Significance of low mTORC1 activity in defining the characteristics of brain tumor stem cells

Abstract

Background.

The significance of mammalian target of rapamycin complex 1 (mTORC1) activity in the maintenance of cancer stem cells (CSCs) remains controversial. Previous findings showed that mTORC1 activation depleted the population of leukemia stem cells in leukemia, while maintaining the stemness in pancreatic CSCs. The purpose of this study was to examine the currently unknown role and significance of mTORC1 activity in brain tumor stem cells (BTSCs).

Methods.

Basal mTORC1 activity and its kinetics were investigated in BTSC clones isolated from patients with glioblastoma and their differentiated progenies (DIFFs). The effects of nutrient deprivation and the mTORC1 inhibitors on cell proliferation were compared between the BTSCs and DIFFs. Tissue sections from patients with brain gliomas were examined for expression of BTSC markers and mTORC1 activity by immunohistochemistry.

Results.

BTSCs presented lower basal mTORC1 activity under each culture condition tested and a more rapid decline of mTORC1 activity after nutrient deprivation than observed in DIFFs. The self-renewal capacity of BTSCs was unaffected by mTORC1 inhibition, whereas it effectively suppressed DIFF proliferation. In agreement, immunohistochemical staining of glioma tissues revealed low mTORC1 activity in tumor cells positive for BTSC markers. In in vitro culture, BTSCs exhibited resistance to the antitumor agent temozolomide.

Conclusions.

Our findings indicated the importance of low mTORC1 activity in maintaining the undifferentiated state of BTSCs, implicating the relevance of manipulating mTORC1 activity when developing future strategies that target BTSCs.

In cancer research, the significance of cancer stem cells (CSCs) in the progression of cancers has long been the subject of intensive study.¹ CSCs undergo slow cell cycling and show tolerance to various stresses, which confer resistance to conventional therapies.² CSCs reside in either hypoxic or perivascular niches, where they maintain their undifferentiated state, which also contribute to their resistance to therapeutic intervention and the recurrence of cancer.^3–5 CSCs share some properties with dormant cancer cells and can survive in a nutrient-deprived microenvironment. Those CSCs, either in primary tumors or in remote sites, eventually undergo active proliferation in some microenvironments, which underlie late recurrence and the metastasis to distant organs.⁶

Many studies have investigated the role of brain tumor stem cells (BTSCs) in the progression of glioma. BTSCs can be isolated from gliomas by methods analogous to those described for the isolation of neural stem cells in developing mouse brains.^7–9 BTSCs show enhanced tumorigenicity compared with their differentiated progenies (non-BTSCs) in xenograft tumor models, which provides an opportunity to investigate the hierarchy of tumorigenicity within heterogeneous populations of tumor cells.^10,11 Some BTSC markers have been identified, including CD133, CD44, sex-determining region Y-box 2 (Sox2), Nanog, Nestin, and Oct4, which enables tracking the presence of BTSCs in tissues, although their expression levels are variable depending on individual BTSC clones, the tumor type, and in vitro culture conditions.^12–14 A hallmark of BTSCs is their ability to undergo changes in intracellular metabolism and gene expression through multiple mechanisms, which is essential for their adaptation to local microenvironments.^15,16

One of the challenges in understanding BTSCs has been the identification of mechanisms by which BTSCs adapt to cues from local niches to maintain their undifferentiated state. Although the mechanisms whereby perivascular and hypoxic niches regulate the maintenance of BTSCs have been extensively studied, the significance of a nutrient-deprived microenvironment in maintaining BTSCs and how BTSCs sense external nutrient levels have just begun to be studied.^16,17 In addition, although many studies have shown that the metabolic state in cancer cells differs from that of normal cells, limited information is available as to how BTSCs undergo metabolic reprogramming to survive in nutrient-deprived microenvironments, where non-BTSCs are destined to undergo necrosis.^18,19

