Glioblastoma multiforme (GBM) is characterized by a pathogenic vasculature that drives aggressive local invasion. Recent work suggests that GBM cells recruit bone marrow–derived progenitor cells (BMDC) to facilitate recurrence after radiotherapy, but how this may be achieved is unclear. In this study, we established the spatiotemporal and regional contributions of perivascular BMDCs (pBMDC) to GBM development. We found an increased recruitment of BMDCs to GBM in response to tumor growth and following radiotherapy. However, in this study, BMDCs did not differentiate into endothelial cells directly but rather provided a perivascular support role. The pBMDCs were shown to associate with tumor vasculature in a highly region-dependent manner, with central vasculature requiring minimal pBMDC support. Region-dependent association of pBMDC was regulated by VEGF. In the absence of VEGF, following radiotherapy or antiangiogenic therapy, we documented an increase in Ang2 that regulated recruitment of pBMDCs to maintain the vulnerable central vasculature. Together, our results strongly suggested that targeting pBMDC influx along with radiation or antiangiogenic therapy would be critical to prevent vascular recurrence of GBM. Cancer Res; 74(14); 3727–39. ©2014 AACR.

Glioblastoma multiforme (GBM) is the most frequently occurring and aggressive of all primary adult brain tumors (1, 2). Despite maximal surgical resection followed by concurrent chemotherapy and radiotherapy, the prognosis for patients with GBM remains extremely poor, with median survival time being 12 to 16 months and an almost 100% recurrence rate (1–3). A major histopathologic hallmark of GBMs is increased vascular endothelial proliferation, making GBMs one of the most highly vascularized solid malignancies (1, 2). Given their hyperproliferative vasculature, novel antiangiogenic therapies are actively being explored (4, 5). Unfortunately, despite some positive early therapeutic benefits, overall patient outcomes have not significantly improved with antiangiogenic therapies (6). Resistance to antiangiogenic therapies indicates that there remains, as yet, an unexplored mechanism of neovascularization that allows GBM cells to evade therapy and continue to grow.

Like other solid cancers, GBM uses complex molecular cascades to create an independent vascular network sustaining growth and recurrence following therapy (3, 6, 7). There has been emerging interest in understanding the role of bone marrow–derived cells (BMDC) in cancer and their contribution to tumor neovascularization (8, 9). The literature remains controversial as work from several laboratories suggests that BMDCs do not incorporate into the vessel lumen to a significant extent but rather stay adjacent to the vessel in the perivascular niche (10), whereas other reports suggests that endothelial cells in glioma models are directly derived from BMDCs (11, 12).

One of the challenges in being able to definitively establish whether circulating progenitor cells differentiate to form endothelial cells lies in the limitations of histopathologic analysis in distinguishing between different cell types and the origin of cells that compose a vascular structure. We have established an experimental strategy that takes advantage of 2-photon laser microscopy (2PLM) coupled with an intracranial window in mouse models of GBM to obtain in vivo real-time longitudinal imaging of normal brain and GBM-associated vasculature (13). We take advantage of this experimental strategy to complement traditional immunohistochemical (IHC) and immunofluorescent analysis by allowing direct visualization of single cells intracranially.

In this study, we demonstrate that BMDCs do not directly differentiate into endothelial cells during brain tumor growth or following radiotherapy but provide a supportive role as perivascular cells. We show that BMDCs have a highly region-dependent association with GBM neovascularization, where beyond a critical tumor volume, BMDCs are not required for maintaining the central vasculature. VEGF regulates the regional recruitment of perivascular BMDCs (pBMDC). Loss of VEGF, either through antiangiogenic therapies or radiotherapy, results in upregulation of angiopoietin-2 (Ang2), which in turn signals recruitment of pBMDCs to support and restore the vulnerable central vasculature following therapy. Blocking Ang2, following loss of VEGF, prevents the influx of pBMDCs in central vessels, confirming that an interplay between VEGF and Ang2 is central to the regional association of pBMDC. These results suggest targeting pBMDCs concurrent with radiotherapy and antiangiogenic therapies will be an important therapeutic strategy for preventing GBM recurrence following therapy.

All animal procedures were carried out according to animal user protocols approved ethically by Institutional Animal Care Committee under the guidelines of the Canadian Council on Animal Care.

For all sections, see Supplementary Methods for full procedure outline.

Cell lines

U87MG and U251MG GBM cell lines and GSC8-18 stem cell lines were used for intracranial xenograft models (Cell lines were obtained from ATCC or MD Anderson, Houston, TX; all lines are regularly authenticated on the basis of viability, recovery, growth, and morphology, most recently confirmed 3 months before use). Cells were grown as previously described (13–15).

Cell transfections

Cell lines (U87MG, U251, GSC8-18) were transfected with either FuGENE HD (Roche) or Lonza-Amaxa electroporation, as per commercial instructions, to express constituently the mCherry fluorescent gene (pCAG:mCherry:Puro courtesy of Dr. Peter Tonge) or PBase-VEGFTrap (Dr. Andras Nagy).

Bone marrow reconstitutions

Bone marrow was extracted from 8- to 10-week-old donor transgenic GFP/RFP mice (Dr. Andras Nagy) and transplanted into 3 irradiated host mice, as previously described (13, 16).

Xenograft models of GBM

For all experimental arms, including control, 10 mice were used, and all experimental arms were performed in triplicate.

Chimeric bone marrow mice underwent intracranial window surgery, as previously described (13, 16). Mice were streamed into cohorts of control (no cell injection); GSC Cherry 2.5 × 104 cells injected; U87/U251 Cherry/VEGFTrap 4 × 105 cells injected.

Transgenic mouse models

GBM transgenic mouse model, RASB8, was generated through integration of a V12 RASB8 mutation under the control of glial fibrillary acidic protein-promoter, leaving mice predisposed to sporadic GBM-like astrocytoma (17). Both normal RASB8 and chimeric bone marrow RASB8 mice were used in the study.

Targeted radiotherapy

Experimental mice undergo targeted radiotherapy on the Xrad225 small animal irradiator at set treatment regimens of single 2, 6, or 15 Gy or fractionated 3 × 2 and 3 × 5 Gy, as previously described (13, 16).

Chemotherapy/Drug therapy

Temozolomide (Sigma) was administered at 60 mg/kg daily for 5 days administered 1 hour before all radiotherapy doses. Doxycycline (Bioserve) was administered in the mouse chow at 3 mg/mL 1 week following tumor implantation to turn on VEGFTrap. AMG386 (Amgen) was administered twice weekly at 10 mg/kg as an intravenous injection at the beginning of either radiotherapy or VEGFTrap therapy.

Real-time in vivo imaging

Experimental mice were imaged on an inverted confocal microscope through the intracranial window generated, as previously described (13, 16). Intravenous fluorescent dextran (Invitrogen) or CD31 antibody (BD Pharmingen) was administered 5 minutes before the imaging session to highlight vasculature.

MRI

Serial multiparametric MRI protocol was carried out on a 7-Tesla Bruker MRI system to look at anatomy and perfusion, as previously described (18).

Histologic analysis: IHC and IF

Mice at end point were sacrificed with paraformaldehyde perfusion, as previously described. Optimal cutting temperature blocks of the brain were made and sectioned at 5-μm intervals onto coated glass slides. Sections were air dried and inhibited for peroxidase and biotin activity. Slides were blocked before primary incubation and then washed three times before secondary incubation. IHC slides then underwent a peroxidase labeling treatment before DAB color development. Slides were dehydrated and mounted.

Semiquantification image analysis

For 2PLM data quantification, 10 high-powered fields were analyzed on ImageJ. IHC and IF image analyses were carried out on 10 high-powered fields in Kati-Kat cell counting software. All were carried out by three blinded counters.

Laser capture microscopy

Frozen tissue was sectioned onto laser capture microscopy (LCM) slides and dehydrated through alcohols before loading onto the microscope. Vessels were isolated through GFP content, a minimum of 15 per sample.

Quantitative PCR analysis

RNA was extracted from LCM samples as per the Picopure RNA Extraction Kit (Applied Biosciences), and cDNA synthesized using superscript VILO kit (Invitrogen). cDNA was diluted 1:3 before use in 50 μL SYBR Green qPCR reactions (Invitrogen).

Statistical analysis

All experiments were performed in triplicate with mean and SE reported. ANOVA was followed by a post-Dunnett test or post-Tukey. Significance was defined as *, P < 0.05.

