Introduction
No effective therapy currently exists for glioblastoma (GBM), which is the most common primary brain tumour and one of the most aggressive types of cancers. Challenges in tackling these tumours are manifold, including their inter‐ and intratumour heterogeneity, the limited accessibility of systemically administered drugs, their infiltrative growth pattern and the complexity of the microenvironment in which they are embedded (Aldape
et al,
2019).
The contribution of the tumour microenvironment (TME), which is shaped by the communication between tumour cells and non‐malignant cells, is undisputed in GBM pathogenesis (Quail & Joyce,
2017). Particular emphasis has been placed on immune infiltrates, including tumour‐associated microglia/macrophages (TAM), which are the most numerous infiltrating cell population in GBM (Szulzewsky
et al,
2015; Chen
et al,
2017; Darmanis
et al,
2017; Roesch
et al,
2018). These cells engage in a bidirectional interaction with tumour cells to promote several aspects of glioma development, including proliferation, angiogenesis, immune evasion and therapeutic resistance (Hambardzumyan
et al,
2016; Chen & Hambardzumyan,
2018). TAM in GBM are pro‐tumourigenic, with increased accumulation in high‐grade gliomas that correlates with poor prognosis (Komohara
et al,
2008; Hambardzumyan
et al,
2016; Sorensen
et al,
2018). Moreover, TAM produce low levels of pro‐inflammatory cytokines and lack key molecular mechanisms necessary for T‐cell stimulation, suggesting a suppression of T‐cell activation capacity in GBM (Hussain
et al,
2007; Quail & Joyce,
2017). TAM can be classified into tumour‐associated microglia (TAM‐MG), endogenous to central nervous system (CNS) tissue and tumour‐associated macrophages (TAM‐BMDM), originating from bone marrow‐derived monocytes that infiltrate the tumour from the periphery (Muller
et al,
2015; Bowman
et al,
2016; Haage
et al,
2019). The functional contribution of TAM to GBM pathogenesis is well documented; however, it is unclear how each of these two ontogenetically distinct populations differentially contribute to the GBM phenotype.
Efforts are being invested in therapeutically depleting immune cells from the TME as well as altering cytotoxic potential with immunomodulation (Seoane,
2016). Targeting chemokines and their receptors such as the CCR2/CCL2 axes has been explored as a therapeutic strategy to inhibit infiltration of TAM (Ruffell & Coussens,
2015; Vakilian
et al,
2017). For the re‐education of TAM immune activity, inhibition of colony‐stimulating factor 1 receptor (CSF1R) has shown promising result in preclinical GBM models by blocking tumour growth and progression (Pyonteck
et al,
2013; Yan
et al,
2017). However, acquired resistance and tumour relapse emerge following long‐term exposure to these therapies (Quail
et al,
2016). To design successful re‐education strategies targeting TAM, a better characterisation of their signalling mechanisms is essential.
The mTOR pathway has been extensively studied in the context of cell growth, proliferation and survival in many cancers, including GBM (Li
et al,
2016; Jhanwar‐Uniyal
et al,
2019). The central component of the pathway, the mTOR protein kinase, forms the catalytic subunit of the protein complexes known as mTOR complex 1 (mTORC1) and mTOR complex 2 (mTORC2), which regulate different branches of the mTOR network (Shimobayashi & Hall,
2014; Yuan & Guan,
2016). mTORC1 signalling integrates inputs from inflammatory and growth factors as well as amino acids, energy status, oxygen levels and cellular stress pathways. The two major substrates of mTORC1 are p70 ribosomal protein S6 Kinase 1 (p70S6K1) and the eukaryotic translation initiation factor (eIF)‐binding protein 1 (4EBP1; LoRusso,
2016). These signalling molecules impact on cell growth and metabolism, in part by increasing the biosynthesis of the cellular translational apparatus (Thoreen
et al,
2012). The small GTPase Ras homolog enriched in brain (Rheb) is the only known direct activator of mTORC1. Conversely, the signalling pathways that lead to mTORC2 activation are not characterised in such detail. mTORC2 is known to regulate cell cycle entry, cell survival and actin cytoskeleton polarisation through its most common downstream substrates: AKT, SGK and PKC (Yang
et al,
2013). Despite their biochemical and functional differences, crosstalk has been reported between the two complexes, which contributes to the modulation of their activity (Xie & Proud,
2014). In GBM, altered mTORC1 signalling activity correlates with increased tumour grade and is associated with poor prognosis (Duzgun
et al,
2016). Consequently, mTOR kinase inhibitors targeting both mTORC1 and mTORC2 are considered promising anti‐cancer therapies and are being tested in clinical trials, in combination with radiation and chemotherapy (Zhao
et al,
2017; Mecca
et al,
2018).
