Differentiation of human neuroblastoma cells toward the osteogenic lineage by mTOR inhibitor

Current hypothesis suggest that tumors can originate from adult cells after a process of 'reprogramming' driven by genetic and epigenetic alterations. These cancer cells, called cancer stem cells (CSCs), are responsible for the tumor growth and metastases. To date, the research effort has been directed to the identification, isolation and manipulation of this cell population. Independently of whether tumors were triggered by a reprogramming of gene expression or seeded by stem cells, their energetic metabolism is altered compared with a normal cell, resulting in a high aerobic glycolytic 'Warburg' phenotype and dysregulation of mitochondrial activity. This metabolic alteration is intricately linked to cancer progression.The aim of this work has been to demonstrate the possibility of differentiating a neoplastic cell toward different germ layer lineages, by evaluating the morphological, metabolic and functional changes occurring in this process. The cellular differentiation reported in this study brings to different conclusions from those present in the current literature. We demonstrate that 'in vitro' neuroblastoma cancer cells (chosen as experimental model) are able to differentiate directly into osteoblastic (by rapamycin, an mTOR inhibitor) and hepatic lineage without an intermediate 'stem' cell step. This process seems owing to a synergy among few master molecules, metabolic changes and scaffold presence acting in a concerted way to control the cell fate.

Cancer stem cells are currently viewed as the cells capable of generating cancer (tumor-initiating cells), owing to their intrinsic features of self-renewal and longevity. 1 However, emerging evidence suggests a surprising ability of normal committed cells to act as reserve stem cells upon reprogramming following tissue damage resulting from inflammation and wound healing. This brings to the alternative hypothesis that tumors may originate from differentiated cells that have recovered stem cell properties owing to genetic or epigenetic reprogramming. 1,2 Possibly, both models are correct, and consequently there is a continuum of cells capable of generating cancer, ranging from early primitive stem cells to committed progenitor or even terminally differentiated cells.
The development of methods for reprogramming somatic cells to induced pluripotent stem cells (iPSCs) through ectopic expression of a few pluripotency factors holds the promise for disease modeling, drug screening studies and treatment of several diseases. 3 Generating iPSCs from cancer cells might also clarify the mechanisms that underlie oncogenic transformation. 4,5 Thus reprogramming and oncogenic transformation are processes that have interesting common steps, while iPSCs generated from cancer cells could give clues to molecular mechanisms underlying the pathogenesis of human cancer. 6 To date, there are only few reports demonstrating a successful reprogramming of human primary cancer cells. Only one report describes the reprogramming of human primary cancer cells 7 while the remaining studies used established cell lines. [8][9][10][11][12] Carrete et al. 9 generated iPSCs from the chronic myeloid leukemia (CML) cell line KBM7 carrying the BCR-ABL fusion oncogene by expressing four ectopic reprogramming factors (OCT4, KLF4, SOX2, and c-Myc (OKSM)). Conversely, Choi et al. 10 reprogrammed EBV-immortalized B lymphocytes to pluripotency using non-integrative episomal vectors. Lin et al. 11 reprogrammed human skin cancer cell lines to pluripotency using the microRNA miR-302. Miyoshi et al. 12 reprogrammed gastrointestinal-transformed cell lines using retroviral vectors expressing c-Myc and BCL2. Finally, Hu et al. 8 successfully reprogrammed primary human lymphoblasts from a BCR-ABL + CML patient using transgene-free iPSC technology to ectopically express OKSM and LIN28. In addition, Ramos-Mejia et al. 4 in a recent review emphasize the importance of deciphering the barriers underlying the reprogramming process of primary cancer cells to obtain information on the links between pluripotency and oncogenic transformation that would be instrumental for therapy development.
Cancer cells show distinct metabolic features. In fact, neoplastic cells adapt their metabolic pathways to face the demands of abnormal proliferation. For example, cancer cells increase glucose uptake and the rate of glycolysis even under normoxic conditions; this process of aerobic glycolysis was first described by Warburg et al. 13,14 and thence called Warburg effect. Recent studies are increasingly highlighting the importance of metabolic manipulation in cancer cells and how bio-energetic and biosynthetic changes could be exploited to stop tumor cells progression. [15][16][17] Reprogramming, pluripotency, oncogenic transformation and metabolic changes are therefore connected processes that share interesting similarities. 18,19 The fact that the same alterations driving tumorigenesis can influence the reprogramming of non-cancer somatic cells is a double-edged sword. It poses safety concerns for the cell therapy applications with iPSCs, while at the same time it promotes further studies aimed to analyzing the mechanisms and barriers underlying the direct reprogramming of cancer cells. This is a fundamental attempt to acquire valuable new insight on reprogramming and cell transformation. 20 Along these concepts, here we have investigated the possibility to revert cancer progression by targeting cancer cells, seen as deprogrammed cell and therefore similar to adult stem cells. Using a human neuroblastoma cell line (SH-SY5Y) as model, our work has been designed to experimentally explore different aspects. We show how these cells can differentiate toward a germ line different from the original one, modifying their morphology and acquiring metabolic changes which are distinctive of a more normal phenotype.

