Detection of cancer stem cells by EMT-specific biomarker-based peptide ligands

The occurrence of epithelial-mesenchymal transition (EMT) within tumors, which enables invasion and metastasis, is linked to cancer stem cells (CSCs) with drug and radiation resistance. We used two specific peptides, F7 and SP peptides, to detect EMT derived cells or CSCs. Human tongue squamous carcinoma cell line-SAS transfected with reporter genes was generated and followed by spheroid culture. A small molecule inhibitor-Unc0642 and low-dose ionizing radiation (IR) were used for induction of EMT. Confocal microscopic imaging and fluorescence-activated cell sorting analysis were performed to evaluate the binding ability and specificity of peptides. A SAS xenograft mouse model with EMT induction was established for assessing the binding affinity of peptides. The results showed that F7 and SP peptides not only specifically penetrated into cytoplasm of SAS cells but also bound to EMT derived cells and CSCs with high nucleolin and vimentin expression. In addition, the expression of CSC marker and the binding of peptides were increased in tumors isolated from Unc0642/IR-treated groups. Our study demonstrates the potential of these peptides for detecting EMT derived cells or CSCs and might provide an alternative isolation method for these subpopulations within the tumor in the future.

www.nature.com/scientificreports/ as stabilizing cytoskeletal interactions 11 . Recently, vimentin is also regarded as a standard biomarker for EMT. The level of vimentin in tumor cells correlates well with increased tumor growth, invasion and poor prognosis 3 . Likewise, a SP peptide (SAHGTSTGVPWP) that specifically binds cell surface vimentin is identified by screening a phage display peptide library on human vascular endothelial cells under hypoxic conditions 12,13 . Herein, F7 and SP peptides are potential biomarker-based peptides to detect CSCs or EMT derived cells.
In the current study, we generated two human tongue squamous carcinoma derived cell lines with the fluorescent and luciferase genes, SAS-EGFP-Fluc and SAS-E2-crimson-P2A::ttksr39 cells, to investigate the binding ability and specificity of F7 and SP peptides. A small molecule inhibitor-Unc0642 and low-dose ionizing radiation (IR) were used to simulate the cancer recurrence after chemotherapy and radiotherapy, thereby inducing EMT. The in vitro data showed that the F7 and SP peptides were capable of detecting EMT derived cells with high vimentin and nucleolin expressions. In addition, a tumor xenograft mouse model with EMT induction was established to evaluate the binding ability of F7 peptide and SP peptide. Our results suggest that F7 peptide and SP peptide can serve as a tool to detect CSCs or EMT derived cells.

Results
Evaluation of peptide binding activity to CSCs. To track the EMT-derived cells both in vitro and in vivo, we transduced the reporter genes to the human tongue squamous carcinoma derived cell line-SAS using lentivirus, which were referred as SAS-EGFP-Fluc and SAS-E2-crimson-P2A::ttksr39 cells, respectively. The expression of E2-Crimson (red fluorescence) and EGFP fluorescence in stable cell lines were visualized under a confocal fluorescence microscopy (Fig. S1a). Fluorescence-activated cell sorting (FACS) analysis also revealed over 90% of stable lines exhibiting the higher fluorescence intensity compared to SAS-wt cells (Fig. S1b). SAS-EGFP-Fluc cells showed higher luciferase activity as demonstrated by luciferase assay (Fig. S1c). Furthermore, there was no significant difference in doubling time between SAS-wt, SAS-EGFP-Fluc and SAS-E2-crimson-P2A::ttksr39 cells, indicating that these transgenes did not affect cell growth (Fig. S1d). To evaluate whether F7 and SP peptides are able to bind to SAS cells, Lissamine Rhodamine (red fluorescence) labeled F7 peptide and FITC-labeled SP peptide, were used. As shown in Fig. 1a, the fluorescent signal derived from F7 and SP peptides were found in SAS cells that are known to express vimentin and nucleolin. HEK293 cells without detectable vimentin and nucleolin were used to be a negative control for the specificity of peptides. For further confirmation, the kinetics of F7 and SP peptide binding to SAS-EGFP-Fluc and SAS-E2-crimson-P2A::ttksr39 cells respectively were monitored at 37 °C for various periods of time. Within 30 min, either F7 peptide or SP peptide was observed to adhere to the cell membrane. One hour later, the fluorescent signal appeared in cytoplasm gradually and accumulated with the incubation time ( Fig. 1b and c). At 24 h, F7 and SP peptides remained at the edge of nucleus and no longer moved. These results indicated that both F7 and SP peptides penetrated into cytoplasm of SAS cells specifically and then accumulated within cells over time rather than bound randomly.
