The stem cell–determining transcription factor Sox2 is required for the maintenance of normal neural stem cells. In this study, we investigated the requirement for Sox2 in neural cancer stem-like cells using a conditional genetic deletion mutant in a mouse model of platelet-derived growth factor–induced malignant oligodendroglioma. Transplanting wild-type oligodendroglioma cells into the brain generated lethal tumors, but mice transplanted with Sox2-deleted cells remained free of tumors. Loss of the tumor-initiating ability of Sox2-deleted cells was reversed by lentiviral-mediated expression of Sox2. In cell culture, Sox2-deleted tumor cells were highly sensitive to differentiation stimuli, displaying impaired proliferation, increased cell death, and aberrant differentiation. Gene expression analysis revealed an early transcriptional response to Sox2 loss. The observed requirement of oligodendroglioma stem cells for Sox2 suggested its relevance as a target for therapy. In support of this possibility, an immunotherapeutic approach based on immunization of mice with SOX2 peptides delayed tumor development and prolonged survival. Taken together, our results showed that Sox2 is essential for tumor initiation by mouse oligodendroglioma cells, and they illustrated a Sox2-directed strategy of immunotherapy to eradicate tumor-initiating cells. Cancer Res; 74(6); 1833–44. ©2014 AACR.
The cancer stem cell (CSC) hypothesis provides a novel point of view on the mechanisms of tumor development and on therapeutic approaches. CSC can reinitiate tumor development following conventional therapeutic approaches (to which they are resistant) and following experimental transplantation into mouse brain. Neural tumors were among the first tumors in which CSCs were identified (1–4). Defining the gene regulatory networks that control the maintenance of the malignant phenotype of CSCs is thus a fundamental objective to understand tumor pathogenesis and to develop novel approaches to targeted therapy.
Neural CSCs are proposed to originate from normal neural stem cells (NSC) or from the “reprogramming” of more differentiated cells to a stem cell state (5–7).
Sox2 is a transcription factor functionally important for normal stem cells, including pluripotent and tissue-specific stem cell types (8–12). In the nervous system, Sox2 is active in NSCs and progenitors of the embryo and adult (8, 12). By conditional knockout in mouse, we found that Sox2 is required to maintain self-renewal of NSCs in vitro and in vivo within specific postnatal brain regions (e.g., hippocampus; ref. 13). Sox2 also “reprograms” differentiated fibroblasts to pluripotent iPS cells, acting together with a small number of other molecules (14, 15).
Sox2 is expressed in many neural tumors (including gliomas/glioblastomas, medulloblastomas, and ependymomas), and its expression is consistently detected in the cell fraction displaying properties of CSCs (2, 3, 16, 17).
Oligodendroglioma is a type of glioma consisting primarily of cells resembling oligodendroglia and it is the second most common malignant brain tumor in adults (18); patients affected by high-grade (anaplastic) oligodendroglioma have a median survival of 3 to 4 years (19). Alteration of platelet-derived growth factor (PDGF)-B signaling is a common molecular lesion in oligodendrogliomas, and in gliomas in general, and PDGF-B experimental overexpression in neural stem/progenitor cells of the mouse brain generates neural tumors, in particular oligodendrogliomas (20, 21). We generated a mouse genetic model of oligodendroglioma by overexpression of PDGF-B in mouse embryonic neural precursors, by retroviral transduction of embryonic brains in utero (22); these tumors (PDGF-induced high-grade gliomas, pHGG hereafter) display a homogeneous character of oligodendroglioma, express Sox2 (IA/PM, unpublished data; see Figs. 1 and 2), and reproducibly develop after a latency of several weeks (22, 23). pHGGs contain CSCs that will reform the same tumor type following in vivo transplantation of dissociated tumor tissue or in vitro cultured pHGG cells (22, 23).
