Establishment of multicellular tumor spheroids (MCTSs) that recapitulate important elements of hepatic fibrosis for high-throughput screening of potential liver fibrosis inhibitors
In our previous study, we found that the interaction between HCC cells and various non-parenchymal cells affected the compactness of the spheroids as well as cell migration through accumulation of collagen and EMT-related proteins . In order to generate a fibrosis model in vitro, various HCC cell lines (Huh7 cells, SNU449 cells, and HepG2 cells) were grown together with fibroblasts (WI38), hepatic stellate cells (LX2), and endothelial cells (HUVEC) in MCTS models. Despite the fact that both SNU449 cells and HepG2 cells innately formed loose aggregates, these cells acquired the rigidness of the spheroids following co-culture with stromal cells in spheroids. Similarly, although Huh7 cells formed a relatively solid spheroids, co-culture with stromal cells enhanced rigidness in spheroids [Figure 1-A]. This result showed that crosstalk between stromal cells and HCC cells in MCTS models was an important determinant of rigidness of spheroids, emphasizing the importance of culturing these cells as a system rather than as individual components.
Next, gene expression profiling was performed on the MCTS model systems to compare against the expression profiles observed in tumor spheroids. In the MCTS models, genes that were involved in the production of ECM structural constituents were significantly enriched. In particular, we found increased relative expression of MMP1, COL6A1, COL6A3, and TGFB1 in MCTS relative to tumor spheroids [Figure 1-B].
Because the process of EMT leads to organ fibrosis, we compared the expression of mesenchymal markers such as vimentin, α-smooth muscle actin (α-SMA), Snail, and N-cadherin between the MCTS models and tumor spheroids. Mesenchymal markers were generally upregulated in MCTS models relative to tumor spheroids alone. Because TGF-β1, which is a critical regulator of fibrosis, stimulates the EMT and EndMT processes through the activation of Smad, we measured the expression and activation of Smad2 and Smad3. Expression of Smad2/3 and relative abundance of p-Smad2, p-Smad3 were upregulated in Huh7-, HepG2-MCTSs. SNU449 MCTSs also displayed upregulated p-Smad3 expression [Figure 1-C, Supplementary Fig. 1].
The MCTSs provided a useful in vitro model of liver fibrosis, where increased size of spheroids, which results from the loss of tight cross-linking among cells, indicates decreased expression of fibrosis-related proteins and thus provides a reliable morphometric indication of the reversal of liver fibrosis [Figure 1-D]. Next, we sought to establish a MCTS-based drug screening platform for the evaluation of potential liver fibrosis inhibitors. To obtain reproducible results, we plated single, homogenously sized and configured spheroids in 384-well plates for HTS screening.
A library comprised of 4,763 drug compounds with known molecular targets was tested for potentially promising candidates as inhibitors of fibrosis. All compounds were screened at an initial concentration of 10 µM in duplicate to confirm the reproducibility of the observed effects. A correlation coefficient of 0.89 for replicate screens indicated that the assay was reliable [Figure 1-E]. In that screening, we identified 12 positive compounds (HITs) including four compounds involved in the cAMP/PKA pathway, five retinoic acid analogs, an anti-diabetic drug, a regulator of cholesterol, and a NMDA receptor modulator [Table I].
Because nintedanib and pirfenidone were recently authorized for the treatment of idiopathic pulmonary fibrosis, we evaluated the effects of both drugs on the size of HCC-MCTSs. Surprisingly, nintedanib and pirfenidone did not alter the morphology of HCC-MCTSs relative to control solvent (2% DMSO) [Supplementary Fig. 2]. On the other hand, treatment with 10 µM concentration of the 12 HITs significantly increased of size of HCC-MCTSs, according to morphometric analyses, suggesting inhibition of fibrosis.
Next, dose-response studies were performed to find the most effective compounds among the 12 HITs identified from HTS of HCC-MCTSs. We found that the retinoic acid analogs and modulators of cAMP/PKA pathway, particularly retinoic acid and forskolin, led to the most significant increases of HCC-MCTSs at concentrations as low as 0.1µM, in a dose-dependent manner [Figure 1-F, Supplementary Fig. 3].
