Author contributions: M.M.S.: conception and design, collection of data, data analysis and interpretation, manuscript writing, final approval of manuscript; A.R.: collection of data, data analysis and interpretation, final approval of manuscript; J.K.W.: provision of study material and patients, final approval of manuscript; B.H.C.: conception and design, collection of data, data analysis and interpretation, manuscript writing, final approval of manuscript.
Disclosure of potential conflicts of interest is found at the end of this article.
First published online in STEM CELLS EXPRESS August 5, 2009.
Signal transducer and activator of transcription 3 (STAT3) regulates diverse cellular processes, including cell growth, differentiation, and apoptosis, and is frequently activated during tumorigenesis. Recently, putative glioblastoma stem cells (GBM-SCs) were isolated and characterized. These cells can self-renew indefinitely in culture, are highly tumorigenic, and retain the ability to differentiate in culture. We have found that treatment of GBM-SCs with two chemically distinct small molecule inhibitors of STAT3 DNA-binding inhibits cell proliferation and the formation of new neurospheres from single cells. Genetic knockdown of STAT3 using a short hairpin RNA also inhibits GBM-SC proliferation and neurosphere formation, confirming that these effects are specific to STAT3. Although STAT3 inhibition can induce apoptosis in serum-derived GBM cell lines, this effect was not observed in GBM-SCs grown in stem cell medium. Markers of neural stem cell multipotency also decrease upon STAT3 inhibition, suggesting that STAT3 is required for maintenance of the stem-like characteristics of these cells. Strikingly, even a transient inhibition of STAT3 leads to irreversible growth arrest and inhibition of neurosphere formation. These data suggest that STAT3 regulates the growth and self-renewal of GBM-SCs and is thus a potential target for cancer stem cell-directed therapy of glioblastoma multiforme. STEM CELLS 2009;27:2383–2392
Glioblastoma multiforme (GBM), the most common adult brain tumor, is a highly malignant and aggressive disease. GBM tumors are invasive and highly vascularized and patients diagnosed with GBM have a mean survival time of only 12–14 months . GBM can arise de novo or from lower grade astrocytomas. GBMs are composed of multiple cell types, including cells expressing astrocytic, neuronal, or both astrocytic and neuronal lineage markers, suggesting they may originate from a multipotent stem cell.
Recent work has led to the identification, in several cancer types, of a putative tumor “stem cell” with distinct properties from the bulk tumor and from traditional serum-derived lines. Tumor stem cells display an undifferentiated phenotype and an enhanced ability to initiate tumor formation relative to other cells from the bulk tumor in mouse xenograft models. Tumor stem cells have been isolated from human GBM. These cells share many properties with normal neural stem cells (NSCs) [2–5]. GBM-derived stem cells (GBM-SCs) can self-renew, proliferate, and differentiate to form multiple cell types, including cells expressing neuronal and glial markers. Unlike normal NSCs, GBM-SCs are highly tumorigenic in mice and display aberrant proliferative capacity and gene expression patterns . Tumors initiated by GBM-SCs recapitulate the phenotype of the original tumor from which they are isolated, and microarray analysis has shown that GBM-SCs have a gene expression signature that more closely resembles the tumor of origin than do serum-derived cell lines from the same tumor .
Signal transducer and activator of transcription 3 (STAT3), a member of the STAT family of transcription factors, is important in GBM, tumorigenesis, central nervous system development, and embryonic stem (ES) cell biology. STAT3 is activated by a wide variety of cytokines or growth factors. Upon tyrosine phosphorylation by receptor-associated tyrosine kinases, STAT3 translocates to the nucleus and regulates transcription of target genes . STAT3 target genes regulate many cellular processes, including proliferation and apoptosis [8–10]. Constitutive activation of STAT3 has been observed in many human cancers, including breast, head and neck, prostate, melanoma, and thyroid cancer . Knockout of STAT3 in the mouse epithelium completely abrogates the induction of skin tumors by the carcinogen 7,12-dimethylbenz[a]anthracene . Mice overexpressing constitutively activated STAT3 in alveolar epithelial cells develop spontaneous lung tumors .
STAT3 is also activated in a high percentage of GBMs . We previously used RNA interference (RNAi) knockdown of STAT3 in serum-derived GBM cell lines to demonstrate that STAT3 knockdown induces apoptosis in GBM cells . We have also shown that STAT3 knockdown inhibits the expression of telomerase, Bcl-xl, and survivin in serum-grown GBM cell lines [9, 15]. Thus, STAT3 plays an antiapoptotic role in GBM cell lines.
