T-cell TGF-β signaling abrogation restricts medulloblastoma progression

Significance

Medulloblastoma (MB) is a tumor of the cerebellum that primarily forms in pediatric patients during brain development. The immune system ultimately fails to eradicate MB because it is “blind” to tumor cells as a result of poor brain immune surveillance caused by the existence of the blood–brain barrier and the brain’s immune privileged status. Another mechanism of tumor escape is immune suppressors that act as a “smokescreen,” blocking effective antitumor immunity. We show that blockade of the TGF-β signaling pathway promotes memory T cell development, conferring antitumor immunity to the smoothened A1 mouse model of MB. Our data lay the cellular immune mechanistic framework for blocking T cell TGF-β signaling in pediatric brain cancer.

Abstract

Cancer cell secretion of TGF-β is a potent mechanism for immune evasion. However, little is known about how central nervous system tumors guard against immune eradication. We sought to determine the impact of T-cell TGF-β signaling blockade on progression of medulloblastoma (MB), the most common pediatric brain tumor. Genetic abrogation of T-cell TGF-β signaling mitigated tumor progression in the smoothened A1 (SmoA1) transgenic MB mouse. T regulatory cells were nearly abolished and antitumor immunity was mediated by CD8 cytotoxic T lymphocytes. To define the CD8 T-cell subpopulation responsible, primed CD8 T cells were adoptively transferred into tumor-bearing immunocompromised SmoA1 recipients. This led to generation of CD8⁺/killer cell lectin-like receptor G1 high (KLRG1^hi)/IL-7R^lo short-lived effector cells that expressed granzyme B at the tumor. These results identify a cellular immune mechanism whereby TGF-β signaling blockade licenses the T-cell repertoire to kill pediatric brain tumor cells.

CNS tumors are the leading cause of death among pediatric cancer patients, and medulloblastoma (MB) is the most common pediatric CNS malignancy (1). Following surgical resection, these patients require aggressive multimodal therapy, resulting in long-term treatment-related complications, including developmental, neurological, neuroendocrine, and psychosocial deficits (1–3). Variants of MB have been defined (e.g., classic, desmoplastic, anaplastic, and extremely nodular) pathologically and by gene expression profiling (4–7). Approximately one third of all MBs are characterized by up-regulation of genes in the Sonic hedgehog (Shh) pathway and are predominantly desmoplastic (5, 7, 8). Constitutive activation of the Shh pathway typically results from mutations in the Shh receptor [Patched (Ptc)] or downstream Smoothened (Smo), leading to hyperproliferation of granule neuron precursors (GNPs) and tumorigenesis (5, 8, 9). Smoothened A1 (SmoA1) transgenic mice harbor a Smo activating mutation under regulatory control of the CNS specific neuronal differentiation 2 (Neurod2) promoter, which drives expression in GNPs (10–12). These animals have constitutively active Shh signaling in the cerebellum, resulting in highly penetrant MB (10).

Although the CNS was once thought to be excluded from immune surveillance, CNS neoplasms are now known to affect local and systemic immune responses (13, 14). Patients with brain tumors often present with T-cell lymphopenia (15–17), an immunodeficiency condition that promotes tumor growth and recurrence (18). Despite the presence of tumor infiltrating lymphocytes (TILs), the immune response to CNS tumors is severely impaired (15, 17). This phenomenon is at least partly a result of tumor secretion of immunosuppressive cytokines such as TGF-β (16, 17, 19), a pleiotropic cytokine that dampens responses of innate and adaptive immune cells (20, 21). In particular, TGF-β can block differentiation of naïve T cells into T helper (T_h) CD4, and CD8 cytotoxic T lymphocytes (CTLs) (22–25). Because they are chiefly responsible for antitumor immunity, evasion of CTLs is critical for tumor survival.

TGF-β is also involved in the generation and function of T regulatory cells (T_regs) (26, 27), which suppress physiological and pathological immune responses and prevent autoimmune disease (28–30). Interestingly, previous studies have shown greater immunosuppressive activity of T_regs associated with CNS tumors (31–33). The transcription factor Foxp3 plays a crucial role in immune homeostasis and T_reg development (34–36), as demonstrated by the dramatic autoimmune phenotype of the Foxp3 forkhead domain-deficient Scurfy mouse (37), and autoimmune disease in humans bearing Foxp3 mutations known as immunodysregulation polyendocrinopathy enteropathy X-linked syndrome. Foxp3 expression itself is regulated by cooperative activation of multiple transcription factors by cytokine receptors, including the TGF-β receptor (26, 38–41). Specifically, TGF-β receptor binding causes the transcription factors Smad3 and nuclear factor of activated T cells (NFAT) to promote Foxp3 expression through histone acetylation at one of two known enhancer elements (26). Therefore, oversecretion of TGF-β by CNS tumors can potentially undermine effective antitumor immunity by stimulating Foxp3⁺ T_regs (42, 43).

To determine the role of TGF-β T-cell immunosuppression in MB progression, we used transgenic mice with blockade of T-cell TGF-β signaling. These mice express a dominant-negative (DN) form of TGF-β receptor type II under CD4 promoter regulatory control (TGF-βRII-DN). This DN retains the extracellular and transmembrane domains, but has a truncated intracellular kinase domain, thereby allowing it to outcompete endogenous receptors for TGF-β ligands (23). Because this construct lacks the CD8 silencer, the transgene is expressed by CD4 and CD8 T cells (44). We crossbred TGF-βRII-DN and SmoA1 animals, and demonstrate that blockade of T-cell TGF-β signaling promotes CD8 T-cell differentiation into killer cell lectin-like receptor G1 high (KLRG1^hi)/IL-7R^lo short-lived effector cells (SLECs), which militate against MB progression. Furthermore, CD8 T cells derived from tumor-bearing TGF-βRII-DN⁺/SmoA1⁺ mice are efficient tumor cell serial killers in vitro and in vivo. In this scenario, CTL antitumor immunity can be propagated from one animal to another when adoptively transferred into tumor-bearing Rag2^−/−γc^−/−/SmoA1⁺ hosts. These results show how TGF-β signaling affects adaptive immunity in the fight against pediatric brain cancer.