Retinoic acid and forskolin reversed EMT and EndMT in stromal cells in multicellular tumor spheroids (MCTSs), but not multicellular hepatocyte spheroids (MCHSs)
Generally, hepatic fibrosis is associated with upregulated expression of α-SMA via EMT and EndMT. To investigate the architectural changes observed in MCTSs following treatment with retinoic acid and forskolin, we used immunohistochemistry to evaluate the expression of F-actin in spheroid structures. Interestingly, when 5 µM of retinoic acid and forskolin were added to MCTSs for 48 hr, spheroid size was bigger than DMSO-treated MCTSs, without increasing of cell size, demonstrating loss of tight cell-cell interactions and decreasing F-actin intensity among the cells [Figure 2-A].
Western blot analysis also showed that elevation of α-SMA expression in MCTSs was sufficiently attenuated by treatment with 1 µM retinoic acid and forskolin, whereas expression of CD31 was increased under the same conditions. These results indicated that retinoic acid and forskolin inhibit the EndMT process as well as fibrotic properties in MCTSs [Figure 2-B, Supplementary Fig. 4]. The retinoic acid analogs AM580 and TTNPB and the water-soluble forskolin derivative NKH477 also inhibited α-SMA expression in MCTSs, but did not alter expression of CD31, in contrast to retinoic acid and forskolin in MCTSs [Supplementary Fig. 5-A].
Upregulation of CD133 facilitates EMT in various cancers [20–22]. Interestingly, expression of CD133 in MCTSs was also inhibited by treatment with 1 µM retinoic acid and forskolin [Figure 2-C]. However, AM580, TTNPB and NKH477 did not inhibit CD133 expression as effectively as retinoic acid and forskolin [Supplementary Fig. 5-B].
Next, we were curious whether the replacement of HCC cells with normal hepatocytes in the MCHSs would result in the same phenotypic effects. Instead of Huh7, we used Fa2N-4, a well-known normal hepatocyte cell line, to generate MCHSs. As shown in Fig. 2-D, when stromal cells were mixed together, they formed compact spheroids. Similar to MCTSs, MCHSs also showed increasing expression of vimentin, α-SMA, collagen I, and Snail as well as decreasing E-cadherin and CD31 as was observed in the MCHS models [Figure 2-E, Supplementary Fig. 6]. However, the hit compounds identified from HTS, retinoic acid and forskolin, did not change the size of spheroids created with normal hepatocytes. This suggests that MCTSs composed of HCCs exist in a severe inflammatory environment that is treatable with these compounds, making it a more suitable model for screening compounds than MCHSs with normal hepatocytes [Figure 2-F].
Liver fibrosis is a complex phenomenon orchestrated by numerous cellular actors in tumor microenvironments. These results suggested that retinoic acid and forskolin may inhibit hepatic fibrosis through reversing EMT and EndMT processes of stromal cells in MCTSs and suggested that an MCTS model-based morphometric screening approach may be a good strategy for the screening of novel effective therapies for fibrosis.
Retinoic acid and forskolin depolarized hepatic stellate cells in a fibrotic environment
To confirm the potential efficacy of retinoic acid and forskolin in reprogramming activated HSCs, which are the main collagen-producing cells in liver fibrogenesis, we conducted cellular phenotype-based assays. Increasing production of α-SMA [23, 24] and F-actin stress fibers are associated with HSC activation when HSCs are stimulated with TGF-β1. To define distinctive morphometric signatures before and after TGF-β1 treatment, we focused on the expression pattern of F-actin and α-SMA after treatment with TGF-β1. Treatment with TGF-β1 increased the intense cytoplasmic α-SMA and F-actin of LX2 cells in a dose-dependent manner [Figure 3-A]. When the intensity of α-SMA and F-actin were analyzed by Harmony 3.5.1® high-content imaging and analysis software, we found that treatment with 5 ng/ml TGF-β1 increased the intensity of α-SMA more than 1.5-fold compared to the control, whereas intensity of F-actin increased only slightly [Figure 3-B]. Therefore, we selected α-SMA as a marker of fiberized hepatic stellate cells. Western blot analysis also displayed similar results in agreement with the cellular phenotype-based assays. Expression of fibroblast markers, α-SMA, fibroblast activation protein (FAP) and collagen I were increased after TGF-β1 treatment in LX2 cells [Figure 3-C].
