In this study we have evaluated the cellular host responses to orthotopic xenografts of glioma cell lines and GBM biopsy implants in rodent models. We observed that host stem/progenitor cells identified by rodent nestin expression displayed tropism for human glioma xenografts in both mice and rats, and that host cells originated and migrated from the ipsilateral SVZ to xenografts. We and others have previously reported specific tropism of exogenous and endogenous NSPCs to glioma xenografts 
. During normal brain development, tissue regeneration and learning in rodents, NSCs originate in the SVZ and follow the rostral migratory stream to reach the olfactory bulb. This process is tightly regulated, and the cells do not disperse into the brain but rather follow a well-defined route 
. In the pathological brain, the migratory patterns of NSCs change, with fewer new cells found in the olfactory bulbs and more cells found in the diseased cortex and corpus callosum 
. NSCs were recruited from the SVZ ipsilateral to the tumor implant, while the contralateral NSC pools did not appear to be affected. In GBM biopsy xenografts in rats, the tumor border contained host nestin-positive cells with various morphologies: some showed a bipolar morphology characteristic of migrating cells with very thin elongated cell bodies, whereas others had several processes and assumed a star-like appearance characteristic of astrocytes. In low-generation invasive glioma lesions, nestin-expressing single cells with astrocytic or bipolar morphology were richly interwoven between infiltrating tumor cells. Previous investigations have shown that nestin-positive neural precursor cells accumulate at the site of mechanical brain injury, and by day 14 post-surgery, these cells display the branched morphology of reactive astrocytes 
. On the other hand, NSCs around tumors displayed a bipolar morphology and persisted for at least 30 days, although in lower numbers, around the lesions. In comparison, our high generation biopsy spheroid-derived lesions grew in an obstructive way, creating a mass effect similar to the growth pattern of the glioma cell lines transplanted by Glass et al. (2005). Similarly, we generally observed a bipolar morphology in the nestin-positive host population of these lesions. In contrast, in low-generation xenografts, the tumor cells seamlessly infiltrated the brain, and we generally observed nestin-positive cells with an astrocytic morphology, comparable to what was seen at the site of mechanical injury (Glass et al., 2005). Possibly, in low-generation biopsy xenografts, after repair/regeneration of the injection site, the dispersed single glioma cells failed to stimulate further influx of neural progenitors from the SVZ. The previously recruited progenitors may have differentiated and displayed morphology similar to mature astrocytes.
The high (fifth) generation glioma implants were double-stained for GFAP and rat-specific nestin to evaluate the glial nature of the infiltrating host cells. GFAP stained both mature astrocytes and glioma cells. At the tumor border, the vast majority of nestin-positive cell bodies also expressed GFAP. These cells had the elongated cell bodies of invasive bipolar cells, although some also had several end-processes, resembling astrocytes. The bipolar cells may have been incoming neural progenitors, whereas those with branched morphology may have been reactive astrocytes. In the tumor, cells with elongated morphology, as well as cells with astrocytic morphology stained positively for both rat nestin and GFAP.
In both mice and rats, a host nestin-positive cell population (endothelial progenitors) contributes to human xenograft growth by assembling into the glioma microvasculature. In contrast, non-pathological brain microvessels showed little or no nestin expression. This confirms xenograft data from other tumor types, where nestin positive cells in tumor vessels have been identified as endotheliocytes 
. In the high (fifth) generation GBM spheroid xenografts, we observed glomerulus-like microvascular proliferations, which stained intensely for host (rat) nestin. Such glomerulus-like vessels are a hallmark of GBM pathology 
, and likely represent non-functional vessels that might be responsible for generating the hypoxic environment in gliomas 
. It is important to note that vascular endothelial growth factor (VEGF), when processed by matrix metalloprotesases (MMPs), causes generation of dilated, glomerulus-like vessels in tumors, whereas in the absence of MMPs more vascular sprouting is observed, which resembles the normal blood vasculature 
. Because glioma pathology, invasion and angiogenesis are closely tied to the activity of various MMPs 
, and based on our current data, it is likely that recruitment of rat nestin-positive host cells by glioma reflects the intricate processes of glioma biology, including tumor hypoxia and angiogenesis.
In the mouse xenografts, staining for both host nestin and SMA showed similar staining pattern, illuminating the vascular network in the tumor bed. These two markers did not co-localize, but were expressed by two distinct, closely associated cell populations that together generated tumor microvessels. Based on their elongated morphology, luminal localization and relative abundance, host nestin-expressing cells are suggested to be endothelial cells or endothelial progenitors 
. Indeed, the endothelial cell marker CD31 has been found co-expressed with nestin in the tumor vasculature of orthotopic xenografts of the pancreas in the nestin-GFP mice 
. On the other hand, SMA is a marker of smooth muscle cells and mature pericytes, which may be derived from hematopoietic stem cells or bone marrow-derived pericyte progenitor cells that are recruited to the brain during glioma angiogenesis, suggesting a ‘long-distance’ effect in recruitment of cells that contribute to tumor angiogenesis 
. Furthermore, direct contacts between endothelial cells and smooth muscle cells have been described as ‘peg-socket’ contacts, where the basement membrane is absent. We detected structures resembling such contacts between SMA-positive cells and mouse nestin-positive cells (). These cell-to-cell communication points have been described to contain tight, gap, and adherence junctions 
. The contact between endothelial cells and pericytes leads to activation of latent TGF-β, which promotes pericyte differentiation via the ALK5 Smad2/3 signaling pathway 
. Endothelial cells line the inside of blood vessels, whereas pericytes line the outside surface and interact with the basement membrane that surrounds the blood vessels 
. In an analysis of the spatial localization of endothelial cells (CD31+) and pericytes (SMA+) during angiogenesis in spontaneous tumors of the pancreatic islet as well as in transplanted mammary and lung carcinomas, Morikawa et al found that, similar to our data, there was SMA-positive pericytic coverage or ‘sleeves’ at the leading edges of vascular sprouts 
. In their study, pericytes in capillary microvessels formed by angiogenesis uniformly expressed SMA, in contrast to corresponding normal tissue, in which only arterioles and venules had SMA-positive adventitial cells. Our data show that in the brain, tumor capillaries were richly populated by SMA-positive accessory (pericytic/mural) cells, whereas the normal brain vasculature did not express SMA. In a recent study of 40 GBM patients Sica et al reported on the presence of nestin-positive, SMA-positive and CD105-positive cell populations associated with microvasculature in the peritumor area 
. High microvascular density in the peritumor area correlated with a shorter survival time of the GBM patients, which suggests that the cell populations attracted by glioma may have an influence on the clinical progression of the disease.