In addition to its role in tumorigenesis, STAT3 is also an important regulator of stem cells and the developing nervous system. Mouse ES cells are dependent on leukemia inhibitory factor (LIF), a potent activator of STAT3, and dominant negative STAT3 leads to ES cell differentiation and loss of pluripotency [16–18]. In both ES cells and NSCs, STAT3 is important in maintaining self-renewal. Deletion of STAT3 in murine embryonic NSCs inhibits neurosphere formation and self-renewal [19, 20]. In contrast, other data indicate that STAT3 plays a critical role in the differentiation of NSCs into astrocytes [21, 22]. Thus, STAT3 may have multiple roles in NSC function.
The function of STAT3 in GBM-SCs, however, has not been determined. Here, we show that inhibition of STAT3 in GBM-SCs irreversibly abrogates neurosphere formation and inhibits proliferation. In addition, we found that inhibition of STAT3 causes the downregulation of genes associated with the NSC phenotype, providing evidence that STAT3 regulates multipotency in these cells. These results suggest that STAT3 could be an effective therapeutic target for the treatment of GBM.
MATERIALS AND METHODS
Establishment and Culture of GBM-SC Lines
GBM tumor samples were obtained from recent surgical resections in accordance with the Tufts Medical Center institutional review board. Tumor samples were dissociated to form a single cell suspension, and plated in serum-free Dulbecco's modified Eagle's medium (DMEM)/F12 supplemented with B27 (Invitrogen, Carlsbad, CA, http://www.invitrogen.com), epidermal growth factor (EGF) (20 ng/ml; Peprotech, Rocky Hill, NJ, http://www.peprotech.com), and fibroblast growth factor (FGF)-2 (20 ng/ml; Peprotech). Neurosphere cultures were passaged approximately every week after initial neurosphere formation was observed. Sphere dissociation was performed according to an acid base dissociation protocol . When indicated, differentiation of neurosphere cultures was obtained following protocols established for the differentiation of neural progenitor cells . In short, cells were plated overnight on poly-L-ornithine- and laminin-coated plates in DMEM/F12 supplemented with B27 but lacking EGF and FGF-2. After 2 days of growth factor withdrawal, DMEM/F12 was replaced with DMEM supplemented with 2% fetal bovine serum (FBS). Cells were lysed for western blotting or reverse transcription quantitative polymerase chain reaction (RT-qPCR) analysis after 5 days in 2% FBS/DMEM. Inhibitor experiments were performed using 100 μM S3I-201 (National Cancer Institute [NCI] NSC 74859) or 30 μM STA-21 (BioMol International, Plymouth Meeting, PA, http://www.biomol.com; NCI, NSC 628869) unless otherwise indicated.
Immunoblotting and Immunocytochemistry
Cells were harvested by centrifugation and lysed in radioimmunoprecipitation assay buffer—0.15 mol/l NaCl, 1% NP40, 0.01 mol/l desoxycholate, 0.1% SDS, 0.05 mol/l Tris-HCl (pH, 8.0). Lysates were electrophoresed on SDS-PAGE gels and transferred to nitrocellulose membranes. Membranes were probed with each antibody according to the manufacturer's instructions and immunoreactive proteins were visualized using Western Lightning chemiluminescence reagent (PerkinElmer Life and Analytical Sciences, Waltham, MA, http://www.perkinelmer.com). Anti-β-actin (1:2,000) was obtained from Sigma-Aldrich (St. Louis, MO, http://www.sigmaaldrich.com); anti-STAT3 (1:1,000), anti-pSer-STAT3 (1:200), anti-pTyr STAT3 (1:500), anti p-AKT (1:1,000), and anti–phosphatase and tensin homologue (PTEN) (1:1,000) were obtained from Cell Signaling Technology (Beverly, MA, http://www.cellsignal.com). Anti-nestin (1:2,000) and anti-CD133 (1:300) were obtained from Abcam (Cambridge, MA, http://www.abcam.com). Anti p-extracellular signal–regulated kinase (ERK) (1:1,000) was obtained from Santa Cruz Biotechnology Inc. (Santa Cruz, CA, http://www.scbt.com). For immunocytochemistry, cells were allowed to attach to laminin-coated coverslips, fixed with 4% parafomaldehyde for 20 minutes, washed, permeabilized, and blocked in 1% bovine serum albumin (BSA)/0.2% Tween-20/10% goat serum/phosphate-buffered saline (PBS), and stained with anti-pTyrSTAT3 (1:50; Cell Signaling Technology). After washing, cells were stained with a secondary goat anti-rabbit antibody (1:50; Jackson Immunoresearch Laboratories, West Grove, PA, http://www.jacksonimmuno.com) and counterstained with 4′,6-diamidino-2-phenylindole (DAPI).
