DCLK1 induces spheroid growth of primary human hepatocytes in 3D suspension culture
We previously demonstrated that normal human liver parenchyma stains negatively for DCLK1. However, when primary human hepatocytes from normal livers are cultured in Matrigel, which contains several growth factors and extracellular matrix, some cells form spheroids containing numerous DCLK1 + cells16. These spheroids upon further growth contain hepatic cell lineages, such as AFP+ hepatoblasts, progenitor/stem-like cells marked by AFP/CK19 co-staining, and albumin-expressing mature hepatocytes. In the present study, we tested whether DCLK1 overexpression induces anchorage-independent spheroid-forming ability in the untransformed primary human hepatocytes in the absence of matrix. Hepatocytes derived from normal human liver were cultured on collagen-1-coated plates and infected with lentiviruses expressing GFP (Lenti-GFP) or GFP-tagged human DCLK1 (Lenti-GFP-DCLK1). FACS-based analysis suggested that 12–15% of hepatocytes were transduced after the lentiviral infections and expressed the GFP marker within 48 h (not shown). Similar transduced and subsequently trypsinized cultures formed spheroids in a magnetic levitation assay in which newly formed spheroids grow in suspension culture29. As shown in Fig. 1a (upper panel), Lenti-GFP-DCLK1 hepatocytes formed anchorage-independent spheroids growth within one week (highlighted in Fig. 1b). A similar culture of hepatoma cells harboring a GFP tagged HCV NS5A-expressing replicon11 that also overexpress DCLK1 was used as a positive control (Fig. 1c). Under similar conditions, Lenti-GFP-infected hepatocytes showed aggregation but without spheroid development (Fig. 1a, lower panel). These observations suggest that DCLK1 overexpression induces anchorage-independent spheroid growth in untransformed primary human hepatocytes.
DCLK1 overexpression generates a short form of active β-catenin, which increases cyclin D1 expression
β-catenin is an important regulator of clonogenic properties, epithelial cell polarity, and stemness of cells30. Therefore, we investigated β-catenin signaling to better understand the role of DCLK1-induced clonogenic properties in liver cells. Hepatoma (Huh7) cells were infected with lentiviruses expressing either RFP or N-terminus RFP-tagged human DCLK1 as described previously16. We previously demonstrated that the expressed RFP-DCLK1 protein extensively colocalizes with microtubules, stimulates inflammatory signals, and promotes Huh7 cell migration16,19. In this study, live cell imaging revealed the dynamic distribution of RFP-DCLK1 with GFP-labeled microtubules and centrosomes during cell division (Supplementary Fig. S1a) as expected for active DCLK126.
Using lentiviruses, we generated Huh7 cells expressing RFP (Huh7-RFP) and RFP-DCLK1 (Huh7-RFP-DCLK1). The cells were immunostained with anti-active β-catenin [ABC(8E7)]. This monoclonal antibody is highly specific and recognizes the N-terminus unphosphorylated HSGATTTAP (aa 36–44) motif of the protein. The β-catenin form detected by this antibody has been shown to be transcriptionally active31. Confocal microscopy revealed that active β-catenin was mostly stained in the cytoplasm and nuclear regions of RFP-DCLK1+ cells (Fig. 2a, upper panel). However, similar staining pattern for active β-catenin in the nucleus or cytoplasm was not observed for RFP + (lower panel) or Huh7 (not shown) cells, suggesting that the ABC(8E7) staining for active β-catenin was specific for RFP-DCLK1 cells. Growing colonies of Huh7-RFP-DCLK1 cells also showed this staining pattern for active β-catenin (Supplementary Fig. S1b). Quantitation of nuclear staining intensities revealed considerably higher presence of active β-catenin in Huh7-RFP-DCLK1+ cells than in Huh7-RFP cells (Fig. 2c). Parallel confocal microscopy was performed for total β-catenin expression in these cells using PLA0230 antibody. This antibody binds to the C-terminus distal region (aa 760–781) of β-catenin and detects full-length (92 kDa) in active and inactive forms. Total β-catenin localization was observed mostly in the cell membrane regions in both cell types (Fig. 2b).