Studies have indicated that cultured tumor spheres exhibit the characteristics of CSCs including stemness, self-renewal and resistance to drugs and radiotherapy 14,15 . Thus, a non-adhesive culture system was used in this study to generate tumor spheroids of SAS cells for enrichment of CSCs cells or EMT-derived cells. The spheres were prepared and the morphological alteration was observed using optical microscope for 3 consecutive days (Fig. 2a). The spheres showed a round shape, smooth contour and clustered in a three-dimensional configuration growing gradually over time in comparison with adherent cells. To further investigate whether the cells undergo the process of EMT, the representative CSC/EMT markers CD44, E-cadherin, vimentin and nucleolin were used to differentiate the spheres from adherent cells. Immunostaining results showed the expressions of CD44 and vimentin increased in spheres, whereas the expression of E-cadherin was decreased (Fig. 2b). It is suggested that there was a remarkable increase in EMT after spheroid formation. Notably, the expression of nucleolin remained unchanged in spheres and adherent cells. After incubating with fluorescein-labeled peptides for 24 h, the red fluorescence and green fluorescence were observed in CD44 expressing spheres, suggesting that both F7 and SP peptides are able to bind and detect CSCs (Fig. 2c). Considering the tumor heterogeneity, both the spheres and adherent cells were incubated with fluorescein-labeled peptides respectively to evaluate the binding activity of peptides (Fig. 2d). The confocal images showed that F7 and SP peptides bound to cancer cells grown either in adherent or in spheroid form. Moreover, some of the two peptides were found to colocalize in cells (Fig. 2d, white arrows).
In vitro induction of EMT and stemness with Unc0642/IR. To increase the number of EMT-derived cells, Unc0642 and low-dose IR were used to simulate the cancer recurrence after chemotherapy and radiotherapy in our study. Small molecule inhibitor specifically targeting G9a (a lysine methyltransferase) has been found to disrupt DNA repair, thereby increasing sensitivity to radiation 16 . In this study, Unc0642 was chosen to inhibit G9a function. Accumulating evidence indicates that radiation can accelerate the spread of primary tumors and development of distant metastasis, attributing to the induction of EMT and the presence of CSCs with radioresistance 17,18 . At first, the IC 50 (50% inhibition concentration) was determined for Unc0642 and combined treatment with low-dose IR (Fig. 3a). The IC 50 value for SAS cells exposed with Unc0642 for 72 h was 12.09 μM. When combining with 4 Gy of IR, the IC 50 was found to decrease and occur at 4.79 μM. Of note, there was a peak at a dose of 1.25 μM in either single treatment or combined treatment, indicating this dose of Unc0642 could induce cell proliferation. Transforming growth factor-β1 (TGF-β1), as a multifunctional cytokine, has been reported to enhance cell migration and invasion by inducing EMT 19 . Thus, we ensured that the treatment of TGF-β1 upregulated the expression of vimentin but downregulated the expression of E-cadherin, as judged by immunoblotting (Fig. S2). Next, the effects of Unc0642 and IR on the EMT on SAS cells were explored. As shown in Fig. 3b, the expression of vimentin was increased in SAS cells exposed to IR (2 and 4 Gy) and 1. 25  www.nature.com/scientificreports/ exposed to 2 Gy of IR and 1.25 μM Unc0642 showed ~ twofold higher vimentin expression compared to control SAS cells (p < 0.01). The level of nucleolin was slightly increased in the presence of 1.25 μM Unc0642 (p < 0.05). In addition, we found a diminished protein expression level of E-cadherin in SAS exposed to Unc0642 or IR alone. Corresponding to the critical steps in EMT, high vimentin expression and loss of E-cadherin expression were found in the combined Unc0642-and IR-treated SAS cells. These results showed that the combination of Unc0642 and radiation therapy could be an effective method to promote cancer stemness and EMT phenotype.