BMDCs are recruited to GBMs in a tumor growth stage-dependent manner

We examined the longitudinal response of BMDCs to GBM growth in both intracranial xenograft models of GBM (U87 and glioma stem cells; GSC) and in the spontaneous mouse models of high-grade gliomas, RasB8 (15, 17, 19, 20). In vivo 2PLM images of BMDCs in GBM xenografts show that BMDCs are recruited via the intracranial vasculature specifically to the site of tumor growth (Fig. 1A and B) as proven by minimal GFP+BMDCs seen in control non–tumor-bearing mice (Fig. 1C) and contralateral brain. Similarly, BMDCs are recruited specific to the tumor foci in transgenic RASB8 models (Fig. 1D). BMDCs populate three distinct anatomical distribution: circulating intravascular, pBMDC, and stromal BMDC (Fig. 1E–G). pBMDCs are seen as those BMDCs that migrate out of the vasculature but remain intimately associated with the vessels (Fig. 1F). In both xenograft and transgenic models of GBMs, we observe recruitment, migration, and infiltration of BMDCs, which increases significantly in a longitudinal manner in response to the tumor growth (Fig. 1H).

Figure 1.

Spatiotemporal response of BMDCs to GBM growth. A–C, in vivo real-time 2PLM images demonstrating recruitment, migration, and infiltration of BMDCs in intracranial xenograft models of GBM. Day 21: post-intracranial injection of GSC (A) and U87 (B) and in control normal brain (C). GFP+BMDC, green; mCherry+ GBM tumor cells, red; CD31+ endothelial cells, blue. D, ex vivo imaging of GBM transgenic tumors, RASB8, confirms recruitment of BMDC specific to the site of tumor formation. i, hematoxylin and eosin-stained coronal section of astrocytoma foci (black box). ii, magnified corresponding IF image sections demonstrate increased levels of BMDCs within central tumor foci (C) compared with periphery (P) and normal brain (NB). BMDCs demonstrate three distinct anatomical distribution: circulating intravascular (gray arrow; E), pBMDCs (yellow arrow; F), and stromal BMDCs (red arrow; G). H, quantification of BMDCs using 2PLM images demonstrates a statistically significant increase in BMDC recruitment in response to tumor growth longitudinally in both U87 and GSC intracranial xenograft models. ***, P < 0.0001. Data represent mean ± SEM, n = 10. Scale bars, μm.

Figure 1.

Spatiotemporal response of BMDCs to GBM growth. A–C, in vivo real-time 2PLM images demonstrating recruitment, migration, and infiltration of BMDCs in intracranial xenograft models of GBM. Day 21: post-intracranial injection of GSC (A) and U87 (B) and in control normal brain (C). GFP+BMDC, green; mCherry+ GBM tumor cells, red; CD31+ endothelial cells, blue. D, ex vivo imaging of GBM transgenic tumors, RASB8, confirms recruitment of BMDC specific to the site of tumor formation. i, hematoxylin and eosin-stained coronal section of astrocytoma foci (black box). ii, magnified corresponding IF image sections demonstrate increased levels of BMDCs within central tumor foci (C) compared with periphery (P) and normal brain (NB). BMDCs demonstrate three distinct anatomical distribution: circulating intravascular (gray arrow; E), pBMDCs (yellow arrow; F), and stromal BMDCs (red arrow; G). H, quantification of BMDCs using 2PLM images demonstrates a statistically significant increase in BMDC recruitment in response to tumor growth longitudinally in both U87 and GSC intracranial xenograft models. ***, P < 0.0001. Data represent mean ± SEM, n = 10. Scale bars, μm.

Close modal

Response of BMDCs following chemotherapy and radiotherapy of GBM

We next examined the response of BMDCs to GBMs treated with radiotherapy and the clinically relevant alkylating chemotherapy, temozolomide. Radiotherapy demonstrated a statistically significant decrease in final tumor volume (Fig. 2A and B) together with a statistically significant increase in BMDC recruitment in both 2PLM and ex vivo IF analysis of xenografts and transgenic tumors (Fig. 2C, i and ii, D, and E). In comparison, irradiation of normal brain results in recruitment of BMDCs specifically to the site of radiotherapy, with minimal migration into the surrounding microenvironment (Fig. 2F; ref. 13). Furthermore, extent of BMDCs seen in response to radiotherapy in GBMs is significantly greater than that seen in radiotherapy alone and untreated GBMs (Fig. 2G). Chemotherapy, using temozolomide did not result in a change in the pattern or extent of BMDC recruitment (Fig. 2H, i and ii).

Figure 2.

Spatiotemporal response of BMDCs following treatment. A, T1+ contrast MRI images of untreated U87 GBM xenografts (i) and radiotherapy-treated U87 GBM (ii), 3 × 2 Gy. B, volumetric analysis demonstrates a statistically significant reduction in final volume (*, P = 0.0256). Data represent mean ± SEM, n = 10. C, i and ii, ex vivo IF images show an increase in BMDC recruitment in response to radiotherapy (3 × 2 Gy; ii) compared with untreated U87 GBM xenograft (ii). GFP+BMDC, green; mCherry+ U87 GBM tumor cells, red; nuclei, blue. D and E, in vivo 2PLM images also confirm recruitment, migration, and infiltration of the same three BMDC anatomical distributions in response to radiotherapy (3 × 2 Gy). GFP+BMDC, green; mCherry+ U87 GBM tumor cells, red; vascular EC CD31+APC, blue. F, radiotherapy (3 × 2 Gy) of normal brain results in recruitment of morphologically undifferentiated, rounded BMDCs that remain specific to site of radiotherapy, with minimal migration beyond site of radiotherapy. GFP+BMDC, green; mCherry+ GBM tumor cells, red; vascular EC CD31+APC, blue. G, quantification of BMDC on 2PLM images demonstrates a statistically significant increase in BMDC recruitment in response to radiotherapy compared with a control (***, P < 0.0001). Radiotherapy results in increased recruitment of BMDCs beyond what is seen in response to tumor growth. Data represent mean ± SEM, n = 10. H, i and ii, in vivo 2PLM images demonstrate that the addition of temozolomide (TMZ; i) and TMZ + radiotherapy (ii) have no effect on recruitment, migration, and infiltration of BMDCs when compared with non–TMZ-treated GBM (D) and GBM + radiotherapy (E). GFP+BMDC, green; RFP+ U87 GBM tumor cells, red; vascular EC CD31+, blue. Scale bars, μm. RTx, radiotherapy.

Figure 2.

Spatiotemporal response of BMDCs following treatment. A, T1+ contrast MRI images of untreated U87 GBM xenografts (i) and radiotherapy-treated U87 GBM (ii), 3 × 2 Gy. B, volumetric analysis demonstrates a statistically significant reduction in final volume (*, P = 0.0256). Data represent mean ± SEM, n = 10. C, i and ii, ex vivo IF images show an increase in BMDC recruitment in response to radiotherapy (3 × 2 Gy; ii) compared with untreated U87 GBM xenograft (ii). GFP+BMDC, green; mCherry+ U87 GBM tumor cells, red; nuclei, blue. D and E, in vivo 2PLM images also confirm recruitment, migration, and infiltration of the same three BMDC anatomical distributions in response to radiotherapy (3 × 2 Gy). GFP+BMDC, green; mCherry+ U87 GBM tumor cells, red; vascular EC CD31+APC, blue. F, radiotherapy (3 × 2 Gy) of normal brain results in recruitment of morphologically undifferentiated, rounded BMDCs that remain specific to site of radiotherapy, with minimal migration beyond site of radiotherapy. GFP+BMDC, green; mCherry+ GBM tumor cells, red; vascular EC CD31+APC, blue. G, quantification of BMDC on 2PLM images demonstrates a statistically significant increase in BMDC recruitment in response to radiotherapy compared with a control (***, P < 0.0001). Radiotherapy results in increased recruitment of BMDCs beyond what is seen in response to tumor growth. Data represent mean ± SEM, n = 10. H, i and ii, in vivo 2PLM images demonstrate that the addition of temozolomide (TMZ; i) and TMZ + radiotherapy (ii) have no effect on recruitment, migration, and infiltration of BMDCs when compared with non–TMZ-treated GBM (D) and GBM + radiotherapy (E). GFP+BMDC, green; RFP+ U87 GBM tumor cells, red; vascular EC CD31+, blue. Scale bars, μm. RTx, radiotherapy.