To confirm β-catenin status in these cells, we performed Western blot analysis of total lysates derived from Huh7-RFP and Huh7-RFP-DCLK1 cells (Fig. 2d). While full-length total β-catenin (92 kDa) levels were similar in the lysates, a 48-kDa active β-catenin form was significantly enhanced (2.5 ± 0.5 fold increase) in Huh7-RFP-DCLK1 cells (lane 2) compared with the Huh7-RFP control (lane 1). This increase was accompanied by an enhanced level of cyclin D1 (lane 2), a direct target of active β-catenin. These data suggest that DCLK1 activation may promote cyclin D1-mediated effects in HCC.
Cell fractionation analysis showed a significant increase in 48-kDa active β-catenin in the cytoplasmic and nuclear fractions of Huh7-RFP-DCLK1 cell lysates (Fig. 2e, lanes 3, 6) compared with control (Huh7-RFP, lanes 2, 5) and parent Huh7 lysates (lanes 1, 4). Quantitative analysis of the band intensities revealed a 4.7-fold and twofold increase of 48-kDa active β-catenin in the cytoplasmic and nuclear fractions of Huh7-RFP-DCLK1 cells respectively compared with the corresponding Huh7-RFP cell lysates (Fig. 2f). This 48-kDa band was not detected by either anti-total β-catenin antibodies (PLA0230) (Fig. 2e, second panel from top) or anti-β-catenin (phosphor Y654) antibody (ab59430) (not shown) in the cytoplasmic and nuclear lysates. Of note, the ABC(8E7) did not detect 52–56 kDa β-catenin32 in the lysates of these cells (Fig. 2e). These data suggest that the observed 48-kDa active β-catenin represents a major portion of unphosphorylated or hypophosphorylated N-terminus, but lacks the C-terminus regulatory region of the full-length active β-catenin. Nuclear fractions of these cells showed enriched TCF-4 transcription factors compared with the cytoplasmic fractions in all the cell types (Fig. 2e, lower panel). The lamin B1 band intensities for all three cell-types in the nuclear fractions were similar but it was absent in the cytoplasmic fractions (Fig. 2e, fourth panel from top). These results confirmed that the cell fractionation and loading were appropriate.
DCLK1-overexpressing hepatoma cells exhibit clonogenicity and produce dedifferentiated lineages
Huh7-RFP-DCLK1 and Huh7-RFP cells were prepared as single cell suspensions and added to the wells of chambered glass covers containing matrigel to determine clonogenicity of DCLK1-overexpressing cells and cell phenotypes of spheroids. Live images of spheroids derived from these cells were recorded at 4, 8, and 12 days after plating. The results revealed that RFP-DCLK1 overexpression in Huh7 hepatoma cells led to an increase in both the number (2.5 fold) and size of the spheroids (Fig. 3a and b) compared with the Huh7-RFP-derived spheroids. (Fig. 3a, lower panel). Immunofluorescence staining of spheroids directly in the matrigel culture using confocal microscopy showed cells expressing RFP-DCLK1, active β-catenin, α-fetoprotein, and SOX9 (Fig. 3c, upper panel). Total β-catenin staining was mostly localized to the cell membrane (left panel) and was similar to that usually observed in epithelial cells. The results suggest that Huh7-RFP-DCLK1 cells possess clonogenic properties and are capable of producing α-fetoprotein + bipotent progenitor and SOX9 + dedifferented/tumor stem-like cells. The Huh7-RFP cells (control) also formed spheroids, but at lower growth rate, and reduced number of spheroids than Huh7-RFP-DCLK1 (Fig. 3a lower panel). In addition, these spheroids also produced cell lineages with much lower levels of active β-catenin, α-fetoprotein, and SOX9 (Fig. 3c, lower panel), indicating differentiated cell phenotypes.
