kit−CD133+ subpopulations that expressed also Sca-1 (Fig. 6d); however, there were no differences in content of these subpopulations between tested models (Fig. 6e).
Finally, we tried to explain why some of the described populations expressing stem cell markers were changed in spheroid-plug model, while other did not differ. One possibility is that in vitro spheroid culture itself induces cells with cancer stem cell phenotype, which then persist during in vivo tumor development. Flow cytometry analysis confirmed that at least part of populations expressing stem cell markers were induced already by 3D culture conditions in vitro. We observed that spheroid culture increased frequency of CXCR4+CD49f+ B16 cells (Supplemental Fig. S3A) consistently with changes observed in vivo in spheroid-plug model (Fig. 6c). Similarly, spheroid culture led to increased number of CXCR4+CD49f− LLC cells (Supplemental Fig. S3B), in concert to cell profile observed in vivo in spheroid-plug model (Fig. 6b).
In contrast, in in vitro spheroid culture, we did not detect CXCR4+CD49f− population in B16 cells as well as CXCR4+CD49f+ and c-Kit+CD133−Sca-1+ subsets in LLC cells (data not shown). Moreover, c-Kit−CD133+Sca-1+ population was induced in LLC spheroids in vitro (Supplemental Fig. S3C), what was not observed in spheroid-plug LLC tumors (Fig. 6e). Thus, the altered frequency of these populations in vivo in spheroid-plug model cannot be explained by the changes caused by spheroid formation in vitro.
Spheroid-plug model is suitable for testing the efficacy of antiangiogenic drugs
To compare the efficacy of known antiangiogenic drug in classical and spheroid-plug model, we chose axitinib which is a new type of small molecule tyrosine kinase inhibitor  with evidenced effectiveness in B16 tumor model [35, 36]. The treatment with axitinib was started 7 days after injection of tumors (Fig. 7a).
Firstly, we compared B16 tumor growth kinetics between treated and control groups in both models with the use of 3D-USG. Again, we could observe fast logarithmic growth of tumor volume in classical model (13.89 ± 1.68-fold increase at day 21 comparing to day 7 in control group) (Fig. 7b). In mice dosed with axitinib changes in tumor volume were smaller (2.08 ± 0.31-fold increase) (Fig. 7b). The fast growth of tumors in classical model forced us to terminate experiment at day 21.
In comparison, control tumors from spheroid-plug model grew slower (2.86 ± 0.92-fold increase at day 21 when comparing to day 7), and we were able to prolong experiment till day 28 (Fig. 7b). Similarly to classical model, we observed significant difference in tumor volumes between control and treated groups at the end of experiment (684.6 ± 162.2 mm3 in control group vs. 64.34 ± 13.56 mm3 in axitinib group, p < 0.0001) (Fig. 7b).
We also monitored tumor development by IVIS. Although axitinib significantly decreased the size of tumors at the end of the experiment in both models, there was no significant difference in IVIS signal in classical model (Fig. 7c).
On the other hand, in spheroid-plug model, the axitinib treatment not only decreased tumor size but also resulted in substantial differences in IVIS signal between treated and non-treated groups (37.01-fold lower signal in axitinib group at day 28, p < 0.01) (Fig. 7c). Additionally, calculated necrosis factor values showed that axitinib significantly influenced necrosis content only in spheroid-plug model, whereas it had no effect on tumors in classical model (Fig. 7d).
Axitinib is an antiangiogenic drug; therefore, we evaluated tumor vascularization as an important parameter reflecting the drug efficiency. Measurements performed using 3D-USG with perfusion analysis revealed different effect of axitinib on tumor vasculature in examined models. Axitinib treatment had bigger effect on tumors in spheroid-plug approach: treated tumors were significantly less vascularized than those from control group (3.18 ± 0.68 vs. 7.47 ± 1.11 % at day 21, p < 0.01; 4.38 ± 1.33 vs. 9.08 ± 0.92 % at day 28, p < 0.01) (Fig. 7e). In contrast, differences between groups in classical model were smaller and did not reach statistical significance at the end of experiment (4.5 ± 0.77 % in treated group vs. 6.38 ± 0.83 % in control at day 21, NS) (Fig. 7e). Consistently, we observed significantly less endothelial cells (defined by flow cytometry as CD45−CD31+Sca-1+c-Kit−) after treatment with axitinib in spheroid-plug model, what was not visible in classical model (Fig. 7f).
Moreover, we also examined axitinib influence on metastasis occurrence in both models (Fig. 8a). We were able to observe that 60 % of spheroids administered subcutaneously in the control group formed metastasis after 28 days. Axitinib treatment completely inhibited that process in spheroid-plug model (Fig. 8b). In classical model, all mice from control group developed metastasis already after 21 days, but treatment with axitinib showed only 50 % efficacy in metastasis inhibition (Fig. 8b).