The SDF-1/CXCR4 signaling axis has been shown to be a major player in the dissemination and metastasis of tumor cells, e.g., in breast cancer metastasis to the bone marrow 
, VEGF-induced neovascularization and retention of myeloid cells in normal and cancerous tissues 
, recruitment of vascular progenitors to glioma 
, and glioma invasion and growth 
. This chemokine/receptor system may be also involved in the homing of bone marrow-derived hematopoietic cells or neural stem or progenitor cells to tumors 
. In GBM, the expression levels of both SDF-1 and its receptor correlate positively with tumor grade, with CXCR4 being mainly localized to tumor endothelial cells, which likely use the SDF-1/CXCR4 axis to migrate during angiogenesis 
. Exogenous NSCs that target human glioma xenografts in mouse brain localize to the tumor (including the hypoxic regions) and tumor edge, which displayed high levels of SDF-1 expression 
. In normal brain, the expression pattern of SDF-1 and CXCR4 has been shown to coincide with that of NPCs, with virtually all nestin-positive cells in the SVZ also expressing CXCR4 
. Functional studies have confirmed the role of the SDF-1/CXCR4 system in NPC migration after neurological damage. For example, in cerebral stroke, the migration of NSCs to the injury site was attenuated by adding anti-SDF-1 blocking antibodies 
. SDF-1 also promotes adult NPC proliferation as well as migration 
, and the SDF-1/CXCR4 signaling axis is involved in trafficking of normal stem cells as well as in metastasis of cancer cells 
. We found that the location (peritumor halo, scattered single infiltrating cells in the tumor, and expression in tumor-associated blood vessels) and morphology (elongated single cells, cells with astrocytic morphology and small vessels) of SDF-1-expressing cells were identical to those of mouse nestin-expressing cells. This suggests that NPCs may express SDF-1, possibly indicating an autocrine stimulatory loop, or ‘trapping’ of SDF-1 by the recruited progenitors and vascular structures that arise from these cells.
migration of NSPCs is induced by vascular endothelial and astrocytic expression of SDF-1 
. Similarly, we observed SDF-1 expression in cells with astrocytic as well as endothelial morphology (), in addition to invasive cells that resembled infiltrating NPCs. CXCR4 expression localized to the same areas as that of SDF-1, but seemed to be less abundant both at the tumor-bordering brain and the tumor core. The exception was stronger expression of CXCR4 around necrotic foci as compared to SDF-1. Necrotic foci and the pseudopalisading regions are hypoxic and are the main sites of HIF-1α and VEGF expression in glioblastoma 
, which may result in high expression of CXCR4. Indeed, over-expression of CXCR4 has been shown to be mediated by HIF-1 and VEGF 
, which is the main hypoxia-induced angiogenic pathway present in both vascularized cell line-based xenografts and in the high-generation lesions in the biopsy xenograft model 
Finally, the nestin-positive host cells recruited by gliomas may have effects other than those discussed above. For example, glioma cells generate considerable amounts of glutamate, which can lead to excitotoxity of neurons in the brain parenchyma surrounding the glioma mass, thus promoting the invasive migration and growth of glioma 
. The presence of nestin-positive stem or progenitor cells in and around glioma may have protective effects against glutamate cytotoxicity. On the other hand, gliomas may co-opt some of the physiological properties of neural and mesenchymal stem cells for their advantage. Because neural and mesenchymal stem cells can release large amounts of TGF-β, a cytokine with immosuppressive potential, the anti-glioma immune response may be attenuated, resulting in tumor ‘escape’ from normal immunological surveillance 
. It should be noted that glioma cells may communicate with host cells in even more intricate ways; for instance, cell-cell fusion or horizontal gene transfer may result in exchange of genetic material between glioma and host cells 
. Furthermore, microvesicles (exosomes) that contain mRNA, microRNA and angiogenic proteins have been shown to be released by glioma cells, which can be taken up by normal host cells, such as brain microvascular endothelial cells 
. A more recent and striking finding has been that glioblastoma stem-like cells can give rise to tumor endothelium 
. These data, together with our findings presented here, underscore the complex nature of malignant gliomas. Indeed, in addition to the cell types described in our study, the repertoire of cells attracted by or interacting with glioma is extremely diverse and includes neurons, oligodendrocytes, astocytes, microglia, cells of the immune system (B lymphocytes, various populations of T lymphocytes, macrophages), mast cells, and numerous others 
. These cell types contribute to the microenvironment in which the tumor cells can proliferate, invade and destroy the normal brain parenchyma. However, some of these tumor-targeting host cells may serve also a protective function of the host by inhibiting tumor growth and invasion. It is hoped that further elucidation of the tumor and host interactions may aid the development of novel treatments for glioma, one of the most dreaded brain cancers.