In the last decade, extensive work has been carried out to characterise mTOR‐dependent signalling in innate immune cells and its role in regulating the expression of inflammatory factors, antigen presentation, phagocytic activity, cell migration and proliferation (Weichhart
et al,
2008; Jones and Pearce,
2017). mTOR signalling is known to regulate the balance between pro‐ and anti‐inflammatory responses and may be responsible for the dysregulated inflammatory response in TAM, which display a shift towards anti‐inflammatory activity. Interestingly, increased mTOR phosphorylation at Ser‐2448 is present in nearly 40% of TAM in human GBM (Lisi
et al,
2019); however, the functional impact of this mTOR deregulation and its molecular mechanism have never been characterised.
Here, we have used GBM orthotopic allografts in genetically engineered mice in which mTORC1 signalling has been silenced in TAM‐MG, as well as human expanded‐potential stem cells (EPSC)‐derived microglial‐like cells and matched GBM cells to study the role of the mTOR pathway in TAM‐MG in the GBM microenvironment.
Discussion
We show that GIC induce mTOR signalling in TAM‐MG but not TAM‐BMDM in in vivo and in vitro mouse models of GBM as well as in a human GIC/iMGL in vitro assay. The mTOR‐dependent regulation of STAT3 and NF‐κB activity promotes an immunosuppressed phenotype in TAM‐MG, which hinders effector T‐cell proliferation and immune reactivity and contributes to tumour immune evasion.
We describe increased mTOR signalling in TAM‐MG but not in TAM‐BMDM in mouse models of GBM. mTOR signalling positively correlates with TAM‐MG enrichment in human GBM samples but not with TAM‐BMDM at the transcriptomic level, supporting the translational relevance of our findings in mouse models. Our data imply that ontogeny affects the way microglia and monocyte‐derived macrophages respond to GIC‐secreted factors, which extends previous work showing that the transcriptomic profiles of TAM‐MG and TAM‐BMDM differ when exposed to the same tumour microenvironment (Bowman
et al,
2016; Muller
et al,
2017). Recent reports have identified mixed transcriptional states in the TAM population in glioma patients (Szulzewsky
et al,
2015,
2016; Gabrusiewicz
et al,
2016), and here, we identify mTOR signalling as a key driver of this intratumoral TAM heterogeneity.
We demonstrate that the pro‐tumourigenic role of TAM‐MG in GBM is mediated by mTOR, as reduced tumour growth and increased survival were observed upon genetic silencing of the pathway in these cells in GL261 allografts. Although a contribution to the observed phenotype by the small proportion of TAM‐BMDM also targeted by the Cx3cr1‐Cre driver cannot be entirely excluded, it seems unlikely as both the transcriptomic analysis in the in vivo model and the signalling analysis in the in vitro model highlight the lack of significant mTOR activity in TAM‐BMDM. The transcriptomic profile of Cx3cr1‐Rheb1Δ/Δ tumours demonstrated a shift in the immune landscape with an overall decrease in the negative regulation of T cells in the TME and a change in T‐cell state from an exhausted to an active profile, suggesting a capacity to mount an anti‐tumour adaptive immune response. This is further demonstrated by a change in the immune composition of Cx3cr1‐Rheb1Δ/Δ tumours, defined by reduced numbers of microglia, while immune cells that have infiltrated from the peripheral circulation are more numerous, including effector T cells and TAM‐BMDM. Transcriptomic analysis of Rheb1Δ/Δ GL261 TAM‐MG revealed a re‐education of these cells to an immune reactive and anti‐tumour profile, with an enrichment for pathways linked to the regulation of Th1, Th2 and IFN signalling as well as pathways linked to recruitment, proliferation and priming of APC and cytokine signalling pathways. The predicted impact of this transcriptional deregulation is an increase in the stimulation of the adaptive immune system by innate immune cells and consequently an increase in effector and cytotoxic T cells within the tumour; an effect which was confirmed by tissue and flow cytometry analyses, which revealed an increase in CD4+ and CD8+ T cells, and no significant changes in FoxP3+ cells in the Cx3cr1‐Rheb1Δ/Δ TME as well as an increase in infiltration, proliferation and effector function of CD4+ and CD8+ T cells.
