Elimination of head and neck cancer initiating cells through targeting glucose regulated protein78 signaling

Previously, we have demonstrated the existence of HN-CICs [13]. To further elucidate the molecular mechanisms by which to mediate the self-renewal ability and tumorigenicity of HN-CICs, molecular targets specifically expressed in HN-CICs were to be identified. The differential expression profile between HN-CICs and HNSCCs was examined by either systemic transciptome analysis or two-dimensional differential gel electrophoresis (2-D DIGE) followed by mass spectroscopy analysis. We noticed that the transcripts and protein level of GRP78 were significantly up-regulated in enriched HN-CICs (Additional file 1 and Figure 1A). To further validate the results from Affymatrix microarray and proteomic analyses, western blotting was performed. Immunoblotting analyses showed that antibody against GRP78 detected more GRP78 protein in crude cell extracts of enriched HN-CICs than in that of parental HNSCCs (Figure 1B).

Recent findings of GRP78 on plasma membrane of cancer cells but not on normal cells suggest that targeted therapy against surface GRP78 of cancer cells may be effective [24]. Compared to parental HNSCCs, we found more membrane-associated GRP78 positive (^memGRP78⁺) cells in HN-CICs by FACS analyses (Figure 1C). In addition, it has been demonstrated that aldehyde dehydrogenase 1 (ALDH1) activity could be used as a selection marker to isolated breast cancer CICs and head and neck CICs [25, 26]. Consistent with tumor spheres formation ability, ALDH1⁺ HNSCCs also displayed more ^memGRP78⁺ cells (Figure 1D). Finally, HN-CICs showed elevated co-expression of either CD133 or Cripto-1 with ^memGRP78 in comparison to parental HNSCCs (Figure 1E and 1F), where both CD133 and Cripto-1, the well known CICs markers, have been used to identify CICs [13, 27, 28]. Taken together, we hypothesized that up-regulation of GRP78/^memGRP78 is pivotal for maintenance cancer stemness characteristics of HN-CICs.

To test whether ^memGRP78⁺ HNSCCs had the CICs characteristics, SAS cells were sorted into ^memGRP78⁺ and ^memGRP78^- cells by flow cytometry (Additional file 2A). Compared with ^memGRP78^- SAS cells, the ^memGRP78⁺ SAS cells displayed higher levels of protein and mRNA of stemness genes (Oct-4 and Nanog) (Figure 2A and Additional file 2B). We next performed tumor spheres assay for evaluating the self-renewal ability of ^memGRP78⁺ and ^memGRP78^- cells, respectively. Interestingly, ^memGRP78⁺ cells had higher tumor spheres-forming capability than ^memGRP78^- HNSCCs (Figure 2B). When isolated ^memGRP78⁺ and ^memGRP78^-cells were first cultivated within 10% serum for 10 days, then the cell surface GRP78 expression profile was further analyzed by flow cytometry, respectively. We observed that ^memGRP78⁺ cells regenerated both ^memGRP78⁺ and ^memGRP78^- cells, whereas, ^memGRP78⁺ cells were not detectable from cultivated ^memGRP78^- cells (Figure 2C). These data indicate that ^memGRP78⁺ HNSCCs could re-differentiate into ^memGRP78^- cells. To address whether the tumorigenic activity differed between ^memGRP78⁺ and ^memGRP78^- cells, in vitro tumorigenic properties including matrigel invasion and anchorage independent growth, and in vivo xenografts assay were performed. The colony/invasion formation abilities of ^memGRP78⁺ HNSCCs were significantly higher than those of the ^memGRP78^- HNSCCs (Figure 2D and 2E). To further evaluate the correlation between ^memGRP78 expression profile and radioresistance, we established radioresistant (R) HNSCCs (R1, R2, and R3) by serially fractionated irradiation (see details from Material and methods). We found that the expression profile of ^memGRP78 was significantly enhanced in radioresistant HNSCCs (Figure 2F; R3>R2>R1>Parental OECM1). For in vivo xenotransplantation assay, we observed that 10000 GRP78^- cells did not induce tumor formation but 100 GRP78⁺ HNSCCs resulted in the generation of visible tumors 4 weeks after injection in xenotransplanted mice (Figure 2G, H, and 2I, Additional file 2D, 2E and 2F). Collectively, ^memGRP78 positive cells possess the capabilities for self-renewal, differentiation, radioresistance and high in vivo tumorigenicity.