The mammalian target of rapamycin (mTOR) is a serine/threonine kinase that forms a large protein complex named mTOR complex 1 (mTORC1), which is activated downstream of oncogenic pathways as well as by nutrients.^20,21 mTORC1 phosphorylates its downstream targets 40S ribosomal protein S6 kinase (S6K) and eukaryotic initiation factor 4E-binding protein 1 (4E-BP1) to regulate protein synthesis and cell growth, both of which are essential for cancer cell proliferation.^20,21 However, despite the fact that several inhibitors targeting the mTORC1 pathway are being developed, with many showing promising antitumor activities in animal tumor models, most of them showed limited efficacy in the treatment of cancer patients.^20,21 Also, controversy remains regarding the significance of the mTORC1 pathway in the maintenance of CSCs. For example, the mTORC1-specific inhibitor rapamycin inhibited the growth of xenografted CSCs derived from various solid cancers,^22–24 whereas the deregulated activation of mTORC1 depleted the population of leukemia stem cells,²⁵ indicating that mTORC1 activity needs to be tightly controlled to limit the differentiation of stem cells. Other evidence showed that mTORC1 inhibition led to the compensatory activation of the Akt pathway, which attenuated their therapeutic effects.²⁶ These data suggest the need to focus on cancer cell plasticity and heterogeneity in terms of the role of the mTORC1 pathway in cancer progression.

To the best of our knowledge, the significance of mTORC1 in the maintenance of BTSCs in gliomas has not been reported thus far. In this study, we investigated mTORC1 activity and its kinetics in response to nutrient deprivation and refeeding in 3 BTSC clones isolated from glioblastoma patients, and compared these results with those from non-BTSCs, which were termed here as DIFFs. We also tested the effects of mTORC1 inhibition on the proliferation of BTSCs and DIFFs. We found that BTSCs exhibited lower mTORC1 activity and sensitivity to mTORC1 inhibition than did DIFFs, which was in agreement with our histological analysis on glioma tissues showing an inverse correlation between mTORC1 activity and BTSC marker expression. These data suggested heterogeneous mTORC1 activities and variable nutrient requirements among the tumor cells, which accounted for the resistance of BTSCs residing in low-nutrient environments to conventional therapies.

Methods

Isolation and Culture of BTSCs and DIFFs

Three independent BTSC clones were isolated from glioblastoma tissues of patients who underwent surgery at the Nagoya University Hospital after obtaining informed consent. This study was approved by the Ethics Committee of Nagoya University Graduate School of Medicine. The procedures for BTSC isolation, culture, and differentiation into DIFFs were previously described.²⁷ Briefly, freshly dissociated tumor cells were incubated with BTSC medium comprising Neurobasal medium and the B-27 and N-2 supplements (Invitrogen), supplemented with human recombinant basic fibroblast growth factor and epidermal growth factor (20 ng/mL each; R&D Systems). To passage BTSCs, BTSC spheres were dissociated to single cells with the NeuroCult Chemical Dissociation Kit (StemCell Technologies). The differentiation of BTSCs was induced by culturing them in high-glucose Dulbecco’s modified Eagle’s medium (DMEM) (Wako) containing 10% fetal bovine serum (FBS). DIFFs cultured for 10 generations were used in all experiments, after confirming stable differentiation by monitoring the expression of stem cell markers by western blot analysis and immunofluorescence (Fig. 1B, D). DMEM without amino acids and glucose was purchased from the Cell Science & Technology Institute. The protocols for sphere formation assays with BTSCs are described in the Supplementary material.

Fig. 1.