Results

T_regs Infiltrate TGF-β–Producing Human MBs.

TGF-β ligand is initially synthesized as a precursor protein that undergoes proteolysis. The mature, dimeric form of TGF-β is composed of two ∼12.5-kDa monomers stabilized by hydrophobic interactions and disulfide bonds, forming a protein of 25 kDa that initiates intracellular signaling (45). We Western blotted for TGF-β in MB samples, control pediatric cerebral cortex, and glioma tissue homogenates. Precursor TGF-β protein was detected in all nine MB samples assayed, and there was approximately eightfold greater abundance of this protein vs. control cortical and glioma tissue by densitometry (Fig. 1A). Interestingly, MB samples contained similar monomeric TGF-β abundance as control glioma tissue, which was not detected in control cerebral cortex (Fig. 1A). The mature, dimeric form of TGF-β was found in five of nine MBs, but was not detected in cortical or glioma tissue (Fig. 1A). There was no relationship between TGF-β1 abundance and MB subtype.

Fig. 1.

TGF-β pathway and T_regs in human MBs. (A) TGF-β1 protein expression in human MB, glioma, and normal cerebral cortex samples shows precursor, dimeric, and monomeric forms of TGF-β1. Actin is shown as a loading control. (B) A maximum projection confocal image of a malignant MB depicts multiple CD4⁺TGF-βRII⁺ T cells (arrowheads). (C) Single confocal optical sections of a malignant MB shows colocalization of TGF-β1 and TGF-βRII on CD4⁺ T cells (Upper) and lack of nuclear Zic1 expression in a CD4⁺ T cell (Lower). (Scale bars: 10 μm.) (D) Flow cytometry analysis of freshly isolated human MB cells shows Foxp3⁺CD25⁺ T_regs. (E) Quantification of seven freshly isolated primary pediatric MB samples demonstrates that ∼5–10% of CD25⁺ T cells express Foxp3.

Next, we investigated whether human MB-infiltrating CD4 T cells expressed TGF-β signaling components. Immunohistochemistry (IHC) and confocal imaging were carried out on freshly isolated human malignant MB tissue biopsies using CD4 and TGF-βRII antibodies. CD4 T cells expressed TGF-βRII at the cell surface (Fig. 1B), and coexpression of TGF-β1 ligand and TGF-βRII was frequently observed (Fig. 1C). CD4⁺TGF-βRII⁺ T cells lacked nuclear expression of the MB/GNP-enriched transcription factor Zic1 (46) (Fig. 1D), indicating that CD4⁺TGF-βRII⁺ cells were not CNS-derived. Collectively, these results suggest that human MB-infiltrating T cells express canonical TGF-β signaling machinery. To ascertain whether T_regs (defined by Foxp3 and CD25 positivity) were present in human MBs, flow cytometric analyses were carried out on single cell suspensions generated from seven primary human MB samples. Interestingly, Foxp3⁺ cells represented ∼5–10% of CD25⁺ T cells in human MBs (Fig. 1 D and E).

T_regs Are Present in Shh-Driven Mouse MB.

Shh pathway mutations are thought to drive MB (4, 8, 9, 47). The Shh ligand undergoes autocatalytic processing and modification that is critical for signaling activity. Specifically, the 45-kDa precursor protein is cleaved to yield a 19-kDa N-terminal signaling domain (Shh-N) and a 25-kDa C-terminal domain (48). By using an Shh-N antibody, Shh precursor protein was detected in 10 of 11 human MBs and the N-terminal signaling domain was found in 9 of 11 samples (Fig. S1A). Shh-N binds to a multicomponent receptor complex containing Ptc (a 12-pass transmembrane protein) and Smo (a seven-transmembrane G protein-coupled receptor). Binding of Shh to Ptc releases basal repression of Smo; therefore, Smo activity is widely used as a surrogate for Shh pathway activation (49–51). We detected the higher molecular weight, activated form of Smo in all human MB samples (Fig. S1A). Interestingly, control pediatric cortex tissue contained Shh precursor, yet lacked detectable Shh-N protein (Fig. S1). Moreover, whereas Shh-N was observed in control human glioma tissue, activated Smo was not (Fig. S1), likely because Shh does not drive forebrain oncogenesis (52). Significantly higher active-to-inactive Smo ratios were detected in each of the desmoplastic MB samples (Fig. S1), the form of human MB most often associated with Shh pathway oncogenesis (4, 8). Collectively, these results indicate that Shh pathway activity is elevated in multiple human MB variants.

