A microglia-specific gene cluster discriminates MBM

Microglia are a unique population of antigen-presenting cells in the central nervous system (CNS) that are capable of clearing the brain of microbes, dead cells and protein aggregates [59]. Besides, microglia play a crucial role during injury repair and display an exceptional role in immune surveillance and tumor clearance [9, 13]. Although the role of tumor-associated microglia and macrophages (TAMs) in primary brain tumors such as glioblastoma [3, 8, 68, 69] has been intensively studied, their role in the progression of brain metastases remains poorly understood.

We performed immunohistochemistry (IHC) of our MBM cohort (Supplementary Table 1 and [49]) to determine the levels of activated TAMs expressing Iba1 (AIF1). Although Iba1 serves as a well-established marker, reactive microglia cannot be distinguished from brain-infiltrated macrophages [34]. We observed that Iba1 levels classified MBM into highly and lowly TAM infiltrated tumors (Fig. 1a, Supplementary Fig. 1a). Moreover, we observed overlapping patterns of infiltration of Iba1high TAMs and CD3+ T cells (Fig. 1b). As CD3 only provided information about levels of T cell infiltration, we used the ESTIMATE algorithm [75] to gain insight into the overall degree of immune cell infiltration of MBM. In line with our previous observation, tumors with intensive TAM and T cell infiltration exhibited a high immune score (Pts 3, 4, 10, 12) whereas MBM with low levels of Iba1high/ CD3+ cell infiltration (Pts 1, 2) or low expression of Iba1 (Supplementary Fig. 1b, c) showed low immune scores (Fig. 1c). As expected, brain metastases derived cell lines (BMCs) with absence of immune cells featured lowest scores (Supplementary Fig. 1b, c). The brain has long been considered a sanctuary where tumor cells can grow undisturbed and protected from attack by immune cells. We observed a high correlation between Iba1 and the immune score by analyzing the expression levels of Iba1 in an independent dataset of brain (MBM) and extracranial metastases (EM) (Fig. 1d), suggesting a relationship between the degree of infiltration of TAMs and immune cells not only in the brain.

Fig. 1

Transcriptome and methylome profiling of Iba1high and Iba1neg MBM revealed the identification of subset-specific genes. a Immunohistochemistry (IHC) for Iba1 (red) of MBM of indicated patients. b Representative IHC for levels of CD3 in Iba1high (Pat 4) and Iba1low/neg (Pat 1) MBM. c Immune score of MBM (study EGAS00001005976, n = 16) indicating different immunologic (color coded) subsets of tumors. d Dot plot showing the significant correlation of Iba1/AIF1 expression and immune score of brain metastases (BM, R = 0.86, p < 2.2e-16) and extracranial metastases (EM, R = 0.78, p = 5.5e-13). e Survival analysis of patients with MBM (study, EGAS00001003672), featuring high or low level of Iba1/AIF1 expression revealed no significant difference (p = 0.11). f Survival analysis of TCGA melanoma patients (n = 459), featuring a high or low level of Iba1/AIF1 expression revealed a significant difference (logrank p = 1.3e-07) and Cox-regression analysis showed association with favorable disease course (HR = 0.46). g Schematic representation of candidate identification by methylome and transcriptome profiling of n = 16 MBM of study. Methylome (850 k) profiling of Iba1high (n = 5) or Iba1low/neg (n = 2) identified 416 differentially methylated regions (DMRs), within the 5´-UTR of 316 corresponding genes of which 294 were expressed in MBM with 56 genes (77 DMRs), significantly (p ≤ 0.05) discriminating Iba1high and Iba1low/neg MBM. h Heat map representation of 77 DMRs (left panel) and top expressed (right panel) genes (n = 31). Analysis identified a panel of 12 genes that clustered with expression of microglia/TAM-associated genes AIF1, SYK and HCK. i Correlation analysis of cluster genes with association to immune/TAM regulated processes, the strength of the correlation is color coded. j Comparative t-SNE representation of brain cell subclasses microglia, neurons and oligodendrocytes (left) and expression of APBB1IP (Amyloid Beta Precursor Protein Binding Family B Member 1 Interacting Protein), expression level (log2 RPKM) is color coded. k Dot plot showing the significant correlation of APBB1IP expression and immune score of brain metastases (BM, R = 0.86, p < 2.2e-16) and extracranial metastases (EM, R = 0.92, p < 2.2e-16). Significance was determined by unpaired, two-sided t-test (d, g, k)

