DBTRG-05MG, U87, and U118 are human glioma cell lines originally purchased from American Type Culture Collection (ATCC, Manassas, VA). DBM2 is a subclone of DBTRG-05MG derived through lung metastases after mouse tail vein injection as described below. U251 cells were provided by Dr. Han-mo Koo of the Van Andel Research Institute. All cells were grown in Dulbecco's Modified Eagle's Medium (DMEM) (GibcoTM, Invitrogen Corporation, Carlsbad, CA) supplemented with 10% fetal bovine serum (FBS) (Invitrogen Corporation) and penicillin and streptomycin (Invitrogen Corporation).

DBTRG-05MG, U251, U87 and U118 cells (10⁶) in 100 μl PBS were injected into nude mice via the tail vein. Individual mice were euthanized when moribund; the pulmonary lesions were collected at necropsy and transplanted subcutaneously into the flank of fresh host mice to propagate the tumors. To generate primary cultures, subcutaneous tumors were harvested at necropsy, washed in PBS, minced, and treated with 0.25% trypsin (Invitrogen Corporation) for 45 min. Released cells were collected at 1500 rpm and resuspended in complete DMEM containing 10% FBS. This procedure was repeated twice to obtain GBM-M2 cell lines. U251-M1 cells were harvested after 1 cycle of selection.

To compare the metastatic potential of GBM cell lines, 10⁶ cells in 100 μl PBS were injected intravenously into nude mice. By time of necropsy, lungs were harvested and a scoring system was established as follows. If no visible lesions were observed in lungs or other organs, mice were scored as (-); if visible and/or hematoxylin and eosin (H&E)-stainable lung lesions were confined to ≤ 50% of the tissue section area, animals were scored as (+); if lesions in the lung exceeded 50% of tissue section area, animals were scored as (++); and if most of the lung was involved and a lesion was present in at least one other organ, animals were scored as (+++).

To prepare GBM-conditioned media, 5 × 10⁵ cells were seeded into 10-cm dishes and grown to 80% confluency. Cells were washed with PBS twice, and complete medium was replaced with DMEM lacking serum. After culture for an additional 24 hrs medium was collected and spun at 13,000 × rpm for 5 min (Sorvall RT7 Plus) and the supernatant fraction was collected and stored at -80C for Multi-Analyte Profile (MAP) testing (Rules-Based Medicine, Austin, TX). To do the data analysis, the concentration levels of cytokines and growth factors from each cell line was normalized based on cell numbers. The fold change in expression of 89 cytokines and proteins are determined by comparing expression levels of GBM-M2 sub-lines to their parental DBTRG-05MG, U87 and U251 cell lines. R version 2.6.1 was used to generate the heat-map of the expression level fold change.

Immunocompromised [athymic nude (nu/nu)] mice at about six weeks of age were used for intracerebral injections. Mice were anesthetized using isoflurane gas anesthesia (~2%) and placed into the ear bars of a stereotaxic frame. A burr hole was created through the skull 2 mm posterior to the bregma, and 5 × 10⁵ cells in 5 μl PBS were injected into the brain at 3 mm depth.

Tumor tissues were harvested, fixed with formalin, and embedded in paraffin. Paraffin blocks were sectioned to perform H&E and immunohistochemistry (IHC) staining for microscopic evaluation. IHC was performed using the Discovery XT Staining Module (Ventana Medical Systems, Inc., Tucson, Arizona). Briefly, deparaffinized sections were incubated in Tris/Borate/EDTA, pH 8 at 95°C for 8 minutes and at 100°C for 36 minutes for antigen retrieval. For Met staining, slides were then incubated with primary antibodies MET4, a mouse monoclonal antibody (mAb) against the extracellular domain of human MET 21 at 1:250 dilution (8 μg/ml), anti-uPAR (R&D, Minneapolis, MN) at 1:200, and anti-CD31 (Neomarkers, Fremont, CA) at 1:200 for 60 minutes. The slides were then incubated with a universal secondary antibody, which is an anti-mouse and rabbit cocktail (Ventana Medical Systems, Inc.) for 30 minutes followed by diaminobenzidine (DAB) staining (Ventana Medical Systems, Inc.).

17AAG was purchased from LC Laboratory (Woburn, MA). 17AAG was first dissolved in 100% DMSO and stored at -80°C and then freshly diluted with vehicle PBST (PBS with 0.05% Tween 80) just prior to injection 22. For all tumor models, host mice (6-week old female nude mice) were given vehicle alone (control), 17AAG in vehicle at a daily dose of 20 mg/kg (single injection daily), or 60 mg/kg body weight (administered as two divided doses 6 hrs apart), all administered by intraperitoneal injection 22. For drug testing in the GBM subcutaneous xenograft model, tumor volume (V_t) was measured with manual calipers twice a week (V_t = length × width × depth). Results are expressed as mean ± SE.