Active Smo and Shh signaling in human MB samples prompted investigation into whether SmoA1⁺ mouse MBs contained T_regs. We hypothesized that this might be the case because tumor-bearing SmoA1 mice had elevated dimeric, mature TGF-β1 compared with tumor-free mice of the same genotype (Fig. 2A). To determine this, we bred SmoA1 animals with mice coexpressing Foxp3 and EGFP under endogenous regulatory control (Foxp3^EGFP mice) (53, 54). This mouse strain has normal Foxp3 and T_reg function (53), and was used to detect T_regs during the primary immune response to MB. As Foxp3 is located on the X chromosome, random X-inactivation results in fewer circulating EGFP⁺CD4⁺ T cells of heterozygous females vs. hemizygous males. Therefore, males were separately analyzed from females. Confocal microscopy of Foxp3^EGFP/SmoA1⁺ MBs revealed 80–140 Foxp3^EGFP+CD4⁺CD25⁺ cells per sagittal brain section in tumor-bearing 10-mo-old Foxp3^EGFP/SmoA1⁺ males and females (Fig. 2 B and C). Importantly, these populations were absent in brains of control male and female Foxp3^EGFP/SmoA1⁺ mice lacking tumors or WT Foxp3-expressing SmoA1 (Foxp3^wt/SmoA1⁺) mice. TGF-βRII and TGF-β1 ligand were also found on EGFP⁺ cells in tumor-bearing Foxp3^EGFP/SmoA1⁺ animals (Fig. 2D).

Fig. 2.

SmoA1 mouse MBs contain mature TGF-β1 and T_regs. (A) Western blot comparing TGF-β1 levels between control (tumor-free) and tumor-bearing SmoA1 cerebellar lysates. Protein levels were quantified by densitometry and revealed significantly higher TGFβ1:Actin ratios in tumor-bearing SmoA1 mice vs. tumor-free controls. (B) Confocal imaging reveals EGFP⁺CD4⁺CD25⁺ T_regs in a Foxp3^EGFP/SmoA1⁺ mouse MB. (C) Quantitation of MB-associated T_regs from male and female tumor-bearing and tumor-free mice. Exhaustive counts of total EGFP⁺CD4⁺CD25⁺ T_regs per hindbrain were averaged for three sections per animal. (D) Confocal images show EGFP⁺ cells coexpressing TGF-βRII (Upper) and TGF-β1 (Lower). (Scale bars: 20 μm.)

To assess whether tumor malignancy impacted peripheral T_regs, we immunostained spleens from Foxp3^EGFP/SmoA1⁺ mice with antibodies against EGFP, CD4 and CD25. Compared with control animals, tumor-bearing Foxp3^EGFP/SmoA1⁺ mice had increased numbers of splenic CD4⁺CD25⁺Foxp3^EGFP+ cells that predominantly localized to germinal centers (Fig. S2A). These data were verified by flow cytometric analysis of T-cell receptor (TCR)α/β gated splenocytes; and there was ∼2.5-fold increased splenic T_regs in both male and female Foxp3^EGFP/SmoA1⁺ animals (Fig. S2 B and C).

T-Cell TGF-β Signaling Blockade Mitigates MB Progression in SmoA1 Mice.

To assess the impact of abrogating T-cell TGF-β signaling on MB progression, we bred TGF-βRII-DN⁺ with SmoA1⁺ mice and performed survival analyses. Animals were followed for 10 mo, and TGF-βRII-DN⁺/SmoA1⁺ mice survived ∼40% better than SmoA1⁺ littermates (Fig. 3A). Advanced SmoA1⁺ MBs have phenotypic changes including enlarged posterior fossa, tilted head, and hunched posture. These symptoms—and tumor size—were all reduced in tumor-bearing TGF-βRII-DN⁺/SmoA1⁺ animals (Fig. 3B). Additionally, SmoA1⁺ mice had smaller spleens vs. WT littermate controls, and this effect was reversed in TGF-βRII-DN⁺/SmoA1⁺ animals (Fig. 3C). TGF-βRII-DN⁺/SmoA1⁺ mice also had markedly improved motor activity, including hind limb function and balance as assayed by pole and wire tests, and lower measures of ataxia (by splay and wobble scores) compared with SmoA1⁺ mice (Fig. 3D).

Fig. 3.

The TGF-βRII-DN transgene restricts MB progression. (A) Survival analysis of SmoA1⁺ and TGF-βRII-DN⁺/SmoA1⁺ littermates is shown. (B) SmoA1⁺ mice with MB have an enlarged posterior fossa, tilted head, and hunched posture. All these symptoms (Upper), and tumor area (by H&E histologic analysis; Lower), are mitigated in TGF-βRII-DN⁺/SmoA1⁺ animals. Black outlines indicate tumor areas. (C) Hyposplenia in SmoA1⁺ tumor-bearing animals is mitigated in TGF-βRII-DN⁺/SmoA1⁺ animals; a WT spleen is shown as a control. (D) TGF-βRII-DN⁺/SmoA1⁺ mice have improved motor activity by pole and wire tests and normal splay and wobble scores compared with SmoA1⁺ mice (*P < 0.05 and ***P < 0.001).

The TGF-βRII-DN Transgene Diminishes MB Infiltrating T_regs.

