EPEN-05. Adenosine receptor expression in paediatric ependymoma

PURPOSE: Paediatric ependymoma is associated with dismal outcomes. Whilst understanding of its underlying biology has advanced, there has been little progress in treatment and clinical outcomes. Acting through four G protein-coupled receptors (encoded by ADORA1, ADORA2A, ADORA2B and ADORA3), adenosine is a signalling molecule often present at high levels in tumours. Adenosine signalling can aid tumour proliferation and invasiveness via mechanisms including suppression of tumour-infiltrating immune cells. Adenosine receptors therefore represent a potential therapeutic target in paediatric ependymoma, however neither levels nor patterns of expression have been previously reported. We hypothesised that adenosine receptors would be expressed in paediatric ependymoma and that this expression would vary between molecular subgroups. METHODS: Three publicly available gene expression datasets were analysed for adenosine receptor expression using Kruskal-Wallis, Mann-Whitney U and chi-square tests. RNAscope assays for adenosine receptors and CD68 were then performed on ten full-face ependymoma FFPE sections from posterior fossa A (PFA1 and PFA2) tumours to understand patterns of expression within ependymomas with the highest levels of expression identified by the gene expression datasets. RESULTS: Statistically significant differences were identified between adenosine-related genes across ependymoma subgroups of differing anatomical origin (supratentorial ZFTA-positive versus posterior fossa A and B (PFA/PFB)), with median adenosine-related gene levels generally higher in the PFA subgroup. Particularly, ADORA1, 2A, 2B and 3 gene expression was higher in PFA tumours than other subgroups. Analysis of the ten cases demonstrated measurable expression of all four adenosine receptors by RNAscope and patterns in the distribution and relative levels of expression of the adenosine receptors across PFA1 and PFA2 tumours were described. CONCLUSION: Using two different techniques we demonstrated that adenosine receptors are expressed in paediatric ependymomas. There are significant differences in level of expression between tumour subgroups. Adenosine receptors therefore represent a potential therapeutic target which should be explored further.

MODL-26. Development of humanized immune system, posterior fossa A ependymoma patient-derived xenograft model

Cellular interactions between tumor and immune cells are critical in ependymoma biology. We have shown distinct immunobiology phenotypes by ependymoma molecular subgroups, with PFA2 developing an anti-tumor immune phenotype and in contrast PFA1 tumor immune cells being pro-tumor. We recently established two fully characterized pediatric PFA1 intracranial xenograft models in NSG mice. These models, while critical for advancing PFA studies, lack the ability to make lymphocytes. To address this we have established a humanized orthotopic model of PFA1 ependymoma that are grafted to produce functional human lymphocytes. To do this, CD34+ human umbilical cord blood was injected into the cranial facial vein of newborn, irradiation immunodepleted, BRGS mice. Human immune chimerism was determined at 10 weeks by flow cytometry of peripheral blood. Next, we injected existing PDX model MAF-928_XF cells into the 4th ventricle at 12 weeks age and tumors were monitored by MRI. Initial scanning found the tumors were delayed in developing, consistent with other humanized solid tumor mouse models. Mice were euthanized between 32-34 weeks age and necropsies were performed to isolate brain, lymph nodes, blood and spleen. Human chimerism was detectable, by flow cytometry, in spleen, lymph nodes and blood, and most cells were T-cells. Human T-cells were detectable in all tumors, in proportions consistent with human disease (0.05-0.5% of CD45+). Tumors also had 35-50% mouse myeloid infiltration. Tumors were MHC I negative and MHC II positive, and PD-L1 low. Histology was consistent with human ependymoma. We are using spatial proteomics to determine cellular location and phenotype of infiltrating immune cells. Initial studies indicate this model will be usable for modeling the critical tumor-immune interactions and pre-clinically testing the next therapies in PFA ependymoma and can be easily adapted to other pediatric brain tumors in which immune factors have a critical role.