Glioblastoma multiforme are lymphocyte depleted with a high infiltration of TAM (Mirzaei
et al,
2017; Thorsson
et al,
2018; Woroniecka
et al,
2018). Within the tumour‐infiltrating lymphocytes, Tregs are the most numerous population and can suppress T helper cell and CTL responses (El Andaloussi and Lesniak,
2006; Mirzaei
et al, 2017), while CD8
+ and CD4
+ T cells are exhausted (Thorsson
et al, 2018; Woroniecka
et al, 2018). Moreover, CD8
+ and CD4
+ T cells which do infiltrate the tumour seem unable to mount an anti‐tumour effector response in GBM (Learn
et al,
2006). T‐cell exhaustion is known to result from an excessive and continuous stimulation by APC and cytokines, resulting in sustained expression of inhibitory receptors and the lack of a productive anti‐tumour effector response (Mirzaei
et al, 2017). TAM contribute to T‐cell dysfunction in GBM via their immunosuppressed phenotype, characterised by reduced expression of pro‐inflammatory factors, antigen‐presenting machinery and T‐cell activation factors (Poon
et al,
2017). In our
Cx3cr1‐Rheb1Δ/Δ model, T‐cell effector profiles were stimulated, as shown by the increased proliferation and expression of cytotoxic factors such as IFNγ, granzyme b and perforin, in keeping with a scenario where mTOR significantly contributes to TAM‐mediated T‐cell dysfunction in GBM. Moreover, a distinct deregulation of cytokine signalling pathways was identified in our
Cx3cr1‐Rheb1Δ/Δ model, most notably those regulated by STAT3 (anti‐inflammatory cytokines) and NF‐κB (pro‐inflammatory cytokines). The survival benefits of STAT3 inhibition have been shown in a GL261 model, where the expression of cytokines promoting tumour growth, such as IL‐10 and IL‐6, was blocked (Zhang
et al,
2009). Further work by Hussain
et al (
2007) illustrates the potential effect of these STAT3‐regulated cytokines, expressed by TAM, on the proliferation of effector T cells and TCR‐mediated signalling. Importantly, STAT3 is upregulated in TAM in human GBM and considered an attractive therapeutic candidate (Heimberger and Sampson,
2011; Wei, Gabrusiewicz and Heimberger,
2013; Chang
et al,
2017; Poon
et al, 2017). We show here that mTOR signalling increases STAT3 activity and inhibits NF‐κB in TAM‐MG in different GBM models, therefore hampering APC immune reactivity as well as effector T‐cell proliferation and immune response via the expression of anti‐inflammatory cytokines.
In our study, we have taken advantage of methodologies to derive induced microglia (iMGL) from EPSC (Muffat
et al,
2016) to develop a new
in vitro assay. GIC were established from human GBM, and hGIC‐CM obtained therefrom were incubated with the syngeneic iMGL to assess the relevance of the results of our mouse models in humans on a patient‐specific basis. mTOR signalling positively correlated with TAM‐MG enrichment but not with TAM‐BMDM at transcriptomic level in the TCGA samples, a finding that was most prevalent in mesenchymal tumours, and thus, we applied the assay to hGIC/iMGL derived from GBM classified as belonging to the mesenchymal subtype. We reason that as no significant differences in the levels of TAM‐MG were previously observed across the molecular subgroups (Bowman
et al,
2016), the strong correlation we observed in mesenchymal GBM was not due to a higher number of TAM‐MG in this subgroup. We showed that hGIC derived from mesenchymal GBM triggered activation of mTOR signalling in iMGL and mTOR‐dependent regulation of STAT3 and inhibition of NF‐κB signalling, in line with our findings in mouse models. Interestingly, a key characteristic of the mesenchymal subgroup of GBM is its strong association with immune‐related genes and an enrichment of infiltrating immune cells (Chen & Hambardzumyan,
2018; Behnan
et al,
2019), thereby raising the possibility that mesenchymal‐specific features of the hGIC phenotype might be responsible for inducing mTOR signalling in TAM‐MG.