To assess whether tumor-infiltrating T_regs were impacted by T-cell TGF-β abrogation, we crossed Foxp3^EGFP reporter animals with TGF-βRII-DN⁺/SmoA1⁺ mice and examined tumor sections from sex-matched 10-mo-old Foxp3^EGFP/SmoA1⁺ and Foxp3^EGFP/TGF-βRII-DN⁺/SmoA1⁺ littermates. Interestingly, EGFP⁺CD4⁺ T_regs were morphologically distinct: those associated with Foxp3^EGFP/SmoA1⁺ tumors had an activated phenotype consisting of multiple elaborate processes in comparison with the naïve T_reg morphology in Foxp3^EGFP/TGF-βRII-DN⁺/SmoA1⁺ tumors (Fig. 4A, Insets). Importantly, T_reg numbers were significantly decreased in Foxp3^EGFP/TGF-βRII-DN⁺/SmoA1⁺ vs. Foxp3^EGFP/SmoA1⁺ MBs (Fig. 4 A and B). These IHC results were further evaluated by flow cytometric analysis of hindbrain leukocyte T_regs, revealing ∼50% decreased proportion of CD4⁺EGFP⁺CD25⁺ cells in sex-matched animals (Fig. 4C). Further, EGFP⁺CD4⁺ T cells expressed the proliferative marker Ki67 in Foxp3^EGFP/SmoA1⁺ tumors, but not in Foxp3^EGFP/TGF-βRII-DN⁺/SmoA1⁺ tumors (Fig. 4D). We also found increased proportions of splenocytes in TGF-βRII-DN⁺/SmoA1⁺ mice that expressed TCRα/β at early (5 mo of age; Fig. 5A) and late stages (10 mo of age) of tumor growth (Fig. S3). Furthermore, splenocytes from TGF-βRII-DN⁺/SmoA1⁺ mice at both ages had increased CD8 and CD4 T-cell effector memory (T_EM) and central memory (T_CM) populations and reduced naïve T cells compared with appropriate controls (Fig. 5 B and C and Fig. S3). Similar results were observed in thymocytes from the same animals (Fig. S4). Further confirmation of enhanced T-cell memory in TGF-βRII-DN⁺/SmoA1⁺ mouse spleens came from increased expression of CD44, T-bet, and STAT3 phosphorylation by Western blot (Fig. 5D).

Fig. 4.

Reduced MB-associated T_regs in Foxp3^EGFP/TGF-βRII-DN⁺/SmoA1⁺ mice. (A) Confocal image stacks show reduced EGFP⁺CD4⁺ T_regs in Zic1⁺ MBs from Foxp3^EGFP/TGF-βRII-DN⁺/SmoA1⁺ vs. Foxp3^EGFP/SmoA1⁺ mice. (Scale bars: 50 μm.) (Insets) Only EGFP and CD4 signals to demonstrate T_reg phenotypic differences. (Scale bars: 10 μm.) (B) Quantification of hindbrain EGFP⁺CD4⁺ T_regs (averages of three sections per animal) in female and male Foxp3^EGFP/TGF-βRII-DN⁺/SmoA1⁺ and Foxp3^EGFP/SmoA1⁺ mice is shown. (C) Flow cytometry results are shown of pooled hindbrain leukocytes from three male mice per genotype. Cells were gated on CD4 expression, and EGFP⁺CD25⁺ T_regs are reduced in Foxp3^EGFP/TGF-βRII-DN⁺/SmoA1⁺ hindbrains. Numbers indicate proportions of total CD25⁺ populations. (D) Confocal image stacks demonstrate proliferative EGFP⁺CD4⁺ T_regs in a Foxp3^EGFP/SmoA1⁺ MB. Single optical section insets are shown to demonstrate nuclear Ki67 expression in CD4⁺EGFP⁺ cells in a Foxp3^EGFP/SmoA1⁺ MB. Proliferative EGFP⁺CD4⁺ T_regs were not detected in Foxp3^EGFP/TGF-βRII-DN⁺/SmoA1⁺ MBs.

Fig. 5.

T_EM and T_CM splenocytes are increased in TGF-βRII-DN⁺/SmoA1⁺ mice. (A) Elevated numbers of TCRα/β-expressing splenocytes from tumor-bearing 5-mo-old TGF-βRII-DN⁺/SmoA1⁺ vs. SmoA1⁺ animals are shown. (B) TCRα/β-expressing splenocytes gated by CD8 and (C) CD4 expression have increased percentages of T_EM and T_CM cells. (D) Western blots of spleen homogenates from tumor-bearing TGF-βRII-DN⁺/SmoA1⁺ mice show increased CD44 and T-bet expression, as well as increased phosphorylated STAT3 (pSTAT3)-to-STAT3 ratio. Actin is shown as a loading control. Data are averages of six mice per genotype for each experiment (*P < 0.05 and ***P < 0.001).

TGF-β Signaling Blockade Promotes CD8 T-Cell Differentiation.

We next determined whether TIL populations differed between TGF-βRII-DN⁺/SmoA1⁺ and SmoA1⁺ MBs. To start, IHC for the pan-leukocyte marker CD45 and the MB protein Zic1 revealed copious tumor infiltrating leukocytes in TGF-βRII-DN⁺/SmoA1⁺ vs. SmoA1⁺ mice (Fig. 6A). To further characterize these immune cells, isolated hindbrain leukocytes from TGF-βRII-DN⁺/SmoA1⁺ and SmoA1⁺ animals were used as input for flow cytometric analysis of TCRα/β and CD45. These markers revealed three distinct populations of cells: TCRα/β^loCD45^lo, TCRα/β^hiCD45^hi, and TCRα/β^negCD45^hi (Fig. 6B). Hindbrains from TGF-βRII-DN⁺/SmoA1⁺ mice contained significantly higher numbers of TCRα/β^hiCD45^hi and TCRα/β^negCD45^hi TILs, and fewer TCRα/β^loCD45^lo TILs than SmoA1⁺ mice (Fig. 6C). Increased expression of TCRα/β and CD45 by TGF-βRII-DN⁺/SmoA1⁺ TILs suggests greater activation of the T-cell repertoire. To examine this further, we gated CD45⁺ TILs from TGF-βRII-DN⁺/SmoA1⁺ and SmoA1⁺ hindbrains and observed a CD8⁺CD44^hiCD62L^lo population of TILs in TGF-βRII-DN⁺/SmoA1⁺ MBs (Fig. 6D). CD8 T-cell subsets were next analyzed for expression of KLRG1 and IL-7 receptor α-chain (IL-7Rα) to assess whether T-cell TGF-β blockade influenced CD8 T-cell differentiation into SLECs (KLRG1^hiIL-7Rα^lo) or memory precursor effector cells (MPECs; KLRG1^loIL-7Rα^hi). Significantly increased proportions of KLRG1^hiIL-7Rα^lo and KLRG1^loL7Rα^hi CD8 cells were observed in TGF-βRII-DN⁺/SmoA1⁺ TILs (Fig. 6E) and splenocytes (Fig. S5). These cells also expressed higher amounts of the CTL markers IFN-γ and CD45 (Fig. 6E). Together, these results indicate that blocking T-cell TGF-β signaling during MB malignancy promotes CD8 T-cell differentiation into SLECs and MPECs.