IMMU-10. TUMOR ASSOCIATED MYELOID CELLS DRIVE THE IMMUNOBIOLOGY OF HIGH RISK PEDIATRIC EPENDYMOMA

Molecular profiling of pediatric ependymoma (EPN) has previously identified discrete neoplastic subpopulations, of which Mesenchymal EPN Cells (MEC) characterize Posterior Fossa A tumors (PFA). MECs are associated with tumor immunosuppression. Here we further characterize the EPN immune environment using single-cell sequencing, spatial phenotyping and cytokine analyses to better define infiltrating myeloid subpopulations. We hypothesize that neoplastic and myeloid cells interact to propagate an immune suppressive environment conferring resistance to traditional therapies. We delineated myeloid cell subpopulations from single-cell RNA-seq of 26 pediatric EPNs and validated them through deconvolution of bulk gene expression profiling (n=299). To define subpopulation spatial distribution, we interrogated a range of tumor and myeloid markers using multiplex immunofluorescence (mIF). Finally, using single-cell cytokine analyses, we gained further insight into myeloid subpopulation function. Eight distinct myeloid subpopulations were identified, relating to macrophages, microglia and monocytes. A subpopulation of cells with wound healing ontologies and characterized by TREM1 expression, demonstrated features of myeloid derived suppressor cells, including IL6/STAT3 pathway activation. We called these hypoxia-M. Like MEC neoplastic cells, hypoxia-M was associated specifically with PFA1 subgroup EPN in both single-cell and bulk tumor gene expression profiling (p<0.001). Additionally, the presence of MEC and hypoxia-M correlated strongly in gene expression (r2=0.92, p<0.001) and IHC analyses, where they co-located to borders between necrosis, blood vessels and viable tumor. Analysis using mIF (n=54) confirmed MEC/hypoxia-M co-location and highlighted that all types of immune cell corralled in significant numbers around areas of vasculature and necrosis. Single cell cytokine analyses demonstrated that hypoxia-M secrete IL-8 which, we hypothesize, amplify the pro-tumor phenotype in PFA1 tumor microenvironment. EPN is characterized by discrete myeloid cell subpopulations which contribute to the tumor microenvironment. Treatment strategies must focus on modifying this pro-tumor, immunosuppressive microenvironment to deliver more effective treatment for childhood ependymoma.

EPEN-11. TUMOR DIFFERENTIATION IMPACTS THE BIOLOGY OF RECURRENCE IN CHILDHOOD POSTERIOR FOSSA EPENDYMOMA

Ependymoma (EPN) of childhood is curable in only 50% of cases, with recurrences in the remainder that are refractory to treatment. In recent years significant advances have been made in understanding the molecular and cellular biology of EPN. Recent studies show that PFA subgroup EPN are comprised of multiple neoplastic subpopulations that show undifferentiated, differentiated and mesenchymal characteristics. These studies focused on tumor at presentation, with recurrent EPN being less well understood. In the present longitudinal study we examine changes in neoplastic cell heterogeneity in serial presentations of PFA EPN using deconvolution (Cibersort) of bulk RNAseq data. Analysis of a cohort of 48 PFA EPN presenting at Children’s Colorado showed survival and PFA1/PFA2 subtype assignment was associated with the proportion of individual neoplastic subpopulations as determined by deconvolution. Tumors that subsequently regrew had a significantly higher estimated proportion of undifferentiated EPN cells (UEC) at presentation, than those that were non-recurrent after 5 years follow-up. This outcome association potentially age related, as UEC proportions are significantly higher in PFA arising in children &lt; 1 year old who have a particularly poor prognosis. Changes in PFA neoplastic subpopulations at recurrence was performed in two cohorts of patients from Children’s Colorado (n=23) and Nottingham, UK (n=15). As a whole, no subpopulation proportion was significantly changed at recurrence. However, separation of PFA into subtypes PFA1 and PFA2 revealed an increase in the proportion of the cilia-differentiated EPN cell subpopulation is more frequent event in PFA1 (15/24), and rare in PFA2 (2/11). Changes in other neoplastic subpopulations at recurrence were smaller and only seen in PFA1, both UEC and mesenchymal subpopulations being lower at recurrence. In summary, only PFA1 showed dynamic changes in neoplastic subpopulation proportions at recurrence, with potential impacts on transcriptomic based-subgroup assignment, whereas PFA2 proportions remained largely stable.

