Background
Hepatocellular carcinoma (HCC) is the most frequently diagnosed liver cancer and the sixth most common neoplasm worldwide [
1,
2]. Sorafenib, a Food and Drug Administration (FDA)-approved first-line targeted agent, lengthens the median survival time from 7.9 months to 12.3 months [
3,
4]. However, a majority of individuals show primary or secondary drug resistance. Because the mechanisms underlying resistance to sorafenib are rather unclear, it is vital to explore the causes of resistance and to implement a corresponding treatment strategy. Tumor-initiating cells (TICs), a minority subpopulation of cells that possess properties of progenitor cells, are present in numerous solid tumors [
5,
6]. Studies based on human samples and in vitro models have demonstrated that TICs are highly aggressive and responsible for tumor metastasis, relapse and drug resistance, leading to disease-related mortality [
7]. For example, Victoria Tovar and colleagues used transcriptomic analysis to uncover an increased number of TICs in tumors resistant to sorafenib [
8], indicating that TICs might play an important role in sorafenib resistance. TICs are mainly identified through tumor formation in vivo, sphere formation in vitro, and specific cell markers. In HCC, CD44 is TIC marker most associated with the epithelial-mesenchymal transition (EMT), drug resistance and tumor formation in immunodeficient mice [
9‐
11], suggesting that CD44 expression might be a crucial sign to guide clinical treatment. In addition, various stemness-related signals, such as Notch, Hippo, Wnt and Hedgehog signaling, are usually aberrantly activated in TICs, and drug resistance can be overcome by blocking these signals [
12‐
14]. Among them, Hedgehog signaling, which is related to embryogenesis, participates in the differentiation of hepatocytes from endodermal progenitors [
15]. It has been reported that activation of the Hedgehog pathway occurs in both the initiation and early progression stages of hepatocarcinogenesis and is associated with a poorly differentiated histopathology and relatively invasive phenotype [
16,
17]. Thus, therapeutic strategies aimed at targeting TICs and stemness-related signals are particularly attractive, as they may hinder the development of drug resistance.
Nevertheless, another barrier in overcoming drug resistance is the lack of proper tumor models to evaluate the treatment efficacy of candidate agents. The currently commonly used cell line models of HCC cannot completely capture the features of primary tumors, such as cellular heterogeneity, cell-cell interactions and three-dimensional architecture. In contrast, recently developed patient-derived organoid (PDO) models, which have been described for liver [
18], prostate [
19], breast [
20], and bladder [
21] cancers, can overcome these limitations [
22]. Moreover, these models conserve all parts of primary tumor cells, including TICs, and form functional organ-like structures [
23,
24], providing an unparalleled strategy for studying tumor heterogeneity and creating an opportunity to tailor therapies for individual patients.
In this study, we established PDO models using samples from untreated HCC patients, and these HCC PDOs maintained the features of the original tumor, providing useful information for drug testing. Furthermore, we found that CD44-positive PDOs were significantly resistant to sorafenib. Blocking Hedgehog signaling significantly reduced cell viability and increased sensitivity to sorafenib, especially for CD44-positive HCC PDOs. Therefore, we propose that the combination of Hedgehog signaling inhibitors and sorafenib is a promising strategy for improving the curative effect in CD44-positive HCC patients.
Methods
Human HCC specimens
Tumor specimens were acquired from HCC patients who underwent surgical resection at the Department of Hepatobiliary Surgery, Peking University People’s Hospital, Beijing, China. All patients were first diagnosed with HCC and had not yet received therapy. The specimens were cut into two pieces of approximately 1 cm3: one random piece was fixed in formalin for histopathological analysis, and the other was processed into single cells for PDO culture.