RNA was isolated using the RNeasy mini-kit (Qiagen, Valencia, CA, http://www1.qiagen.com) according to the manufacturer's instructions and treated with DNAse I (Invitrogen) for 30 minutes in order to digest genomic DNA. RNA was reverse transcribed with Superscript III (Invitrogen). qPCR was performed using the Quantitect SYBR Green PCR Kit (Qiagen) and the MX4000 real time cycler (Stratagene, LA Jolla, CA, http://www.stratagene.com). Data presented are the average of three individual experiments; within each experiment, technical triplicates were performed. Fold changes were calculated relative to control by the 2−ΔΔCt method . All primer sets amplify a single product of expected size; the following sequences were used: Olig2 forward, GGACAAGCTAGGAGGCAGTG; Olig2 reverse, ATGGCGATGTTGAGGTCGTG; glial fibrillary acidic protein (GFAP) forward, GGCAAAAGCACCAAAGACGG; GFAP reverse, GGCGGCGTTCCATTTACAAT; βIII-tubulin forward, GCCTCTTCTCACAAGTACGTG; βIII-tubulin reverse, CCCCACTCTGACCAAAGATGAA; β-actin forward, CCTGGGCATGGAGTCCTGTGG; and β-actin reverse, CTGTGTTGGCGTACAGGTCTT.
GS7-2 or GS6-22 cells were treated for 24 hours with dimethylsulfoxide (DMSO) control or STAT3 inhibitor and lysed in a high salt lysis buffer (500 mM NaCl, 50 mM Tris, 1% Triton X-100, 10% glycerol, 1 mM EDTA). Lysates were treated with STAT3 inhibitors (STA-21, 30 μM; S3I-201, 100 μM) for 30 minutes at 37°C. Twelve micrograms total protein was incubated with 32P-end-labeled human sis-inducible element (hSIE) STAT3 binding probe (approximately 40,000 cpm per reaction), 5 μg/μl poly dIdC, 0.1M dithiothreitol, 10 mg/ml BSA, and 50% glycerol, and incubated for 30 minutes at 30°C. Lysates were loaded on a 5% PAGE gel and electrophoresed; the gel was fixed and exposed to film overnight.
GS7-2 and GS6-22 cells were treated with STA-21, S3I-201, or DMSO control for 7 days. On day 7, the cells were trypsinized, washed, and fixed in paraformaldehyde in suspension. Cells were incubated with Stem Cell Technologies mouse anti-nestin antibody (1:50) for 30 minutes, washed, and incubated with Jackson fluorescein isothiocyanate (FITC)-rabbit anti-mouse secondary antibody (1:50) for 20 minutes in the dark. Cells were washed four times in BD Biosciences Perm/Wash buffer (BD Biosciences, San Diego, http://www.bdbiosciences.com), resuspended in PBS, and analyzed via flow cytometry.
Lentiviral Short Hairpin RNA Infection
Cells were infected with either PLKO.1 nontargeting control lentivirus (Sigma) or shSTAT3 PLKO.1 lentivirus (clone number, TRCN0000020843; Sigma) at an approximate multiplicity of infection of 35 in the presence of 8 μg/ml polybrene. Twenty-four hours postinfection the virus-containing medium was removed and replaced with fresh stem cell medium for 24 hours. Cells were then treated with 2.5 μg/ml puromycin. For 5-bromo-2′-deoxyuridine (BrdU) assays, cells were selected in puromycin for 2 days, plated in stem cell medium without puromycin for 24 hours, then pulsed with 10 μM BrdU for 3 days. Cells were stained using the FITC BrdU flow kit (BD Biosciences) and analyzed on a flow cytometer.
Establishment of GBM-Derived Stem Cell Lines
In order to determine what role STAT3 plays in GBM-SCs, we first established GBM-SC lines from human GBM tumor specimens. Primary tumor samples were dissociated and plated as single cells in serum-free medium containing FGF-2 and EGF, culture conditions that favor the growth of GBM-derived stem cells . Although we have established several GBM-SC lines, this work focuses on two lines, which we have designated GS7-2 and GS6-22. Both stem cell lines are derived from right temporal GBMs; GS7-2 was derived from the tumor of an 83-year-old male whereas the GS6-22 cells were isolated from an 82-year-old female. These cell lines have the expected characteristics of GBM-SCs [4, 6]. Both cell lines grow as nonadherent neurospheres in culture and can be continuously passaged (Fig. 1A). Both GBM-SC lines express the NSC markers CD133 and nestin (Fig. 1B). Ninety-eight percent of GS6-22 cells and 94% of GS7-2 cells express CD133, as determined by immunostaining (data not shown). When plated on poly-L-ornithine/laminin in the presence of low (2%) serum and without EGF and FGF-2, conditions established for the differentiation of NSCs and GBM-SCs, both GBM-SC lines grow as adherent monolayers and begin to exhibit morphology consistent with that of astrocytes and neurons (Fig. 1A). Importantly, under differentiation conditions, the expression of the NSC markers CD133, nestin, and olig2 in both the GS6-22 and GS7-2 cells decreased (Fig. 1B, 1C). Concurrent with this decrease in stem cell markers, we observed an induction of the astrocytic marker GFAP and the neural marker βIII-tubulin (Fig. 1C). This demonstrates that the tumor-derived cell lines retain the ability to differentiate, a defining characteristic of both normal and tumor-derived NSCs. We have confirmed that the GS7-2 cell line gives rise to tumors when as few as 1,000 cells are implanted in the brains of nude mice. Additional characterization of these lines reveals that neither stem cell line harbors the epidermal growth factor receptor vIII mutation present in a subset of GBMs (supporting information Fig. 1). The loss of PTEN expression, which is often found in GBM patient samples, is observed in the GS6-22 line but not in the GS7-2 stem cell line (supporting information Fig. 2) [25, 26].