BTSCs exhibited lower basal mTORC1 activity than DIFFs. (A) Characterization of BTSC clones used in the study. Sphere formation of BTSCs and fibroblastic morphology of DIFFs are shown. Scale bar, 30 µm. (B) Expression profiles for stem cell and differentiation markers in the 3 BTSC clones and their respective DIFFs. Mr, molecular marker. (C) Quantification of mRNA expression of stem cell and differentiation markers by qPCR. The expression levels of differentiation and stem cell markers in BTSCs and DIFFs are presented as relative values compared with those in DIFFs and BTSCs, respectively, which were assigned a value of 1. A.U., arbitrary unit. (D) Immunofluorescence staining of BTSCs (clone 222) showed significant expression for Sox2, Nanog, and Nestin, whereas the expression of differentiation markers was detected in DIFFs. In far bottom panel, the BTSCs and DIFFs were double-stained for Sox2 and phospho-S6. Note that the BTSCs were heterogeneous for Sox2 expression, where we observed an inverse correlation between Sox2 and phospho-S6 (arrowheads). (E) Lysates from BTSCs and DIFFs 48 hours after passaging were subjected to western blot analysis. Higher activity of mTORC1 was found in DIFFs than in BTSCs.

Quantitative Reverse Transcription–Polymerase Chain Reaction

Quantitative (q)PCR analysis was performed using an ABI7300 thermal cycler and Thunderbird SYBR qPCR Mix (Toyobo), according to the manufacturers’ instructions. The data were analyzed by the comparative threshold cycle method and normalized against glyceraldehyde-3-phosphate dehydrogenase expression. The sequences of the primers used are described in the Supplementary material.

Antibodies, Western Blot Analysis, and Immunofluorescence

The following antibodies were used in this study: anti-mTOR, anti-S6K, anti-phospho-S6K (Thr389), anti-S6 ribosomal protein (S6), anti-phospho-S6 (Ser235/236), anti-phospho-S6 (Ser240/244), anti-4E-BP1, anti-phospho-4E-BP1 (Thr37/46), anti-Sox2, anti-Oct4, phospho–signal transducer and activator of transcription 3 (Y705), and anti-phospho–extracellular signal-regulated kinase (Thr202/Tyr204) (all from Cell Signaling Technology); anti-Nanog and anti–oligodendrocyte lineage transcription factor 2 (Olig2) (Abcam); anti-Nestin and anti-oligodendrocyte specific protein (OSP) (Millipore); anti–glial fibrillary acidic protein (GFAP) and anti–β-actin (Sigma). Protocols for western blot and immunofluorescence analyses are described in the Supplementary material.

Immunohistochemistry

All tumor samples (5 cases each for glioblastoma, anaplastic astrocytoma, and diffuse astrocytoma samples) were obtained at the time of surgery, after patients gave informed consent. This study was approved by the Ethics Committee of Nagoya University Graduate School of Medicine. Protocols for immunohistochemical analyses, double staining, and the quantification are described in the Supplementary material.

Results

Characterization of the Properties of BTSCs Used in the Study

We first characterized 3 BTSC clones (222, 316, and 1228) that we previously isolated from patients with glioblastoma of diverse genetic background (Supplementary Table 1).²⁷ As described previously, all the BTSC clones satisfied the following criteria: (i) BTSCs can be maintained in the BTSC media for at least 3 months and (ii) 1000 BTSCs can form tumors in the brains of non-obese diabetic mice with severe combined immunodeficiency disease (NOD/SCID mice), but 100000 DIFFs cannot.²⁸ All BTSCs examined exhibited a sphere-like morphology and could be maintained in the BTSC medium (Fig. 1A). In vivo tumorigenicity in NOD/SCID mice was again proven using a selected BTSC cell line (clone 222) which showed higher engraftment efficiency of BTSCs than DIFFs (Supplementary Fig. S1). Replacement of the complete growth medium with DMEM containing 10% FBS and subsequent culture for 48 hours induced the differentiation of BTSCs into adherent glioblastoma cells (DIFFs) (Fig. 1A). Western blot, qPCR, and immunofluorescence-staining results showed that the expression of BTSC markers such as Sox2, CD44v9 (a splicing variant of CD44), and Oct4 was downregulated, while that of differentiation markers such as GFAP, OSP, and β3-tubulin was upregulated following differentiation (Fig. 1B–D, Supplementary Figs. S2–S3). Notably, our data conflicted with previous findings with respect to the changes in some BTSC and differentiation markers; CD133 and Nestin, previously well-described markers for BTSCs, were not significantly downregulated upon differentiation in the BTSC clones 316 and 1228, respectively (Fig. 1C). These data indicated that not all of the known BTSC markers perfectly represent the undifferentiated state and raised the possibility that variation occurred among BTSC clones.