EPEN-08. THE TREM1 POSITIVE HYPOXIC MYELOID SUBPOPULATION IN POSTERIOR FOSSA EPENDYMOMA

We have previously shown the importance of immune factors in posterior fossa ependymoma (PF EPN). Recently, we found eight transcriptionally unique subpopulations of myeloid cells infiltrating PF EPN with one population particularly enriched in PFA1 tumors. This subpopulation, denoted as hypoxia myeloid subpopulation, is defined by genes associated with angiogenesis, hypoxia response, wound healing, cell migration, neutrophil activation, and response to oxygen levels. TREM1 (Triggering receptor expressed on myeloid cells 1) was found to be expressed almost exclusively within this hypoxia myeloid subpopulation. TREM1 encodes for a receptor belonging to the immunoglobulin superfamily that is expressed on myeloid cells, and stimulates neutrophil and monocyte inflammatory responses. However, single-cell RNAseq give little data suggesting location of cells within the tumor microenvironment. We performed immunohistochemistry (IHC) on our bank of ~90 FFPE PFA EPN samples using TREM1 to characterize and identify the location of the hypoxia myeloid cells. The TREM1 positive cells have an ambiguous cytomorphology reminiscent of a monocyte with modest cytoplasm and a mono-lobated nucleus. IHC also showed that TREM1+ myeloid cells are largely localized to the interface of necrosis and viable tissue, most frequently in a perivascular and intravascular distribution. The latter finding suggests that the TREM1+ cells are derived from the bone marrow and that they may be associated with the mesenchymal tumor population (MEC), which we have previously described as being enriched in PFA1 tumors and localizing to perinecrotic zones. This is supported by parallel IHC analysis of subpopulation-specific markers in the same cohort of PFA EPN which showed the highest TREM1 correlation was with CAIX, a marker of MEC. In PFA matched primary/recurrent pairs, the proportion of TREM1+ cells were increased at recurrence in the majority of cases, suggesting an evolving interaction between this TREM1+ hypoxia myeloid subpopulation and neoplastic cells over the disease course.

EPEN-07. SINGLE-CELL RNA SEQUENCING IDENTIFIES A UNIQUE MYELOID SUBPOPULATION ASSOCIATED WITH MESENCHYMAL TUMOR SUBPOPULATION IN POOR OUTCOME PEDIATRIC EPENDYMOMA

We have previously shown immune gene phenotype variations between posterior fossa ependymoma subgroups. PFA1 tumors chronically secrete IL-6, which induces secretion of myeloid cell IL-8 and pushes the infiltrating myeloid cells to an immune suppressive function. In contrast, PFA2 tumors have a more immune activated phenotype associated with a better prognosis. The objective of this study was to use single-cell(sc) RNAseq to descriptively characterize the infiltrating myeloid cells. We analyzed approximately 8500 cells from 21 PFA patient samples. Using advanced machine learning, we identified eight myeloid cell subpopulations with unique gene expression profiles. Interestingly, only one subpopulation was significantly enriched in PFA1 tumors. This subpopulation, denoted as the hypoxia myeloid subpopulation, was defined by genes associated with angiogenesis, response to hypoxia, wound healing, cell migration, neutrophil activation and response to oxygen levels. These myeloid cells also share similar gene expression profile to a mesenchymal tumor subpopulation (MEC) enriched in PFA1 and associated with poor outcome in EPN patients. This tumor subpopulation was the only population expressing IL-6. Using immunohistochemistry, we found the hypoxia myeloid located in regions of tumor necrosis and perivascular niches. The MEC cells were also more abundant in these regions. In an independent single-cell cytokine release assay, we identified eight subpopulations of functional myeloid cells. One subpopulation significantly secreted IL-8, which represented the hypoxia subpopulation based on IL-8 gene expression in the scRNAseq dataset. This data suggests the tumor necrosis resulting in the development of MEC tumor subpopulation is driving the immune suppressive myeloid phenotype in PFA1 tumors through polarization of myeloid cells to the hypoxia subpopulation. Further studies are needed to determine how these myeloid cells interact with the lymphocyte subpopulations and whether they contribute to the progression of PFA1 EPN.