Evaluation of peptide binding activity to EMT derived cells induced by Unc0642/IR. To inves-
tigate whether the peptide binding is increased by the induction of EMT, fluorescein-labeled F7 and SP peptides were added into EMT-induced SAS cells for 24 h respectively and FACS analysis was used to monitor peptide binding. The population of F7 peptide-binding cells was increased with increasing dose of IR and Unc0642, reflected by the enhanced red fluorescence (Fig. 4a). Likewise, we found the SP peptide bound to EMT-induced SAS cells and particularly the average fluorescence intensity (FITC) was significantly higher in combined Unc0642-and IR-treated SAS cells (Fig. 4b). The results indicated the potential of F7 and SP peptides to detect EMT derived cells with high expression of vimentin and nucleolin.  3 , the mice received an intraperitoneal injection of Unc0642 followed by IR starting from day 10 to day 16; the procedure for in vivo EMT induction was as shown in Fig. 5a. Tumor growth was monitored by bioluminescence imaging (Fig. 5b). Mice treated with Unc0642 or IR alone showed a clear delay in tumor growth, the Fluc signal decreased in comparison with control group on day 18, indicating that growth inhibition effect persisted for at least 1 week. Interestingly, no significant reduction was observed in Unc0642 plus IR treated mice until Day 21, but drastically dropped down on day 24. The tumor burdens in the mice receiving the IR alone and combined treatment of Unc0642 and IR suddenly increased on day 31, whereas the tumor burden in untreated mice kept growing constantly (Fig. 5c). Body weights were recorded twice a week before and after different treatments (Fig. 5d). Initially, there was no difference between the mice treated with Unc0642/IR alone and those treated with DMSO (the control group). Drastic weight loss occurred only in mice receiving the combined treatment of Unc0642 and IR until the end point of the study (p < 0.001). In light of these results, the tumor xenograft mouse model treated with drug and radiation successfully exhibited the tumor relapse and a malignant tendency of cancer cells.
To evaluate whether these two peptides are able to detect EMT derived cells, the mice implanted with SAS-EGFP-Fluc cells were given an intra-tumoral injection of F7 peptide and the mice implanted with SAS-wt cells were given an intra-tumoral injection of SP peptide, respectively on day 31 (Fig. S3). After 24 h, analysis of peptide binding and EMT markers from the excised tumor tissue was performed by immunofluorescence (IF) staining. The tumors from Unc0642/IR alone or Unc0642 plus IR-treated groups displayed higher CD44 expression than that of the control group ( Fig. 5e and f). The amounts of fluorescein labeled F7 peptide and SP peptide were also increased in tumor tissues respectively. These data were in accordance with the in vitro findings that F7 peptide and SP peptide have the potential to detect EMT derived cells.

Discussions
Our results demonstrated that F7 and SP peptides are able to penetrate into cytoplasm of SAS cells specifically. In addition, in vitro and in vivo induction of EMT studies showed that these two peptides could detect EMT derived cells or CSCs with high nucleolin and vimentin expression.