As high levels of immune (T) cell infiltration are generally associated with good prognosis [55], we determined the probability of survival related to Iba1 expression of patient´s with (study EGAS00001003672) and without (TCGA-SKCM) MBM. We observed beneficial effects of high Iba1 levels in the TCGA cohort (HR = 0.46 (0.35–0.62), logrank p = 1.3e-07) (Fig. 1e, f), but observed no beneficial effect on the survival of MBM patients. Since no data on TAM-infiltrated MBM are available, we performed comparative methylome and transcriptome profiling of a core set of Iba1high (n = 5, methylome profiling; n = 10, transcriptome profiling) and Iba1low (n = 2, methylome profiling; n = 6, transcriptome profiling) tumors which were selected based on IHC. We identified a set of 416 differentially methylated genomic regions (DMRs) that corresponded to 294 MBM expressed genes (Fig. 1g) a core set of markers (n = 31) sufficient to split tumors (Fig. 1h; Supplementary Table 2, 3). Among them, we identified the integrin family member and gut-homing receptor ITGB7 -which we described in our previous study as a marker distinguishing BRAF and NRAS mutant MBM [49]—and APBB1IP (amyloid b precursor protein-binding family b member 1 interacting protein). Both are associated with better prognosis in patients with colorectal cancer [19, 80] and clustered with known TAM-associated genes such as P2RY12 and AIF1 (Fig. 1h). Remarkably, all clustered tumors were associated with a high immune score. A correlation analysis of clustered genes revealed a high degree of correlation among each other (Fig. 1i) and association with hepatocyte growth factor (HGF) that was recently connected with microglia activation [52]. However, only some of the identified markers within the gene cluster such as APBB1IP were specifically expressed in microglia but not in brain-infiltrating macrophages or other brain cells (Fig. 1j). APBB1IP has been identified as a conserved microglial gene [20] and binding partner of amyloid precursor protein (APP), Tau, 14–3-3 g, and glycogen synthase kinase 3 b (GSK3 b) was associated with actin dynamics and retinoic acid signaling [31, 35] and was significantly (MBM: R = 0.86, p < 2.2e-16) correlated with immune score (Fig. 1k) and survival of melanoma patients (Supplementary Fig. 1d-e). Moreover, our survey identified a differentially methylated side (Supplementary Table 4) within the promoter of PD-L2 (PDCD1LG2) that may predict progression-free survival in melanoma patients receiving anti-PD-1 immunotherapy [28]. PD-L2 expression was associated with favorable survival (p = 0.020) of patients with MBM (Supplementary Fig. 1f). We found additional genes among our cluster that were expressed in TAMs and significantly associated with immune score (Supplementary Fig. 1 g–n).

Expression of ITGB7 serves as indicator of immune cell infiltration

Recent studies have shown that ITGB7 plays a critical role in the recruitment of T cells to the intestine and that downregulation of ITGB7 is important in protecting intestinal tumors from attack by activated T cells [10, 80]. Hence, we sought to investigate ITGB7 in more detail. Mining of publicly available immune cell data (studies GSE146771 [79], DICE database [61]) revealed expression of ITGB7 across different immune cell stages including naïve and memory subsets of T cells, B cells and NK cells (Fig. 2a and Supplementary Fig. 2a). We found that ITGB7 was rather expressed in MBM with infiltration of immune cells and particularly within immune cell dense areas (Supplementary Fig. 2b). Co-staining revealed accumulation of CD3+ T cells as well as of Iba1high TAMs (Fig. 2b). In line, we found higher levels of CD4, CD274 (PD-L1) and Sushi Domain Containing 3 (SUSD3) in MBM featuring high ITGB7 expression (Fig. 2c) and validated a potential, previously observed [49] correlation of ITGB7 and SUSD3. ITGB7, SUSD3 and APBB1IP showed expression across different immune cell types except for monocytes and NK cells (Supplementary Fig. 2c–f). Global (850 k) methylome profiling uncovered four epigenetic regulation sites of ITGB7 (Supplementary Table 4) with two sites that were associated with gene expression and immune score (Fig. 2d, left and center panel, Supplementary Fig. 3a), located in a proximal enhancer-like region (probe cg26689077) or nearby the promotor of ITGB7 (probe cg01033299). The latter site was also identified in the TCGA-SKCM cohort. The sites did not correlate with the BRAF mutation status of MBM (Fig. 2d, right panel) in contrast to additional two sides that were found within intergenic regions including an CpG island located between exons 4 and 5 (probes cg11510999 and cg18320160; Supplementary Fig. 3b–e). We suggest that ITGB7 may serve as an indicator of the degree of immune cell infiltration of MBM and is assessable by two newly identified DMRs in the ITGB7 gene.