Taken together, these data (Figs. 2 and 3) suggest that DCLK1-overexpressing cells efficiently generate a 48-kDa form of active β-catenin with preserved unphosphorylated N-terminus, which results in clonogenic and dedifferentiated cellular phenotypes.
DCLK1 enhances β-catenin signaling via increased phosphorylation of GSK3β-Ser9
Since nuclear accumulation of the 48-kDa form of active β–catenin was observed in DCLK1 + cells, we determined its interaction with the transcription factor TCF-4 by immunoprecipitation (IP) using nuclear lysates of Huh7-RFP-DCLK1 cells. IP with anti-TCF-4 antibody followed by probing blots with anti-ABC(8E7) showed binding of the 48-kDa form of active β–catenin with nuclear TCF-4 (Fig. 4a, lane 3), suggesting that the short form retains a TCF-binding site within the armadillo (Arm) repeats (Arm 4 through 8). We next employed a luciferase-based reporter assay to determine whether the 48-kDa β–catenin-TCF-4 interaction activates a downstream target gene. The β-catenin activity in Huh7-RFP-DCLK1 cells were measured by transient transfection of TOPFlash (wild-type promoter-Luc) or FOPFlash (mutant promoter-Luc) luciferase (Luc) reporter plasmids. The results showed that TCF-4/LEF-responsive wild type promoter activities were significantly enhanced (fourfold) compared with the mutant TCF/LEF promoter in Huh7-RFP-DCLK1 cells (Fig. 4b). These observations clearly suggest that the interaction of the small 48-kDa form of β–catenin is capable of activating TCF-4-dependent transcription of target genes.
In the absence of Wnt signaling, GSK-3β phosphorylates the N-terminus of β–catenin at Ser33, Ser37, and Thr45 residues for proteasomal degradation33. However, GSK-3β kinase activity is inhibited due to Ser9 phosphorylation by several kinases (e.g., AKT, p70S6, p90Rsk), leading to stabilization of β-catenin34. To understand this dynamic in our cell lines, we next performed Western blot analysis of total cell lysates derived from the three cell-types described above. Huh7-RFP-DCLK1 cells demonstrated an increased Ser9 phosphorylation of GSK-3β (Fig. 4c, lane 3, top panel) as compared to the corresponding controls (lanes 1, 2), suggesting that overexpression of DCLK1 may inhibit classic proteasome degradation of β-catenin due to downregulation of GSK-3β kinase activity. Total GSK-3α and GSK-3β levels were not affected by RFP or RFP-DCLK1 expression in these cells (middle panel). To confirm whether a DCLK1-dependent mechanism generates the 48-kDa active β-catenin, we repeated this analysis using a validated anti-DCLK1 shRNA as an inhibitor16. Downregulation of DCLK1 resulted in a dramatic reduction of the 48-kDa form of active β-catenin form (Fig. 4d, lane 2).
Taken together, these experiments show that DCLK1 activates β-catenin (48-kDa) signaling. In addition, GSK-3β activities that are required for N-terminus hyperphosphorylation-mediated degradation of β-catenin are downregulated due to increased GSK-3β-Ser9 phosphorylation in DCLK1-overexpressing hepatoma cells.