Close modal

To confirm that results seen with BMDCs in response to tumor growth and treatment are not simply related to tumor proliferation rate, we measured the tumor proliferative index using Ki67 staining. We see no statistically significant difference in early to late stages of tumor growth in Ki67 staining. Similarly, there is no statistically significant change in Ki67 positivity following radiotherapy, unless used at higher than clinically relevant doses (3 × 5 Gy; Supplementary Fig. S1).

BMDCs differentiate to form distinct cell types

Concordant with previous reports by Kozin and colleagues, we found that the majority of BMDCs recruited to the tumor showed a predominantly inflammatory and microglial phenotype, in both xenograft and transgenic GBMs (Fig. 3A–C; refs. 21, 22). Approximately 80% of the recruited BMDCs in untreated and radiotherapy GBMs are MAC3+, CD11b+, and IBA1+ (Fig. 3D). We found that 70% of cells expressing IBA1+ were RFP+BMDCs, indicating that they are not resident inflammatory/microglial cells in the brain, but rather recruited from the bone marrow, suggesting that local microglia are either not activated or sufficient to provide a response to GBM growth (Fig. 3E). Majority of the bone marrow–derived IBA1+ cells take on a perivascular localization. We also found a subpopulation of BMDCs, which could be characterized as Tie2-expressing monocytes (TEM; Tie2+ CD11b+), which have been shown previously to be involved in tumor progression (Fig. 3F; refs. 22, 23). We found no evidence of BMDCs differentiating to other potential cell types, such as fibroblasts (reticulin), glial cells (glial fibrillary acidic protein), and neurons (TUJ1) in any stage of tumor growth. Furthermore, we did not find any new cell types into which BMDCs differentiate in response to radiotherapy or temozolomide.

Figure 3.

Characterization of recruited BMDCs in GBMs. Ex vivo IF staining of U87 GBM intracranial xenograft (i) and transgenic RASB8 mice (ii) demonstrates that BMDCs differentiate to form IBA1+ (A), MAC3+ (B), and CD11b+ (C) cells, as evident by either pink or orange merged cells, highlighted by arrowheads. Xenograft sections: GFP+ GBM tumor, green; RFP+BMDC, red; IF stain, blue; nuclei, cyan. Transgenic sections: RFP+BMDC, red; stain, green; nuclei, blue. Merged images correspond to white dotted box areas at higher magnification. D, quantification of recruited BMDCs on IF images shows that 80% stain for IBA1+ and inflammatory markers (MAC3+, CD11b+) in GBMs with no significant change in the differentiation pattern of BMDCs following radiotherapy (3 × 2 Gy). Data represent mean ± SEM, n = 10. E, majority of microglial phenotype (IBA1+; 80%) seen in GBM microenvironment (either xenograft or RasB8) are derived from bone marrow and not resident brain microglia (i) represented by purple colocalization of red BMDCs and blue IBA1+ stain, as compared with host-derived blue microglia. *, P = 0.0274; Fisher exact test. Data represent mean ± SEM, n = 10. F, ex vivo dual IF costaining of CD11b+ and TIE2+ identifies a small subpopulation of BMDCs in the perivascular niche that shows positive expression of both markers (arrowheads). GFP+ BMDCs, green; IF stain 1, blue; IF stain 2, red. Scale bars, μm.

Figure 3.

Characterization of recruited BMDCs in GBMs. Ex vivo IF staining of U87 GBM intracranial xenograft (i) and transgenic RASB8 mice (ii) demonstrates that BMDCs differentiate to form IBA1+ (A), MAC3+ (B), and CD11b+ (C) cells, as evident by either pink or orange merged cells, highlighted by arrowheads. Xenograft sections: GFP+ GBM tumor, green; RFP+BMDC, red; IF stain, blue; nuclei, cyan. Transgenic sections: RFP+BMDC, red; stain, green; nuclei, blue. Merged images correspond to white dotted box areas at higher magnification. D, quantification of recruited BMDCs on IF images shows that 80% stain for IBA1+ and inflammatory markers (MAC3+, CD11b+) in GBMs with no significant change in the differentiation pattern of BMDCs following radiotherapy (3 × 2 Gy). Data represent mean ± SEM, n = 10. E, majority of microglial phenotype (IBA1+; 80%) seen in GBM microenvironment (either xenograft or RasB8) are derived from bone marrow and not resident brain microglia (i) represented by purple colocalization of red BMDCs and blue IBA1+ stain, as compared with host-derived blue microglia. *, P = 0.0274; Fisher exact test. Data represent mean ± SEM, n = 10. F, ex vivo dual IF costaining of CD11b+ and TIE2+ identifies a small subpopulation of BMDCs in the perivascular niche that shows positive expression of both markers (arrowheads). GFP+ BMDCs, green; IF stain 1, blue; IF stain 2, red. Scale bars, μm.

Close modal

Role of bone marrow–derived inflammatory cells

We explored the potential role of the bone marrow–derived macrophages (MAC3+, CD11b+) and microglial/inflammatory cells (IBA1+) as scavengers in GBMs through analysis of Z-stack images of bone marrow–derived inflammatory cells that were engulfing GBM tumor cells (Supplementary Fig. S2A–S2F). Approximately 10% of the three cell types are involved in scavenging tumor cells, with the percentage of scavenging MAC3+ and CD11b+ cells significantly increasing following radiotherapy (Supplementary Fig. S2G). There was no evidence of BMDCs scavenging endothelial cells. All bone marrow–derived cells that are seen scavenging GBM tumor cells are seen only in the GBM microenvironment and not in perivascular niches, suggesting that pBMDCs have a distinct vascular supportive role. Previous published data support the described phagocytic activity of microglia and immune cells in GBMs. Hussain and colleagues showed that purified glioma-infiltrating microglia have the ability to phagocytose and play a cytotoxic role in the microenvironment (24). In vitro assays have demonstrated the ability of glioma cells to define the response of many immune cells and their ability to phagocytose glioma cells (25–27).

Contribution of BMDCs to GBM vasculature

Using in vivo 2PLM analysis complemented with IHC and IF, we sought to definitively establish whether pBMDCs differentiate to form endothelial cells (CD31 and factor-VIII/vWF) or smooth muscle cells (αSMA, Ang1, PDGFRβ). Intravenous injection of CD31+ antibody allowed specific in vivo identification of mature vessel endothelial cells on 2PLM images. In-depth analysis of pBMDCs and their relationship to the vessel wall and specifically the endothelial cells is possible through use of time lapse footage (Supplementary Video S1) and z-stack imaging (Supplementary Video S2). Dextran was used to highlight vascular channels that had no mature CD31+ endothelial cells but carried a functional blood flow. 2PLM imaging does not demonstrate any pBMDCs differentiating to form endothelial cells, nor do they line the vascular walls or form other potential vascular channels at any stage of GBM growth (Fig. 4A). Similarly, ex vivo IF staining with CD31 and vWF showed no evidence of differentiation or colocalization of pBMDCs with CD31+-positive cells (Fig. 4B, i). A recent study by Cheng and colleagues suggests that due to limited supply of resident pericytes in the brain, GBMs modulate support of the tumor vasculature through differentiation of GSCs into pericytes (28). We see a small percentage of pBMDCs differentiating to form pericytes through colocalization with SMA staining, although we show that the majority of vascular pericytes are not differentiated from BMDCs (Fig. 4B, ii). We examined whether association of pBMDCs varied on the basis of tumor vessel size and similarly association with pericytes; however, no difference was seen. The majority of pBMDCs show macrophage and microglial differentiation (Fig. 4B, iii). This similar pattern is evident in the transgenic mouse model RASB8 (Fig. 4C).

Figure 4.

Characterization of perivascular BMDCs in GBMs. A, in vivo 2PLM z-stack images very clearly demonstrate that there is no colocalization of BMDCs with endothelial cells in intracranial xenograft models of GBMs. GFP+BMDC, green; vascular EC CD31+, blue). B, i, IF analysis at day 21 post-intracranial shows no evidence of colocalization of BMDCs with CD31+ ECs. ii, ex vivo IF staining of perivascular smooth-muscle actin (SMA+) shows minimal BMDC colocalization in GBM xenograft (arrowheads). iii, ex vivo IF staining of IBA1+ demonstrates that microglia derived from BMDCs remain in the perivascular niche (GFP+GBM tumor cells, green; RFP+ BMDC, red; IF stain, blue; nuclei, cyan). C, a similar staining pattern was seen in the transgenic RASB8 GBMs where there was no colocalization of BMDCs with CD31+ cells (i) and minimal colocalization with SMA (arrowheads; ii). RFP+BMDC, red; IF stain green; nuclei, blue. Scale bars, μm.