To further investigate the crucial role of GRP78 up-regulation in maintaining biological properties of HN-CICs and HNSCCs, we performed the loss-of-function approach to evaluate the effect of GRP78 knockdown on HNSCCs derived HN-CICs. First, the HNSCCs derived HN-CICs were generated by cultivating HNSCCs under defined serum-free medium as described [13]. Then, the enriched HN-CICs were infected with lentivirus expressing either small hairpin RNA (shRNA) targeting GRP78 (shGRP78) or shRNA against luciferase (shLuc), respectively. HN-CICs infected with shLuc lentivirus were used as control cells. Successful infected HN-CICs was validated as the Green Fluorescence Protein (GFP) positive cells since GFP was co-expressed as a reporter marker for cell transduction (data not shown). Western blot analyses confirmed that both sh-GRP78-1 and sh-GRP78-2 markedly repressed GRP78 protein expression in both HN-CICs and HNSCCs (Figure 3A and Additional file 3A). ^memGRP78⁺ cells were also reduced in shGRP78-expressing HN-CICs and HNSCCs (Figure 3B and Additional file 3B). Differential levels of GRP78 suppression between membrane and cytosol in head and neck cancer initiating cells by western blotting and flow cytometry results were examined in Additional file 3C.

Tumor-derived side population (SP) cells also have been found to have characteristics of cancer stemness [29].GRP78 depletion significantly decreased the side population in HN-CICs and HNSCCs, respectively (Figure 3C and Additional file 3D). To further investigate whether GRP78 expression plays a role in maintaining self-renewal or cancer stem-like properties in HN-CICs directly, the HNSCCs-derived tumor spheres, afterward transduction with Sh-GRP78 lentivirus, did not maintain floating spheres but show more attached epithelial-like cells (Figure 3D). In opposite, HN-CICs after Sh-GRP78 lentiviruses infection displayed decreased expression of "cancer stemness" genes (Oct-4, Nanog, and Nestin) but enhanced expression of epithelial differentiation marker, CK18 and Involucrin (Figure 3E and 3F). To determine whether the reduction in tumor sphere formation efficiency with GRP78 down-regulation is due to decreased HN-CICs survival, we determined the percentage of apoptotic cells using Annexin V staining. HN-CICs transduced with Sh-GRP78 lentivirus significantly increased the percentage of Annexin V-positive cells (Figure 3G). Together, these results further support that the loss of GRP78 resulted in a decrease of CICs properties due to up-regulation differentiation and apoptotic activity.

To elucidate the direct effect of GRP78 knockdown on in vitro tumorigenic properties including cell migration, matrigel invasion and anchorage independent growth of HN-CICs, single cell suspension of control- or GRP78- knockdown HN-CICs were plated onto transwell chamber (Figure 4A), onto transwell chamber coated with matrigel (Figure 4B) or into soft agar (Figure 4C), and analyzed as described in Materials and Methods, respectively. The migratory/invasion/colony formation abilities of GRP78 knockdown HN-CICs were significantly reduced than those of the control HN-CICs (Figure 4A, B, and 4C). We next sought to determine if down-regulation of GRP78 expression could attenuate the tumor initiating activity of HN-CICs in vivo. Strikingly, GRP78-knockdown HN-CICs gave rise to a new tumor at 5x10⁵ in one of six mice, however, HN-CICs control cells generated tumor when 1x10⁴ cells were injected into nude mice (three out of three mice)(Figure 4D). In addition, knockdown of GRP78 expression in HN-CICs and HNSCCs significantly reduced the tumor volumes (Figure 4E and Additional file 3E). Overall, our data indicate that down-regulation of GRP78 inhibited in vitro tumorigenicity and in vivo tumor-initiating activity of HN-CICs.

To evaluate whether overexpression of GRP78 could enhance tumorigenic properties of HNSCCs, we generated HNSCCs with transient overexpression of GRP78 by transfection with plasmids overexpressing GRP78 protein into HNSCCs. Total proteins from 293T cells or HNSCCs (SAS) with transfection of GRP78 expressing plasmids displayed elevated expression of GRP78 (Additional file 4A). Furthermore, we demonstrated that GRP78 overexpression also resulted in increased ability on in vitro cell migration (Additional file 4B). To evaluate whether overexpressios of GRP78 on promoting ^memGRP78⁺ cells in HNSCC, SAS cells were co-transfected with plasmids expressing green fluorescence protein (GFP) and GRP78. We discovered GFP positive cells (meaning cells under successful transfection) showed more ^memGRP78⁺ in co-transfected cells than control cells (Additional file 4B). Together, our data demonstrated that overexpression of GRP78 not only enhanced in vitro malignancy but also expression profile of ^memGRP78⁺ in HNSCCs.