STAT3 Is Present and Activated in GBM-SCs
Next, we characterized the expression and phosphorylation status of STAT3 in GBM-SCs. STAT3 is expressed in these cells as shown by immunoblotting (Fig. 2A). Immunoblotting with phosphospecific antibodies demonstrated that STAT3 is phosphorylated on both tyrosine-705 and serine-727 in the GS6-22 and GS7-2 GBM-SC lines. Because phosphorylation of tyrosine-705 is required for DNA-binding, and serine-727 phosphorylation has been shown to stimulate maximal transcriptional activity [27, 28], we conclude from our immunoblots that STAT3 is present and activated in GBM-SCs under NSC culture conditions. We found that STAT3 is also activated in three additional GBM-SC lines that we have isolated, as shown by immunostaining with antiphospho-Y705 staining (supporting information Fig. 3). Thus, STAT3 activation is a common feature of GBM-SCs.
Interestingly, differentiation of the GS6-22 and GS7-2 lines for 7 days led to a decrease in phosphotyrosine levels, whereas phosphoserine STAT3 became undetectable (Fig. 2A). A cell line derived from the GS7-2 tumor sample in the presence of serum had comparable levels of total STAT3, but undetectable phosphotyrosine or phosphoserine STAT3. Thus, both the differentiated GBM-SCs and a matched serum-derived cell line have reduced STAT3 phosphorylation.
In order to assess the function of STAT3 in GBM-SCs, we first analyzed two structurally distinct small molecule inhibitors of STAT3 DNA binding. These inhibitors, STA-21 and S3I-201, both target the SH2 domain of STAT3, thereby preventing STAT3 dimers from binding DNA and activating transcription of their target genes [29, 30]. To confirm their STAT3 inhibitory activity in GBM-SCs as measured by DNA binding, each drug was used in a bandshift assay with a radiolabeled SIE probe that binds STAT3 with high affinity . As shown in Figure 2B, both STA-21 and S3I-201 abolished the ability of STAT3 to bind the SIE probe in GS7-2 cells. STA-21 and S3I-201 also inhibited STAT3 DNA binding in GS6-22 cells (supporting information Fig. 4). This confirms that STA-21 and S3I-201 can inhibit STAT3 DNA binding in our GBM-SC lysates, which is consistent with published reports of their efficacy against STAT3 in other cell lines in vitro and in vivo [29, 30]. Neither STA-21 nor S3I-201 affect ERK or AKT activation, as shown by immunoblotting of the U251 GBM serum line (supporting information Fig. 5).
Previously, both STA-21 and S3I-201 have been shown to inhibit STAT3 preferentially as compared with STAT1 and STAT5 [29, 30]. From immunoblots of GS6-22 and GS7-2 GBM-SCs, it is clear that although these cell lines express STAT5 and STAT1, STAT3 is the major activated STAT. Phosphotyrosine STAT1 and STAT5 were undetectable under NSC culture conditions (supporting information Fig. 6). Thus, STAT3 is the principal activated STAT protein in these cells.
Because STA-21 and S3I-201 directly target the SH2 domain of STAT3 and not the upstream activating kinases, their efficacy is not necessarily correlated with a decrease in STAT3 tyrosine phosphorylation, in contrast to DNA-binding activity. However, in some cases, STAT3 tyrosine phosphorylation may decrease with these drugs if targeting of STAT3 to receptors by the SH2 domain is disrupted. In fact, a time course of both GS6-22 and GS7-2 GBM-SCs treated with each inhibitor reveals that STAT3 tyrosine phosphorylation is more strongly inhibited in the GS6-22 line than in the GS7-2 line (supporting information Fig. 7). This is probably a result of distinct modes of STAT3 activation in the two cell lines .