Figure 4.

Characterization of perivascular BMDCs in GBMs. A, in vivo 2PLM z-stack images very clearly demonstrate that there is no colocalization of BMDCs with endothelial cells in intracranial xenograft models of GBMs. GFP+BMDC, green; vascular EC CD31+, blue). B, i, IF analysis at day 21 post-intracranial shows no evidence of colocalization of BMDCs with CD31+ ECs. ii, ex vivo IF staining of perivascular smooth-muscle actin (SMA+) shows minimal BMDC colocalization in GBM xenograft (arrowheads). iii, ex vivo IF staining of IBA1+ demonstrates that microglia derived from BMDCs remain in the perivascular niche (GFP+GBM tumor cells, green; RFP+ BMDC, red; IF stain, blue; nuclei, cyan). C, a similar staining pattern was seen in the transgenic RASB8 GBMs where there was no colocalization of BMDCs with CD31+ cells (i) and minimal colocalization with SMA (arrowheads; ii). RFP+BMDC, red; IF stain green; nuclei, blue. Scale bars, μm.

Close modal

Tumor region–dependent association of BMDCs to GBM vasculature

BMDCs demonstrated a distinct regional pattern of association with the vasculature, where peripheral and central vessels show a differential association of pBMDCs. We defined the periphery of the tumor as the region between tumor and normal brain within 300 μm of nonneoplastic brain in both xenografts and transgenic models of GBMs (Figs. 5A and 1D). We quantified the extent of pBMDCs associated with the vasculature in the center versus periphery of GBMs at early stages (<7 days) when tumor volume, based on volumetric MRI analysis, is less than 1 mm3 and at late (>14 days) stages of tumor growth when tumor volume exceeds 1 mm3. At early stages of GBM growth, pBMDCs were detected in 60% to 80% of the vasculature throughout all regions of the tumor, with no distinguishable difference between center and periphery or small and large vessels (Fig. 5B, i, ii, and C), whereas in late stages of tumor growth, the vessels in center have less than 10% association with pBMDCs (Fig. 5B, iii and C). In late stages of tumor growth, peripheral vessels continued to show approximately 80% to 90% association of pBMDCs (Fig. 5B, iv and C). This similar regional association of pBMDCs was evident in the RASB8 model (Fig. 5D, i and ii).

Figure 5.

Region-dependent localization of perivascular BMDCs. A, ex vivo imaging of U87 GBM intracranial xenograft sections stained CD31 (i) and hematoxylin identifies the tumor. ii, the overlay image of the tumor and red fluorescent BMDC. iii, IF images demonstrate red fluorescent BMDCs recruited specifically to the site of tumor growth. White inner circle defines the central core of the tumor versus the periphery defined as the next 300 μm (CD31+, brown; nuclei, blue; RFP+BMDCs, red). iv, higher magnification demonstrates that RFP+BMDCs do not colocalize with CD31+ hollow lumen vessel identified by brown stain on IHC (arrowhead) in the central core region of the tumor. B, ex vivo IF images of U87 GBM xenografts demonstrate BMDCs associate with vasculature in both the tumor center (i) and periphery (ii) equally at early stages of tumor growth (<1 week). While at later stages of GBM growth (beyond 14 days), the vessels in the center have few to no associated BMDCs (iii), whereas the vessels in tumor periphery continue to have localization of pBMDCs throughout all stages of tumor growth (iv; GFP+ BMDC, green; CD31+, red). C, quantification demonstrates the statistically significant difference between central vessels and peripheral vessels at late-stage GBM growth (***, P = 0.0002). Data represent mean ± SEM, n = 10. D, a similar regional pattern of localization of pBMDCs can be seen in the RASB8 GBM where central vessels (i) have minimal localization of pBMDC compared with peripheral vessels (ii) that retain localization of pBMDCs throughout all stages of tumor growth. (Colors reversed from previous U87 model; CD31+, green; RFP+BMDC, red; nuclei, blue). E, results of real-time PCR performed on vessels isolated using LCM from the center versus the periphery of U87 GBM intracranial xenografts demonstrate a differential expression of human (i) and mouse (ii) angiogenic factors. At late stages of GBM growth (>14 days), human Ang2 decreases in both the center (*, P = 0.0140) and the periphery (***, P < 0.0001), whereas human Ang1 increases (*, P = 0.0401). Mouse Ang2 is increased in peripheral vessels at late stages of GBM growth (*, P = 0.0212). Mouse Ang1 is undetermined and presumed low in all samples. Data represent mean ± SEM, n = 5, triplicate. Scale bars, μm.

Figure 5.

Region-dependent localization of perivascular BMDCs. A, ex vivo imaging of U87 GBM intracranial xenograft sections stained CD31 (i) and hematoxylin identifies the tumor. ii, the overlay image of the tumor and red fluorescent BMDC. iii, IF images demonstrate red fluorescent BMDCs recruited specifically to the site of tumor growth. White inner circle defines the central core of the tumor versus the periphery defined as the next 300 μm (CD31+, brown; nuclei, blue; RFP+BMDCs, red). iv, higher magnification demonstrates that RFP+BMDCs do not colocalize with CD31+ hollow lumen vessel identified by brown stain on IHC (arrowhead) in the central core region of the tumor. B, ex vivo IF images of U87 GBM xenografts demonstrate BMDCs associate with vasculature in both the tumor center (i) and periphery (ii) equally at early stages of tumor growth (<1 week). While at later stages of GBM growth (beyond 14 days), the vessels in the center have few to no associated BMDCs (iii), whereas the vessels in tumor periphery continue to have localization of pBMDCs throughout all stages of tumor growth (iv; GFP+ BMDC, green; CD31+, red). C, quantification demonstrates the statistically significant difference between central vessels and peripheral vessels at late-stage GBM growth (***, P = 0.0002). Data represent mean ± SEM, n = 10. D, a similar regional pattern of localization of pBMDCs can be seen in the RASB8 GBM where central vessels (i) have minimal localization of pBMDC compared with peripheral vessels (ii) that retain localization of pBMDCs throughout all stages of tumor growth. (Colors reversed from previous U87 model; CD31+, green; RFP+BMDC, red; nuclei, blue). E, results of real-time PCR performed on vessels isolated using LCM from the center versus the periphery of U87 GBM intracranial xenografts demonstrate a differential expression of human (i) and mouse (ii) angiogenic factors. At late stages of GBM growth (>14 days), human Ang2 decreases in both the center (*, P = 0.0140) and the periphery (***, P < 0.0001), whereas human Ang1 increases (*, P = 0.0401). Mouse Ang2 is increased in peripheral vessels at late stages of GBM growth (*, P = 0.0212). Mouse Ang1 is undetermined and presumed low in all samples. Data represent mean ± SEM, n = 5, triplicate. Scale bars, μm.

Close modal

Given the distinct region-dependent pattern of association of pBMDCs, we aimed to establish the quantitative PCR expression profile of key angiogenic factors of the vasculature, including the perivascular niche, isolated by LCM from the center and periphery of both early- and late-stage GBMs, reporting only significant changes in each profile. Mouse Ang1 was the only factor to remain undetermined, low, in all samples when compared with the control (Fig. 5E). Human Ang2 decreased significantly in both central and peripheral vasculature in response to tumor growth (Fig. 5E, i), whereas mouse Ang2 increased significantly in only the peripheral vessels (Fig. 5E, ii). Human Tie2 did not change in response to tumor growth (Fig. 5E, i), a result that was expected as Tie2 is considered to be endothelial cell-specific and derived from host (mouse) vasculature. Similarly, there was a significant increase in mouse Tie2 in peripheral vessels in response to GBM growth (Fig. 5E, ii).

Region-dependent association of BMDCs in response to chemotherapy and radiotherapy

In response to temozolomide, the region-dependent association of pBMDCs was not changed compared with untreated tumors. However, following radiotherapy, there was an influx of pBMDCs to the central vessels. There remained no evidence of pBMDC differentiation into endothelial cells or de novo vessel formation in response to radiotherapy (Fig. 6A). Following radiotherapy, there is a statistically significant increase in number of central vessels that had associated pBMDCs (Fig. 6A, iii and B). The loss of regional association of pBMDCs cannot be simply explained by a decrease in GBM tumor volume as a consequence of radiotherapy as the effect is seen immediately upon radiotherapy treatment and sustained until GBM reoccurrence.