PDO culture
Briefly, each HCC specimen was minced on ice and digested with liberase (TM) (Roche) and 0.1% DNase I (Sigma) for 0.5 h–2 h at 37 °C on an orbital shaker. The incubation continued until the digestion preparation was visually inspected, after which the suspension was strained though a 75-μm nylon filter, and approximately 10 ml of Advanced DMEM/F12 (Gibco) with 10% fetal bovine serum (FBS) was added before centrifugation for 5 min at 300×g. The supernatant was removed, and the pellet was washed in precooled phosphate buffer saline (PBS), mixed with a growth factor-reduced (GFR) Matrigel matrix (Corning) and seeded in a 24-well or 96-well plate (Corning). After polymerization of the GFR Matrigel matrix, PDO culture medium (advanced DMEM/F12 supplemented with 1× GlutaMAX (Sigma), 10 mM HEPES (Sigma), 1× B-27 (Sigma), 1× N-2 (Sigma), 1.25 mM N-acetyl-l-cysteine (Sigma), 10 mM nicotinamide (Sigma), 10 nM recombinant human (Leu15)-gastrin I (Sigma), 50 ng/ml recombinant human EGF (Invitrogen), 100 ng/ml recombinant human FGF10 (PeproTech), 25 ng/ml recombinant human HGF (PeproTech), 10 μM forskolin (Tocris), 5 μM A8301 (Sigma), 10 μM Y27632 (Sigma) and 3 nM dexamethasone (Sigma)) was added [
18,
25]. The culture medium was changed twice per week. The PDOs were passaged every 1–2 weeks after dissociation with Dispase (Corning). For storage, the PDOs were dissociated and resuspended in recovery cell culture freezing medium (Stem Cell) and frozen according to standard procedures.
Cell lines and reagents
The HCC cell lines BEL7402, SNU423 and Hep3B were purchased from China Type Culture Collection (Shanghai, China) and cultured in RPMI-1640 medium (Sigma) supplemented with 10% FBS (Gibco), 100 U/ml penicillin and 100 mg/ml streptomycin. The following reagents were used: dibenzazepine (Selleck, S2711); Wnt-C59 (MCE, HY-15659); GANT61 (MCE, HY-13901); verteporfin (MCE, HY-B0146); sorafenib (Selleck, S1040) and purmorphamine (Selleck, S3042).
Immunohistochemistry (IHC) and immunofluorescence (IF)
Tumor specimens or HCC PDOs were fixed in formalin or 4% paraformaldehyde, respectively, followed by dehydration and paraffin embedding. The sections were subjected to hematoxylin and eosin (H&E), IHC and IF staining. IHC was performed with Vectastain Elite ABC Kit (Vector Laboratories) following the manufacturer’s protocol. The Citrate buffer (pH 6.0) was used for antigen retrieval, and 0.3% NaHB4 was used for immunoperoxidase labeling. After washing with PBS three times, the sections were incubated with a primary antibody at 4 °C overnight. Incubation with the corresponding secondary antibody and a peroxidase-antiperoxidase complex was carried out for 1 h at room temperature. Immunoreactive sites were visualized with 3, 30-DAB. For IF, all procedures were carried out as performed for IHC, except that the cells were incubated with an IgG antibody conjugated with Alexa Fluor® 488 (1:1000) (Thermo Scientific) or an IgG antibody conjugated with Alexa Fluor® 555 (1:1000) (Thermo Scientific). The resulting signals were visualized using a confocal laser scanning microscope (Olympus BX61, Tokyo, Japan). The following primary antibodies were used: anti-AFP (Proteintech, 14,550–1-AP), anti-GPC3 (Abcam, ab129381), anti-EpCAM (Proteintech, 21,050–1-AP), anti-Ki-67 (CST, 9449), anti-CD44 (CST, 3570), anti-Sonic Hedgehog (Abcam, ab53281), anti-PTCH1 (Immunoway, YT3598), anti-Smoothened (Abcam, ab113438), and anti-GLI1 (Abcam, ab49314). To detect apoptosis, paraffin-embedded sections were deparaffinized and treated with DeadEnd Fluorometric TUNEL System (Promega) according to the manufacturer’s instructions.