Inhibition of STAT3 Prevents Neurosphere Formation and Decreases Proliferation in GBM-SCs
We next investigated the effect of STAT3 inhibition on GBM-SCs. Cultures of both GS7-2 and GS6-22 cells were dissociated and plated as single cells in the presence of either STA-21, S3I-201, or DMSO control. After several days, a dramatic difference was observed between the drug and control-treated cultures. Whereas neurospheres formed in the control wells, very few spheres formed from cells treated with either of the STAT3 inhibitors (Fig. 3A, supporting information Fig. 8). Whereas an average of five neurospheres could be seen in each field of view in DMSO-treated cultures, fewer than one neurosphere per field could be seen in wells treated with either STA-21 or S31-201 (Fig. 3B). Cells were viable, however, as determined by trypan blue exclusion (data not shown). Thus, inhibition of STAT3 abrogates neurosphere formation in GBM-SCs. Interestingly, treatment of intact neurospheres with either STAT3 inhibitor caused the dissociation of pre-existing neurospheres (Fig. 3C).
To determine whether the inhibition of neurosphere formation could be explained by decreased proliferation, GS6-22 and GS7-2 cells were cultured in the presence of STA-21, S3I-201, or DMSO for 24 hours and then pulsed with BrdU for 16 hours. Cells were analyzed by flow cytometry following staining with an anti-BrdU antibody to assess the percentage of cells that incorporated BrdU during the 16-hour pulse. Treatment with either STAT3 inhibitor resulted in significantly lower BrdU incorporation, as compared with the DMSO control cells (Fig. 3D). In the GS6-22 cell line, an average of 46% of DMSO-treated cells incorporated BrdU, compared with only 3% of STA-21-treated cells and 2% of S3I-201-treated cells. A similarly lower level of BrdU incorporation was observed in the GS7-2 line treated with inhibitors. The GS7-2 cells continued to proliferate under differentiation conditions, and upon STAT3 inhibition with STA-21 or S3I-201, a lower rate of BrdU incorporation—13% and 10% for inhibitor-treated cells, respectively, versus 32% for DMSO-treated cells—was observed (supporting information Fig. 9A). The GS6-22 cells, in contrast, did not continue to proliferate under differentiation conditions. Interestingly, BrdU incorporation of the serum-derived cell line derived from the GS7-2 tumor was not affected by STAT3 inhibitor treatment (supporting information Fig. 10A). This is consistent with the observation that the serum line does not have detectable activated STAT3, as determined by immunoblotting for phospho-STAT3 (Fig. 2A). This further confirms that these STAT3 inhibitors are not nonspecifically inhibiting cellular proliferation.
To confirm that STAT3 is required for the growth of GBM-SCs, the effects of knockdown of STAT3 by RNAi were investigated. Cells infected with a lentivirus containing a short hairpin (sh)RNA to STAT3 showed lower BrdU incorporation than cells expressing a control shRNA. In the GS6-22 cells, infection with the STAT3 knockdown virus resulted in a 65% lower BrdU incorporation than in cells infected with a nontargeting shRNA control virus (Fig. 3E). In the GS7-2 cells, BrdU incorporation was 50% lower in the cells infected with the STAT3 knockdown virus. Consistent with STAT3 small molecule inhibition, knockdown of STAT3 by RNAi decreased the ability of dissociated GS6-22 cells to form neurospheres (supporting information Fig. 8B, 8C). Although this phenotype is not as pronounced as the inhibition caused by the STAT3 small molecule inhibitors, this is likely a result of incomplete knockdown of STAT3 protein expression by this shRNA (supporting information Fig. 11). It is clear, however, that treatment with either of the two distinct small molecule inhibitors as well as genetic knockdown of STAT3 inhibits GBM-SC proliferation.
To confirm that the STAT3 inhibitors do inhibit cellular proliferation, GS7-2 and GS6-22 cells were treated with each inhibitor, and the cell number was counted 3 and 6 days later. STA-21 or S3I-201 treatment dramatically inhibited proliferation over the 6-day period (Fig. 3F). In the GS6-22 cells, STA-21- and S3I-201-treated cell numbers remained constant over the 6-day period. This suggests that STAT3 inhibition blocks cell growth rather than induces cell death. In the GS7-2 cells, a slightly lower cell number was observed in STAT3 inhibitor-treated cells. Although this difference was statistically significant, no difference in live cells between control and STAT3 inhibitor-treated cells was observed by trypan blue exclusion of cells before counting (data not shown). Overall, however, it is clear that both small molecule inhibitors and shRNA knockdown of STAT3 lead to a dramatic decrease in GBM-SC proliferation.