Figure 6.

Loss of region-dependent localization of BMDCs following radiotherapy. A, ex vivo IF images of GBM intracranial xenografts treated with radiotherapy (3 × 2 Gy) show a loss in distinct regional localization of pBMDCs in GBM vasculature compared with control untreated xenografts. Untreated late tumors demonstrate minimal BMDC recruitment to central vessels (i), while periphery recruitment is high (ii). Following radiotherapy (3 × 2 Gy), pBMDCs are seen in the tumor center and are evident in the perivascular niche (iii) as well as in the tumor periphery (iv; GFP+ BMDC, green; CD31+, red). B, quantification of pBMDC localization in central and peripheral vasculature confirms a significant increase in region-dependent association of pBMDCs in response to radiotherapy (3 × 2 Gy) within the center (**, P = 0.001) and the periphery (*, P = 0.0045). Data represent mean ± SEM, n = 10. C, differential expression pattern of human (i) and mouse (ii) angiogenic factors in both central and peripheral vessels of untreated and radiotherapy-treated (3 × 2 Gy) U87 GBMs. Radiotherapy significantly decreases both human and mouse Tie2 in peripheral vessels (*, P = 0.0528; **, P = 0.0049). There is an increase in mouse Ang2 in central vessels with radiotherapy (*, P = 0.010). Mouse Ang1 is undetermined and presumed low in all samples. Data represent mean ± SEM, n = 5, in triplicate. D, real-time PCR data confirm an increase in mouse Ang2 in response to radiotherapy (6 Gy). E, Western blot analysis also shows an increase in mouse Ang2 expression by BMDCs with exposure to radiotherapy (6 Gy). F, ex vivo IF analysis also verifies an upregulation of Ang2 expression seen predominantly on tumor vasculature in response to radiotherapy. The increase in Ang2 is associated with an increased colocalization of Ang2 to pBMDCs (white arrowhead; GFP+ BMDC, green; Ang2+, red). G, quantification of the percentage of vessels positive for Ang2 expression demonstrates a significant increase in expression in late-stage GBM growth (***, P < 0.0001) and similarly with addition of radiotherapy (*, P = 0.0134). Data represent mean ± SEM, n = 10. Scale bars, μm.

Figure 6.

Loss of region-dependent localization of BMDCs following radiotherapy. A, ex vivo IF images of GBM intracranial xenografts treated with radiotherapy (3 × 2 Gy) show a loss in distinct regional localization of pBMDCs in GBM vasculature compared with control untreated xenografts. Untreated late tumors demonstrate minimal BMDC recruitment to central vessels (i), while periphery recruitment is high (ii). Following radiotherapy (3 × 2 Gy), pBMDCs are seen in the tumor center and are evident in the perivascular niche (iii) as well as in the tumor periphery (iv; GFP+ BMDC, green; CD31+, red). B, quantification of pBMDC localization in central and peripheral vasculature confirms a significant increase in region-dependent association of pBMDCs in response to radiotherapy (3 × 2 Gy) within the center (**, P = 0.001) and the periphery (*, P = 0.0045). Data represent mean ± SEM, n = 10. C, differential expression pattern of human (i) and mouse (ii) angiogenic factors in both central and peripheral vessels of untreated and radiotherapy-treated (3 × 2 Gy) U87 GBMs. Radiotherapy significantly decreases both human and mouse Tie2 in peripheral vessels (*, P = 0.0528; **, P = 0.0049). There is an increase in mouse Ang2 in central vessels with radiotherapy (*, P = 0.010). Mouse Ang1 is undetermined and presumed low in all samples. Data represent mean ± SEM, n = 5, in triplicate. D, real-time PCR data confirm an increase in mouse Ang2 in response to radiotherapy (6 Gy). E, Western blot analysis also shows an increase in mouse Ang2 expression by BMDCs with exposure to radiotherapy (6 Gy). F, ex vivo IF analysis also verifies an upregulation of Ang2 expression seen predominantly on tumor vasculature in response to radiotherapy. The increase in Ang2 is associated with an increased colocalization of Ang2 to pBMDCs (white arrowhead; GFP+ BMDC, green; Ang2+, red). G, quantification of the percentage of vessels positive for Ang2 expression demonstrates a significant increase in expression in late-stage GBM growth (***, P < 0.0001) and similarly with addition of radiotherapy (*, P = 0.0134). Data represent mean ± SEM, n = 10. Scale bars, μm.

Close modal

In response to radiotherapy, there was a pronounced change in expression profile of angiogenic factors in central and peripheral vasculature as determined by real-time PCR. There is an increase in mouse Ang2 in central vessels with radiotherapy (Fig. 6C, ii) and a significant decrease in both human and mouse Tie2 in peripheral vessels (Fig. 6C, i and ii). PCR of angiogenic factors on systemic bone marrow shows an increase in expression of mouse Ang2 but not Ang1 in response to radiotherapy (Fig. 6D). Western blot analysis confirmed an increase in Ang2 but not Ang1 protein expression by bone marrow in response to radiotherapy (Fig. 6E). IF analysis confirmed an upregulation of Ang2 protein expression, seen predominantly in tumor vasculature and colocalized to pBMDCs, in response to radiotherapy (Fig. 6F and G).

Interplay of VEGF and ANG2 regulates the regional association of pBMDCs

To establish the molecular mechanism responsible for regulating the regional association of pBMDCs, we investigated the role of VEGF. VEGF is known to be a pivotal angiogenic factor in regulating GBM neovascularization. We inhibited VEGF using doxycycline-inducible expression of VEGFTrap (29, 30) in U87 or GSC GBM intracranial xenografts. The blockade of VEGF through VEGFTrap activation leads to an increase in invasiveness of the cells at the tumor border (Fig. 7A), a characteristic previously described by Paez-Ribes and colleagues (31). In addition, the loss of VEGF was able to decrease the microvascular density and increase vessel diameter similar to that previously seen by Chae and colleagues (32) (Fig. 7B–D). VEGF inhibition did not change the levels of recruitment and migration of BMDCs to GBMs but did result in a clear loss of the region-dependent association of pBMDCs mirroring the effect of radiotherapy (Fig. 7E and F). Discontinuation of VEGFTrap resulted in tumor recurrence and restoration of the regional association of pBMDCs and loss of pBMDCs in the central vasculature in GBM xenografts. These results implicate VEGF as a key molecular regulator in triggering the regional association of pBMDCs in GBM vasculature.

Figure 7.

Molecular regulator of regional localization of perivascular BMDCs. A, hematoxylin and eosin staining demonstrating the infiltrative edge of the tumor being more invasive in the VEGFTrap-treated tumor (i) compared with the U87 control tumor (ii). B, i, CD31+ IHC staining of ex vivo normal brain VEGFTrap-treated U87 GBM intracranial xenograft (GBM-VEGF); ii, illustrates the tumor vasculature. CD31+, brown; nuclei, blue). C, quantification of vessel diameter shows a significant dilation of vessels following treatment with VEGF inhibition and radiotherapy (***, P < 0.0001). Data represent mean ± SEM, n = 15. D, quantification of microvascular density shows a significant reduction following treatment with VEGF inhibition and radiotherapy (***, P < 0.0001). Data represent mean ± SEM, n = 15. E, ex vivo IF CD31+ staining demonstrates a loss of region-dependent association of pBMDCs with VEGF inhibition; central and peripheral have equal association of pBMDCs (i and ii). Combination of radiotherapy and VEGF continues to show a region-independent association of pBMDCs, with equal numbers seen in center versus periphery (iii and iv). GFP+ GBM tumor, green; RFP+ BMDC, red; vascular CD31+, blue). F, quantification of vessels that have association of pBMDCs confirms the loss of region-dependent association of pBMDCs to GBM vasculature (**, P = 0.0043). Addition of radiotherapy does not exceed the effects seen with VEGF inhibition (**, P = 0.0039). Data represent mean ± SEM, n = 10. G, differential expression of human (i) and mouse genes (ii) in both central and peripheral vasculature shows that VEGF inhibition causes a significant increase in Human Ang2 and Ang1 in central vessels (*, P = 0.0489; *, P = 0.0102) as well as human Ang1 in peripheral vasculature (*, P = 0.0136). With the addition of radiotherapy to VEGF inhibition, the same pattern is seen with an increase in central human Ang2 and Ang1 (*, P = 0.02456; **, P = 0.006) and peripheral Ang1 (**, P = 0.0075). Mouse Ang2 is decreased in the peripheral vessels following VEGFTrap and radiotherapy treatment (*, P = 0.0118). Mouse Ang1 is undetermined and presumed low in all samples. Data represent mean ± SEM, n = 5, in triplicate. Scale bars, μm. RTx, radiotherapy.

Figure 7.

Molecular regulator of regional localization of perivascular BMDCs. A, hematoxylin and eosin staining demonstrating the infiltrative edge of the tumor being more invasive in the VEGFTrap-treated tumor (i) compared with the U87 control tumor (ii). B, i, CD31+ IHC staining of ex vivo normal brain VEGFTrap-treated U87 GBM intracranial xenograft (GBM-VEGF); ii, illustrates the tumor vasculature. CD31+, brown; nuclei, blue). C, quantification of vessel diameter shows a significant dilation of vessels following treatment with VEGF inhibition and radiotherapy (***, P < 0.0001). Data represent mean ± SEM, n = 15. D, quantification of microvascular density shows a significant reduction following treatment with VEGF inhibition and radiotherapy (***, P < 0.0001). Data represent mean ± SEM, n = 15. E, ex vivo IF CD31+ staining demonstrates a loss of region-dependent association of pBMDCs with VEGF inhibition; central and peripheral have equal association of pBMDCs (i and ii). Combination of radiotherapy and VEGF continues to show a region-independent association of pBMDCs, with equal numbers seen in center versus periphery (iii and iv). GFP+ GBM tumor, green; RFP+ BMDC, red; vascular CD31+, blue). F, quantification of vessels that have association of pBMDCs confirms the loss of region-dependent association of pBMDCs to GBM vasculature (**, P = 0.0043). Addition of radiotherapy does not exceed the effects seen with VEGF inhibition (**, P = 0.0039). Data represent mean ± SEM, n = 10. G, differential expression of human (i) and mouse genes (ii) in both central and peripheral vasculature shows that VEGF inhibition causes a significant increase in Human Ang2 and Ang1 in central vessels (*, P = 0.0489; *, P = 0.0102) as well as human Ang1 in peripheral vasculature (*, P = 0.0136). With the addition of radiotherapy to VEGF inhibition, the same pattern is seen with an increase in central human Ang2 and Ang1 (*, P = 0.02456; **, P = 0.006) and peripheral Ang1 (**, P = 0.0075). Mouse Ang2 is decreased in the peripheral vessels following VEGFTrap and radiotherapy treatment (*, P = 0.0118). Mouse Ang1 is undetermined and presumed low in all samples. Data represent mean ± SEM, n = 5, in triplicate. Scale bars, μm. RTx, radiotherapy.

Close modal

In response to VEGF inhibition, there was a significant increase in human Ang2 and Ang1 in central vasculature (Fig. 7G, i) together with elevated mouse Tie2 (Fig. 7G, ii). The addition of radiotherapy concomitantly with VEGF inhibition did not enhance the response or angiogenic profile beyond VEGF inhibition alone.

Following addition of AMG386, a therapeutically potent ANG1/2 inhibitor, the influx of pBMDCs to central vasculature seen following radiotherapy or VEGFTrap is lost. In tumors, treated AMG386 and radiotherapy or AMG386 and VEGFTrap pBMDC are seen associated with the peripheral vessels, whereas there is minimal pBMDC seen in the central vessels. These findings lend further support to the proposal that a compensatory upregulation of ANG2 following loss of VEGF (either indirectly through radiotherapy or directly through VEGF inhibition) is a potential driver for pBMDC recruitment following therapy (Supplementary Fig. S3).

Emerging evidence suggests that de novo vessel formation or vasculogenesis can occur in adult life, and it has been argued to provide a potential mechanism for cancer neovascularization, through mobilization of endothelial progenitor cells from the bone marrow or circulation (12, 33–36). This hypothesis remains highly controversial with debate as to whether BMDCs actually differentiate to directly form endothelial cells or instead contribute a more supportive role to neovascularization in a paracrine manner (10, 21).

In this study, we take advantage of real-time in vivo 2PLM imaging to demonstrate that BMDCs do not differentiate to form endothelial cells at any stage of GBM growth, regardless of tumor region, stage of tumor growth, or model investigated. Moreover, we see no evidence of true vasculogenesis occurring by BMDCs at any stage of GBM growth; specifically, no bone marrow–derived vessels or vascular channels were evident in the GBM models. Our results are in keeping with previous studies that have also reported that BMDCs do not incorporate into the tumor vasculature as endothelial cells (21, 37). In our study, the majority of BMDCs remain on the outer lumen of vascular channels throughout all stages of GBM growth in a perivascular niche.

We identified a very distinct region-dependent association of pBMDCs with GBM vasculature in response to tumor growth. At early stages of GBM formation, there is no regional variation in association of pBMDCs with the vasculature, with equal distribution of pBMDCs seen in peripheral and central vessels. However, at later stages of tumor growth, pBMDCs remain intimately associated with the peripheral vasculature and show minimal to no association with central vasculature. The region-dependent localization of pBMDCs suggests that beyond a critical tumor volume, GBMs rely on distinct molecular mechanisms of neovascularization in peripheral versus central regions of GBMs.

It is established that host vessel co-option is the principle mechanism of GBM vascularization at early stages of GBM growth. At this stage of GBM formation, Ang2 is known to be the earliest angiogenic factor to be upregulated (1). Ang2 has previously been shown to promote recruitment of a specific CD11b+ subtype of BMDCs (37–39). Here, we show that at early stages of GBMs, growth there is a significant recruitment of pBMDCs to GBM vasculature that continues throughout all stages of GBM growth in the peripheral vessels. Associated with this, there is an upregulation of Ang2 by the host vasculature, as the source of Ang2 is mouse and not human, suggesting that elevated levels of Ang2 are created as a consequence of host vessel destabilization that is needed for vessel co-option to take place and this most likely involves pBMDCs as suggested by their constant presence in the peripheral GBM vessels.

VEGF is recognized as the central angiogenic factor that regulates neovascularization, through angiogenesis or branching of pre-existing vessels in cancer (1, 17, 40, 41). GBMs have one of the highest levels of VEGF expression of all cancer cells, in particular in pseudopalisading proliferative cells (6, 42), thereby creating a VEGF concentration gradient, highest in central tumor regions and dissipating through peripheral regions and the adjacent normal brain (43, 44). At late stages of GBM growth, we show that there is minimal association of pBMDCs. Along with this observation, we find that there is minimal Ang2 or Tie2 upregulation in central vessels. Taken together, these results suggest that with high levels of VEGF, the central vasculature in GBMs does not depend on pBMDCs for tumor neovascularization and similarly the Ang2/Tie2 pathway is not involved in maintaining the GBM central vasculature. However, with inhibition of VEGF, there is a distinct loss of the region-dependent association of pBMDCs as evidenced by association of pBMDCs in central vasculature of VEGFTrap-treated tumors. In addition, with loss of central VEGF, there is an upregulation of Ang2/Tie2 in the central GBM vasculature, which supports a central role for Ang2 in signaling recruitment of pBMDCs. Furthermore, we show that bone marrow cells themselves provide a source of Ang2, potentially to ensure that a continuous and sufficient supply of pBMDCs is available to promote neovascularization and protection of newly formed tumor vasculature.

Previous studies show that there is an influx of BMDCs into GBMs following radiotherapy to facilitate new vessel formation, thereby providing a mechanism through which GBMs can revascularize (21, 37). In this study, we also see an influx of BMDCs, and furthermore, we demonstrate that there is a clear loss of the regional association of pBMDCs in GBM vasculature following radiotherapy. We propose that following radiotherapy, with the loss of central GBM tumor cells, there is a loss of the primary source of VEGF and overall decrease in VEGF production in the tumor center (Supplementary Fig. S4). In turn, we find that radiotherapy results in an increase in Ang2 and Tie2 in the central vessels, compared with untreated GBMs. Once again, it appears that the upregulation of Ang2 signals recruitment of pBMDCs to support, remodel, and restore the vulnerable central vasculature, following radiotherapy. Concurrent with elevated Ang2, we see no increase in Ang1 following radiotherapy, either on the central vasculature or by the systemic BM. Ang1 is known to provide a stabilizing force on the vasculature, preventing vascular remodeling. Absence of Ang1 upregulation is necessary to allow Ang2-induced vascular remodeling of central vessels to occur following radiotherapy.