Results
Morphological analysis of differentiated SH-SY5Y cells.
In order to test the possibility to differentiate tumor cells, we used the human neuroblastoma SH-SY5Y cell line as experimental model. Using different types of scaffolds, such as hydroxyapatite and acellular bone matrix and rapamycin 5 μM as inductor, these cells differentiate into osteoblasts, that is, from an ectodermal to a mesodermal lineage. Figure 1 shows, by immunocytochemistry analysis, how the osteogenic differentiation was indeed taking place, as shown by the presence of osteocalcin, a protein typical of osteoblasts.
The neuroblastoma cells begin to get in contact with the support already on the early days of differentiation (day 5). Then a gradual colonization of the scaffolds followed by morphological change of the cells was observed along with an increase in size of the supports (Supplementary Figure S1A). The staining with Alizarin Red demonstrated extensive deposits of calcium phosphate (Supplementary Figure S1B) and western blotting analysis further evidenced the presence  Figure S1A), Time-course differentiation experiments showed that metabolic/protein changes were already occurring in the first 5 days from induction (Supplementary Figure S2E). In all the experiments, the SH-SY5Y control cells were plated and maintained in the same culture time and media conditions (without rapamycin). Scanning electron microscopy (SEM) of the differentiated cells shows the colonization of the scaffold ( Figure 2). Accordingly, larger tissue-like structures forming agglomerates of several highly differentiated columnar-shaped cells (20-30 μm in diameter) morphologically similar to osteoblasts can be observed. Finally, we performed a Ki67 proliferation assay to determine the percentage of proliferating and nonproliferating cells ( Supplementary Figures S2A-C).
Differentiated SH-SY5Y cells change their energy metabolism. The energy metabolism of control and differentiated SH-SY5Y cells was checked in parallel to the morphological change. Figure 3a reports the intracellular ATP content and its source (mitochondrial versus extra mitochondrial), the activity of lactate dehydrogenase (LDH) and the L-lactate production in untreated and differentiated SH-SY5Y cells (Figures 3b and c, respectively). Figure 3d shows the pyruvate kinase (PK) activity experiments before and after adding the substrate phosphoenolpyruvate (PEP), and Figure 3e shows the western blotting analysis for PKM1 and PKM2 expression in control and differentiated SH-SY5Y cells. All the experiments were performed in triplicate and repeated three times. The errors reported represent the mean ± S.E.M. of the three independent experiments.
Differentiated SH-SY5Y cells: metabolic and functional analysis. In parallel with the analysis of the energetic metabolism, we also checked the change in protein expression taking place during the differentiation process. The more significant differences detected between controls and SH-SY5Y differentiated cells are reported in Figure 4, showing Akt expression and its phosphorylation ( Figure 4a) and the change in c-Myc, p53, mouse double-minute 2 homolog (Mdm2), hexokinase (HK) 1/2 and BH3 interacting-domain death agonist (Bid) protein expression (Figure 4b). We also checked the changes in the expression of isocitrate dehydrogenase (IDH) 1/2 and mTOR/phosphorylated mTOR. The results of these analysis are reported in Supplementary Figure S2D. A quantitative measurement of Sirtuin-3 (SIRT3) activity performed on SH-SY5Y control and differentiated cells, respectively, is reported in Figure 4c.
Additional evidence of SH-SY5Y cancer cells' plasticity is reported in Supplementary Figure S3, showing that these cells can also differentiate toward an (ectodermal) hepatic phenotype. To demonstrate how this differentiative ability is not limited only to a neuroblastoma cell line, similar experiments were also performed on the human CML K-562 (Supplementary Figure S4). These preliminary results showed how these cells were able to secrete calcium phosphate and synthesize osteocalcin (Supplementary Figures S4A and B, respectively). From a point of view of energetic metabolism, An overall view of the differences between the control and the differentiated cells is shown in Figure 5.