Here, we ask whether Sox2 is required by oligodendroglioma stem cells, mirroring its requirement for normal NSCs. We used our Sox2flox conditional mutation (13), in combination with the pHGG mouse model (22), to address the effects of Sox2 ablation on tumor reinitiation following tumor cell transplantation into brain. Mice transplanted with Sox2-deleted cells remained tumor-free throughout the time window in which controls developed lethal tumors. Loss of tumorigenesis of Sox2-ablated cells is prevented by transduction with a Sox2-expressing virus. Microarray analysis identifies early gene expression changes following Sox2 deletion. Finally, vaccination with Sox2 peptides elicits a response that significantly delays tumor development, pointing to Sox2 itself as a possible therapeutic target.
Materials and Methods
Tumor induction in Sox2flox/flox mice
Lentiviral constructs and infections
Cre-encoding virus was obtained from Cre-IRES-GFP (a gift from S. Brunelli, The University of Milan-Bicocca, Milan, Italy; ref. 24) by GFP deletion (using BstXI/SalI); this avoided GFP toxicity observed while superinfecting oligodendroglioma cells (which synthesize PDGF-B-GFP) with the original GFP-expressing virus. The control mCherry-expression virus (1070.935.hPGK.dNGFR.minhCMV.mCherry.SV40PolyA) was a gift from L. Naldini, San Raffaele Telethon Institute for Gene Therapy, San Raffaele Scientific Institute, Milan, Italy. The Sox2-expressing lentivirus was obtained from a Sox2-IRES-GFP lentivirus (25) by SalI/BstxI GFP deletion.
For Cre- or control-lentiviral transduction, cells were plated in 24-well plates at 50,000 cells per well on Matrigel and transduced 4 hours after plating, with a multiplicity of infection (MOI) of 7. Medium was changed 15 hours after transduction. In some experiments (Fig. 2), pHGG cells were initially transduced with Sox2 lentivirus at MOI 7, tested for expression of transduced Sox2 (Supplementary Material), passaged, and further transduced with the Cre virus, as above.
Sox2 PCR, quantitative reverse transcription PCR, and immunofluorescence
PCR primers and procedures are described in Supplementary Materials. Sox2 immunofluorescence was performed as described in ref. 25.
Transplantation of virally transduced cells into mouse brain
In vitro assays
For in vitro assays, pHGG cells were transduced, collected after 96 hours, and plated: (i) at clonal density (60 cells/100 μL/well) in 96-well plates in normal growth medium (22) without Matrigel, to allow the formation of well-individualized clones, with or without EGF and basic fibroblast growth factor (bFGF; Fig. 3A); and (ii) at a density of 5,000 cells per well in Matrigel-coated 4-well chambered slides, in medium devoid of EGF and bFGF, supplemented with 2% fetal calf serum (Fig. 3B). A total of 10 μmol/L EdU (A10044 Molecular Probes, Invitrogen) was administered for 30 minutes before 4% paraformaldehyde fixation and EdU-positive nuclei were detected by a Click-IT EdU Alexa Fluor 549 HCS Assay Kit (Molecular Probes, Invitrogen). Terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling (TUNEL) analysis and GFAP IF were performed as described in ref. 25. O4 and GalC IF used anti-O4 and anti-GalC hybridomas (undiluted supernatant), a gift from C. Taveggia.
Gene expression analysis
Total RNA was prepared as described (22) from triplicate independent cultures of Sox2flox/flox pHGG cells transduced with Cre virus or control (nontransduced, or transduced, with mCherry virus). Cre-transduced cells were harvested 40 or 96 hours after transduction and mCherry-transduced and nontransduced control cells were harvested 40 hours after transduction.
RNA extraction, microarray hybridization, and analysis were performed as described previously (for details, data analysis and Gene Ontology annotation, see Supplementary Files; ref. 22).