Next, we measured the effects of retinoic acid and forskolin on TGF-β1 -induced HSC activation using cellular phenotype-based assays. As expected, 1 µM retinoic acid and forskolin inhibited the expression of α-SMA after treatment with TGF-β1 in LX2 cells, with efficacy comparable to 10 µM pirfenidone, which served as our positive control [Figure 3-D]. Pirfenidone [25, 26] and nintedanib , which are FDA-approved anti-fibrotic drug, inhibit TGF-β1 -induced fibrogenesis. However, in our system, the intensity of α-SMA was not decreased as much as 2% FBS-treated control when pirfenidone and nintedanib were treated at various concentrations [Supplementary Fig. 7]. When 1 µM retinoic acid [Figure 3-E] and forskolin [Figure 3-F] were added with TGF-β1 to LX2 cells, EMT-related markers N-cadherin and Snail were inhibited, but E-cadherin was elevated, in contrast to EMT-related markers [Figure 3-E, F]. Collectively, this phenotypic-based 2D assay system using LX2 cells appears to be an effective tool for validating anti-fibrosis compounds and suggested that retinoic acid and forskolin can reprogram activated hepatic stellate cells.
Retinoic acid and forskolin suppress the EndMT process in HUVEC
In our previous studies, we established a visual phenomic screening platform to measure radiation-induced EndMT using HUVECs . Herein, this technology was applied to measure TGFβ1-induced EndMT in HUVECs. HUVECs treated with TGF-β1 expressed increasing amounts of F-actin and cytoplasmic α-SMA in a dose-dependent manner [Figure 4-A]. When the intensity of α-SMA and F-actin were analyzed by Harmony 3.5.1® high-content imaging and analysis software, we found that treatment with 20 ng/ml TGF-β1 increased the intensity of α-SMA more than 1.8-fold compared to the control, and intensity of F-actin increased 1.6-fold compared to the control [Figure 4-B]. Expression of fibroblast marker, α-SMA was increased after TGF-β1 treatment in HUVEC cells [Figure 4-C]. Next, we examined the effects of retinoic acid and forskolin on TGFβ1-induced HUVEC activation using the cellular phenotype-based 2D assay system. In this experiment, we found that 1 µM retinoic acid and forskolin decreased the expression of α-SMA after TGF-β1 treatment in HUVECs relative to treatment with 10 µM pirfenidone [Figure 4-D]. Of interest, expression of α-SMA were inhibited when 1 µM retinoic acid and forskolin were added with TGF-β1 to HUVEC cells, [Figure 4-E, F]. These results suggested that anti-fibrotic compounds, such as retinoic acid and forskolin, suppress the EndMT process in HUVECs.
The combination of anti-cancer drugs and anti-fibrosis compounds improves responses by enhancing penetration of anti-cancer drugs
Liver cancer patients typically experience fibrosis, cirrhosis, and liver-related disease. As shown in Fig. 2-A and F, spheroids showed loose compactness after treatment with anti-fibrosis compounds, and cell-cell tight junction interactions were also weak compared to controls. In our previous study , we compared the efficacy of drug penetration by detecting the distribution of doxorubicin using fluorescence microscopy in HepG2 spheroids and HepG2-MCTS grown with LX2 or WI38 cells. In this study, we sought to determine whether the anti-fibrosis compounds may increase the penetration of anti-cancer drugs in MCTSs by decreasing cell-cell interactions. When the MCTSs that were treated with 5 µM retinoic acid or forskolin were treated with 10 µM doxorubicin for 8 hr, the distribution of doxorubicin in MCTSs was highly increased relative to spheroids that were not treated with retinoic acid or forskolin [Figure 5-A]. Indeed, doxorubicin only penetrated the periphery of MCTSs after treatment with 0.5% DMSO. This result was not surprising in light of the observed decreased compactness of MCTSs after treatment with anti-fibrosis compounds. Based on these results, we expected that anti-fibrosis compounds may accelerate anti-cancer effects by enabling delivery of anti-cancer compounds to the center of tumor tissues. In MCTS models, we found that spheroids treated with 1 µM retinoic acid or forskolin combined with 1 or 3 µM of sorafenib had high expression of cleaved caspase-3 (an apoptosis marker), significantly higher relative to spheroids treated with sorafenib alone [Figure 5-B]. From these results, it appears that anti-fibrosis compounds, such as retinoic acid or forskolin, may improve the efficacy of anti-cancer drugs and attenuate tissue compactness and stiffness observed in liver fibrosis.