In cancer research, approaches in identification and isolation of subpopulations within tumors are based on the expression pattern of cell surface markers. Although the selection of unique markers is the first priority, the subsequent detection and isolation methods are another big issue that is related to the efficiency and translation into clinical utility 20 . Targeting selected markers via magnetic beads conjugated antibodies or fluorescein www.nature.com/scientificreports/ labeled antibodies are common methods with high specificity. The use of peptides offers several advantages in comparison with antibodies or ligands, including (1) low molecular weight and poorly immunogenicity, (2) ability to penetrate through cell membranes, (3) ease of synthesis using phage display and low cost, and (4) simple formulation assembly. Thus, the availability of such specific peptides might be suitable for the improvement of the abovementioned methods. In this regard, F7 and SP peptides that specifically bind to CSC marker nucleolin and EMT mesenchymal marker vimentin respectively can serve as a connector in the isolation system, such as Biotin-Avidin system and FACS or magnetic-activated cell sorting (MACS) techniques, to isolate the subpopulations with EMT/CSC properties. In addition, these peptides can be used in diagnostic or therapeutic applications, such as pharmaceutical carriers to selectively deliver imaging agents or chemotherapeutic agents 21 . For example, the liposomal doxorubicin conjugating F7 peptides specifically targets to nucleolin in MCF-7, a breast cancer cell line and accumulates within cells through enhanced permeability and retention effect (EPR effect), thereby reducing side effects and dose-limiting toxicities 22,23 .
Despite the cancer cells captured by F7 or SP peptide were identified as CSCs that expressed CD44, single peptide binding cells (F7+ or SP+) and double peptides binding cells (F7+ /SP+ , as shown in Fig. 2d) were considered as different subpopulations that exist at the different stages of cancer. Additionally, we found that the binding patterns were inconsistent when F7 and SP peptide bound to EMT derived cells (Fig. 4) and the signals representing F7 and SP peptides were partially overlapped with CD44 ( Fig. 5e and f). There are several www.nature.com/scientificreports/ possibilities for these findings. First, the permeability, stability and retention of these peptides in vivo might be taken into consideration. In the described experiments, the mice were given an intra-tumoral injection of F7 and SP peptide respectively 24 h before the IHC analysis of excised tumor samples. These peptides have a high stability at 37 °C in the presence of serum according to previous studies and our results also showed that the F7 and SP signals could be still found when incubation with SAS cells for 24 h (Fig. 1b and c)  www.nature.com/scientificreports/ while CD44 low CD24 high cells are regarded as non-stem cells. Notably, CD44 high CD24 high cells exhibit traits which are transitional between the non-stem and the stem-like, representing an intermediate population. In addition to CD44, other CSC markers might be used to reflect the tumor heterogeneity and meanwhile to evaluate the peptide binding.
In the current study, we successfully established a protocol for EMT induction via the combination of the G9a inhibitor and low-dose IR. G9a, mainly catalyzes histone H3K9 methylation, is shown to be overexpressed in many types of cancers, including gastric cancer, ovarian cancer, head and neck cancer, etc. Higher G9a expression levels represent higher methylation levels, thus leading to the transcriptional repression of tumor-suppressor genes and are also associated with poor prognosis in patients 25 . Liu et al. explored that snail-induced EMT and repression of E-cadherin in head and neck squamous cell carcinoma requires the G9a protein 26 . G9a is also noted to be implicated in mediating the DNA repair through efficient homologous recombination, which make cancer cells more resistant to radiation and chemotherapies; therefore, inhibition of G9a catalytic activity to interfere DNA repair pathways is a rational therapeutic strategy against cancers 27 . Unc0642, as a selective G9a inhibitor, has been optimized to exhibit lower cell cytotoxicity, higher selectivity and improved in vivo pharmacokinetic properties, making it suitable for animal studies 28 . Cao et al. reported that the treatment with UNC0642 decreased the level of histone H3K9me2 and increased the expression of BIM, a proapoptotic protein, in urinary bladder cancer (UBC) cells. Thereby, apoptosis was induced and leaded to