Discussion
A core question in cancer biology is the identity and nature of the cancer 'cell of origin', that is, the target cell where the first oncogenic-driving mutation occurs leading to tumor initiation. The concept of cancer stem cells has recently emerged in view of their intrinsic ability of self-renewal and of their longevity, antiapoptotic and differentiation capabilities that makes them quite similar to the early primitive stem cells. 1 However, new evidence on the plasticity of normal cells, able to acquire stem cell features, suggest that committed progenitor cells or deprogrammed differentiated cells (possibly in response to tissue damage and wound healing) can also trigger tumor initiation. [21][22][23] Consequently, an alternative hypothesis suggests that tumors may originate from differentiated cells that can get back stem cell properties owing to genetic or epigenetic modifications. To date, the term cellular reprogramming is linked to the work of Takahashi and Yamanaka 3 showing the possibility of obtaining in vitro pluripotent stem cells starting from adult cells. In cancer cells, reprogramming is the possibility to obtain iPSCs, by inserting genes of stem cells and then differentiate them into different cell types. 24 This may offer a novel differentiative approach by reprogramming the cancer cells without creating or isolating the stem precursors. We used, as an experimental model, a human neuroblastoma cell line, namely SH-SY5Y, to perform a differentiation protocol leading the cells toward a different germ layer (from ectoderm to mesoderm). To this end, we directed them toward an osteoblastic lineage using rapamycin as inducer, a compound able to promote the in vitro osteogenic differentiation of stem cells by acting on Akt/mTOR pathway. [25][26][27] Compared to the work of Jonhsen et al., 26 we decided to use rapamycin in combination with hydroxyapatite scaffolds, thus allowing a more extended differentiation of the cells in culture. As shown, the SH-SY5Y cells are not affected by the different types of support in the differentiation process, demonstrating that the presence of the scaffold is always required regardless of its characteristics. The concept was to utilize matrices derived from the stem cell studies, where the use of scaffolds allows an improvement in the osteogenic differentiation. The cellular differentiation was visualized by morphological and metabolic analysis. Figures 1 and 2 (and Supplementary Figure 1) show that, already few days after the induction, a colonization of the scaffolds was taking place followed by an increase in size of the supports, the formation of calcium  Figure 2E). We then investigated the metabolic and functional changes occurring during differentiation (Figures 3 and 4). We observed relevant metabolic changes. Although cancer cells display elevated glycolysis under normoxic conditions (Warburg effect), [13][14][15] Figure 3a shows that differentiation causes a switch from glycolytic to oxidative metabolism. However, unexpectedly, the differentiated cells showed higher levels of LDH activity and of L-Lactate production (Figures 3b  and c, respectively), when compared with the control cells growth in the same culture conditions but without the inducer. [28][29][30][31] To explain these data, we studied the PK activity, that is, the final rate-limiting step of glycolysis. Of the four mammalian PK isoforms (L, R, M1, M2), tumor cells mainly express the M2 isoform. In contrast, neuroblastoma cells express all four isoforms. 32 PKM2 is subject to a complex regulation by both oncogenes and tumor suppressors, which allows a fine tuning of its activity, so lowering PK activity it is expected to produce less pyruvate. 33 In this context, an increase of Akt expression could be a key player in the Warburg effect induction (aerobic glycolysis). 34 In 2011, Kosugi et al. 35 observed that an increase in p-Akt was responsible for the activation of PKM2. A recent work confirmed that indeed Akt2 is responsible for the PKM2 upregulation. 36 Furthermore, switching PKM2 expression to PKM1 in tumor cells can reverse the Warburg effect, reducing the ability of the cells to form tumors in nude mice. 37,38 PKM1 expression also correlates with terminal differentiation. 36,39 Figure 3d shows that PK enzymatic activity on its substrate PEP was significantly higher in differentiated cells with respect to controls as shown by the increase of ATP when compared with the respective untreated cells, whereas the PKM2 isoform expression is reduced. 36 Thus we hypothesize that the PKM1 isoform was responsible for the increased PK enzymatic activity observed in the differentiated cells. Accordingly, inhibition of the PK seems responsible for the low levels of L-Lactate and LDH activity in the control cells. 36,40 The activation of PK during differentiation might induce pyruvate channeling to the mitochondria for ATP synthesis and White nuanced indicate the total absence of protein expression found in the performed experiments. In cancer cells, a high rate of glycolysis is observed with highly increased glucose uptake even under normoxic conditions, while respiration remains almost unchanged (aerobic glycolysis). The major mediators of the metabolic shift toward increased glycolytic flux are, in our experiments, c-c-Myc, p53, Akt, PK and IDH. The less PK activity in cancer cells increases the concentrations of glycolytic intermediates. The survival signals lead to translocation of HK to the mitochondrial membrane through the Akt signaling pathway helped by c-Myc and p53. Mitochondria-bound HK confers an advantage to utilize ATP for phosphorylating glucose and, at the same time, mitochondrial HK negatively modulates Bid preventing the release of cytochrome c into cytosol and cell death. 