BTSCs Displayed Lower Basal mTORC1 Activity Compared with Their Differentiated Progenies

To examine the basal activity of mTORC1 in BTSCs and DIFFs, we monitored the phosphorylation of mTORC1 downstream proteins, including S6K, S6, and 4E-BP1, all of which have been used as surrogate markers for mTORC1 activation.^20,21 Exponentially proliferating BTSCs and DIFFs cultured in their respective complete growth media were used. We found that mTORC1 activity in BTSCs was significantly lower than that found in DIFFs across all tested clones (Fig. 1E, Supplementary Fig. S2). These data suggested that BTSCs possess an intrinsic or extrinsic mechanism(s) to maintain low mTORC1 activity. Interestingly, immunofluorescence staining showed variation in Sox2 staining intensity among individual BTSCs, and we observed an inverse correlation between Sox2 expression and mTORC1 activity (Fig. 1D and Supplementary Figs. S3–S4). These data indicated the occurrence of heterogeneity within BTSCs, where mTORC1 is variably regulated.

To exclude the possibility that long-term culturing of cells changes their properties, we freshly prepared BTSCs from glioblastoma tissues by sorting for CD133-positive population, which is known to represent BTSCs (Supplementary Fig. S5).^8,11 The data showed that tumor cells that are positive for CD133 exhibited lower mTORC1 activity, supporting the view that mTORC1 activity is differentially regulated within tumor cell populations.

The Kinetics of mTORC1 Activation and Deactivation in BTSCs Were Different from Those in DIFFs

Based on the above data that mTORC1 basal activity was lower in BTSCs than DIFFs, we hypothesized that BTSCs have an innate mechanism(s) to suppress mTORC1 activation. The major nutrients that activate mTORC1 are growth factors, amino acids, and glucose.^20,21 To test whether the low activity of mTORC1 in BTSCs is achieved intrinsically or extrinsically, we starved the BTSCs and DIFFs of nutrients by replacing the complete growth medium with DMEM lacking serum, amino acids, and glucose (Fig. 2A). In this experiment, we first cultured the cells with fresh complete growth medium for 2 hours to obtain maximum mTORC1 activation, followed by a starvation test. The data showed that mTORC1 activity rapidly declined to very low levels over the first 10 to 20 minutes in BTSCs, whereas that in DIFFs remained high even 1 hour after initiating starvation (Fig. 2A, Supplementary Fig. S6A). The data implied that BTSCs might have an intrinsic mechanism to suppress mTORC1 activity. This possibility was further supported by the finding that mTORC1 reactivation following refeeding with DMEM supplemented only with epidermal growth factor (Fig. 2B, Supplementary Figs. S6B, S7) or fresh complete growth media (Supplementary Fig. S8) after long-term starvation (3 h) was much more modest in BTSCs than in DIFFs.

Fig. 2.

Different kinetics for the activation and deactivation of mTORC1 between BTSCs and DIFFs. (A) The experimental protocol is shown in upper panel. Cells cultured with fresh complete growth media for 2 hours were starved with DMEM without amino acids (AA) and glucose, followed by harvesting cell lysates at the indicated times. Western blot analysis showed that mTORC1 activity in BTSCs rapidly declined by 10–20 minutes after starvation was initiated, whereas mTORC1 was not deactivated during the test period in DIFFs. GFs, growth factors. (B) Cells (BTSCs [clone 222] and corresponding DIFFs) starved with nutrient-free DMEM were stimulated with DMEM supplemented with epidermal growth factor, followed by western blot analysis. DIFFs more rapidly reached peak mTORC1 activation than did BTSCs, and the peak activity of mTORC1 found in BTSCs was lower than that in DIFFs.