Sox2 peptide design and vaccination
For SOX2 immunotherapy, we used four SOX2 peptides: TLMKKDKYTL (26), SGPVPGTAI (Score 21); VSALQYNS (Score 14); GGGGNATA (Score 16), and four OVA control peptides: OVA257-264 SIINFEKL (Sigma Aldrich), OVA55-62 KVVRFDKL (Score 22); OVA107-114 AEERYPIL (Score 22); OVA176-183 NAIVFKGL (Score 22) that were expected to bind the murine MHC class I (H-2Db). The new peptides were designed using SYFPEITHI (http://www.syfpeithi.de/) and BIMAS (http://www-bimas.cit.nih.gov/molbio/hla_bind/) binding-motif algorithms and were synthesized by Primm srl (Milano).
C57BL/6N (5-week-old females) mice were injected (day 0) with 20,000 tumor cells (stereotactic coordinates with respect to the bregma: 1 mm anterior, 1.5 mm left lateral, and 2.5 mm deep).
Injected mice were treated with temozolomide (Sigma Aldrich) and/or peptide vaccinations. Temozolomide was administered by intraperitoneal injections (5 mg/kg). Peptides were emulsified with Montanide ISA 51 VG (1:1; SEPPIC) and administered by subcutaneous injections of the four peptides separately (15 μg/peptide) into different areas of the flank.
We tested four different conditions: group I: vehicle only (montanide); group II: three peptide vaccinations spaced 1 week apart (days 14, 21, 28); group III: temozolomide alone, five daily injections (day 10–14); group IV: peptide vaccination combined with temozolomide. Groups I, II, and IV also received a total of 3 μg of recombinant murine granulocyte macrophage colony-stimulating factor as described (27). Cumulative survival curves were obtained using the Kaplan–Meier method (MedCalc 12.7).
For cytotoxicity assay, isolation of tumor-infiltrating lymphocytes (TIL) and flow cytometry, see ref. 27 and Supplementary Materials.
We generated oligodendrogliomas in mouse by transduction of a PDGF-B-IRES-GFP-encoding retrovirus within the brain at embryonic day (E) 14.5 (22, 23). Embryos were homozygous carriers of a Sox2flox mutation, allowing subsequent Sox2 excision via Cre recombinase (Fig. 1A; ref. 13). Tumors developed, at different times after birth: early-onset, showing low-grade tumor features and late-onset, displaying high-grade glioma characteristics, as expected from our previous data (23). We focused on tumors arising at least 90 days after birth, as our previous analyses showed that low-grade tumors arising before day 90 can be hardly grown in culture and are not tumorigenic (23). Indeed, 4 of 7 tumors appearing after 90 days reinitiated tumorigenesis following transplantation into adult mouse brain. We cultured in vitro three of these secondary tumors (pHGGs), in conditions allowing the long-term maintenance of TICs, i.e., presence of EGF and bFGF and absence of serum (16, 22), and we used for subsequent analyses one cell population derived from such tumor.
Sox2 deletion impairs tumor reinitiation by pHGG cells following in vivo transplantation in the brain
To evaluate the role of Sox2 in tumor initiation, we deleted the “floxed” Sox2 gene from tumor-derived cells by transduction with lentiviruses expressing Cre recombinase or mCherry as a control (Fig. 1A). Transduction of Cre recombinase (but not of control virus) induced efficient deletion of Sox2 (>95% by DNA analysis; Fig. 1B), leading to loss of Sox2 mRNA (>90% by real-time RT-PCR; Fig. 1C) and protein (>95% by immunofluorescence) by 36 hours after transduction (Fig. 1D). We then transplanted Cre-transduced, or control mCherry-transduced, or nontransduced cells into the brain of adult C57/Bl6 mice, 36 hours after viral transduction (Fig. 2). Control mCherry-transduced and nontransduced cells caused the development of tumors (17/20 mice, 85%; of which 8/10 mCherry, 9/10 nontransduced), resulting in an overall median survival of the control mice of 40 days, consistent with previous reports with similar tumor-derived cells (Fig. 2B; Table 1; ref. 23). However, mice injected with Cre-transduced cells were almost all alive (13 of 17; 76.5%) at day 118 after transduction (Fig. 2B; log-rank test, P < 10−4). When sacrificed and analyzed at day 121 ± 3, these mice were found tumor-free. Analysis of the 4 mice injected with Cre-transduced cells that had died (2 by day 50, 1 on day 64, 1 on day 117; Table 1) showed that they had developed tumors that demonstrated a nondeleted status of the Sox2flox gene upon genotyping (not shown), quantitative reverse transcription (qRT)-PCR (Supplementary Fig. S1), and immunofluorescence (Fig. 2C), indicating their likely origin from the few non–Sox2-deleted cells.