58 High IDH expression probably are involved in the early steps of initiating Warburg effect and help the cancer cells to maintain this metabolic state. After induction, in SH-SY5Y differentiated cells the shift in energy metabolism eventually leads to the oxidation of nutrients via oxidative phosphorylation. An increase in the PK activity, due mainly to the PKM1, continuously generates pyruvate, which is transported into mitochondria and further metabolized via the tricarboxylic acid cycle. The almost nonexistent IDH2 expression helps to maintain the citrate in this oxidative metabolic way. The disappearance of c-Myc and p53 with the lowest Akt expression and the increase in SIRT3 activity also induce the detachment of HK and increase the mitochondrial activity address the cellular metabolism of these cells toward the oxidative metabolism. 40 This finding correlates with the extremely low expression of IDH isoforms (IDH1 and IDH2) in the differentiated cells. An increased glycolysis appears to be an advantage for cancer cells, providing not only energy but also the basic metabolites needed to a rapidly dividing cell, 41,42 and IDH1/2 may represent the mechanism for triggering this metabolic shift. [43][44][45] This is a further evidence for the role of Akt and PK as key molecules in the control of glycolysis in cancer cells. 39 Figure 4 shows the different expression of some known tumor master genes in SH-SY5Y cells in comparison to the differentiated cells. An increased expression of Akt, c-Myc, p53, Bid, HK2/HK1, IDH1/2, PKM1/2, a low SIRT3 activity and no Mdm2 expression was observed in SH-SY5Y controls, whereas in SH-SY5Y differentiated cells there was a lower Akt, Bid, HK2/HK1, IDH1/2 and PKM2 expression, no c-Myc and p53 expression but a higher amount of Mdm2 and a strong SIRT3 activity. Collectively, these data are summarized in Figure 5.
In SH-SY5Y control cells, the elevated c-Myc and p53 levels promote the synthesis and/or upregulation of many glycolytic enzymes, including HK 2, owing to an increased transcription 46 and insufficient p53-mediated control, 18 as this protein is abnormally sequestered in the cytoplasm. 32,[47][48][49] HKs have multifunctional roles besides glucose phosphorylation. There is evidence that HKs bound to the mitochondrial membranes participate in the apoptotic pathway. [50][51][52][53][54] The HK2 association with mitochondria is promoted by Akt activity, which also induces aerobic glycolysis and increases the uptake of glucose. 55,56 In the apoptotic process, mitochondrial HK negatively modulates the cleaved form of Bid and may directly inhibit Bad, while Bid is inhibited by Akt. All these activities are able to prevent apoptosis and cell death, thus leading to an immortalized cancer cell. 52,57,58 The differentiated SH-SY5Y cells had a decreased level of Akt associated with an increase in the PK activity leading to a metabolic switch from glycolytic to oxidative. This switch is also facilitated by the decreased expression of IDH2 and consequently leaving the citrate to flow through the tricarboxylic acid cycle. The absence of c-Myc transcriptional activity, and the synthesis of Mdm2 which promotes the p53 degradation, allow a regulation of glycolysis and glycolytic enzymes and might explain the decreased expression of HKs. 47,59-62 An increased mitochondrial activity was also consistent with a greater SIRT3 activity, which also induces dissociation of HK from mitochondria. 63 These data are in agreement with recent reports demonstrating that SIRT3 is a crucial regulator of the Warburg effect. 64 Probably this might be the biochemical basis of the anticancer activity of resveratrol particularly toward neuroblastoma, which has been attributed to the in vitro or in vivo activation of SIRT1 and SIRT3. 65 According to our results, it seems feasible to change the fate of a cancer cell by two different approaches, that is, by differentiating a cancer cell in a germ line different from the original one and by obtaining differentiated cells by acting on glucose metabolism and on the expression of some key proteins working in concert. This reversal to a 'committed' state was until now only suggested by few reports. [66][67][68][69] In conclusion, this paper shows the ability to induce a noncanonical differentiation in cancer cells, accompanied by unexpected metabolic changes. In this model, it is crucial the combination of two components namely rapamycin as inductor and a scaffold to obtain an extensive osteogenic differentiation. In our opinion, the study of these processes can represent an advancement in the understanding of the molecular mechanisms able to cause a reversal state of the tumor cell and possibly a hint to new therapeutic approaches. The cell line of human CML K-562 (ATCC, CCL-243) was grown on tissue culture flask (Corning, Durham, NC, USA, cat. no. 353108) in RPMI1640 (Biowest, Nuaille, France, cat. no. L0501-500) supplemented with 2 mM L-glutamine, 10% (v/v) FBS (Lonza, cat. no. DE14-830F), 1% (v/v) penicillin (100 units/ml)/streptomycin (100 mg/ ml) (Lonza, cat. no. DE17-602E) and 1% of non-essential amino acids (Biowest, cat. no. X0557-100) maintained in an humidified incubator at 37°C with 5% (v/v) CO 2 . Every 3 days, the medium was changed and the cells centrifuged for 10 min at 200 × g and reseeded at a density of approximately 3 × 10 5 cells in complete media.