Inhibition of STAT3 Causes Apoptosis in GBM Serum-Derived Cell Lines, but Not in GBM-SCs
Because we have previously shown that STAT3 regulates survival of traditional serum-derived GBM cell lines, we asked whether STAT3 inhibitor treatment would have a similar effect on GBM-SCs . As judged by annexin V staining (Fig. 4A), GS7-2 and GS6-22 GBM-SCs treated with either STA-21 or S3I-201 did not show a greater rate of apoptosis than untreated or DMSO-treated cells over a 3-day period. Treatment with hydrogen peroxide, however, did result in a higher percentage of annexin V-positive cells, demonstrating that the GBM-SCs are capable of undergoing apoptosis. Thus, the lack of apoptotic induction with STAT3 inhibitor treatment does not reflect a complete resistance of the stem cells to apoptosis. The percentage of dead cells staining positive for propidium iodide also was no different with STAT3 inhibitor treatment (supporting information Fig. 12), further indicating that cell viability is also not affected by drug treatment. The GBM-SC lines, then, do not undergo greater apoptosis upon STAT3 inhibitor treatment. It is possible that the modestly lower cell number in the GS7-2 line in response to STAT3 inhibitors is a result of greater autophagic cell death instead of apoptosis.
Though STAT3 inhibitor treatment failed to induce apoptosis in GBM-SCs, it did cause a greater rate of apoptosis in the A172 serum-derived GBM cell line. Cells treated with STA-21 for 3 days had a twofold greater number of annexin V-positive cells than DMSO-treated and untreated controls (Fig. 4B). The A172 cell line harbors constitutively activated STAT3 and has been shown to undergo apoptosis when STAT3 is knocked down using RNAi . Interestingly, the GS6-22 and GS7-2 stem cell lines became sensitive to apoptosis induced by STAT3 inhibition when differentiated for 7 days (supporting information Figure 9B). The STAT3 inhibitors also induced extensive cell death of GBM-SC when placed into differentiation medium in the absence of growth factors.
In contrast, the serum line derived from the GS7-2 tumor lacked activated STAT3 (Fig. 2) and failed to undergo apoptosis in response to STAT3 inhibitor treatment (supporting information Fig. 10B). Thus, as expected, STAT3 inhibition leads to apoptosis and cell death in serum-derived GBM lines containing activated STAT3, but not in serum-derived lines lacking tyrosine-phosphorylated STAT3. In GBM-SCs, STAT3 inhibitor treatment induced apoptosis only when removed from stem cell medium. This suggests that the role of STAT3 in GBM-SCs is distinct from its role in serum-derived glioma cell lines and in GBM-SC cultures that have been induced to differentiate.
Inhibition of STAT3 Depletes Markers of Multipotency in GBM-SCs
GBM-SCs treated with STAT3 inhibitors failed to proliferate, but did not undergo apoptosis. In order to determine whether this effect signifies a loss of self-renewal potential, we examined the expression of known stem cell markers in GBM-SCs treated with either STAT3 inhibitor. After 7 days of treatment with STA-21 or S3I-201, the expression of the NSC marker olig2 in GS6-22 and GS7-2 cells was lower at the mRNA level (Fig. 5A). Staining for the NSC marker nestin was also lower in GBM-SCs treated with STAT3 inhibitors, consistent with a loss of multipotency (Fig. 5B). Expression of the neuronal marker βIII-tubulin was slightly higher in both cell lines treated with the STAT3 inhibitors (Fig. 5A), suggesting that STAT3 inhibition over a period of 7 days leads to a partial induction of differentiation in addition to the observed loss of multipotency. However, the STAT3-inhibitor-treated cells did not attach to the tissue culture plate and morphologically continued to resemble undifferentiated cells. Other differentiation markers, such as GFAP, are not upregulated (data not shown), suggesting that STAT3 inhibition alone is not sufficient to induce complete differentiation of the GBM-SCs.
Because loss of stem cell markers suggests that inhibition of STAT3 initiates differentiation of GBM-SCs, we sought to determine whether the inhibition observed was reversible. Cells were treated transiently for various periods of time with either STA-21 (Fig. 5C) or S3I-201 (supporting information Fig. 13). A single treatment of GS6-22 cells with STA-21 for as little as 4 hours prevented neurosphere formation from single cells for several weeks despite replating cells in fresh medium lacking STA-21. Transient treatment with S3I-201 also irreversibly prevented neurosphere formation (supporting information Fig. 13). Cells remained viable, as judged by trypan blue exclusion and by luciferase transgene expression (data not shown), but remained single cells. STAT3 activity was not permanently blocked by inhibitor treatment, as demonstrated by the recovery of its nuclear localization as quickly as 1 hour following inhibitor removal (supporting information Fig. 14). The inability to produce neurospheres for as long as 2 weeks after drug removal, along with the downregulation of stem cell markers, suggests that STAT3 inhibition was able to induce a sustained loss of self-renewal with respect to growth in culture. Even transient treatment with STAT3 inhibitor was able to induce this loss, despite reactivation of STAT3 after drug removal. Both the loss of NSC markers and the permanent nature of the growth arrest induced by transient STAT3 inhibition suggest that activated STAT3 is critical for maintaining GBM-SCs in a proliferative and self-renewing state.