|Transplanted cells .||Lethal tumors/transplanted animals .|
|Lenti-Sox2; Lenti-CRE||6/8 (75%)|
|Transplanted cells .||Lethal tumors/transplanted animals .|
|Lenti-Sox2; Lenti-CRE||6/8 (75%)|
aThe four tumors were tested by PCR for the presence of the undeleted and deleted Sox2 locus; all four were Sox2-positive (undeleted Sox2); one also presented a band for the deleted Sox2 locus, indicating that some Sox2 deleted cells are part of the tumor mass.
bThe 13 surviving mice were tumor-free at day 120.
To address whether the loss of TIC properties was specifically due to loss of Sox2, rather than to nonspecific effects (Cre-toxicity, etc.) we performed a control experiment. Before Cre transduction, we transduced the tumor-derived cells with a Sox2-encoding lentivirus (Fig. 2A and B; ref. 13). After subsequent ablation of endogenous Sox2 by Cre (leading to loss of endogenous Sox2 mRNA, as verified with specific primers, see Supplementary Materials), we transplanted the cells into host mouse brains and compared their survival with that of controls, i.e., cells transduced with Sox2 virus but not with Cre, or with the control mCherry virus, or untransduced (Fig. 2B). Sox2-transduced cells demonstrated tumorigenic ability similar to that of controls, with tumors developing in 6 of 8 mice (75%) within 120 days and a median survival of 34 days (Fig. 2B).
We conclude that the tumor-initiating ability of PDGF-B–induced oligodendroglioma cells requires Sox2 function.
Consequences of Sox2 deletion on in vitro growth of pHGG cells
To obtain information on the mechanisms of loss of tumor-initiating ability of the Sox2-deleted pHGG cells, we studied the in vitro growth of intact or Sox2-deleted cells (Figs. 3 and 4). We tested cells in three growth conditions: the first one optimized for maintenance of stem cell properties, and corresponding to the initial condition in which the cultures had been derived (with EGF and bFGF and without serum: +EGF/bFGF); a second one in the same medium with no added growth factors (no factors); and a third one in medium without added factors, but with 2% serum (no factors + serum). The latter represents “differentiating” conditions normally used to obtain terminal differentiation of normal NSC (neurospheres; refs. 1, 25, 28, 29). Cells were plated at clonal density (96 hours after viral transduction) and scored for clone numbers after 7 days. In EGF/bFGF, the number of clones obtained with Sox2-deleted cells was only slightly, although significantly, decreased compared with undeleted controls; however, when plated without factors, Sox2-deleted cells were reduced to less than 50% of controls (Fig. 3A). Cells plated in serum adhered to the substrate and did not form individualized clones. We thus plated the cells, 96 hours after transduction, in 2% serum-containing medium without added factors, at nonclonal density, on Matrigel, allowing more efficient growth. Under these conditions, untreated pHGG cells continue to grow, although to a rate somewhat lower (20%–30% increase in duplication time) than in medium with added growth factors. Following Sox2 deletion, the total cell number at day 7 was strongly reduced relative to controls (<20% that obtained with untransduced or mCherry transduced cells; Fig. 3B). We also assessed proliferative ability (at day 2 and 7) by administering EdU for 30 minutes and measuring the percentage of cells that incorporated EdU (Fig. 4A). While controls (mCherry transduced or nontransduced) cells had similar high levels of EdU incorporation (30%–35%), Sox2-deleted cells showed significant reduction of EdU incorporation (to about 10%; Fig. 4A). We then evaluated apoptosis by the TUNEL assay (Fig. 4B). TUNEL-positive cells were more than 4-fold increased following Sox2 deletion relative to controls (Fig. 4B). Immunofluorescence for differentiation markers of oligodendroglia, O4 and GalC, and astroglia, GFAP, revealed widespread positivity, together with an altered morphology of Sox2-deleted cells, with features suggestive of aberrant differentiation (branching, flattening), as compared with the relatively undifferentiated morphology of undeleted pHGG cells (Fig. 4C). We conclude that in “differentiating” growth conditions, Sox2 ablation leads to progressive exhaustion of in vitro cell proliferation, increased apoptotic cell death, and morphologic changes, suggesting aberrant differentiation.