Fig. 2

Expression of ITGB7 serves as an indicator of lymphocyte infiltration. a Box plot representation of levels of ITGB7 indicates a wide pattern of expression among indicated immune cell populations. Monocytes and neutrophil granulocytes show low levels of ITGB7. b IHC of a representative MBM of a patient with refractory intracranial disease for Iba1 (red, first column) and CD3 (brown, second column) indicating focal enrichment of microglia/macrophages and CD3+ T cells within ITGB7 positive areas (red, second column). Hematoxylin and eosin (H&E) staining shows discrimination of tumor cells and tumor-infiltrating lymphocytes (TILs) c Expression (FPKM, log2) of CD4, PD-L1 (CD274) and SUSD3 in MBM with a high or low level of ITGB7, indicating cellular co-occurrence. d Dot plot showing the significant inverse correlation (R =  – 0.87, p = 5.2e-05) of β-values (probe cg26689077) indicating the methylation level at a side located within the proximal enhancer-like structure of the ITGB7 gene and immune score of MBM (n = 14) of study EGAS00001005976 (first panel). Box plots represent a significant (p = 4.5e-04) or non-significant (p = 0.86) association of ITGB7 methylation (probe cg26689077) or BRAF mutation status (center and right panels) of all MBM investigated (n = 21). e Dot plot showing the significant correlation of ITGB7 expression and immune score of MBM (R = 0.51, p = 1.8e-06) and EM (R = 0.61, p = 1.1e-09) indicating immune-related expression of ITGB7 irrespective of the side of metastasis. f.) Correlation map showing high association (p < 0.05) of ITGB7 with relevant immune cell markers such as PD-1 (PDCD1), PD-L1 (CD274), PD-L2 (PDCD1LG2) but low correlation with tumor cell markers NGFR, MITF, MLANA or SLC45A2. g Dot plot showing the significant correlation of ITGB7 and expression of PD-L2 (BM: R = 0.45, p = 3.4e-05; EM: R = 0.42, p = 1.1e-03) and SUSD3 (BM: R = 0.44, p = 5.2e-05; EM: R = 0.61, p = 2.6e-07). h Dot plot showing the correlation of ITGB7 expression and immune scores of primary (PT; R = 0.59, p = 9.4e-16), metastatic (EM; R = 0.78, p = 2.2e-16) and brain metastatic (BM; R = 0.2, p = 0.61) melanoma (TCGA-SKCM), indicating that expression of ITGB7 is independent of melanoma progression stages. i Survival analysis of TCGA melanoma patients (n = 459), featuring a high or low level of ITGB7 and SUSD3 expression revealed a significant difference (log rank p = 4.0e-04 and p = 6.6e-08) and Cox-regression analysis showed association with favorable disease course (HR = 0.60 and HR = 0.48). Box and whisker plots show the median (center line), the upper and lower quartiles (the box), and the range of the data (the whiskers), including outliers (a, c, d). Significance was determined by unpaired, two-sided t-test (c, d) or one-way ANOVA (a)

A recent study demonstrated that MBM feature a lower T cell content than matched extracranial metastases, however, response rates to ICi of both were comparable [71]. Assuming that ITGB7 expression might be crucial for T cell recruitment, we ascertained the levels in MBM (n = 79) and EM (n = 59; study EGAS00001003672). ITGB7 was expressed in both metastatic subtypes and was significantly correlated (MBM: R = 0.51, p = 1.8e-06; EM: R = 0.69, p = 1.1e-09) with the tumor´s immune scores (Fig. 2e). As we suggest that ITGB7 expression might indicate the degree of immune cell infiltration and possibly serve as indicator of response to ICi, we next performed correlation analysis of ITGB7 and known markers of T cells and B cells. We observed a high concordance with immune cell-related but not tumor cell-related genes (NGFR, MITF, MLANA, SLC45A2) and correlation with expression of PDCD1LG2 and SUSD3, irrespective of the side of metastasis (Fig. 2f, g). In line with previous observations, ITGB7 was expressed in primary and metastatic tumors (TCGA-SKCM) and like SUSD3 was associated with favored survival (Fig. 2i). In summary, our survey identified a set of markers that are potentially associated with the level of TAM/immune cell infiltration, particularly ITGB7 might serve as a marker for a favorable course of the disease.

A signature-based deconvolution revealed MET receptor signaling in microglia-enriched MBM

Our previous survey identified a set of markers that potentially characterize a molecular subset of MBM, likely showing a favorable course and response to ICi therapy [23, 30]. To further characterize the molecular subsets, we performed a single-sample Gene set Enrichment-Analysis (ssGSEA) using signatures that defined immune-related signaling or processes that involved MET receptor or STAT3 signaling (Supplementary Table 5). We found that MBM with high immune score were enriched in genes associated with MET and STAT3 signaling, tumor inflammation, stress and senescence (SenMayo [58]) (Fig. 3a) and featured the presence of reactive microglia, astrocytes and immune cell subsets, among them stem cell-like CD8+ T cells (TCF7) [46] in tumors, absent in BMCs. CD8+ (TCF7) T cells are necessary for long-term maintenance of T cell responses and predicted positive clinical outcomes [14, [57]. Signatures clearly discriminated MBM and BMCs and reinforced the differences between Iba1high (Pts 3, 4) and Iba1low/neg (Pts 1, 2) tumors. We therefore suggest that the activation of MET- or STAT3-mediated signaling processes or those related to stress/senescence or inflammation strongly depends on the composition of the tumor microenvironment, likely determining the response to therapeutic interventions. Although infiltration of TAMs is not evident in all MBM, microglia infiltration seems to be an early occurring process observed ~ 21d after intracranial injection of BMCs into brains of immune-compromised Crl:CD1-Foxn1nu mice [49] (Fig. 3b). The activation of Stat3 signaling in tumor-adjacent cells (Fig. 3c), may propose a rapid response of brain microenvironmental cells to brain colonizing tumor cells. We observed a comparable pattern of enrichment of molecular processes in Iba1high MBM of an independent (study EGAS00001003672 [17], n = 79 MBM) (Supplementary Fig. 4a).

Fig. 3

Signature-based deconvolution identified the parameter of MBM featuring a favorable disease course and identified a role of MET signaling. a Single-sample GSEA (ssGSEA)-based deconvolution of MBM of study EGAS00001005976 using customized gene signatures indicating “Signaling” processes, cellular subsets and stages of microglia and astrocyte and immune cell subsets. ssGSEA demonstrated distinct separation of MBM with high, median or low immune score regarding expression levels of signature genes, BMCs served as controls. ssGSEA uncovered differentially activated pathways and processes such as MET and STAT3 and interferon signaling, senescence (SenMayo), stress response and tumor inflammation in tumors enriched for reactive microglia and astrocytes and innate and acquired immune cells subsets. b Confocal microscopy images of orthotopic tumors established by stereotactic injection of BMC1-M4 or BMC2 cells into brains of Crl:CD1-Foxn1nu mice [49], stained for Iba1 (green, microglia) or Iba1, GFAP (red, astrocytes) and KBA.62 (turquoise, pan-melanoma cell marker). DAPI served as a nuclear counterstain. Markers show distinct areas of tumor (MBM) and microenvironment (TME) and regions of microglia infiltration, 21 days after intracranial injection [49]. MBM-TME boarders are indicated by white, dashed lines. c IHC of tumors investigated in (b) for activation and tyrosine phosphorylation (residue Y705) of STAT3. pSTAT3Y705 is particularly present in microenvironmental cells (astrocytes). Black, dashed lines indicate MBM-TME boarders. In b, c, bars indicate 50 µm. d-e Expression levels of hepatocyte growth factor (HGF) in tumors of studies EGAS00001005976, TCGA-SKCM and EGAS00001003672 demonstrating HGF expression in all tumor subsets. f, g Investigation of HGF expression in immune cell subsets (DICE database [61]) and brain cells (study GSE73721) revealed the highest levels in basophil granulocytes and monocytes (f) and in astrocytes and microglia (g). h UMAP projection of expression profiles from nuclei isolated from 5 neurotypical donors as provided by Seattle Alzheimer’s disease brain cell atlas (https://portal.brain-map.org/explore/seattle-alzheimers-disease), cellular subtypes are color coded (left panel). Log-normalized expression levels of HGF in nuclei isolated from 5 neurotypical donors (center panel). Log-normalized expression levels of HGF in nuclei isolated from 84 aged donors (42 cognitively normal and 42 with dementia), right panel, demonstrating an increased number of HGF expressing microglia and astrocytes as triggered by inflammatory processes. i Dot plot showing the correlation of HGF expression and immune score of BM (R = 0.49, p = 5.3e-06) and EM (R = 0.41, p = 1.5e-03) indicating a potential role of HGF in immune cell-related processes. Box and whisker plots show median (center line), the upper and lower quartiles (the box), and the range of the data (the whiskers), including outliers (d–g). Significance was determined by unpaired, two-sided t-test (e) or one-way ANOVA (g)

HGF or scatter factor (SF) is the only identified ligand of MET and exerts pivotal functions during neural development, regulating the growth and survival of neurons [15, 45], likely serving as an inducer of reactive microglia by an autocrine loop in response to trauma or neurodegenerative disorders [52]. Therefore, MET-expressing melanoma cells infiltrating the brain can benefit from the HGF-regulated systems that naturally occur in the brain and employ them as a survival strategy. We observed HGF expression among tumors of different data sets comprising MBM, EM and primary tumors (studies EGAS00001005976; TCGA-SKCM; EGAS00001003672) with no significant difference in HGF levels of tumor subsets (Fig. 3d, e). Investigation of immune cell and brain cell data (DICE database [61] and study GSE73721 [78]) revealed high expression of HGF in monocytes and astrocytes (Fig. 3f, g), and the exploration of single-cell studies (GSE115978 [32] and GSE186344 [22]) revealed melanoma and MBM associated microglia/macrophages as a source of released HGF (Supplementary Fig. 4b).

Assuming that the mutual interaction of tumor cells and TAMs determines the routes of MBM progression and fosters the activation of MET receptor-related processes, we investigated levels of MET signaling-associated genes. We found that degrees of HGF, PIK3CG, PTK2B, STAT3 and MAP4K1 significantly correlated with microglia score (Supplementary Fig. 4c–e) that was defined as average expression (log2 FPKM) or methylation status (β-value) of microglia markers APBB1IP, SYK, HCK and P2RY12 (Supplementary Table 6). HGF might be released by homeostatic and reactive microglia (RM) [1, 52] or reactive astrocytes (RA) [37, 64] but transcriptome profiling data on MBM-related RM/RA are not available. We surveyed the Seattle Alzheimer´s Disease Brain Atlas which is implemented in the Allen brain atlas database (https://portal.brain-map.org/). Dementia fostered the expansion of microglia and astrocytes with increased expression of HGF (Fig. 3h, center and right panels). Reactive microglia and immune cell released HGF might hence be responsible for the activation of growth factor/survival signaling in adjacent tumor cells. The level of HGF expression significantly correlated with an immune score in the brain (BM, R = 0.49, p = 5.3e-06) and extracranial metastases (EM, R = 0.41, p = 1.5e-03), (Fig. 3i).

Expression and activation of MET receptor classifies a molecular subset of MBM

Understanding the molecular mechanisms that establish cellular dependencies and thus control the development and maintenance of brain metastases is critical for their therapeutic manipulation. Recently, we identified that the expression of Ecad and NGFR sufficiently discriminated molecular subsets of MBM [49] likely to exhibit different responses to therapeutics and mechanisms of interaction with cells in the microenvironment (Fig. 4a). To identify potential druggable targets, we surveyed the pan-MBM, NGFR and Ecad-specific gene sets for cell surface receptors that may serve as crucial key factors controlling tumor cell survival and maintenance. We identified 24 receptors that distinguished Ecad+ and NGFR+ tumors (Fig. 4b). Particularly ADIPOR1 (adiponectin receptor 1, p = 1.9e-02), SIRPA (signal regulatory protein alpha, p = 1.1e-05) and PLXNC1 (plexin C1) showed significantly increased expression in Ecad+ subsets of MBM and EM (Supplementary Fig. 5a). In addition, we found MET receptor predominantly expressed in Ecad + tumors (p = 1.4e-04) and significantly (p = 2.7e-05) higher expressed in MBM than EM (Fig. 4c, left and center panels). MET was enriched in proliferating tumor cells featuring high levels of markers of cell cycle progression such as cyclins b1 and b2 (CCNB1, CCNB2), proliferating-cell nuclear antigen (PCNA) and Ki67 (MKI67) (Fig. 4c, right panel, Supplementary Table 6) and likely define yet another subset of MBM.