DIFF, but not BTSC, Proliferation Depended on mTORC1 Activity

To examine the biological significance of the difference in mTORC1 activity and kinetics between BTSCs and DIFFs, we tested the effect of the mTORC1-specific inhibitor rapamycin on their respective proliferation rates (Fig. 3). We first tested the self-renewal capacity of BTSCs by studying the formation of clonal spheres from dissociated single cells, which is a hallmark of stem cells (Fig. 3A). We also tested secondary sphere formation assay, where the single spheres formed from primary spheres were again dissociated into single cells and replated in the BTSC medium (Supplementary Fig. S9A). In these experiments, we also assessed the self-renewal of BT142 oligodendroglioma cells.²⁹ The data showed that rapamycin treatment did not affect the sphere-formation efficiency of all BTSC clones, which is consistent with results showing no apparent effect of rapamycin on BTSC proliferation (Fig. 3C, Supplementary Figs. S9, 10A–C). Considering that BTSCs comprise heterogeneous populations and the possibility that their proliferation depends on intercellular communications, we also examined the sphere-formation efficiency from multiple (200) dissociated cells (Fig. 3B, Supplementary Fig. S9B). Again, no effect of rapamycin on sphere formation was observed in any of the tested BTSC clones. In contrast, the proliferation of DIFFs was significantly dependent on mTORC1 activity (Fig. 3D, Supplementary Fig. S10B–D), as observed in many types of cancer cell lines.³⁰ These observations were supported by another experimental series, where we used everolimus, a synthetic derivative of rapamycin, to look at the self-renewal of BTSCs and proliferation of DIFFs (Supplementary Fig. S11). These data are consistent with our biochemical analysis (Figs. 1E, 2) and support the hypothesis that DIFFs, but not BTSCs, depend on mTORC1 activity for their proliferation.

Fig. 3.

Inhibition of mTORC1 activity affected the proliferation of DIFFs, but not the self-renewal of BTSCs. (A) Single cells of dissociated BTSCs were cultured for 10 days in 96-well plates, either with or without rapamycin, followed by quantifying the number of spheres. N.S., not significant. (B) Multiple cells (200) of dissociated BTSCs were cultured for 8 days in 96-well plates, either with or without rapamycin, followed by quantifying the number of spheres per well. (C, D) The effect of rapamycin on the self-renewal and proliferation of BTSCs (clone 222) and the corresponding DIFFs, respectively. Asterisks indicate statistically significant differences (P < .05). (E, F, G) The effect of nutrition deprivation on the self-renewal and proliferation of BTSCs (clone 222) and the corresponding DIFFs, respectively. The cells were cultured in growth media that were serially diluted (50%, 10%, and 5%) with nutrient-free phosphate-buffered saline, followed by quantification of the number of living cells. In (G), the rate of change in the proliferation of cells is shown, in which the values are expressed relative to the mean numbers of cells cultured in 100% medium. Asterisks indicate statistically significant differences (P < .05). (H) The effect of nutrient deprivation on the cell sizes of tumor cells. BTSCs (clone 222) and the corresponding DIFFs were cultured in the indicated nutrient-deprived media for 4 days, followed by measuring the diameter of cells. Asterisks indicate statistically significant differences (P < .05).

The data described above led to the hypothesis that BTSCs can survive under nutrient-poor or deprived conditions. Culture with growth medium diluted with phosphate-buffered saline free of amino acids, glucose, and serum more significantly affected the viability of DIFFs than BTSCs (Fig. 3E–G). As might be expected, a long-term (4-day) culture of BTSCs in the nutrient-poor medium resulted in decreased cell viability, while only 2 days of nutrient deprivation significantly impaired the viability of DIFFs. In addition, consistent with the role of mTOR in determining the cell size,^20,25 we found that the cell sizes of DIFFs, but not BTSCs, were impaired by nutrient deprivation (Fig. 3H). These data showed that BTSCs could survive even in nutrient-poor environments, although it is unclear at present whether BTSCs have limited requirements for nutrients or a capacity for high adaptation. We also tested the effects of a synthetic mTORC1 activator MHY1485 (ref 31) on cultured BTSCs (Supplementary Fig. S12). However, we did not observe any effect of the compound on mTORC1 activity or BTSC proliferation, which was suggestive of cell-type selectivity of this compound.