Sox2 deletion causes alterations in the gene expression program of pHGG cells
The dependence on Sox2 of tumor-initiating properties of oligodendroglioma cells raises the hypothesis that Sox2 may act by regulating the transcription of critical downstream genes. We analyzed gene expression in Sox2-deleted and control cells (nontransduced, m-Cherry virus-transduced) by microarray analysis, at 40 and 96 hours following Cre transduction (Fig. 5A; Supplementary Table S1). The gene expression profile of Sox2flox/flox pHGG cells closely matched the “oligodendroblast” program that we previously reported for several independent PDGF-B–induced oligodendrogliomas (Fig. 5B; refs. 22, 23). At 96 hours, the expression of 146 genes was substantially deregulated (more that 2-fold); at 40 hours, few, if any, gene changed its expression level significantly (Fig. 5A). This suggests that our analysis at 96 hours likely includes the earliest changes in gene expression that follow Sox2 loss, presumably including those genes that directly rely on Sox2. Following Sox2 ablation, 12 genes are downregulated, compatibly with an activator function of Sox2; 134 genes are upregulated (Fig. 5A and D), possibly reflecting a repressor function of Sox2 (25), or indirect effects. Gene ontology analysis of the deregulated genes indicated a significant enrichment in the functional categories of Developmental Processes, Response to Stimulus, Cell Proliferation, Communication and Signaling, and Cell Differentiation (Fig. 5C).
“Vaccination” with SOX2 peptides significantly prolongs survival and induces specific antitumor effector response
The requirement for Sox2 by oligodendroglioma CSC raises the possibility that Sox2 itself may qualify as a target for CSC-directed therapeutic strategies. While Sox2 is highly expressed in oligodendroglioma-initiating cells, its expression in the normal brain is very limited. This led us to try an immunotherapy approach, to “vaccinate” immunocompetent mice after the transplantation of TICs (Fig. 6). After cell transplantation into the brain, we administered the mice 4 SOX2 peptides, including one previously shown to elicit T-cell activation against Sox2-expressing glioblastoma cells in vitro (26). Vaccination with peptides was performed on days 14, 21, and 28 (Fig. 6A). Mice also received temozolomide (five daily injections on days 10–14, alone or in combination with peptide vaccination) as an immunologic adjuvant for enhancing immunogenicity of tumor cells (Fig. 6A; ref. 30). Peptide vaccination alone significantly increased survival time, and combined temozolomide and peptide treatment doubled mice survival, as compared with vehicle-treated and non-specific OVA peptide-treated control mice (Fig. 6B). The tumors that eventually developed in vaccinated mice were widely SOX2-positive by immunohistochemistry (Fig. 6C). These observations indicate that Sox2 requirement by oligodendroglioma CSC may be potentially relevant from a therapeutic perspective.