STAT3 is regulated by a wide variety of growth factors and cytokines . It has been shown to be activated in many different human cancers, including GBM [14, 33, 34]. Dominant negative STAT3 can inhibit transformation by v-src and activated STAT3 can transform 3T3 cells [35, 36]. In mice, directed expression of activated STAT3 to the lung induces lung cancer, and knockout of STAT3 in the skin epithelium renders mice resistant to skin tumor induction by phorbol esters [12, 13]. STAT3 regulates a remarkable number of key cancer genes, including the proliferation genes c-myc, p21, and cyclin D, the antiapototic genes bcl-xl and survivin, the angiogenic gene VEGF, the epithelial to mesenchymal transition genes twist and LIV1, and the immortalization gene telomerase [9, 15, 37–40].
Recently, there has been increasing evidence that a minority of cells in a tumor with stem cell-like properties, called cancer stem cells, are responsible for most of the tumorigenic potential. The evidence for this is particularly compelling for GBM, for which most of the tumor-initiating ability typically resides in the CD133+ fraction of cells [2, 6]. These cells have many properties of NSCs and can differentiate in culture, but unlike normal NSCs can be maintained in culture indefinitely and form tumors in xenografts with high efficiency. Interestingly, STAT3 is known to play an important role in stem cells. It is required for the maintenance of pluripotency in murine ES cells [16–18, 41]. There is also evidence that it is necessary for the self-renewal of murine NSCs [19, 20].
Here we examined the function of STAT3 in GBM-SCs. We found that GBM-SCs express STAT3 and that it is phosphorylated on the activating tyrosine and serine residues. Inhibition of STAT3 in these cells with either small molecular inhibitors or RNAi resulted in inhibition of growth and neurosphere formation. Unlike what we have found for serum-derived glioma cell lines, GBM-SCs do not undergo apoptosis upon STAT3 inhibition in stem cell medium. This is likely a reflection of the finding that tissue stem cells are relatively resistant to apoptosis. Consistent with this, the GBM-SC lines became sensitive to apoptosis induced by STAT3 inhibition when they had differentiated. It is possible that STAT3 inhibition sensitizes GBM-SCs to apoptotic stimuli as it does for some nontransformed cell lines, but this remains to be determined [42, 43].
We also observed that, in addition to inhibiting the formation of neurospheres from single GBM-SCs, inhibition of STAT3 led to dissociation of preformed neurospheres. Although it is not clear how self-adhesion of GBM-SCs or NSCs is mediated, STAT3 is known to regulate various intercellular adhesion molecules, including intercellular adhesion molecule-1 (ICAM-1) .
In addition to inhibiting the proliferation of GBM-SCs, we found that inhibition of STAT3 leads to loss of expression of the NSC markers nestin and olig2. However, unlike cells grown in differentiation medium, the STAT3 inhibitor-treated cells failed to adhere to the culture dish and showed no clear signs of morphological differentiation, though there may have been a modest induction of some differentiation markers, such as βIII-tubulin. The astrocytic marker GFAP was not upregulated. This suggests that the GBM-SCs lost their ability to self-renew in culture, but did not fully execute the differentiation program following STAT3 inhibition. An upregulation of the neuronal marker βIII-tubulin but not the astrocytic marker GFAP suggests that the partial differentiation we observed is biased toward the neuronal lineage, which is consistent with reports that STAT3 is necessary for NSC differentiation to astrocytes [21, 22].
Consistent with this, we found that even a transient inhibition of STAT3 led to an irreversible loss of the ability of the GBM-SCs to form neurospheres. Along with growth inhibition and loss of stem cell markers, this suggests that STAT3 inhibition permanently induces a loss of self-renewal potential with respect to the ability to proliferate in culture as undifferentiated cells. This may or may not correlate with the ability of these cells to initiate tumors in mice. A possible mechanism for the permanent loss of this stem cell characteristic is that STAT3 inhibition leads to epigenetic changes that are not readily reversed. This would explain the rapid and irreversible nature of this phenotype, and would be consistent with findings in ES cells and murine NSCs that inhibition of STAT3 leads to loss of self-renewal and induction of differentiation [16–20, 41]. Similarly, in olfactory bulb cells, knockdown of STAT3 induces terminal neuronal differentiation . In contrast, we found that inhibition of STAT3 in the GBM-SCs did not induce a fully differentiated phenotype, likely because of aberrant signaling caused by tumor-specific mutations. Indeed, inhibition of STAT3 in these cells under differentiation conditions led to cell death. This suggests that STAT3 might be a particularly good target for therapeutic inhibition of cancer stem cells. Short-term treatment could possibly be translated into long-term effects on growth, self-renewal, and resistance to apoptosis of the cancer stem cell population within the tumor. However, it will be important to determine whether STAT3 inhibition will deplete the normal, but quiescent, NSC population as well.