Fig. 6.

TGF-β signaling blockade promotes CD8⁺ T-cell differentiation. (A) Confocal image stacks depict increased CD45⁺ leukocyte infiltration in Zic1⁺ MBs from TGF-βRII-DN⁺/SmoA1⁺ vs. SmoA1⁺ animals. (Scale bars: 100 μm.) (B) Flow cytometry analysis and (C) quantification of TCRα/β and CD45-expressing hindbrain cells demonstrates increased numbers of TCRα/β^hiCD45^hi and TCRα/β^negCD45^hi leukocytes in hindbrains of TGF-βRII-DN⁺/SmoA1⁺ vs. SmoA1⁺ mice. (D) Expression of CD44 and CD62L (density plots) among CD4 and CD8 T cells (dot plots) reveals CD8⁺CD44^hiCD62L^lo cells among CD45-gated leukocytes in TGF-βRII-DN⁺/SmoA1⁺ hindbrains (***P < 0.001). (E) Hindbrain TCRα/β-gated CD8 T cells from TGF-βRII-DN⁺/SmoA1⁺ mice differentiate into KLRG1^hiIL-7Rα^lo and KLRG1^loIL-7Rα^hi populations and express IFN-γ and CD45. Leukocytes were pooled from three tumor-bearing mice per genotype (*P < 0.05 and ***P < 0.001).

CD8⁺ CTLs from TGF-βRII-DN⁺/SmoA1⁺ Mice Are Better Tumor Serial Killers.

Knowing whether CD8 T cells are differentiating into CTLs is critical to understanding antitumor immunity in TGF-βRII-DN⁺/SmoA1⁺ mice. To assess CTL activity, purified CD8 T-cell populations were isolated from tumor-bearing TGF-βRII-DN⁺/SmoA1⁺ mice and control TGF-βRII-DN⁺/SmoA1⁺ mice lacking tumors (Fig. 7A). We then established a cell line from SmoA1⁺ mouse MB. These cells express Zic1 and neural stem cell markers Sox2 and Nestin under neural stem cell growth conditions (Fig. S6A). Additionally, these cells are capable of differentiating into neurons that express neuronal nuclei (NeuN) antigen or neuron-specific class III β-tubulin (Tuj1) and also into glial fibrillary acidic protein (Gfap)-positive astrocytes (Fig. S6 B and C). This cell line was used to measure granzyme B (Gran B)-mediated CD8 T-cell killing. In this assay (55), SmoA1⁺ tumor cells uptake target fluorescent label (TFL) before coincubation with CD8 T cells from tumor-bearing TGF-βRII-DN⁺/SmoA1⁺ animals or appropriate controls. Before this, purified CD8⁺ T cells were primed with a fluorescent Gran B substrate. When a CTL binds to its target, Gran B is discharged by exocytosis and cleaves caspase precursors, thereby causing target cell self-destruction by apoptosis. Therefore, cells that are TFL⁺Gran B⁺ are considered to be tumor cells killed by CTLs (55). Flow cytometry data revealed that CD8 T cells from tumor-bearing TGF-βRII-DN⁺/SmoA1⁺ mice were more efficient serial killers vs. appropriate controls (Fig. 7 B and C). Strikingly, live confocal microscopy and 3D surface rendering showed active CD8 T-cell killing of MB target cells (Fig. 7D and Movie S1).

Fig. 7.

Increased MB cell killing by TGF-βRII-DN⁺/SmoA1⁺ CTLs. (A) Flow cytometry analysis of 1 × 10⁴ MACS-purified CD8 T cells shows ∼98% purity from SmoA1⁺ and TGF-βRII-DN⁺/SmoA1⁺ mouse splenocytes. (B) Flow cytometry analysis and (C) quantification of Gran B-mediated T-cell killing of Tg^SmoA1 MB cells shows enhanced cytotoxicity by TGF-βRII-DN⁺/SmoA1⁺ CD8 T cells. Percent killed values for WT and TGF-βRII-DN⁺ CD8 T cells are shown as controls; numbers indicate percentages of total target fluorescent label (TFL⁺) cells. Experiments were performed in triplicate, and error bars are SDs (***P < 0.001). (D) Confocal imaging and 3D surface rendering depicts a live TGF-βRII-DN⁺/SmoA1⁺ CD8 T-cell killing a TFL⁺ Tg^SmoA1 MB cell. (Scale bars: 10 μm.)

CD8 T Cell Antitumor Immunity Can Be Propagated in Vivo.