To examine the direct effects of peptide vaccination on T-cell activation, splenocytes, and infiltrating lymphocytes from freshly harvested tissues of immunized and control mice were characterized by flow cytometry. The frequency of CD8+ (Fig. 6D) and CD4+ T cells (Supplementary Fig. S2) in spleens and tumor-infiltrated brains increased significantly in immunized mice compared with vehicle-treated controls. We also investigated whether pHGG-specific effector cells were generated in response to SOX2 peptide vaccination. Prestimulated splenocytes were assayed for in vitro cytotoxic activity against pHGG cells or NIH 3T3 cells (negative control) using a cytotoxicity MTT assay. The splenocytes from immunized mice, but not from vehicle-treated mice, displayed cytotoxic activity against tumor cells (Fig. 6E). The specificity of the effector immune response was confirmed by the absence of cytotoxicity against NIH 3T3 cells (Fig. 6E).
We report that Sox2 is required by oligodendroglioma stem cells to reinitiate tumor development within the transplanted mouse brain, and, in some conditions, for in vitro growth. Cells cultured from PDGF-B–induced mouse oligodendroglioma will reform a lethal tumor following transplantation in mouse brain; however, the majority of animals transplanted with Sox2-deleted cells remain tumor-free. Transduction of Sox2-deleted tumor cells with a Sox2-expressing lentivirus maintains tumor-initiating capacity, confirming that this is dependent on Sox2 activity. Finally, vaccination against Sox2 significantly delays tumor development, pointing to Sox2 (and its downstream targets) as a potential therapeutic target.
In adult mouse, Sox2 is expressed only in a minority of cells, mainly stem/progenitor cells within various tissues (8). In contrast, Sox2 is expressed in many tumor types, both in the brain and in other organs (mammary gland, lung, esophagus, bone; refs. 31–33).
In the majority of these tumors, Sox2 is not primarily altered/mutated, with the exception of its amplification in lung and esophageal squamous cell carcinoma (31). Sox2 deregulation is, in rare cases, the immediate downstream consequence of the primary lesion (34); more frequently, it is part of the altered transcriptional program of the tumor. As Sox2 is important for pluripotency and for reprogramming, these observations suggests an analogy between the role of Sox2 in CSC and in the normal development of stem cells (6).
Sox2 is required for the propagation of CSC in oligodendroglioma
In our oligodendroglioma model, Sox2 is necessary for the maintenance of CSC, in agreement with its requirement in normal NSC (13). Previous work showed that in human glioblastoma-derived cell lines, Sox2 downregulation by shRNAs impaired tumorigenesis following transplantation (35). In other patient-derived glioblastoma cells, Sox2 was described to be downstream to Sox4 in a TGF-β signaling-dependent tumorigenicity pathway and was required for in vitro maintenance of tumorigenic cells although its in vivo requirement was not tested (36). Interestingly, TGF-β promotes proliferation of tumors, including gliomas and osteosarcomas, through induction of PDGF-B (37).
While our results agree with those of Gangemi and colleagues (35), in that both glioblastoma and oligodendroglioma require Sox2 for in vivo tumorigenicity, some important differences should be noted. Sox2 ablation in glioblastoma (by shRNA) causes significant loss of cell proliferation in vitro, in media with added growth factors; instead, in oligodendroglioma, we noted only a small decrease in the presence of added growth factors in the proliferation of Sox2-ablated cells, both in clonal tests (Fig. 3) and in mass culture (not shown). However, omission of growth factors, and particularly combined addition of serum, a condition favoring NSC differentiation in vitro (25, 28, 29), strongly decreased Sox2-ablated oligodendroglioma cell proliferation, increased cell death, and caused important morphologic changes. These culture conditions might mimic conditions more similar to those encountered by tumor cells in the brain, with absence of abundant amounts of EGF/bFGF and presence of various cytokines and factors. Sox2-deleted cells showed morphologic changes (branching and flattening), together with marked positivity for differentiation markers (Fig. 3). Immunopositivity for oligodendrocyte differentiation markers was also observed in vivo within the very small number of Sox2-deleted tumor cells found 10 days after transplantation (Supplementary Fig. S3). “Priming” by Sox2 of “differentiation” genes in NSCs was reported (38), and NSCs expressing reduced levels of Sox2 (from mouse hypomorphic mutants) showed morphologic and gene expression abnormalities when induced to differentiation (25). Thus, as in normal NSCs (13, 25), Sox2 may be required in pHGGs in the presence of differentiation stimuli to prevent abnormal differentiation and apoptosis. Prodifferentiative stimuli (bone morphogenetic proteins, BMP, especially BMP4) efficiently antagonize glioblastoma development in mice (39) and targeting of molecules maintaining an undifferentiated state, such as EphA2 receptor, induced differentiation and loss of tumor-initiating capacity in mouse glioblastoma (40). Also in pHGGs, we previously documented a correlation between loss of tumor-initiating ability (following Pax6 overexpression) and the acquisition of differentiated features (41).