Bortezomib increases 48-kDa active β-catenin but not the full-length protein in DCLK1-overexpressing cells
β-catenin is a target of proteasomal processing and degradation. Bortezomib inhibits proteasomal activities, and triggers cell-death by inducing apoptosis and autophagy pathways35. To determine whether the 48-kDa form of active β-catenin is generated by proteasomal activities, Huh7-RFP-DCLK1 cells were treated with increasing amounts (5 µM to 125 µM) of bortezomib or equivalent amounts of solvent (DMSO) for 48 h. Bortezomib treatment of Huh7-RFP-DCLK1 cells caused increased accumulation of 48-kDa active β-catenin (Fig. 5a, lanes 2–5) compared with untreated control (lane 1). This effect was not observed for DMSO controls (lanes 6–9). The biological effects of bortezomib on these cells was confirmed by the activation of autophagy, as indicated by cleavage of LC3B-I (third panel from top, lanes 2–5) compared with the untreated (lane 1) and DMSO-treated cells (lanes 6–9). Full-length total β-catenin (92 kDa)/actin ratios were similar in the bortezomib-treated and untreated cells (Fig. 5b, lanes 2–4), suggesting that proteasome inhibition did not affect levels of full-length β-catenin. At all the DMSO treatment concentrations, total β-catenin (92 kDa) levels were similar (lanes 6–9). Bortezomib is also known to induce caspase 3 cleavage, which is indicative of apoptosis. However, the levels of full-length caspase 3 were similar in the bortezomib-treated and untreated Huh7-RFP-DCLK1 cells. These results suggest that the level of 48-kDa active β-catenin is regulated by a proteasome-independent mechanism.
DCLK1-positive cells express 48-kDa active β-catenin in livers of patients with cirrhosis and HCC
A large percentage (60%-70%) of patients with cirrhosis and HCC express DCLK1 in their livers and this expression correlates with a worse prognosis14,16. To assess whether DCLK1-mediated alteration in β-catenin correlated with clinical activity, we first analyzed the TCGA database for β-catenin mRNA expression levels in HCC, cholangiocarcinoma, and colorectal adenocarcinoma. The expression levels were normalized using normal tissues. There was increased β-catenin mRNA in colorectal cancers (n = 288, p < 0.001) compared with normal samples (n = 41, Fig. 6a). However, β-catenin mRNA levels in HCC (n = 370, p = 0.544) and cholangiocarcinoma (n = 36, p = 0.018) showed variable expression compared with levels found in normal tissues. The survival patients with HCC was significantly reduced with high expression (n = 76, Fig. 6b), suggesting that levels of β-catenin is likely to influence clinical outcomes.
To determine whether active β-catenin and DCLK1 are expressed in the same liver cells, we performed co-immunohistochemical staining of liver tissues from patients with cirrhosis and HCC (n = 20). The immunohistochemistry results revealed specific co-staining of active β-catenin (red) and DCLK1 (brown) in the same epithelial cells (Fig. 7a, five representative cases are shown). Healthy/normal liver did not display staining for DCLK1 or active β-catenin (top left panel) (also see 1 normal and 4 HBV cases in Supplementary Fig. S2). Next, we determined relative expression of DCLK1, and active and total β-catenin in the total lysates prepared from liver tissues of patients with cirrhosis and HCC (Fig. 7b). We observed a clear, upregulated expression of DCLK1 and 48-kDa active β-catenin in these patients (lanes 3–7). Although C3 lysate (lane 4) showed weak DCLK1 band, staining of a similar blot with an another anti-DCLK1 antibody (ab31704) revealed 48-kDa and 45 kDa DCLK1 bands, most likely representing alternatively spliced isoforms (not shown). Patients’ liver tissues exhibited downregulation of full-length total β-catenin (92 kDa) compared with normal liver (lane 1). Interestingly, patient C1 (lane 2) did not exhibit any correlation between DCLK1 and active β-catenin, suggesting that DCLK1 expression may have influenced other tumorigenic stimuli in this case.