Western blotting
Protein lysates were obtained by homogenizing cells with RIPA (Merck with the Roche Complete protease inhibitor mixture). Proteins (20–40 μg) were separated by SDS-PAGE (10%) and transferred to a polyvinylidene difluoride (PVDF) membrane. After blocking with 5% skim milk, the membranes were incubated with a primary antibody at 4 °C overnight, followed by incubation with a horseradish peroxidase (HRP)-conjugated secondary antibody (CST,7074 and 7076) at room temperature for 1 h. Chemiluminescent HRP substrates (Millipore, Billerica, MA, USA) were used to visualize antibody binding. The following antibodies were used: anti-Sox2 (Abcam, ab92494), anti-c-Myc (CST, 5605), anti-Nanog (Abcam, ab109250), anti-Oct4 (Abcam, 181,557), anti-MEKK2 (Abcam, ab33918), anti-MEKK3 (Abcam, ab40756) and anti-GAPDH (Proteintech, 60,004–1-Ig).
RNA isolation and quantitative real-time PCR (qPCR)
Total RNA was extracted from PDOs or HCC cells with TRIzol reagent (Invitrogen) following the manufacturer’s instructions. qPCR was performed to assess mRNA expression using the Bio-Rad CFX96 real-time PCR detection system.
Drug treatment
HCC PDOs were seeded in 10 μl of GFR Matrigel matrix droplets in a 96-well plate (Corning) and cultured for 6 days. The PDOs were then treated with culture medium containing different drugs. Cell viability was detected using CellTiter-Glo 3D reagent (Promega,G9681). For HCC cell lines, cell viability was tested after treatment with purmorphamine, GANT61 or sorafenib, as indicated by alamarBlue™ (Invitrogen, DAL1025). The results were normalized to those for vehicle treatment (DMSO). All experiments were performed in technical (same screening run) and biological (different passages) duplicates and were subjected to stringent quality-control measures.
Chou-Talalay analysis
Cell viability assays for the 20, 40, 60, 80 and 100% IC50 of each drug alone or in combination treatment were performed and tested with alamarBlue™. The data were input into the CompuSyn program (ComboSyn Inc., Paramus, NJ), as were the concentration ratios, to find the combination index for each combination to determine synergism or antagonism. All data were collected and analyzed as directed by the Chou-Talalay method [
26]. CI is a quantitative measure of the degree of drug interaction, whereby CI < 1 indicates synergism, CI > 1 indicates antagonism, and CI = 1 indicates an additive effect.
For colony formation, HCC cell were plated in regular medium in 6-well plates at 37 °C with 5% CO2. After 1–2 weeks of incubation, the cells were fixed with 4% paraformaldehyde, stained with 0.1% crystal violet, and washed with PBS before analysis. The relative numbers of colonies were counted under a microscope. A cell invasion assay was performed using 8-μm pore-size hanging cell inserts (Corning). First, the bottom of the upper chamber was coated with Matrigel matrix (BD Biosciences) (10 mg/ml) according to the manufacturer’s protocol, and 1 × 105 cells in serum-free DMEM were seeded in the upper chamber; the lower chamber was filled with 20% FBS. After a 48 h incubation, the noninvaded cells on the upper surface of the filter were removed with a cotton swab, and the invasive cells were fixed with 4% paraformaldehyde and stained with 0.1% crystal violet. The number of invasive cells on the lower surface of the membrane was counted under a microscope.
HCC cells were seeded in an Ultralow-attachment 6-well plate (Corning) in spheroid formation medium (advanced DMEM/F12 supplemented with 1× N2, 1× B27, 20 ng/ml hEGF, and 20 ng/ml bFGF (PeproTech)). After 7 days of incubation, spheres were evaluated by microscopy and extracted for protein analysis.