Suppression of mTORC1 Activity Affected the Differentiation of BTSCs

We next examined the effect of mTORC1 inhibition on the differentiation of BTSCs (Fig. 4). Adding rapamycin to the differentiation medium attenuated complete differentiation of BTSCs to DIFFs, as shown by the remaining expression of stem cell markers Sox2, CD133, and Nestin, as well as downregulation of β3-tubulin and GFAP after inducing differentiation (Fig. 4). No significant effects on Oct4 or Nanog expression were observed, indicating that rapamycin treatment did not fully inhibit the differentiation of BTSCs or that these markers do not perfectly cosegregate with the status of the tumor cells. No compensatory changes of other pathways that led to the activation of signal transducers and activators of transcription and extracellular signal-regulated kinase were observed by mTORC1 suppression, indicating that the inhibition of BTSC differentiation by rapamycin seems not to be mediated by cytokines or growth factors (Supplementary Fig. S13).

Fig. 4.

The effect of mTORC1 inhibition on BTSC differentiation BTSCs (clone 222) were cultured in differentiation medium (DMEM supplemented with 10% FBS) for 5 days in either the presence or absence of rapamycin. The expression of stem cell and differentiation markers was monitored by qPCR. *P < .05, **P < .01.

Low mTORC1 Activity Was Associated with the Expression of BTSC Markers in Tumor Cells in Human Glioma Tissues

We next examined the correlation between mTORC1 activity and the expression of BTSC markers by immunohistochemical staining of human glioma tissues. In preliminary experiments in which we stained for Oct4, Nanog, Sox2, and phospho-S6, we observed heterogeneity in Sox2 and phospho-S6 production (Fig. 5A). We found that Oct4, Nanog, and CD133 were poorly detectable by immunohistochemistry, which, however, might result from a low sensitivity of the antibodies used. Considering our biochemical data showing that Sox2 expression stably represented the undifferentiated state of cultured BTSCs (Fig. 1B, C), we first selected Sox2 as a marker to visualize BTSCs in tissues. Interestingly, we found some glioblastoma regions (although not in all regions), where Sox2 expression was enriched in phospho-S6-negative cells (Fig. 5A). These data suggested the possibility that Sox2 expression and mTORC1 activation are mutually exclusive, which is consistent with our immunofluorescence data from cultured BTSCs (Fig. 1D, Supplementary Figs. S3–S4).

Fig. 5.

Low mTORC1 activity was associated with the expression of BTSC markers in human glioma tissue. (A) Images from the serial sections of a glioblastoma sample that showed spatial segregation of Sox2 expression and mTORC1 activation. The region within the dotted line was enriched for Sox2-positive cells, but negative for phospho-S6, which presumably represented BTSCs in tissues. The regions in the boxed areas are magnified in the insets. Scale bar, 100 µm. (B, C) Representative double-staining images of Sox2 and phospho-S6 production in glioma tissues. In (B), Sox2 and phospho-S6 were separately visualized by DAB (brown) and AP (pink), respectively. The same results were obtained when the combination of secondary antibodies was reversed, as shown in (C). The regions in the boxed areas are magnified in the insets. Note that Sox2 (arrowheads) was positive in the nucleus, whereas the signal for phospho-S6 (arrows) was detected in the cytoplasm, which made it possible to differentiate both signals. Scale bars, 50 µm. (D) In the experiments shown in (B) and (C), Sox2 or phospho-S6-positive cells in all brain tumor cases were classified according to positivity for phospho-S6 and Sox2. An asterisk indicates statistically significant difference (P < .05). (E) The number of Sox2-positive (+) and phospho-S6-negative (−), and Sox2 (−) and phospho-S6 (+) tumor cells were counted and quantified in each type of glioma tissue. The numbers of Sox2 (−) and phospho-S6 (+) cells, which corresponded to DIFFs in vitro, but not in Sox2 (+) and phospho-S6 (−) cells, correlated with the grade of the gliomas. * P < .05, ** P < .01. (F) Representative double-staining images of CD44v9 (arrowheads) and phospho-S6 (arrows) production in glioma tissues.