We demonstrate that GIC‐secreted factors are sufficient to increase mTOR activity in microglia, although this does not exclude the possibility that factors secreted by other cells, including non‐GIC tumour cells, may contribute to this phenotype. The secretome of GIC has been characterised (Formolo
et al,
2011; Polisetty
et al,
2011), but only little is known on the functional impact of specific secreted factors on TAM phenotypes. We show here that GIC‐CM contains factors capable of inducing mTOR pathway activation in TAM‐MG in both humans and mice. A study comparing conditioned media of GIC and healthy NSC identified several inflammatory and growth factors, including potential mTOR stimuli (Okawa
et al,
2017). While it is likely that a combination of these factors is responsible for the phenotype, osteopontin and lactate emerge as strong candidates (Lamour
et al,
2010; Okawa
et al,
2017). Osteopontin acts via several integrins known to influence PI3K/AKT/mTOR signalling (Ahmed & Kundu,
2010). It regulates migration, phagocytosis and the expression of inflammatory factors in microglia (Yu
et al,
2017), including in a GBM context (Ellert‐Miklaszewska
et al,
2016; Wei
et al,
2019). Osteopontin expression correlates with poorer survival in GBM (Atai
et al,
2011; Wei
et al,
2019), and its expression is enriched in mesenchymal as compared to classical and pro‐neural tumours (Wei
et al,
2019). Strikingly, remarkably similar findings to those seen in our mouse model were described in the TME of a GL261 allograft GBM model upon depletion of osteopontin (Wei
et al,
2019). Lactate was also shown to be highly expressed by GIC and has been proposed as a prognostic marker for GBM (Marchiq & Pouyssegur,
2016). It contributes to acidification of the TME, which polarises TAM (Colegio
et al,
2014; Romero‐Garcia
et al,
2016; Mu
et al,
2018) therefore promoting immune evasion and tumour growth (Lui & Davis,
2018), possibly via interaction of lactate with the GPR65 receptor on TAM (Lailler
et al,
2019). A study examining the effect of lactate and hypoxia on macrophages demonstrated an increase in mTOR signalling, which is suggested to be responsible for acquired M2‐like phenotype with the inhibition of pro‐inflammatory cytokines production (Zhao
et al,
2019). It is therefore conceivable that differences in the threshold of lactate‐ and/or osteopontin‐dependent mTOR activation in TAM‐MG and TAM‐BMDM may explain the different phenotype and function of these two populations in the TME.
Despite the importance of the deregulation of mTOR signalling in driving GBM growth, drugs aimed at targeting this pathway have so far failed in clinical trials (Jhanwar‐Uniyal
et al,
2019). Our results raise the possibility that tumour cells should not be the primary target of mTOR inhibition. Infiltration of CD8
+ and CD4
+ T cells but not FoxP3
+ Treg cells, as observed in our
Cx3cr1‐Rheb1Δ/Δ tumours, correlates with long‐term survival in GBM patients (Heimberger
et al,
2008; Yang
et al,
2010; Abedalthagafi
et al,
2018), hence providing the rationale for immunotherapies aimed at modifying the infiltration or immune reactivity of the T‐cell population, such as drugs targeting inhibitory checkpoints. However, immune checkpoint inhibitors such as anti‐CTLA‐4 and anti‐PD‐1 antibodies have had little success as monotherapies in the treatment of GBM (Chen & Hambardzumyan,
2018), suggesting that blockade of immune checkpoints alone is not sufficient to restore anti‐tumour immune functions in the GBM TME. Our observation that increased CD8
+ and CD4
+ tumour‐infiltrating lymphocytes, induced by mTOR inhibition in Cx3cr1
+ TAM, correlates with reduced tumour growth supports further exploration of this approach and raises the possibility that precision targeting of the mTOR pathway, for example by nanoparticle‐based drug delivery, the efficacy of which have already been demonstrated in liver and breast cancer (Huang
et al,
2012; Singh
et al,
2017), could be a viable approach in combination with existing T cell‐targeted immunotherapies to condition the TME towards a pro‐inflammatory state, which is potently anti‐tumourigenic.