We have been reported that HNSCC patients with abundant Nanog protein expression are more likely to have poor survival outcomes [13]. Overexpression of GRP78 also correlates with poor HNSCC prognosis [30]. To further investigate the correlation between GRP78 and Nanog levels in human cancers, we established the ontogeny of GRP78 and Nanog co-expression by tissue immunohistochemical staining with a panel of specimens array of 46 HNSCC patients. Two representative cases with double-positive or double-negative of GRP78 and Nanog were shown in Figure 5A. We found co-expression of GRP78 and Nanog in the moderate to poor-differentiated HNSCC tissues rather than in well-differentiated HNSCC tissues (Figure 5A). The significant correlation between the expression of GRP78 and Nanog in HNSCC tissues was determined (Figure 5B, p < 0.05). To investigate the prognostic significance of the expression GRP78 and Nanog patterns in HNSCC, we divided patients into four groups: GRP78 (+)Nanog (+), GRP78 (+), Nanog (+), and GRP78 (-)Nanog(-) HNSCC patients. The Kaplan-Meier analyses showed that co-expression of GRP78 and Nanog predicted the worse overall survival than all other HNSCC patients (Figure 5C).

To identify the systemic differential gene expression profile by down-regulation of GRP78 in HN-CICs, we performed Affymetrix microarray analyses. Upon the knockdown of GRP78, we identified 434 probes consistently induced or repressed and mapped them onto the human PPIs. We filtered the mapped PPIs among the differentially expressed genes by their co-expression of the reactants in the GRP78-knockdown HN-CICs (PCCs > 0.5). As shown in Figure 6A and 6B, 79 genes and 64 interactions were retained in the final networks. The direction and strength of co-expression were depicted in Figure 6B. Highly correlated genes were CTNNB1 v.s. PTPN11, E2F1 v.s. CDC6, E2F1 v.s. RECQL, and MCM5 v.s. RPA2, with positive PCCs, as well as CHEK1 v.s. E2F1, PSMA1 v.s. DLEU1, and HSPA8 v.s. NFKBIB, with negative PCCs. Topologically, 24 inter-modular hubs, 4 intra-modular hubs, and 51 periphery genes. Functional annotation of the 79 genes in the networks of GRP78 knockdown in HN-CICs was summarized in Figure 6C. To further study the possible mechanisms involved in GRP78-mediated cancer stemness properties, we found out knockdown of GRP78 enhanced the expression of PTEN, BAX and Caspase3 but reduced the expression of p-MAPK in HN-CICs (Figure 6D). These results support PTEN-PI3K-Akt and ERK signaling is regard as crucial pathways in mediating CICs characteristics [31, 32]. Additionally, GRP78 might regulate survival pathways to modulate HN-CICs behaviors.

Originally, SAS was grown in DMEM, and OECM1 was grown in RPMI supplemented with 10% fetal bovine serum (FBS) (Grand Island, NY), respectively. The two cell lines were then cultured in tumor sphere medium consisting of serum-free DMEM/F12 medium (GIBCO), N2 supplement (GIBCO), 10 ng/mL human recombinant basic fibroblast growth factor-basic (FGF) and 10 ng/mL Epidermal Growth Factor (EGF) (R&D Systems, Minneapolis, MN) (. Cells were plated at a density of 7.5 × 10⁴ live cells/10-mm dish, and the medium was changed every other day until the tumor sphere formation was observed in about 4 weeks [13].

RNA wasextracted from cells using Trizol reagent (InvitrogenLife Technologies), purity confirmed by OD260:280 ratio and analyzed using formaldehyde gel electrophoresis. For Affymetrix GeneChipanalysis, RNAeasy kit (Qiagen, Valencia, CA) was used forfurther RNA purification. Gene profiling was performedusing Affymetrix Human Genome U133 plus 2.0 (containing 47,000 transcripts and variants, including 38,500 well-characterized human genes) for the microarrays hybridization at the genomic core facilities at the National Yang-Ming University Genome Research Center.