Overall, these results suggest that mechanisms causing tumor cell loss after Sox2 ablation may differ between different tumors (glioblastoma and oligodendroglioma), pointing to multiple molecular mechanisms of action of Sox2 in these cells.
On the other hand, Sox2 expression in tumor cells does not always correlate with a strict functional requirement for tumorigenesis. Sox2 is expressed in medulloblastoma, a cerebellar tumor most frequent in childhood; medulloblastoma CSCs express Sox2 in humans, and in mouse models (3, 42, 43). A class of medulloblastomas is associated with mutations activating the SHH pathway; these include SmoM2, a mutation in the SHH-receptor Smo leading to its constitutive activation; in mice, Cre-mediated activation of a SmoM2 transgene leads to medulloblastoma development (44). In these mice, we concomitantly deleted Sox2 (Sox2flox) by Cre; yet, Sox2-negative medulloblastoma still developed (42). The discrepancy with our present work might be explained in several ways. First, Sox2 might act upstream to Smo signaling; indeed, Sox2 was found to activate SHH expression in NSCs and neural cells (13, 45). Second, the close homolog Sox3 is expressed in medulloblastoma and might act redundantly with Sox2 in CSC maintenance (42). Third, in this system, SmoM2-induced medulloblastoma development in vivo may likely arise from multiple SmoM2-expressing cells and it is possible that additional mutations in a subset of these cells allow to bypass Sox2 requirement. Finally, although Sox2 activity was not strictly required for medulloblastoma development, experimental increase of Sox2 levels was found to correspondingly affect medulloblastoma cell proliferation (42).
The differences between neural tumors with respect to the degree of their Sox2 requirement are reminiscent of the differences in Sox2 requirement between different regions of the normal, developing nervous system. Sox2 is expressed ubiquitously in neural stem/progenitor cells, yet its deletion in vivo has region- and stage-specific effects in the brain (hippocampus, ventral telencephalon; refs. 13, 46). These observations point to specificities in the downstream gene expression networks controlled by Sox2 in tumorigenic as well as in normal neural (stem) cells.
Oligodendrogliomas may arise within the committed oligodendrocyte lineage, by “reprogramming” to a CSC state. Oligodendrocyte precursor cells (OPC) can be “reprogrammed” to a neural stem-like state, by sequential treatment with PDGF and EGF, and this process requires Sox2 reactivation (47). A future in-depth molecular investigation of Sox2 function in our model system may uncover if Sox2 regulates genes critical for reprogramming committed cells to a stem cell status, acting as a pioneer factor in ways related to its action in iPS cell generation (6).