the suppression of tumor growth in UBC xenograft model 29 . Of note, there were no significant weight changes between the vehicle-treated group and UNC0642 treatment groups (5 mg/kg body weight, used in their study) at any time point post-treatment, indicating that the dose of UNC0642 used in vivo was effective and safe. In our study, the half dose was administered in mice based on the purpose of EMT induction rather than tumor eradication. Radiotherapy is widely used prior to surgery and other therapeutic strategies. IR induces single strand breaks or double strand breaks, which trigger the activation of DNA damage response signaling pathways to arrest cell-cycle progression and activate appropriate DNA repair mechanisms 30 . Previous findings indicate that IR can promote EMT mediated metastasis and gain of CSC phenotypes by TGF-β signaling 17,18,31,32 . In our mouse model, significant reduction of Fluc signal was observed after Unc0642/IR alone or Unc0642 plus IR treatments in the mice as compared with the mice treated with the same amount of DMSO ( Fig. 5b and c). However, the tumor burdens in these treated mice suddenly increased on day 31 almost near that in control group, representing the occurrence of relapse. The results also demonstrated that Unc0642 and IR induced the EMT phenotype, decreased E-cadherin and increased vimentin expressions as well as CSC markers, CD44 (Figs. 3b and 5e, f). According to the results of our study, we would recommend that 1.25 μM Unc0642 plus 2 Gy of IR is the suitable dosage for the occurrence of EMT in SAS cells.
Although our results showed the potential of the detection approach for EMT derived cells or CSCs by using specific peptides, there were also several limitations. First, in vivo evaluation of peptide binding ability was carried out by intra-tumoral injection rather than systemic delivery, suggesting that certain subpopulations like circulating tumor cells (CTCs) were ignored. Since CTCs with CSC-like phenotype transport via the blood stream to distant sites and thereby form metastatic colonies, it is necessary to further investigate whether this detection approach using specific peptides can track CTCs. Second, in the current study, F7 and SP peptide were labeled with fluorescein followed by confocal microscopic imaging and FACS analysis for the evaluation of peptide binding activity. However, more detailed information regarding the binding of F7 and SP peptides to nucleolin and vimentin respectively is required for the potential application of these tumor targeting peptides in detection and isolation approaches. For instance, the isothermal calorimetry (ITC) is a widely used label-free quantitatively measurement of the binding affinity and thermodynamics for biomolecular interactions, such as protein-protein, protein-small molecules or protein-DNA, might enable us to understand functions and mechanisms at a molecular level 33 . Third, whether the peptide binding to its target expressed in cancer cells causes any cellular alteration by activating the signaling pathway pertaining to angiogenesis, tumorigenesis or apoptosis is our concern. In fact, SP peptide targeting to vimentin was shown to induce angiogenesis under hypoxic conditions 13 . It might be a potential risk of tumorigenesis if applied for in vivo molecular imaging and targeted therapeutic delivery. Thus, evaluations of the course of treatment, administered dose and tumor microenvironments are warranted.
In summary, our study has provided in vitro and in vivo evidences suggesting that the F7 and SP peptides could penetrate into cell cytoplasm and detected EMT derived cells and CSCs with high nucleolin and vimentin expression. Furthermore, the binding of these peptides was increased after the induction of EMT by the combination of small molecule inhibitor and low-dose IR. Our study suggests that these peptides may serve as a tool to detect EMT derived cells or CSCs and might aid the development of isolation methods for understanding these subpopulations within tumors in cancer research.

Methods
Cell culture. The human tongue squamous carcinoma derived cell line, SAS was obtained from the ATCC (a kind gift from Dr. Muh-Hwa Yang) and HEK293, a human kidney fibroblasts cell line, was provided from Dr. Yi-Jang Lee (Department of Biomedical Imaging and Radiological Sciences, National Yang Ming Chiao Tung University, Taipei, Taiwan). Cells were cultured in Dulbecco's modified Eagle's medium (Thermo Fisher Scientific, Inc., Waltham, MA, USA) supplemented with 10% fetal bovine serum (Biological industries, Israel) and 1% penicillin-streptomycin in a 5% CO 2 humidified incubator at 37 °C.