We have found that STAT3 in GBM-SCs is phosphorylated on both tyrosine and serine residues. Interaction of the STAT3 pY705 with the STAT3 SH2 domain is required for high-affinity DNA binding of the STAT3 dimer . Phosphorylation of serine-727 is generally thought to be necessary for maximal transcriptional activity . Previously, serine phosphorylation of STAT3 has been implicated in the self-renewal of NSCs . Indeed, recent evidence indicates that there are important functions of the unphosphorylated form of STAT3 as well . Our experiments do not distinguish clearly among any of these possibilities, though the loss of serine phosphorylation upon GBM-SC differentiation is consistent with a relationship between self-renewal potential and STAT3 serine-727 phosphorylation. The two drugs we used, STA-21 and S3I-201, both bind to the STAT3 SH2 domain, presumably preventing the STAT3 phosphotyrosine residue from interacting with it and thus abrogating the DNA-binding activity. This suggests a requirement for STAT3 tyrosine phosphorylation, but does not directly address whether serine phosphorylation is necessary for GBM-SC proliferation.
Previous studies have implicated STAT3 in the differentiation of neural precursor cells into astrocytes [21, 22, 48]. Our data here do not conflict with these data, but suggest an additional function for STAT3 in maintaining GBM-SC self-renewal. Our findings are consistent with previous work in mice that indicates that STAT3 maintains NSC self-renewal [19, 20]. From our data, this function of STAT3 appears to be maintained in cancer stem cells. It has been suggested that STAT3 promotes self-renewal of NSCs by regulating the Notch ligand DLL1 . However, we did not observe changes in DLL1 expression in GBM-SCs after STAT3 inhibition, suggesting another mechanism for maintaining self-renewal of these cells (data not shown). Future experiments will examine the mechanism by which STAT3 regulates self-renewal in GBM-SCs in culture.
Previous work has suggested that epigenetic changes that occur during astrocytic differentiation lead to changes in STAT3 function during NSC differentiation. For instance, there is a critical STAT3 binding site in the GFAP promoter that is methylated in precursor cells and that becomes demethylated and therefore STAT3 accessible during differentiation . Lee et at.  similarly found that induction of astrocytes in GBM-SC culture depends partly on demethylation and expression of the bone morphogenetic protein receptor (BMPR) 1B, and that this differentiation depends on STAT3 activation. Thus, it may be that STAT3 regulates different genes in astrocytes than it does in GBM-SCs or NSCs because of epigenetic differences between the cell types. This is consistent with our observations that STAT3 inhibition has a distinct phenotype in GBM-SCs and in differentiated cells. Whereas STAT3 inhibition blocks proliferation and self-renewal of GBM-SCs, upon differentiation these cells are sensitized to apoptosis upon STAT3 inhibition. Thus, our findings are in agreement with studies in normal NSCs and astrocytes, which have both been shown to depend on STAT3 activation for distinct cellular processes.
For most human tumors, STAT3 is believed to be oncogenic [11, 32, 50]. Most studies have also found STAT3 to be pro-oncogenic in serum-derived glioma cell lines as well [9, 14, 45, 51–54]. However, two recent papers suggest that, in some cases, STAT3 may be a tumor suppressor gene in GBMs that have lost PTEN expression [55, 56]. In contrast, our GS6-22 stem cell line has lost PTEN expression but remains sensitive to STAT3 inhibition. One possible explanation of these seemingly conflicting observations may be that some glioma cell lines are no longer in the STAT3-dependent stem cell state, but are blocked from terminal astrocytic differentiation by the loss of STAT3 and thereby stuck in a proliferative progenitor cell state. This would be consistent with the phenotype of the aberrant GBM-SC line of Lee et al.  that proliferates instead of differentiates in response to BMP signaling. The relative importance or frequency of these two distinct functions of STAT3 (stem cell self-renewal versus astrocytic differentiation) in primary human GBM remains to be determined, but is an important therapeutic question. Nonetheless, our finding that STAT3 is required for tumor stem cell self-renewal in GBM suggests that it might be an effective target for tumor stem cell-directed therapy.
This work was supported in part by a grant from the Brain Tumor Society and a gift from Catherine Pappas to B.H.C. M.M.S. was supported by NIH training grant T32 DK07542. We thank the NCI/DTP Open Chemical Repository (http://dtp.cancer.gov) for the STA-21 (NSC 628869) and S3I-201 (NSC 74859) inhibitors. We would like to thank the Tufts Neuroscience core for use of the Stratagene qPCR machine, P30 NS047243. We thank Anton Cochran, Forrest Jones, and James Krasker for their help with the proliferation experiments and Hans-Peter Biemann for critical reading of the manuscript.
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest.