Co-expression modules of NF1, PTEN and sprouty enable distinction of adult diffuse gliomas according to pathway activities of receptor tyrosine kinases

Abstract

Inter-individual variability causing elevated signaling of receptor tyrosine kinases (RTK) may have hampered the efficacy of targeted therapies. We developed a molecular signature for clustering adult diffuse gliomas based on the extent of RTK pathway activities. Glioma gene modules co-expressed with NF1 (NF1-M), Sprouty (SPRY-M) and PTEN (PTEN-M) were identified, their signatures enabled robust clustering of adult diffuse gliomas of WHO grades II-IV from five independent data sets into two subtypes with distinct activities of RAS-RAF-MEK-MAPK cascade and PI3K-AKT pathway (named RMPAhigh and RMPAlow subtypes) in a morphology-independent manner. The RMPAhigh gliomas were associated with poor prognosis compared to the RMPAlow gliomas. The RMPAhigh and RMPAlow glioma subtypes harbored unique sets of genomic alterations in the RTK signaling-related genes. The RMPAhigh gliomas were enriched in immature vessel cells and tumor associated macrophages, and both cell types expressed high levels of pro-angiogenic RTKs including MET, VEGFR1, KDR, EPHB4 and NRP1. In gliomas with major genomic lesions unrelated to RTK pathway, high RMPA signature was associated with short survival. Thus, the RMPA signatures capture RTK activities in both glioma cells and glioma microenvironment, and RTK signaling in the glioma microenvironment contributes to glioma progression.

Keywords: glioma; molecular classification; receptor tyrosine kinase.

Figure 1. RMPA^high and RMPA^low gliomas were associated with distinct survival and ages at diagnosis

Based on the signatures of SPRY-M, NF1-M and PTEN-M, 575 adult diffuse gliomas of WHO grades II-IV from the CGGA and TCGA mRNA-seq data sets were clustered using NMF. Distribution of morphologically diagnosed glioma subgroups among the RMPA glioma subtypes was analyzed using the Pearson χ2 test. Patients with RMPA^low gliomas showed better survival and younger ages at diagnosis compared to patients with RMPA^high gliomas. A II, O II, OA II, A III, O III, OA III and GBM are the abbreviations of morphologically diagnosed astrocytoma grade II, oligodendroglioma grade II, oligoastrocytoma grade II, astrocytoma grade III, oligodendroglioma grade III, oligoastrocytoma grade III and glioblastoma, respectively. Samples with no information of morphological diagnosis from the source data are indicated with NULL.

Figure 2. Distinct overall activities in RAS-RAF-MEK-MAPK cascade and PI3K-AKT pathway in RMPA glioma subtypes

Formalin-fixed and paraffin-embedded RMPA^high (n = 10) and RMPA^low (n = 9) glioma samples of the CGGA mRNA-seq database were sectioned and stained for p-ERK and p-AKT. Representative staining results of p-ERK and p-AKT (at a magnification of x40), the staining results summarized as immunoreactive score (IRS), the age at diagnosis and survival of the corresponding patients are shown.

Figure 3. Distinct chromosomal alterations in RMPA^high and RMPA^low gliomas

The SNP6.0 data from 334 gliomas in the TCGA mRNA-seq data set and the 50k HindIII SNP array data from 205 gliomas in the Rembrandt data set were analyzed using GISTIC2.0 at an amplitude threshold of ± 0.2. The arm-level chromosomal gaines or losses in the RMPA^high and RMPA^low gliomas are depicted. In the TCGA mRNA-seq dataset, ~80% of the RMPA^high gliomas showed amplification of Chr 7 accompanied with loss of chr 10, and ~50% of the RMPA^low gliomas showed co-deletion of chr 1p and chr 19q. A similar trend was observed in the Rembrandt data set.

Figure 4. Enriched expression of angiogenic RTKs in vessel endothelial cells and infiltrating immune cells in RMPA^high gliomas

Single cells from RMPA^high or RMPA^low gliomas were co-stained with APC-conjugated anti-CD45 or anti-CD105 mAbs, in combination with one of the PE-conjugated anti-RTK mAbs, or with isotype-matched control antibodies. Dot plots of the bottom rows were the results of co-staining of APC-conjugated anti-CD45 and anti-CD105 mAbs together with one of the indicated PE-conjugated anti-RTK mAb. Living cells excluding 7-AAD staining were gated and analyzed for the co-expression of CD45, CD105 and RTKs. Dot-plots of representative RMPA^high A. or RMPA^low glioma sample B. are presented. The expression of EGFR and PLXNB2 was seen in CD45⁻CD105⁻ cells in both RMPA^high or RMPA^low gliomas. The expression of MET, VEGFR1, KDR, EPHB4 and NRP1 was observed in both CD45⁺ and CD105⁺ cells, but not in CD45⁻CD105⁻ glioma cells. Results of isotype control and other control stainings are depicted in Supplementary Figure S10.

Figure 5. Activation of RTK signaling in vessel endothelial cells and TAMs in RMPA^high gliomas

Sections of the classified RMPA^high or RMPA^low gliomas were co-stained with mAbs against CD31 or CD68 in combination with mAbs towards p-AKT or p-ERK. The staining of anti-CD31 or anti-CD68 was detected with Alexa Fluor^® 555 conjugated secondary antibody, and anti-p-AKT or anti-p-ERK with Alexa Fluor^® 647 conjugated secondary antibody. Sections were further stainined with DAPI and evaluated using a confocal microscope (ZEISS LSM 700). Images of representative RMPA^high A. or RMPA^low. B. gliomas are shown. The control stainings were performed in the absence of the primary antibody. Staining results of other RMPA classified or non-classified samples are summarized in Supplementary Figure S11.

Figure 6. Poor prognosis of PM gliomas with high RMPA signature

Unsupervised hierarchical clustering between RMPA classifiers and PM gliomas was performed on all data sets analyzed. The data on copy number alterations of EGFR and PTEN, co-deletion of 1p and 19q, IDH1 mutation, and the age at the diagnosis are presented. In four indicated data sets, PM gliomas with a high RMPA signature showed poor overall survival compared with PM gliomas with a low RMPA signature.