We confirmed our findings by further staining glioma specimens, including diffuse astrocytoma, anaplastic astrocytoma, and glioblastoma (Fig. 5B, C). We employed double staining for Sox2 and phospho-S6 by using a serial detection system, based on the combination of diaminobenzidine (DAB, brown) and alkaline phosphatase (AP, red) as chromophores. The data showed that no apparent cross-reactivity occurred between either double-staining procedure (Fig. 5B), which was confirmed by the reverse combination of secondary antibodies (Fig. 5C). The data showed an almost complete inverse correlation between Sox2 expression and S6 phosphorylation. Consistent with the data found with cultured BTSCs, most Sox2-positive cells were negative for phospho-S6, supporting the view that BTSCs exhibit low mTORC1 activity even in tumor tissues (Fig. 5D). Staining with another BTSC marker, CD44v9, and another readout for mTORC1 activity, phospho-4E-BP1, also resulted in a low correlation between BTSC marker expression and mTORC1 activation in tumor cells (Fig. 5F, Supplementary Fig. S14). Intriguingly, the frequency of Sox2/CD44v9-negative and phospho-S6-positive tumor cell populations, but not Sox2/CD44v9-positive and phospho-S6-negative BTSC populations, correlated with the grade of glioma (Fig. 5E, Supplementary Fig. S14), suggesting that the number of BTSCs with low mTORC1 activity could not necessarily recapitulate the aggressive behavior of glioma in patients.

DIFFs, but not BTSCs, Exhibited Sensitivity to Temozolomide

Finally, to give some insights into clinical relevance of our findings, we treated the selected clones of BTSCs and DIFFs with temozolomide, a standard chemotherapeutic agent for glioma, and monitored their proliferation (Fig. 6A, B). Interestingly, BTSCs were resistant to temozolomide at any given concentrations, in contrast to DIFFs, the proliferation of which was declined by temozolomide. The data suggested a hypothesis that BTSCs with low mTORC1 activity show resistance to conventional therapies (Fig. 6C), although the conclusion cannot be drawn from this in vitro study.

Fig. 6.

DIFFs, but not BTSCs, exhibited sensitivity to temozolomide. (A, B) The effect of temozolomide (TMZ) on the viability of BTSCs (clones 222 and 316) and the corresponding DIFFs, respectively. Asterisks indicate statistically significant differences (P < .01). (C) A proposed model to explain the results of the present study. Our data on cultured tumor cells showed that BTSCs exhibited low basal mTORC1 activity, and their self-renewal was resistant to rapamycin treatment, indicating that they can survive even in a nutrient-poor condition. In contrast, non-BTSCs (DIFFs) with high mTORC1 activity are sensitive to rapamycin. Although not proven in the present study, it is plausible that the low mTORC1 activity in BTSCs contribute to their resistance to mTORC1-targeted and conventional chemotherapies.

Discussion

In this study, we showed that BTSCs exhibited low basal mTORC1 activity and different kinetics during activation and deactivation from DIFFs, in an in vitro culture system. BTSCs showed less dependency on mTORC1 activity for their proliferation than did DIFFs at least in in vitro conditions, although the significance of the data in the context of human glioma has remained unproven at present. Nonetheless, it was interesting to observe an inverse correlation between the expression of BTSC markers and mTORC1 activation in tumor cells of human glioma tissues. Although preliminary, we believe that these data shed new insight into the different metabolic states and nutrient dependencies between BTSCs and DIFFs and provide clues for the development of novel strategies that specifically target BTSCs (Fig. 6C).