Recruitment of pBMDCs is associated with an upregulation of ANG2 that is seen with VEGF inhibition, either directly with use of VEGFTrap or indirectly following radiotherapy. We show that blocking ANG2 in turn prevents influx of pBMDC into the central vessels following therapy, lending support to the proposal that an interplay between ANG2 and VEGF is central to the regional association of pBMDC.

We established that approximately 80% of BMDCs in the GBM microenvironment are inflammatory cells: macrophages and microglia. The majority of inflammatory cell population in GBM are bone marrow–derived and not resident brain cells, suggesting that the normal resident brain inflammatory response is not sufficient to meet the demands, of either host or cancer cell, to respond to GBM growth. At any given stage of GBM growth, only 10% to 20% of BMDC in the tumor microenvironment, and not in perivascular niche, provide a scavenging role as activated anti-inflammatory cells directed against the cancer cells. The remaining BMDCs appear to provide alternate roles, potentially tumor-protective roles, that is driven by cancer cells to aid in their progression, and given their perivascular localization, possibly tumor vascular protective roles. The region-specific contribution of pBMDCs suggests that their role varies with different stages of tumor growth and in response to therapy. The literature supports a dual role for macrophages in cancer.

In a recent study, Pyonteck and colleagues show that inhibition of CSF-1 increases survival and GBM regression, however, not through depletion of tumor-associated macrophages (TAM) but through ‘re-education’ of TAM to M1-like macrophages that display increased phagocytic properties. This switch is thought to be cancer cell driven through expression of granulocyte macrophage colony-stimulating factor (GM-CSF) and IFNγ (45). In addition, most recently, Casazza and colleagues have shown that Neuropilin-1 (Nrp1) guides TAMs to hypoxic areas to elicit proangiogenic and immunosuppressive functions (46). Our results placed in the context of these recent studies would suggest a better understanding of how these factors regulate the regional association of pBMDCs and are important in being able to design effective targeted treatments concurrent with current standard therapy.

In conclusion, we demonstrate a distinct region-dependent association of pBMDCs in GBM vasculature. We established that BMDCs do not directly contribute to new vessel formation either by differentiating into endothelial cells or creating vascular channels in GBMs. Rather, pBMDCs provide a vascular supportive role, either contributing to vessel co-option in peripheral regions or through facilitating restoration of vasculature following radiotherapy and antiangiogenic therapies (VEGF inhibition). In central GBMs, where there is abundance of proliferative and viable GBM tumor cells and high levels of VEGF, pBMDCs are not needed to support the angiogenic process that is required for tumor neovascularization. However, loss of VEGF, either through antiangiogenic therapies or radiotherapy, results in vulnerable vasculature in the central region, triggering an upregulation in Ang2, which in turn triggers recruitment of pBMDCs to create a supportive vascular environment that allows restoration of the vulnerable GBM vasculature and ultimately tumor survival. Confirming the central role of Ang2 in regional localization of pBMDCs is our data that show that inhibition of Ang2, following loss of VEGF, prevents influx of pBMDCs in the central vasculature. Therefore, targeting influx of BMDCs, particularly pBMDCs, concurrent with radiotherapy and VEGF inhibition, can potentially offer a novel strategy for preventing GBM recurrence following therapy.

No potential conflicts of interest were disclosed.

Conception and design: K. Burrell, R.P. Hill, G. Zadeh

Development of methodology: K. Burrell, S. Singh, R.P. Hill, G. Zadeh

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Burrell, S. Jalali, G. Zadeh

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Burrell, S. Singh, R.P. Hill, G. Zadeh

Writing, review, and/or revision of the manuscript: K. Burrell, S. Singh, R.P. Hill, G. Zadeh

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Burrell, S. Jalali, G. Zadeh

Study supervision: K. Burrell, G. Zadeh

The authors thank the Advanced Optical Microscopy Facility and the Spatio-Temporal Targeting and Amplification of Radiation Response (STTARR) program. They also thank Drs. Peter Tonge, Iacovos Michael, and Andras Nagy for plasmids, fluorescent mice, and for feedback on the article. Dr. Agnihotri provided graphical assistance and scientific input for the article. They thank Dr. Frederick Lang for GSC 8-18 used in this project. The continued support from the Brain Tumor Research Center (BTRC) has been invaluable, and the authors would like to commemorate Dr. Abhijit Guha for his scientific input.