Animals model
For the xenograft experiment, 1 × 106 BEL7402 or Hep3B cells were injected subcutaneously into to 5 week-old BALB/c nu/nu mice. Tumor growth was monitored twice a week by measuring the length and width of the tumor. The tumor volume was calculated according to the formula V (cm3) = 1/2 × Length × Width2. For drug treatment, tumors were allowed to reach approximately 60 mm3 in size before drug treatment. The mice were randomized into 4 groups: control (vehicle), sorafenib (30 mg/kg/day, administered orally), GANT61 (40 mg/kg, twice a week, administered intraperitoneally), and sorafenib+GANT61 treatment. After approximately 4 weeks, the mice were killed under anesthesia. The tumors were collected and fixed in 10% formalin and embedded in paraffin. H&E staining and IF were performed on sections from embedded samples.
Statistical analysis
Data are shown as the mean ± SD Statistical evaluations between two groups were performed by Student’s t-test. Experiments with more than three groups were evaluated by one-way ANOVA followed by Bonferroni’s test. Overall survival curves were plotted by the Kaplan-Meier method and compared using the log-rank test. A P value less than 0.05 was considered statistically significant.
Discussion
This study for the first time explored a crucial role of Hedgehog signaling in sorafenib resistance in a subpopulation of CD44-positive HCC. We established four HCC PDOs from different patients, demonstrating that these in vitro models completely maintain the features of the original tumors and respond to drug treatment. Furthermore, CD44-positive HCC PDOs were more prone to sorafenib resistance. Compared with other classic TIC-related inhibitors, the inhibitor of Hedgehog signaling obviously decreased cell viability and increased apoptosis in HCC PDOs. In addition, cotreatment with sorafenib and Hedgehog signaling inhibitors had a dramatic synergistic effect of inhibiting the proliferation of HCC cells, especially CD44-positive cells, both in vitro and in vivo. Moreover, properties of colony formation and invasiveness were attenuated following Hedgehog signaling blockade. Finally, high levels of CD44 were relatively frequently accompanied by Hedgehog signaling activation. Sorafenib treatment increased CD44 levels, which was reversed by Hedgehog signaling inhibition. The above results suggest that the sensitivity of HCC PDOs to sorafenib can be indicated by CD44 levels, which are related to the activation of Hedgehog signaling. Thus, Hedgehog signaling is promising as a combinational therapeutic strategy for patients with high CD44 levels.
HCC exhibits high molecular heterogeneity, owing to the coexistence of different subsets with different sensitivities to targeted therapies, and accurate individualized treatment is urgently needed [
28]. Nonetheless, preclinical tumor research tools show substantial limitations. Several cell lines have been used as normal in vitro models to represent different tumors, but they fail to reflect the heterogeneous tumor context and predict the clinical outcomes of different individuals [
29]. Another classic tumor model is the patient-derived xenograft (PDX) model, which preserves tumor heterogeneity to some extent but is labor and time intensive, cannot be expanded and exhibits inefficient generation [
30]. The appearance of PDOs overcomes many of these limitations [
22]. PDO models consist of different subsets within tumors and maintain the histological features and expression profiles of the tumors from which they were derived. In this study, the success rate of PDOs was approximately 50%, higher than that reported in a study of biopsied HCC specimens (33%) [
25], possibly due to the small sample size and hepatocyte contamination. All HCC PDOs were generated from moderate or poorly differentiated tumors, which was in line with previous studies [
18], suggesting that cell viability is required for the generation of PDOs. In addition, expression of the TIC marker CD44 persists in HCC PDOs, providing an opportunity to study the differentiation of drug efficacy in primary HCC cells with different CD44 levels.