The pLV-RNAi vector was purchased from Biosettia Inc. (Biosettia, San Diego, CA). The method of cloning the double-stranded shRNA sequence is described in the manufacturer's protocol. Lentiviral vectors expressing short hairpin RNA (shRNA) that targets human GRP78 (oligonucleotide sequence: Sh-GRP78-1:5'- AAAAGCCTAAATGTTATGAGGATCATTGGATCCAATGATCCTCATAACATTTAGGC -3';Sh-GRP78-2:5'-AAAAGGAGCGCAUUGAUACUAGATTTTGGATCCAAAATCTAGTATCAATGCGCTCC-3') were synthesized and cloned into pLVRNAi to generate a lentiviral expression vector. Lentivirus production was performed by transfection of plasmid DNA mixture with lentivector plus helper plasmids (VSVG and Gag-Pol) into 293T cells using Lipofectamine 2000 (LF2000, Invitrogen, Calsbad). Supernatants were collected 48 hours after transfection and then were filtered; the viral titers were then determined by FACS at 48 hours post-transduction. Subconfluent cells were infected with lentivirus in the presence of 8 μg/ml polybrene (Sigma-Aldrich). The GFP is expressed in lentivirus-infected cells as the marker to indicate that the cells express the shRNA for silencing GRP78.

To measure and isolate cells with ALDH activity, the Aldefluor assay was performed according to manufacturer's (Stemcell Technologies, Durham, NC, USA) guidelines. Dissociated single cells were suspended in Aldefluor assay buffer containing the ALDH substrate, Bodipy-aminoacetaldehyde (BAAA) at 1.5 mM and incubated for 40 min at 37°C. To distinguish between ALDH-positive and ALDH-negative cells, a fraction of cells was incubated under identical condition in the presence of a 10-fold molar excess of the ALDH inhibitor, diethylaminobenzaldehyde (DEAB). This results in a significant decrease in the fluorescence intensity of ALDH-positive cells and was used to compensate the flow cytometer.

Cells were resuspended at 1 × 10⁶/mL in pre-warmed DMEM with 2% FCS. Hoechst 33342 dye was added at a final concentration of 5 μg/mL in the presence or absence of verapmil (50 μM; Sigma) and was incubated at 37°C for 90 min with intermittent shaking. At the end of the incubation, the cells were washed with ice-cold HBSS with 2% FCS and centrifuged down at 4°C, and resuspended in ice-cold HBSS containing 2% FCS. Propidium iodide at a final concentration of 2 μg/mL was added to the cells to gate viable cells. The cells were filtered through a 40-μm cell strainer to obtain single cell suspension before analysis. The Hoechst 33342 dye was excited at 357 nm and its fluorescence was dual-wavelength analyzed (blue, 402-446 nm; red, 650-670 nm). Analyses were done on FACSAria (BD, San Diego, CA).

Cells were seeded on 75T flask at a density of 2 × 10⁵ in medium; kept culturing part of the cells for next radiation treatment after ionizing irradiation and repeat three times. The radiation resistant (R1, R2 and R3) cells were for further experiments. The g-radiation (ionizing irradiation) was delivered by Theratronic cobalt unit T-1000 (Theratronic International) at a dose rate of 1.1 Gy/min (SSD = 57.5 cm).

For transwell migration assays, 2 × 10⁵ cells were plated into the top chamber of a transwell (Corning, Acton, MA) with a porous transparent polyethylene terephthalate membrane (8.0 μm pore size). Cells were plated in medium with lower serum (0.5% FBS), and medium supplemented with higher serum (10% FBS) was used as a chemoattractant in the lower chamber. The cells were incubated for 24 h and cells that did not migrate through the pores were removed by a cotton swab. Cells on the lower surface of the membrane were stained with Hoechst 33258 (Sigma-Aldrich) to show the nuclei; fluorescence was detected at a magnification of 100× using a fluorescence microscope (Carl Zeiss, Oberkochen, Germany). The number of fluorescent cells in a total of five randomly selected fields was counted.

The 24-well plate Transwell^® system with a polycarbonate filter membrane of 8-μm pore size (Corning, United Kingdom) was employed to evaluate the invasion ability of cells. The membrane was coated with Matrigel™ (BD Pharmingen, NJ, USA). The cancer cell suspensions were seeded to the upper compartment of the Transwell chamber at the cell density of 1 × 10⁵ in 100 μl within serum-free medium. The lower chamber was filled with media with 10% serum. After 24 hours of incubation, the medium was removed and the filter membrane was fixed with 4% formalin for 1 hour. Subsequently, the remaining cells of the filter membrane faced the lower chamber was stained with Hoechst 33258 (Sigma-Aldrich). The migrated cancer cells were then visualized and counted from 5 different visual areas of 100-fold magnification under an inverted microscope.

Each well (35 mm) of a six-well culture dish was coated with 2 ml bottom agar (Sigma-Aldrich) mixture (DMEM, 10% (v/v) FCS, 0.6% (w/v) agar). After the bottom layer was solidified, 2 ml top agar-medium mixture (DMEM, 10% (v/v) FCS, 0.3% (w/v) agar) containing 10⁴ cells were added, and the dishes were incubated at 37°C for 4 weeks. Plates were stained with 0.005% Crystal Violet then the colonies were counted. The number of total colonies with a diameter = 100 μm was counted over five fields per well for a total of 15 fields in triplicate experiments.