The coating plate protocol required a 0.1 mg/ml collagen type I rat tail high concentration solution (BD Pharmigen, BD, Bedford, MA, USA, cat. no. 354249) added to the 24 multi-well plates and incubated for 1 h at 37°C. After incubation, the plates were washed three times with PBS and kept at +4°C until use.
Cancer cell-a new differentiative approach A Carpentieri et al Alizarin Red staining. The presence of calcium phosphate deposits in SH-SY5Y and K562 cells following differentiation compared with control cells was evaluated by staining with Alizarin Red S (Sigma, cat. no. A5533-25G). The cells were fixed with 4% (v/v) formaldehyde overnight, washed with distilled water, stained with 1% (v/v) Alizarin Red S and incubated at room temperature in the dark for 10 min. The excess of dye was removed with distilled water, and the cells were observed by optical microscopy.
Enzymatic assays. To evaluate the cellular metabolic pathway, for the ATP levels, LDH activity, L-Lactate production, Akt expression/phosphorilation and PK activity, the cells were collected, centrifuged for 10 min at 200 × g, washed in PBS and counted by TC10 Automated Cell Counter (Bio-Rad, Hercules, CA, USA, cat. no. 145-0009), while for SIRT3 activity the cells were lysed by performing three cycles of freezing/thawing. All assays were performed in triplicate on 96-well microplates, and the results were repeated in three separate experimental sessions.
The ATP levels were determined by the luciferin/luciferase method using the luciferase ATP Assay Kit (Sigma, cat. no. FLASC) according to the manufacturer's instructions. For the evaluation of ATP content under strict glycolytic conditions, the cells (3 × 10 5 ) in each well were incubated for 1 h at 37°C in the presence of rotenone (1 μmol/l), a specific inhibitor of complex I, and antimycin A (1 μmol/l), which is a specific inhibitor of complex III, washed with PBS twice and suspended in PBS. The ATP content was measured using a luminometer (Victor X, Perkin Elmer) with high sensitivity (0.01 pmol). An ATP calibration curve was made using titrated ATP solutions following the manufacturer's instructions. To determine the LDH activity the LDH Assay Kit (Abcam, cat. no. ab102526) was used, for the L-Lactate production the L-Lactate Assay Kit (Abcam, cat. no. ab65331) was used and for SIRT3 activity the SIRT3 Direct Florescent screening Assay Kit (Cayman Chemical Company cat. no. 10011566) was utilized. All the assays were performed following the manufacturer's instructions. To measure Akt expression and its phosphorylation, the Phospho-TracerAkt 1/2/3 pThr308 + , pSer473 + Akt (Total) Assay Kit (Abcam, cat. no. ab119675) was used following the manufacturer's instructions. This kit detects endogenous levels of Akt 1/2/3 in cellular lysates. Furthermore, it is able to detect Akt 1/2/3 when phosphorylated at Thr308 and when phosphorylated at Ser473. All assays were performed in triplicate.
For the PK activity, the cells were plated on a 96 multi-well microplates and incubated 10 min in: 150 mM triethanolamine, 75 mM ADP, and 0.04 mM PEP. After incubation, a mixture of luciferase-luciferin was added to the reaction and the plate read by a luminometer (Bedwildbadi, Germany). The amount of ATP produced was compared with the total amount of ATP. For all the assays, the measurements obtained have been normalized for the protein content. All assays were performed in triplicate, and the results were repeated in three separate experimental sessions.
Statistical analyses. Data are expressed as means ± S.E.M. and statistically analyzed by Student's t test.