Our next goal was to address whether the TGF-βRII-DN transgene alone was sufficient to endorse CTL activity in mouse MBs, or whether previous exposure to MB was additionally required. To achieve this, CD8 T cells isolated from tumor-bearing TGF-βRII-DN⁺/SmoA1⁺ and control tumor-free TGF-βRII-DN⁺/SmoA1⁺ spleens were adoptively transferred into immunocompromised tumor-bearing Rag2^−/−γ_c^−/−/SmoA1⁺ recipients. Flow cytometric analysis revealed that 10.8% of CD8 T cells expressed KLRG1 from tumor-bearing TGF-βRII-DN⁺/SmoA1⁺ vs. only 1.2% from tumor-free TGF-βRII-DN⁺/SmoA1⁺ donors. After adoptive transfer, flow cytometric analysis on host TILs revealed marked infiltration of CD8⁺KLRG1⁺ T cells from tumor-bearing TGF-βRII-DN⁺/SmoA1⁺ donors into MB targets. By contrast, virtually none of these cells were detected in MBs of Rag2^−/−γ_c^−/−/SmoA1⁺ recipients (Fig. 8A). To measure CTL activity, we performed confocal microscopy on brain sections immunostained with antibodies against CD8, Gran B, and Zic1. These data revealed enhanced CTL activity in TGF-βRII-DN⁺ T cells isolated from tumor-bearing TGF-βRII-DN⁺/SmoA1⁺ hosts as readout qualitatively and quantitatively by Gran B immunoreactivity surrounding donor CD8 T cells in Zic1⁺ tumors (Fig. 8 B and C). These results demonstrate that anti-MB immunity can be propagated by primed TGF-β signaling-inhibited CD8 T cells.

Fig. 8.

Propagation of CD8 T-cell immunity into Rag2^−/−γ_c^−/−/SmoA1⁺ hosts. (A) Flow cytometry analysis of purified donor CD8 T cells shows increased expression of KLRG1 in TGF-βRII-DN⁺/SmoA1⁺ mouse splenocytes isolated from a tumor-bearing animal (Upper). Flow cytometry analysis of host Rag2^−/−γ_c^−/−/SmoA1⁺ TILs shows increased infiltration of CD8⁺KLRG1⁺ T cells from tumor-bearing TGF-βRII-DN⁺/SmoA1⁺ vs. control tumor-free TGF-βRII-DN⁺/SmoA1⁺ donor cells. (B) Confocal imaging of Zic1⁺ tumor areas in Rag2^−/−γ_c^−/−/SmoA1⁺ hosts shows enhanced CTL activity of T cells from tumor-bearing TGF-βRII-DN⁺/SmoA1⁺ donors as read-out by Gran B immunoreactivity of donor CD8 T cells. (Scale bars: 100 μm.) (C) Quantification of confocal images of Gran B immunoreactivity in tumor areas in Rag2^−/−γ_c^−/−/SmoA1⁺ hosts shows increased Gran B immunoreactivity (IR) in tumors of mice injected with T cells from tumor-bearing TGF-βRII-DN⁺/SmoA1⁺ vs. control tumor-free TGF-βRII-DN⁺/SmoA1⁺ donor cells (*P < 0.05).

Discussion

In this report, we show that blockade of T-cell TGF-β signaling directs expansion and activation of CTLs against MB. Specifically, blockade of this signaling pathway during MB malignancy promotes memory T-cell formation, conferring antitumor immunity. A previous study demonstrated that TGF-β functions in vivo to limit CD8 T-cell expansion and activation after high-affinity peptide–MHC interactions (56). Here, we have shown that human and mouse MBs abundantly produce TGF-β, making these tumors well-suited to evade CTL immune surveillance. Indeed, our results indicate that TGF-β–mediated immunosuppression endorses MB malignancy. Interestingly, it has been shown that TGF-β signaling controls CD8 effector T cell numbers by lowering Bcl-2 abundance and by selectively promoting apoptosis of SLECs (57). These results dovetail with ours; collectively suggesting that TGF-β–mediated apoptosis of T-cell subpopulations likely regulates clonal expansion and contraction during immune responses to CNS malignancies.

Presence of the TGF-βRII-DN transgene drastically reduced Foxp3⁺CD25⁺ T_regs in SmoA1⁺ MB mice. In tumor-bearing Foxp3^EGFP/SmoA1⁺ mice, these cells were negative for the naturally occurring T_reg marker neuropilin-1 (58) (Fig. S7), suggesting that they were induced T_regs (iT_regs). Multiple recent reports have elucidated the molecular mechanisms leading to TGF-β-mediated Foxp3 induction in iT_regs in a variety of disease settings (41, 59, 60). By using a Foxp3 reporter, our data support that TGF-β is required for T_reg homing to mouse MBs. These T_regs critically rely on Foxp3 expression and likely function to suppress antitumor T effector function. The regulatory regions for Foxp3 consist of a promoter, two enhancers and a conserved noncoding sequence region. At the enhancer 1 region, Smad3 and NFAT coordinately induce Foxp3 (26), leading to iT_reg differentiation via signals through TCR and TGF-β receptors (26, 40, 41). Elevated STAT3 activity in TGF-βRII-DN⁺/SmoA1⁺ animals may also explain reduced T_reg numbers, as Foxp3 enhancer 2 activity is negatively regulated by STAT3. Given changes in T_reg and CD8 T-cell populations, it is likely that disruption of TGF-β signaling serves a dual role in promoting the development of T effector responses by (i) limiting iT_reg activity and (ii) promoting expansion and activation of CD8 T cells. This paradigm is summarized in Fig. S8.

Although it is important to recognize that mouse models do not always predict human disease biology, data presented in this report provide a mechanistic framework for a pharmacotherapeutic strategy aimed at blocking T-cell TGF-β signaling in pediatric brain cancer. Specifically, we have shown that human MBs contain high abundance of TGF-β1 and T cells expressing TGF-βRII. A likely possibility is that MBs secrete TGF-β1 to evade T-cell–mediated antitumor immunity. The results herein demonstrate the role that TGF-β plays to dampen CTL responses to mouse MB. This TGF-β–dependent balance between T_regs and CTLs is paramount when considering immunotherapeutic approaches against these devastating childhood tumors.

Materials and Methods

Mice.