Cleavage of full-length E-cadherin (120-kDa) by a number of proteinases at the cell surface produces soluble 80-kDa ectodomain, which reduces epithelial features and induces migratory phenotypes due to disruption of cell–cell interactions36,37,38. We determined soluble E-cadherin levels in the cell culture supernatant of DCLK1-overexpressing hepatoma cells to determine whether DCLK1 overexpression in the liver diseases is related to change in epithelial phenotypes of the cells. Within 24 h of cell passage, we observed a modest increase in the cleaved/soluble E-cadherin level in the media supernatant of Huh7-RFP-DCLK1 cells compared with the media of Huh7-RFP cultures (Fig. 7c). Western blot analysis revealed two predominant forms (120-kDa and 80-kDa) of E-cadherin in cirrhotic and HCC liver lysates (Fig. 7d, lanes 2–7). This 80-kDa E-cadherin form was absent from normal liver tissues (lane 1). These results show that hepatic epithelial cells acquire DCLK1 + /active β-catenin + phenotypes in cirrhosis and HCC, and is accompanied by appearance of cleaved E-cadherin. In H1 and H2 cases (Lanes 6, 7), the expression of both forms of E-cadherin was milder than cirrhosis because of extensive loss of epithelial cells as determined by H&E staining (not shown). These changes would suggest a loss of the epithelial polarity, acquisition of migratory phenotypes, and enhanced neoplastic characteristics of the cells after DCLK1 overexpression.
DCLK1 mediates activation of β-catenin in humanized FRG mouse livers
We next investigated DCLK1 and β-catenin expression in vivo using an animal model that recapitulates liver injury and propagation of human hepatocytes. Fumarylacetoacetate hydrolase (FAH) deficiency in the immunodeficient FRG mouse results in liver damage due to accumulation of toxic tyrosine catabolites. However, NTBC (nitisinone) prevents liver damage in these mice by blocking 4-hydroxyphenylpyruvate (4-HPD) two steps upstream of FAH in the tyrosine catabolic pathway39. Using this model, we cyclically withdrew NTBC from their drinking water and intrasplenically transplanted primary human hepatocytes anticipating that these cells would expand and repopulate mouse livers with the human hepatocytes (scheme in Fig. 8a). This process created a chimeric mouse model in which to investigate the role of DCLK1 expression in vivo.
RFP or RFP-DCLK1 expressing primary human hepatocytes were transplanted into FRG mice (5 mice per group). The RFP fluorescence was clearly detected mostly in the abdominal region 10 days post-transplantation for both RFP and RFP-DCLK1 groups (Fig. 8b, indicated with red arrow). After euthanasia, the mouse livers were subjected to immunohistochemistry for human albumin (hAlb), DCLK1, and active β-catenin. RFP-DCLK1 chimeric mice showed extensive staining for hAlb and DCLK1 (Fig. 8c, right panel, only one representative section is shown). The RFP chimeric liver exhibited hAlb but rare DCLK1 + cells with weak staining (middle panel). In RFP-DCLK1 chimeric liver, we also noticed proliferative nodules staining positive for DCLK1 and hAlb (Fig. 8e). The control (untransplanted) mouse livers were negative for both DCLK1 and hAlb (Fig. 8c, left panel), indicating that liver injury alone due to FAH deficiency is not sufficient to induce DCLK1 in these mice. Together, these results suggest that DCLK1-expressing human hepatocytes can repopulate into the FRG liver microenvironment.
To further determine whether the human hepatocytes exhibit the potential to adopt more immature or dedifferentiated phenotypes, we examined active β-catenin expression in the DCLK1-expressing cells by co-immunostaining. A number of DCLK1 + hepatocytes exhibited cytoplasmic and nuclear accumulation of active β-catenin in RFP-DCLK1 chimeric livers but not in RFP control livers (Fig. 8d, highlighted in lower panel). Staining of the FRG livers for full-length total β-catenin revealed localization mostly in the cell membrane in both RFP-DCLK1 and RFP chimeric mice (Supplementary Fig. S3, left panel). However, active β-catenin staining was only seen in the nucleus and cytoplasm of RFP-DCLK1 chimeric liver (right panel), suggesting that the observed active β-catenin in RFP-DCLK1 chimeric liver lacks C-terminus but retains a hypophosphorylated N-terminus. This finding is reminiscent of the results from the in vitro experiments. Taken together, these results clearly suggest that DCLK1-overexpressing primary human hepatocytes exhibit a propensity to adapt undifferentiated phenotypes due to activation of β-catenin and can propagate in the liver microenvironment.