EPEN-23. A COMPUTATIONAL ANALYSIS OF THE TUMOUR IMMUNE MICROENVIRONMENT IN PAEDIATRIC EPENDYMOMA

Ependymoma is the third commonest childhood brain tumour. Relapse is frequent, often fatal and current therapeutic strategies are inadequate. Previous ependymoma research describes an immunosuppressive environment with T-cell exhaustion, indicating a lack of response to T-cell directed immunotherapy. Understanding the immune microenvironment is therefore critical. We present a computational analysis of ependymoma, gene expression derived, immune profiles. Using 465 ependymoma samples from gene expression datasets (GSE64415, GSE50385, GSE100240) and two RNA-seq databases from UK ependymomas, we applied bulk tumour deconvolution methods (CIBERSORT and xCell) to infer immune cell populations. Additionally, we measured checkpoint blockade related mRNAs and used immunohistochemistry to investigate cell populations in ependymoma sections. CIBERSORT indicated high proportions of M2-like macrophages and smaller proportions of activated natural killer (NK) cells, T follicular helper cells, CD4+ memory T-cells and B-cells. xCell overlapped with the M2-like macrophage and CD4+ memory T-cell signatures seen in CIBERSORT. On immunohistochemistry, T and B cells were scarce, with small numbers of CD8+, CD4+ and CD20+ cells in the parenchyma but greater numbers in surrounding regions. CD68 was more highly expressed in the parenchyma. Analysis of nine checkpoint ligands and receptors demonstrated only the TIM3/GAL9 combination was reliably detectable. GAL9 is implicated in tumour interactions with T-cells and macrophages elsewhere, possibly contributing to poorer outcomes. Our study supports the presence of myeloid cells being leading contributors to the ependymoma immune microenvironment. Further work will delineate the extent of myeloid contribution to immunosuppression across molecular subtypes. Modulation of tumour immunity may contribute to better clinical outcomes.

Modified prone position to dissect the popliteal fossa

Introduction

Traditionally, the prone position is used for dissecting the popliteal fossa, which requires endotracheal intubation. Access to the airway in this position is limited, hence the complications. It is not surprising that the prone position is not favoured by the anaesthetists, especially in patients with a high body mass index. We describe a safe and novel alternative to the prone position.

Methods

The modified prone position (MPP) is described as an alternative position that facilitates access to the airway.

Results

Between October 2007 and May 2010, 12 patients underwent popliteal fossa dissection using the MPP. All patients had general anaesthesia using a laryngeal mask airway with the exception of one, who had an epidural anaesthesia. There were no airway or haemodynamic complications. The surgical access to the popliteal fossa was as good as with the traditional prone position.

Conclusions

The MPP was satisfactory for both the surgeon and the anaesthetists. The authors now use this position routinely for dissecting the popliteal fossa.

Radiological classification of optic pathway gliomas: experience of a modified functional classification system

Optic pathway gliomas (OPGs) in childhood are associated with neurofibromatosis type 1 (NF1) and since 1958 have been classified anatomically using the Dodge classification (DC). MR scanning permits a more detailed anatomical description than can be classified by this historical system. A modified Dodge classification (MDC) has been applied to MRI scans from a cohort of 72 patients (36.1% NF1-positive) from 4 centres participating in an international clinical trial. The MDC was feasible, applicable and more detailed than the original DC. NF1-positive cases more commonly involved both optic nerves (p = 0.021) and other multiple locations (p = 0.001). NF1-negative tumours more commonly involved the central chiasm (p = 0.005) and hypothalamus (p = 0.003). Fewer hypothalamus-positive tumours were associated with optic nerve involvement (p = 0.009), whereas more were associated with central chiasm involvement (p<0.001). From diagnosis to follow-up, there was concordance between DC and MDC in 51/72 cases (70.8%). The MDC is therefore proposed for use in clinical trials of new treatments for OPGs.