Sox2 as a potential therapeutic target
The requirement for Sox2 by CSC raises the possibility that Sox2 itself may qualify as a target for therapeutic intervention. Targeting CSC may be a strategy to increase the potential efficacy of immunotherapy (27). Sox2 vaccination significantly prolongs survival enhancing systemic and local immune response (Fig. 6). Sox2 is localized in the nucleus, and is thus not, a priori, the most accessible molecule to target. However, recent data suggest that intracellular oncoproteins can be targeted by vaccination, as some intracellular antigens may be released and expressed on the surface of cancer cells (48). Antibodies and T-cell immune responses against SOX2 have been detected in patients with monoclonal gammopathy (MGUS), a premalignant condition to myeloma, where Sox2 expression marks the clonogenic compartment (49), and, recently, in about 50% of patients with non–small cell lung carcinoma (NSCLC; ref. 50). Cellular anti-SOX2 immunity inhibited the growth of MGUS cells in vitro and the presence of anti-SOX2 T cells predicted favorable clinical outcome (49); in NSCLC, T-cell response against SOX2 was associated with NSCLC regression upon immunotherapy with anti-PD-1 antibodies (50). These observations suggest that the immune system may be able to “discover” tumor-associated SOX2. Furthermore, an immune reaction by T cells elicited by SOX2-derived peptides (one of which was used here) was reported to lyse human glioblastoma-derived cells in culture (26). Finally, we previously found that vaccination against GLAST, a protein retaining significant expression in the adult brain, elicited an immune reaction specifically targeted to the tumor, not damaging the surrounding tissue (27).
The fact that late-arising tumors that eventually developed in vaccinated animals were widely Sox2-positive (Fig. 6) is consistent with the hypothesis of a failure of the immune system to completely eradicate Sox2-positive tumor cells, rather than with escape mechanisms developed by the tumor, allowing it to develop without Sox2. Collectively, these observations suggest that targeting Sox2-expressing cells may provide a basis for therapeutic approaches. Complementing Sox2 immunotherapy with action directed against some downstream Sox2 targets in oligodendroglioma might further increase the efficacy of this approach.
The observations about Sox2 requirement in neural tumors are extended by the reported requirement for Sox2 in a wider sample of tumor types. These include tumors of the osteoblast lineage, as shown in osteosarcoma cell lines (32); here, Sox2 is required also in the normal tissue stem cell counterpart, osteoblast stem/progenitor cells (10), as seen with neural cells. CSC from mammary tumors cultured as tumorigenic “mammospheres” express Sox2 and Sox2 knockdown impairs mammosphere formation and delays tumor formation following transplantation (33).
We conclude that targeting Sox2, likely in combination with selected downstream targets, may provide an effective strategy to antagonize the development of oligodendroglioma, and, perhaps, other tumor types.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: R. Favaro, S. Ottolenghi, P. Malatesta, S.K. Nicolis
Development of methodology: R. Favaro, G. Finocchiaro, S.K. Nicolis
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R. Favaro, I. Appolloni, S. Pellegatta, A.B. Sanga, P. Pagella, E. Gambini, F. Pisati, M. Foti, G. Finocchiaro, P. Malatesta, S.K. Nicolis
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R. Favaro, I. Appolloni, A.B. Sanga, M. Foti, G. Finocchiaro, P. Malatesta, S.K. Nicolis
Writing, review, and/or revision of the manuscript: R. Favaro, I. Appolloni, S. Pellegatta, S. Ottolenghi, M. Foti, G. Finocchiaro, P. Malatesta, S.K. Nicolis
Study supervision: S.K. Nicolis
The authors thank L. Naldini for the mCherry-encoding lentivirus, S. Brunelli for the Cre-encoding lentivirus, and C. Taveggia for the anti-GalC and anti-O4 antibodies.
This study was supported by grants from Associazione Italiana per la Ricerca sul Cancro (AIRC, grant IG-5801) and Fondazione Cariplo (grant 2010-0673; S.K. Nicolis), AIRC grant NUSUG 1180 and Italian Ministry of Health grant (GR-2008-1135643; P. Malatesta). A.B. Sanga was the recipient of a Fondazione Cariplo Interbiomed PhD fellowship.