Fig. 4

Ecad+ MBM are defined by expression of MET receptor. a Schematic summary of the initial screen of MBM expression data of our recent study (EGAS00001005976; n = 16 MBM) for subset expressed receptors. MBM contains Ecad+ and NGFR+ subsets and admixed cells such reactive microglia, labeled by expression of Iba1/AIF1 and or P2RY12. The initial survey yielded 24 receptors that potentially establish cell survival/growth of MBM. b.) Correlation map (Spearman, p < 0.05) showing the relationship of identified receptors expressed in MBM of our previous study, emphasizing the distinct pattern of Ecad+ and NGFR+ molecular subsets. The value of the correlation coefficient is color coded. c Box plots depicting the levels of MET in Ecadhigh and Ecadlow subsets of MBM and EM (left panel, p = 1.4e-04/p = 0.41) or in all subtypes of MBM and EM (p = 2.7e-05) in high and low proliferating tumor cell subsets (right panel, p = 9.1e-03). d Comparative principal component (PCA) representation of primary tumors (PT), extracranial metastases (EM) and MBM (MBM_TCGA) of the TCGA-SKCM cohort as well as MBM of our study (MBM_CHA, EGAS00001005976) depicting gradual levels of MET and Ecad (CDH1) expression. The panels below show a comparison of levels of MET, Ecad, MITF and NGFR of selected tumors showing distinct and overlapping cell states. Expression levels (log2, FPKM) are color coded. e IHC of selected MBM for MET and MITF validated the two subsets. f, g Expression and activation status of MET in BRAF wildtype (wt; Pts 14, 36) and BRAFV600E/R mutated MBM (Pts 28, 29, 31). Phosphorylation of MET at residues Y1234/1235 is critical for kinase activation. h IHC of indicated tumors for co-localization of pMETY1234/1235 (brown) and Iba1 (red) demonstrating potential activation of MET receptor signaling tumor cells by stromal cell-secreted HGF. i Heat map representing expression levels of regulators and targets of interferon signaling and immune-related genes showing clustering according to the level of ITGB7 expression. Box and whisker plots show median (center line), the upper and lower quartiles (the box), and the range of the data (the whiskers), including outliers (c, d). Significance was determined by unpaired, two-sided t-test (c, d)

We investigated the distribution of MET expression and of Ecad (CDH1) and MITF, the transcriptional regulator of MET in melanocytes as well as marker of pigmented melanoma cells among primary tumors, extracranial metastases (TCGA-SKCM), and brain metastases (MBM_CHA). We observed distinct as well as shared tumor cell states (Fig. 4d). Moreover, we detected that the MET receptor was not only present in Ecad+/MITF± tumors but rather showed gradual distribution across tumors and co-expression in a minority of NGFR+ cells (Fig. 4d and Supplementary Fig. 5b). To additionally unravel the cell states, we next explored single-cell transcriptome data (study GSE115978) for distribution of MET expression in melanoma cell subsets comprising immune cells, cancer-associated fibroblasts (CAF), endothelial cells, macrophages, tumor cells and natural killer (NK) cells. We observed that MET is primarily expressed in tumor cells and to a lesser extent in immune (T cells) cells/NK cells, macrophages, CAF or endothelial cells (Supplementary Fig. 5c, d). Tumor cells showed discrete and overlapping expression of MET, Ecad and MITF (Supplementary Fig. 5d–f) and were co-expressed with NGFR in a small subset of tumor cells (Supplementary Fig. 5 g, h). The investigation of BMCs (BMC1-M1) validated the co-occurrence of MET and NGFR expression in cellular subsets (Supplementary Fig. 5i, j), suggesting that the MET tyrosine kinase receptor pathway may serve as a potent survival and maintenance mechanism in MBM and that the therapeutic targeting by small molecule inhibitors [77] may eliminate several tumor cell fractions including NGFR+ cells poten