Between 1994 and 1997, 46 consecutive patients with operable head and neck cancer underwent surgery at the Department of Oral and Maxillofacial Surgery, Mackay Memorial Hospital. This research follows the tenets of the Declaration of Helsinki and all samples were obtained after informed consent from the patients. None of the subjects received radiation therapy or chemotherapy before surgery.Forty-six patients' tissue samples with different stages of oral cancer were spotted on glass slides for immunohistochemical stainings. After deparaffinization and rehydration, the tissue sections were processed with antigen retrieval by1X Trilogy diluted in H₂O (Biogenics) and heat. The slides were immersed in 3% H₂O₂ for 10 minutes and washed with PBS 3 times. The tissue sections were then blocked with serum (Vestastain Elite ABC kit, Vector Laboratories, Burlingame, CA) for 30 minutes, followed by incubating with the primary antibody, anti-GRP78 (BD Transduction Laboratories™) in PBS solution at room temperature for 2 hours in a container. Tissue slides were washed with PBS and incubated with biotin-labeled secondary antibody for 30 minutes and then incubated with streptavidin-horse radish peroxidase conjugates for 30 minutes and washed with PBS 3 times. Afterwards, the tissue sections were immersed with chromogen 3-3'-diaminobenzidine plus H₂O₂ substrate solution (Vector^® DBA/Ni substrate kit, SK-4100, Vector Laboratories, Burlingame, CA) for 10 minutes. Hematoxylin was applied for counter-staining (Sigma Chemical Co., USA). Finally, the tumor sections were mounted with a cover slide with Gurr^® (BDH Laboratory Supplies, U.K.) and examined under a microscope. Pathologists scoring the immunohistochemistry were blinded to the clinical data. The interpretation was done in five high-power views for each slide, and 100 cells per view were counted for analysis. (-, 0-10% positive cells; +, more than 10% positive cells)

All the animal practices in this study were in accordance with the institutional animal welfare guideline of National Yang-Ming University, Taiwan. HNSCCs or HN-CICs subject to treatment were injected subcutaneously into BALB/c nude mice (8 weeks). Tumor volume (TV) was calculated using the following formula: TV (mm3) = (Length × Width ²)/2 and then analyzed using Image Pro-plus software.

All CEL files were pre-processed using method justRMA and standardized with mean of zero and SD of 1. First, modified t-test of the 'limma' package was used for differential gene expression analysis between the control- or GRP78- knockdown HN-CICs, controlled for FDR < 0.05 [55]. The analysis focused on precompiled calcium, migratory [56, 57]and stemness related gene lists [58, 59]. Second, we further filtered out differential expression gene signatures with any inconsistent direction of regulation between any pair of control- v.s. GRP78- knockdown HN-CICs. Third, differentially expressed probes were mapped onto the human PPIs downloaded from the NCBI Gene Portal (HPRD, BioGrid, and BIND). PPIs would be retrieved if and only if both of the interactants were listed as of those differentially expressed. Fourth, absolute values of Pearson correlation coefficients (PCCs) of the mapped PPIs were calculated to identify cut-off threshold at 0.5 to filter out possible false-positive interactions. Finally, network topological analyses and classification of genes were performed according to methods previously published [60]. Analytical computation, hierarchical clustering and heatmap were performed and displayed using R statistical software [61]. Functional enrichment clustering of genes in the final mapped human PPIs was analyzed by DAVID (Database for Annotation Visualization and Integrated Discovery, NIH) [62].

To overexpress the GRP78 protein in HNSCCs, a plasmid (pCMV-GRP78; a gift from Dr. Ann-Joy Cheng, Chang Gung University, Taipei, Taiwan) which can overexpress the GRP78 in mammalian cells under CMV promoter, was introduced HNSCCs transiently by transfection. In the meanwhile, plasmids encoding green fluorescence protein were co-introduced into host cells to identify the successful transfection cells.

The independent Student's t-test was used to compare the continuous variables between groups, whereas the χ² test was applied for the comparison of dichotomous variable. Statistical Package of Social Sciences software (version 13.0) (SPSS, Inc., Chicago, IL) was used for statistical Kaplan-Meier analysis. The Kaplan-Meier estimate was used for survival analysis, and the log-rank test was selected to compare the cumulative survival durations in different patient groups. The level of statistical significance was set at 0.05 for all tests.