Resistance to targeted drugs, such as sorafenib, in HCC patients is a major problem. Over the last decade, acknowledged mechanisms of resistance to targeted drugs have included continuous activation of target genes owing to secondary mutations [
31], enhanced gene expression [
32] or abnormal regulation of compensatory signaling, such as aberrant activation of PI3K/AKT by MEK or mTOR inhibition [
33,
34]. Another resistance mechanism, gain of stem cell phenotypic features, within the heterogeneous tumor cell subgroup is receiving increasing attention [
7]. CD44 is acknowledged as a vital TIC marker and is associated with a poor survival rate in numerous types of tumors [
35]. Bera and colleagues demonstrated that gemcitabine treatment induces pancreatic cancer cell lines to undergo an EMT process and convert from CD44 negativity to CD44 positivity [
36]. Similarly, we found that the level of CD44 can indicate the sensitivity of HCC PDOs to sorafenib. Furthermore, a switch in the expression levels of CD44 may be observed both in HCC PDO and in cell lines after sorafenib treatment. Thus, sorafenib treatment might induce conversion to a CD44-positive phenotype, including the acquisition of TIC features and increased insensitivity to drugs. On the other hand, our results showed that GANT61 treatment significantly decreased levels of CD44 in HCC PDOs and cotreatment of sorafenib and that GANT61 inhibited the tumor growth both in vitro and in vivo. Therefore, suppression or reversion of this CD44 profile conversion might be a crucial strategy for enhancing treatment efficacy.
In addition to ITC markers, stemness-related signals, including Wnt, Hippo, Notch and Hedgehog pathways, also play vital roles in the acquisition of ITC phenotypes. In this study, we found that inhibition of Hedgehog signaling maximally reduced the proliferation of HCC PDOs. Activation of Hedgehog signaling is mainly caused by binding of the Hedgehog ligand to the receptor Ptch1, which in turn relieves inhibition of the Smoothened (Smo). Smo then activates transcription of Gli1, which enters the nucleus to promote expression of target genes [
37]. HBV infection, a well-known risk factor for HCC, has been reported to regulate posttranslational activation of Hedgehog signaling, which leads to hepatocarcinogenesis [
38]. It has been confirmed that cell proliferation and self-renewal can be inhibited and that drug sensitivity to temozolomide can be increased by blocking Hedgehog signaling in glioma [
39]. Although Hedgehog signaling activation has been demonstrated in HCC [
40,
41] and has been shown to be a key regulator of autophagy in HCC cells [
42], there is little data on the exact roles of Hedgehog signaling in HCC TIC conversion and sorafenib resistance. In this study we first demonstrated that Hedgehog activation may occur in a subpopulation of HCC patients who are CD44 positive and that blocking this activation can significantly increase sorafenib sensitivity. By activating Hedgehog signaling, purmorphamine reversed the inhibitory of cotreatment of sorafenib and GANT61 in HCC. Gli1 is a pivotal component of Hedgehog signaling; its expression is precisely regulated in the process of cell proliferation, and its deregulation leads to tumorigenesis. Gli1 typically binds to with Sufu to form a complex; when Hedgehog signaling is activated, the Sufu-Gli1 complex dissociates, and Gli1 is activated [
43]. Recently, it was reported that silencing MEKK2/3 promotes expression of Hedgehog signaling-related proteins [
44]. Further research has found that MEKK2/3 enhances the Sufu-Gli1 interaction, resulting in Gli1 cytoplasmic retention, which occurs mechanistically via Gli1 phosphorylation by MEKK2/3 at multiple Ser/Thr sites [
27]. Interestingly, our results indicate that sorafenib obviously decreases levels of MEKK2/3 in cell lines in a dose-dependent manner. Combined with the ability of sorafenib to induce activation of Hedgehog signaling, this observation indicates that sorafenib might suppress MEKK2/3 to promote Hedgehog signaling activation, resulting in the upregulation of CD44, acquisition of a progenitor cell phenotype and sorafenib resistance in HCC cells. Therefore, we hypothesize that sorafenib suppresses MEKK2/3, after which Gli1, which is not phosphorylated by MEKK2/3, enters the nucleus and promotes expression of stemness-related genes, converting the cell to the stem/progenitor cell phenotype and resulting in resistance to sorafenib (Fig.
6g).
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.