C57BL/6-Tg(Neurod2-SmoA1) mice (10) were provided by J. M. Olson (Fred Hutchinson Cancer Research Center, Seattle, WA). TGF-βRII-DN mice (23) were provided by R. A. Flavell (Yale University, New Haven, CT). B6.Cg-Foxp3^tm2(EGFP)Tch/J mice (53) and WT C57BL/6 mice were obtained from the Jackson Laboratory. All mice were housed in a 12-h light/dark cycle, and the Cedars-Sinai Medical Center Institutional Animal Care and Use Committee approved all experiments, which were conducted in accordance with institutional and National Institutes of Health guidelines.

Tissue Handling.

Adult mice (5 or 10 mo of age) were killed with isoflurane and perfused intracardially with ice-cold PBS solution. Brains were dissected down the midline and fixed for 24 h in 4% (wt/vol) paraformaldehyde (PFA), rinsed, and embedded in 2% (wt/vol) agarose. Tissue was sectioned in the sagittal plane on a Leica VT1200S Vibratome at 70-μm thickness and applied to Superfrost Plus Gold slides (Fisher Scientific). Fresh human samples were immediately frozen on dry ice and stored at −80 °C for protein isolation or separately fixed for 24 h in 4% (wt/vol) PFA, rinsed, and embedded in 2% (wt/vol) agarose for IHC.

IHC and Histology.

For IHC, coverslips or sections were incubated in protein block (PBS solution containing 10% (vol/vol) FBS and 0.3% Triton X-100) for 1 h at 25 °C. Antibodies to the following proteins were variously used: rat anti-human CD4 (1:500; Thermo Scientific), goat anti–TGF-βRII (1:500; R&D Systems), rabbit anti-human TGF-β1 (1:1,000; Torrey Pines Biolabs), rabbit anti-human/mouse Zic-1 (1:500; Rockland), chicken anti-GFP (1:5,000; Abcam), mouse anti-mouse CD4 (1:200; Thermo Scientific), rat anti-mouse CD25 (1:200; eBioscience), rat anti-mouse CD45 (AbD Serotec), chicken anti-mouse Nestin (1:2,000; LifeSpan Biosciences), goat anti-mouse Sox2 (Santa Cruz), mouse anti-mouse NeuN (1:1,000; Millipore), mouse anti-mouse Tuj1 (1:500; Covance), and rabbit anti-mouse Gfap (1:2,000; Dako). Antibodies were diluted in protein block and incubated overnight at 4 °C. After three rinses for 5 min each in PBS solution, samples were incubated for 1 h at 25 °C with appropriate Alexa Fluor-conjugated secondary antibodies. After an additional three rinses for 5 min each with PBS solution at 25 °C, samples were air-dried in the dark and mounted with Prolong Gold media containing DAPI (Invitrogen–Molecular Probes). Fluorophores were imaged in separate channels with a Nikon A1R confocal microscope mounted onto a Ti-E inverted system equipped with a 40×/1.3 oil immersion Apochromat objective. For bright-field histology, nuclei were stained with alum hematoxylin for 4 min, rinsed in tap water, reacted with 0.3% acid alcohol, rinsed, and then stained with eosin for 2 min, rinsed, and mounted. Hindbrains were imaged at 20× using a Leica SCN 400 slide scanner.

MB Subtype Classification.

Human MB samples were independently analyzed by the Department of Pathology at Cedars-Sinai Medical Center. Freshly isolated tumors were fixed overnight in formalin and then embedded in paraffin blocks. Information regarding MB subtypes was taken from pathology reports under Cedars-Sinai Medical Center Institutional Review Board guidelines.

Flow Cytometry.

Single-cell suspensions of tumor cells, splenocytes, and thymocytes were rinsed twice with PBS solution, centrifuged, and then resuspended in flow cytometry buffer containing PBS solution with 5% (vol/vol) FBS and 0.5% saponin. Cells were incubated at 4 °C for 30 min, rinsed twice with PBS solution, and filtered through flow cytometry tubes (BD Biosciences). The following antibodies were variously used: GFP-FITC (1:1,000; Abcam), mouse IL-7Rα-PerCP (1:200; BioLegend), mouse KLRG1-APC (1:200; BioLegend), mouse CD8-PE-Cy7 (1:500; BD Pharmingen), mouse CD25-PerCP (1:200; Biolegend), mouse CD45-APC-Cy7 (1:200; BD Biosciences), mouse TCRα/β-FITC (1:200; AbD Serotec), mouse CD44-PE-Cy7 (1:200; BD Biosciences), mouse CD62L-Pacific Blue (1:200; Invitrogen), mouse CD4-Pacific Orange (1:500; Invitrogen), human CD4-PE (1:200; BD Pharmingen), and human CD25-V450 (1:500; BD Biosciences). Foxp3 staining was performed using a Foxp3 staining kit (Imgenex) with antibodies to mouse Foxp3 (APC-conjugated; 1:200) and human Foxp3 (Alexa Fluor 647-conjugated; 1:200; BD Pharmingen) according to the manufacturers’ instructions. Flow cytometric analysis was conducted with a CyAn instrument (Dako), and data were analyzed using FlowJo.

Biochemical Analysis.