One issue that needs to be addressed is that our present findings seem to contradict previous data that showed essential roles for mTORC1 in CSCs in other types of cancers.^22,23 The high activity of mTORC1 in CSCs appears to be compatible with the concept of a perivascular niche for CSCs that is rich in nutrients and growth factors that help them proliferate.⁵ In contrast, data from another study showed that the forced activation of mTORC1 in leukemic stem cells led to their aberrant differentiation and depletion, suggesting that mTORC1 activity may be kept low during the maintenance of CSCs.²⁵ Indeed, the therapeutic efficiency of mTORC1-inhibition strategies turned out to be less effective against most cancers than expected in preclinical studies, indicating that a limited number of cancer cells exhibit addiction to mTORC1 activity.^20,32 An attractive hypothesis that could explain these clinical data and our present findings is that some CSCs reside in nutrient-poor (hypoxic or necrotic) niches, where oxygen and nutrient supplies obtained by diffusion from nearby vessels are depleted by the growing cancer cells, and show low mTORC1 activity and resistance to mTORC1-targeted or conventional therapies (Fig. 6C).

Considering this hypothesis, a provocative strategy that merits further study would be to examine the effect of forced mTORC1 activation in CSCs or BTSCs to increase their proliferation and sensitize them to chemotherapeutic drugs. Supporting this notion, one study showed evidence that glioblastoma cells that underwent differentiation following exposure to all-trans retinoic acid became more sensitive to mTORC1 inhibition.³³ An additional issue in targeting brain tumors is the plasticity and reversibility of BTSCs and their differentiated progenies, which have recently been the subjects of extensive studies.^34–36 Clearly, two challenges that should be addressed in future studies are to understand how the plasticity of tumor cells can be controlled and to develop efficient combination therapies that drive BTSCs to become differentiated populations that are more susceptible to mTORC1 inhibition and conventional therapies.

At present, there is no single BTSC marker that unequivocally reflects their undifferentiated state, both in vitro and in tissues.^12–14 For example, the specificity of CD133, which is one of the most widely used BTSC markers, is unclear because it has been shown that CD133-negative tumor cells are also able to form spheres and CD133 expression was dynamically regulated depending on the microenvironments.^37–39 We also found variable changes in the expression of all the tested markers of BTSCs during FBS-induced differentiation (Fig. 1B–D, Supplementary Fig. S3) and rapamycin treatment (Fig. 4). Among those, however, Sox2 and CD44v9 expression was stably regulated between all BTSC clones and their respective DIFFs, and those were also detectable in brain tumor tissues by immunohistochemistry (Fig. 1B–D, 5). We believe that DIFFs should be targeted by conventional radiation and temozolomide treatments, and BTSCs (Sox2-positive and phospho-S6-negative cells) are an important future target. Intriguingly, our data showed that the prevalence of the BTSC population with low mTORC1 activity was comparable between high-grade glioblastoma and low-grade astrocytoma, whereas the differentiated tumor cell population that has high mTORC1 activity was predominant in glioblastoma (Fig. 5E). However, an unaddressed issue is whether this heterogeneity among tumor cells is relevant to the progression and clinical outcome of the disease.

Finally, only limited studies have addressed the mechanism by which BTSCs suppress their own mTORC1 activity. Clearly, extrinsic factors and nutrients derived from the microenvironment are crucial. In addition, given the rapid kinetics for mTORC1 deactivation in BTSCs after serum starvation (Fig. 2A), it is important to identify the intrinsic machinery that promotes low mTORC1 activity in those cells in future studies.

Supplementary material

Supplementary material is available at Neuro-Oncology online.

Funding

This work was supported by a Grant-in-Aid for Scientific Research (S) (to M.T.) commissioned by the Ministry of Education, Culture, Sports, Science and Technology of Japan. Y-P.H. was supported by the China Scholarship Council.