Generation of stable cell lines. To generate SAS cells stably expressing dual reporter genes EGFP-Fluc
or E2-crimson-P2A::ttksr39, cells were transduced with lentiviral vector for 6 h and then replaced with regular growth medium. Two days after the transduction, the cells were replated in 100 mm 2 culture dishes with continually grown in the presence of 500 μg/mL G418 (Sigma-Aldrich) for approximately 2 weeks. G418 resistant colonies were isolated and cultured in G418 containing media. The cell suspensions were filtered through Cancer spheroid culture and immunocytochemistry (ICC). SAS cells were seeded in a 6-well cell culture plate with ultra-low attachment (Corning Costar, Tewksbury, MA, USA) at the density of 1 × 10 5 cells/well and incubated for 3-5 days in a 5% CO 2 humidified incubator at 37 °C. All measurements on the SAS spheroids were carried out with a diameter of approximately 0.75 mm. For ICC-IF staining, the spheroids were fixed in 3.7% paraformaldehyde, washed with PBS and permeated with 0.1% Triton X-100 for 10 min before blocking with 20% FBS for 10 min at 37 °C, followed by the incubation with anti-CD44 antibody (GTX102111, GeneTex, Irvine, CA, USA) at 4˚C overnight. The IF staining was performed as described below. www.nature.com/scientificreports/ relative humidity of 50% to 70% and had a 12-h light/dark cycle. For establishment of xenografts, 2 × 10 6 SAS-EGFP-Fluc cells in 100 μL phosphate-buffered saline (PBS) were injected subcutaneously into the animal's right shoulder and 2 × 10 6 SAS-wt cells were injected into the left leg. Growth of the implanted tumors was monitored twice a week and tumor volumes were estimated from caliper measurements according to the following formula: tumor volume = (lengthÍwidth 2 ) Í0.523. When tumor size reached approximately 10 mm 3 , the mice were randomized into four groups of five animals per group for the following treatments: (1) DMSO (1 ml/kg body weight), (2) Unc0642 (2.5 mg/kg body weight), (3) 2 Gy of radiation, (4) Unc0642 (2.5 mg/kg body weight) and 2 Gy of radiation. Mice were given Unc0642 (cat. SML1037, Sigma-Aldrich, St. Louis, MO) intraperitoneally (i.p) every two days for 3 times. Mice were placed under a RS2000 Series Biological Irradiator and then exposed to 2 Gy of radiation every two days for 3 times. Additionally, all methods were performed in accordance with the relevant guidelines and regulations and the study was carried out in compliance with the ARRIVE guidelines.
In vivo bioluminescence image. Tumor-bearing mice were anaesthetized with 2% isoflurane and then injected D-luciferin (150 mg/kg body weight diluted in PBS) into the mouse peritoneum. Fifteen minutes later, mice were placed in the imaging chamber and photo counts were acquired for 30 s by a bioluminescence imaging system (Perkin Elmer, Waltham, MA, USA). The luminescent intensity at tumor sites was analyzed using living image software 3.2 (IVIS 50 Imaging System, Perkin Elmer, Waltham, MA, USA). Bioluminescent signal was recorded as mean photons/s/centimeter 2 /steradian (photon/s/cm 2 /sr).
Statistical analysis. SigmaPlot12 (Systat Software Inc., San Jose, CA, USA) was used to perform statistical analysis. Student's t-test analysis was performed for normally distributed data. Results were presented as the average ± SEM. A p value of less than 0.05 was considered statistically significant.

Data availability
The data presented in this study are available on request from corresponding authors.