All work was funded by Canadian Institute of Health Research (CIHR) grant MOP-119552.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Fischer
I
,
Gagner
JP
,
Law
M
,
Newcomb
EW
,
Zagzag
D
. 
Angiogenesis in gliomas: biology and molecular pathophysiology
.
Brain Pathol
2005
;
15
:
297
310
.
2.
Louis
DN
,
Holland
EC
,
Cairncross
JG
. 
Glioma classification: a molecular reappraisal
.
Am J Pathol
2001
;
159
:
779
86
.
3.
Hardee
ME
,
Zagzag
D
. 
Mechanisms of glioma-associated neovascularization
.
Am J Pathol
2012
;
181
:
1126
41
.
4.
Brat
DJ
,
Van Meir
EG
. 
Glomeruloid microvascular proliferation orchestrated by VPF/VEGF: a new world of angiogenesis research
.
Am J Pathol
2001
;
158
:
789
96
.
5.
Plate
KH
,
Breier
G
,
Risau
W
. 
Molecular mechanisms of developmental and tumor angiogenesis
.
Brain Pathol
1994
;
4
:
207
18
.
6.
Bergers
G
,
Benjamin
LE
. 
Tumorigenesis and the angiogenic switch
.
Nat Rev Cancer
2003
;
3
:
401
10
.
7.
Jain
RK
,
Carmeliet
PF
. 
Vessels of death or life
.
Sci Am
2001
;
285
:
38
45
.
8.
Grivennikov
SI
,
Greten
FR
,
Karin
M
. 
Immunity, inflammation, and cancer
.
Cell
2010
;
140
:
883
99
.
9.
Kerbel
RS
. 
Tumor angiogenesis
.
N Engl J Med
2008
;
358
:
2039
49
.
10.
Kozin
SV
,
Duda
DG
,
Munn
LL
,
Jain
RK
. 
Is vasculogenesis crucial for the regrowth of irradiated tumours?
Nature
2011
;
11
:
532
.
11.
Machein
MR
,
Renninger
S
,
Lima-Hahn
E
,
Plate
KH
. 
Minor contribution of bone marrow-derived endothelial progenitors to the vascularization of murine gliomas
.
Brain Pathol
2006
;
13
:
582
97
.
12.
Aghi
M
,
Chiocca
EA
. 
Contribution of bone marrow-derived cells to blood vessels in ischemic tissues and tumors
.
Mol Ther
2005
;
12
:
994
1005
.
13.
Burrell
K
,
Hill
RP
,
Zadeh
G
. 
High-resolution in-vivo analysis of normal brain response to cranial irradiation
.
PLoS One
2012
;
7
:
e38366
.
14.
Zadeh
G
,
Koushan
K
,
Baoping
Q
,
Shannon
P
,
Guha
A
. 
Role of angiopoietin-2 in regulating growth and vascularity of astrocytomas
.
J Oncol
2010
;
2010
:
1
7
.
15.
Wei
J
,
Barr
J
,
Kong
LY
,
Wang
Y
,
Wu
A
,
Sharma
AK
, et al
Glioma-associated cancer-initiating cells induce immunosuppression
.
Clin Cancer Res
2010
;
16
:
461
73
.
16.
Burrell
K
,
Agnihotri
S
,
Leung
M
,
DaCosta
R
,
Hill
RP
,
Zadeh
G
. 
A novel high-resolution in vivo imaging technique to study the dynamic response of intracranial structures to tumor growth and therapeutics
.
J Vis Exp
2013
;
e50363
.
17.
Ding
H
,
Roncari
L
,
Shannon
P
,
Wu
X
,
Lau
N
,
Karaskova
J
, et al
Astrocyte-specific expression of activated p21-ras results in malignant astrocytoma formation in a transgenic mouse model of human gliomas
.
Cancer Res
2001
;
61
:
3826
36
.
18.
Chung
C
,
Jalali
S
,
Foltz
W
,
Burrell
K
,
Wildgoose
P
,
Lindsay
P
, et al
Imaging biomarker dynamics in an intracranial murine glioma study of radiation and antiangiogenic therapy
.
Int J Radiat Oncol Biol Phys
2013
;
85
:
805
12
.
19.
Ding
H
,
Shannon
P
,
Lau
N
,
Wu
X
,
Roncari
L
,
Baldwin
RL
, et al
Oligodendrogliomas result from the expression of an activated mutant epidermal growth factor receptor in a RAS transgenic mouse astrocytoma model
.
Cancer Res
2003
;
63
:
1106
13
.
20.
Shannon
P
,
Sabha
N
,
Lau
N
,
Kamnasaran
D
,
Gutmann
DH
,
Guha
A
. 
Pathological and molecular progression of astrocytomas in a GFAP:12 V-Ha-Ras mouse astrocytoma model
.
Am J Pathol
2005
;
167
:
859
67
.
21.
Kozin
SV
,
Kamoun
WS
,
Huang
Y
,
Dawson
MR
,
Jain
RK
,
Duda
DG
. 
Recruitment of myeloid but not endothelial precursor cells facilitates tumor regrowth after local irradiation
.
Cancer Res
2010
;
70
:
5679
85
.
22.
Kozin
SV
,
Duda
DG
,
Munn
LL
,
Jain
RK
. 
Neovascularization after irradiation: what is the source of newly formed vessels in recurring tumors?
J Natl Cancer Inst
2012
;
104
:
899
905
.
23.
De Palma
M
,
Venneri
MA
,
Galli
R
,
Sergi
LS
,
Politi
LS
,
Sampaolesi
M
, et al
Tie2 identifies a hematopoietic lineage of proangiogenic monocytes required for tumor vessel formation and a mesenchymal population of pericyte progenitors
.
Cancer Cell
2005
;
8
:
211
26
.
24.
Hussain
SF
,
Yang
D
,
Suki
D
,
Grimm
E
,
Heimberger
AB
. 
Innate immune functions of microglia isolated from human glioma patients
.
J Transl Med
2006
;
4
:
15
.
25.
Chang
GH
,
Barbaro
NM
,
Pieper
RO
. 
Phosphatidylserine-dependent phagocytosis of apoptotic glioma cells by normal human microglia, astrocytes, and glioma cells
.
Neuro Oncol
2000
;
2
:
174
83
.
26.
Wallace
PK
,
Romet-Lemonne
JL
,
Chokri
M
,
Kasper
LH
,
Fanger
MW
,
Fadul
CE
. 
Production of macrophage-activated killer cells for targeting of glioblastoma cells with bispecific antibody to FcgammaRI and the epidermal growth factor receptor
.
Cancer Immunol Immunother
2000
;
49
:
493
503
.
27.
Rodrigues
JC
,
Gonzalez
GC
,
Zhang
L
,
Ibrahim
G
,
Kelly
JJ
,
Gustafson
MP
, et al
Normal human monocytes exposed to glioma cells acquire myeloid-derived suppressor cell-like properties
.
Neuro Oncol
2010
;
12
:
351
65
.
28.
Cheng
L
,
Huang
Z
,
Zhou
W
,
Wu
Q
,
Donnola
S
,
Liu
JK
, et al
Glioblastoma stem cells generate vascular pericytes to support vessel function and tumor growth
.
Cell
2013
;
153
:
139
52
.
29.
Holash
J
,
Davis
S
,
Papadopoulos
N
,
Croll
SD
,
Ho
L
,
Russell
M
, et al
VEGF-Trap: a VEGF blocker with potent antitumor effects
.
Proc Natl Acad Sci U S A
2002
;
99
:
11393
8
.
30.
Rudge
JS
,
Holash
J
,
Hylton
D
,
Russell
M
,
Jiang
S
,
Leidich
R
, et al
VEGF Trap complex formation measures production rates of VEGF, providing a biomarker for predicting efficacious angiogenic blockade
.
Proc Natl Acad Sci U S A
2007
;
104
:
18363
70
.
31.
PAez-Ribes
M
,
Allen
E
,
Hudock
J
,
Takeda
T
,
Okuyama
H
,
ViNals
F
, et al
Antiangiogenic therapy elicits malignant progression of tumors to increased local invasion and distant metastasis
.
Cancer Cell
2009
;
15
:
220
31
.
32.
Chae
SS
,
Kamoun
WS
,
Farrar
CT
,
Kirkpatrick
ND
,
Niemeyer
E
,
de Graaf
AMA
, et al
Angiopoietin-2 interferes with anti-VEGFR2-induced vessel normalization and survival benefit in mice bearing gliomas
.
Clin Cancer Res
2010
;
16
:
3618
27
.
33.
Asahara
T
,
Murohara
T
,
Sullivan
A
,
Silver
M
,
van der Zee
R
,
Li
T
, et al
Isolation of putative progenitor endothelial cells for angiogenesis
.
Science
1997
;
275
:
964
7
.
34.
De Palma
M
,
Naldini
L
. 
Role of haematopoietic cells and endothelial progenitors in tumour angiogenesis
.
Biochim Biophys Acta
2006
;
1766
:
159
66
.
35.
Duda
DG
,
Cohen
KS
,
Kozin
SV
,
Perentes
JY
,
Fukumura
D
,
Scadden
DT
, et al
Evidence for incorporation of bone marrow-derived endothelial cells into perfused blood vessels in tumors
.
Blood
2006
;
107
:
2774
6
.
36.
Plate
KH
,
Scholz
A
,
Dumont
DJ
. 
Tumor angiogenesis and anti-angiogenic therapy in malignant gliomas revisited
.
Acta Neuropathol
2012
;
124
:
763
75
.
37.
Kioi
M
,
Vogel
H
,
Schultz
G
,
Hoffman
RM
,
Harsh
GR
,
Brown
JM
. 
Inhibition of vasculogenesis, but not angiogenesis, prevents the recurrence of glioblastoma after irradiation in mice
.
J Clin Invest
2010
;
120
:
694
705
.
38.
Ahn
G-O
,
Brown
JM
. 
Role of endothelial progenitors and other bone marrow-derived cells in the development of the tumor vasculature
.
Angiogenesis
2009
;
12
:
159
64
.
39.
Huang
H
,
Lai
JY
,
Do
J
,
Liu
D
,
Li
L
,
Del Rosario
J
, et al
Specifically targeting angiopoietin-2 inhibits angiogenesis, Tie2-expressing monocyte infiltration, and tumor growth
.
Clin Cancer Res
2011
;
17
:
1001
11
.
40.
Zadeh
G
,
Reti
R
,
Koushan
K
,
Baoping
Q
,
Shannon
P
,
Guha
A
. 
Regulation of the pathological vasculature of malignant astrocytomas by angiopoietin-1
.
Neoplasia
2005
;
7
:
1081
90
.
41.
Carmeliet
P
,
Jain
RK
. 
Molecular mechanisms and clinical applications of angiogenesis
.
Nature
2011
;
473
:
298
307
.
42.
Carmeliet
P
,
Jain
RK
. 
Principles and mechanisms of vessel normalization for cancer and other angiogenic diseases
.
Nat Rev Drug Discov
2011
;
10
:
417
27
.
43.
Plate
KH
,
Breier
G
,
Weich
HA
,
Risau
W
. 
Vascular endothelial growth factor is a potential tumour angiogenesis factor in human gliomas in vivo
.
Nature
1992
;
359
:
845
8
.
44.
Shweiki
D
,
Itin
A
,
Soffer
D
,
Keshet
E
. 
Vascular endothelial growth factor induced hypoxia may mediate hypoxia-initiated angiogenesis
.
Nature
1992
;
359
:
843
5
.
45.
Pyonteck
SM
,
Akkari
L
,
Schuhmacher
AJ
,
Bowman
RL
,
Sevenich
L
,
Quail
DF
, et al
CSF-1R inhibition alters macrophage polarization and blocks glioma progression
.
Nat Med
2013
;
19
:
1264
72
.
46.
Casazza
A
,
Laoui
D
,
Wenes
M
,
Rizzolio
S
,
Bassani
N
,
Mambretti
M
, et al
Impeding macrophage entry into hypoxic tumor areas by Sema3A/Nrp1 signaling blockade inhibits angiogenesis and restores antitumor immunity
.
Cancer Cell
2013
;
24
:
695
709
.

Supplementary data