Human tissue or mouse spleen samples was homogenized using a T10 basic Homogenizer (IKA Works). All samples were lysed in cell lysis buffer (Cell Signaling) containing 1 mM PMSF and subjected to Western immunoblot. Protein concentration was determined by using the bicinchoninic acid method, and an aliquot corresponding to 50 μg of protein was electrophoretically separated on 12% (wt/vol) Nu-PAGE polyacrylamide gels (Invitrogen). Proteins were electrophoretically transferred to Immobilon-P polyvinylidene difluoride membranes (Millipore). Membranes were blocked in blocking buffer [5% (wt/vol) nonfat dry milk in Tris-buffered saline solution containing 1% Tween-20] for 3 h at ambient temperature and incubated overnight at 4 °C with primary antibodies directed against TGF-β1 (1:1,000; Torrey Pines Biolabs), Shh (1:500; R&D Systems), Smo (1:200; Lifespan Biosciences), CD44 (1:500; Cell Signaling), T-bet (1:500; BD Pharmingen), STAT3 (1:500; Cell Signaling), phosphorylated STAT3 (1:500; Cell Signaling), and Actin (1:1,000; Millipore). Membranes were then rinsed three times for 5 min each in dH₂O, and incubated with appropriate secondary antibodies conjugated with horseradish peroxidase (1:2,000; GE Healthcare–Amersham Biosciences) in blocking buffer. After an additional three rinses for 5 min each in dH₂O, membranes were incubated for 1 min at ambient temperature with the enhanced chemiluminescence substrate (Thermo Fisher Scientific), exposed to film, and developed.

Pole Test.

The pole test was conducted according to previously established protocols (61). Briefly, mice were individually placed head-upward on the top of a vertical rough-surface pole (diameter, 8 mm; height, 55 cm). The time until the animal descended to the floor was recorded as locomotor activity time, with a maximum duration of 120 s. If an animal descended partly down the pole and fell the rest of the way, the behavior was scored until it reached the floor. When the mouse was not able to turn downward and instead dropped from the pole, a default locomotor activity time value was taken as 120 s. Each mouse was given three trials with a 30-s recovery period between trials.

Wire Test.

The wire test was performed as described previously (62). Briefly, animals were handled by the tail and allowed to grasp the middle of a 1.5-mm-thick, 60-cm-long metallic wire with their forelimbs. Animals were then gently lowered so that their hind paws grasped the wire. The mice were then gently turned upside-down along the axis of the wire. Upon release of the animal, a timer was started. The average time until the mouse completely released its grasp and fell down was recorded; animals unable to grasp the wire or incapable of hanging were assigned the maximum time (120 s). Each mouse was tested three times with a 30-s recovery period between tests.

Observer-Rated Ataxia Measures.

Mice were evaluated and scored for ataxia during a 45-s test according to an established protocol (63). To assess interrater reliability of the splay and wobble rating scales, mice were rated by one person and also independently rated by a separate observer. The rating scales in Table 1 were used.

Table 1.

Rating scales for splayed hind limbs and animal wobbling

Score	Description	Associated activity
	Splaying
0	Standing; no splaying	Moves freely
1	Brief splaying; single incidents within an episode of movement	Movement slightly impaired
2	Recurrent splaying during one or more episodes of movement	Able to move in straight line
3	Frequent splaying	Uncoordinated movements
4	Persistently splayed	Unable to move effectively
5	Completely splayed	No deliberate movement
	Wobbling	—
0	No wobbling	—
1	Wobbles; no loss of balance	—
2	Wobbles; occasional loss of balance	—
3	Wobbles with repeated falls	—
4	Falls; unable to move	—

“Splaying” indicates hind limbs unable to support body; “wobbling” indicates an erratic gate.

Generation of a Tg^SmoA1 Tumor Cell Line.

C57BL/6-Tg(Neurod2-SmoA1) MB cells were generated as previously reported (64). Single cells and small cellular aggregates were plated in fresh neural stem cell culture medium consisting of Neurobasal media (Invitrogen) containing 20 ng/mL of EGF (Sigma), 20 ng/mL of FGF-2 (R&D Systems), and 5 µg/mL of heparin (Sigma). Cells were then plated onto 75-cm² flasks precoated with CELLstart (Invitrogen). Cells were seeded at 2.5 × 10⁵ cells per well on glass coverslips in 24-well plates. Differentiation was achieved by culturing cells in DMEM supplemented with 10% (vol/vol) FBS and 1% penicillin–streptomycin. Cells were then washed twice with PBS solution and fixed for 12 min with 4% (wt/vol) PFA at 4 °C then rinsed twice with PBS solution, air-dried in the dark, and mounted with Prolong Gold media containing DAPI (Invitrogen–Molecular Probes).

In Vitro Cytotoxicity Assay.

Gran B-based killing assays were performed as previously described (55) using a GranToxiLux PLUS kit (OncoImmune). Tumor cells (1 × 10⁵) SmoA1 were incubated with magnetic activated cell sorting-purified CD8 T cells for 1 h at increasing effector-to-target ratios. Percent killing was calculated for the SmoA1 target cells by the following equation: [(number of TFL⁺Gran B⁺ cells)/(total number of TFL⁺ cells)] × 100%.

Adoptive Transfer of TGF-βRII-DN⁺ CD8 T Cells into Rag2^−/−γ_c^−/−/SmoA1⁺ Hosts.

Tumor-bearing Rag2^−/−γ_c^−/−/SmoA1⁺ host animals (aged 3 mo) received tail vein injections of 2.5 × 10⁵ CD8 T cells isolated from tumor-bearing or tumor-free TGF-βRII-DN⁺/SmoA1⁺ donor mice. Animals were killed 36 h later and one hemibrain was processed for flow cytometry and pooled (n = 5 mice per group). The other hemibrain was processed for IHC. Tumor-bearing animals were selected based on enlarged craniums and hunched posture and were pathologically verified to be tumor-bearing postmortem.

Statistical Analysis.

All experiments were performed by an examiner blinded to sample identities, and code was not broken until analyses were completed. Data are presented as the mean ± 1 SEM. In instances of single mean comparisons, t test for independent samples was performed. In instances of multiple mean comparisons, one-way ANOVA was used, followed by post hoc comparison of the means by Student t test. P values of less than 0.05 were considered to be statistically significant.