With the orthotopic GBM xenograft model, DBM2 cells were inoculated intracranially and tumor growth was monitored by serial high-resolution ultrasound as described in the supplementary figures [Additional Files 1 and 2]. Weekly measured tumor volume was normalized with the initial tumor size upon group to achieve the fold change of tumor volume. Result is expressed as mean ± SE. With lung metastasis model, 28 nude mice were divided into control (n = 8), 20 mg/kg (n = 10) and 60 mg/kg (n = 10) groups. Each mouse received a single intravenous tail vein injection of 10⁶ DBM2 cells in 100 μl PBS. Treatment started the second day after the cells were injected and continued for 8 weeks, by which time most of the control mice were moribund. At necropsy, lungs were harvested and scored as described above; body weight and lung weight of each mouse were also recorded.

Statistical analysis of 17AAG-treated DBM2 intracranial tumor growth was performed with a student's "t" test. Log-rank test was used to analyze survival time. Chi-square test was used for comparison of 17AAG treatments against DBM2 pulmonary metastases.

Primary and metastatic brain tumors are often aggressive and exceedingly difficult to treat. Evaluating the efficacy of the novel targeted agents against brain tumors is problematic due to the inadequacy of relevant pre-clinical models. In contrast to metastasic cancers, GBM is highly invasive into the brain parenchyma and rarely fully resectable. Xenograft mouse models for human GBM inadequately recapitulate the human disease because of slow growth and invasion at the orthotopic location.

We tested if we could enhance the growth and invasiveness of commonly used GBM lines by selecting metastatic cell populations from experimental lung metastasis (ELM). Clark et al. 23 used this approach to enrich for highly metastatic and invasive melanoma tumor cells. GBM extra-cranial metastases are rare 89111213, but surprisingly, most GBM cell lines tested have been shown to be metastatic from subcutaneous (SQ) tumor xenografts 14. Here we show that three out of four GBM tumor lines are metastatic in ELM assays (Figure 1) and are more malignant when orthotopically grown (Table 1).

We started by injecting DBTRG-05MG cells into the tail vein of athymic nu/nu mice. DBTRG-05MG is a human glioma cell line that is highly invasive in vitro in response to hepatocyte growth factor (HGF), but grows poorly as SQ tumor xenografts 2425. Starting at 8 weeks after tail vein injection, we sacrificed mice individually and, when pulmonary tumor lesions were observed, we collected the lesions and propagated them in vivo as SQ tumors followed by a second cycle of ELM selection (M2). These cells, DBM2, were highly invasive and metastatic in ELM assays (Figure 1A, B). Tail vein injection of DBM2 cells produced extensive tumors almost replacing the lungs (Figure 1B, c–d, Table 1) compared to parental DBTRG-05MG cells, which only formed occasional and organ confined lung tumors (Figure 1B, a–b). DBM2 cells also formed extensive metastases in skeletal muscles (Figure 1B, e) diaphragm (Figure 1B, f), lymph nodes along the spine (Figure 1B, g), and in the chest cavity (Figure 1B, h). DBM2 cancer cells invaded skeletal muscle (Figure 1B, k left 2 arrows) and caused an osteolytic bone reaction consistent with the skull-erosion phenotype described below. DBM2 cells also grow more rapidly in vitro compared to parental DBTRG-05MG [Additional File 3] and especially in vivo as a xenograft, even compared to the GBM U251 line [Additional File 3]25.

We questioned whether more metastatic tumor cell populations can be selected by ELM from other commonly used GBM cell lines (U87, U251, U118): We were successful in selecting U87-M2 and U251-M2 cell lines after two ELM cycles. Both lines not only grew more rapidly, but as with DBM2, they showed extensive metastasis to lungs and other organs (Table 1). A comparison of tumor growth of U87 to U87-M2 either orthotopically or by ELM assay showed enhanced aggressive biological behavior of U87-M2 in both assays [Additional File 3]. When tested, all three GBM-M2 ELM lines showed significant growth enhancement in ELM, SQ or orthotopic xenograft mouse models (Table 1). By contrast, U118 GBM cells, which grow well as a SQ xenograft, did not form lung tumors in the ELM assay. Interestingly, when inoculated orthotopically, none of the GBM-M2 lines formed extracranial metastases. Why the metastatic potential of these intercranial tumors is not realized is curious, since these cancers are highly vascularized [Additional File 1;B,b], elicit marked angiogenesis (Figure 3C, e–f), and even display tumor cells in the tumor-associated vasculature (Figure 3C, d).

The expression of a number of factors and interleukins is increased in patient GBM and is associated with glioma stage and aggressive tumor behavior 15161718. Of note are pro-angiogenic cytokines and interleukins that are responsible for the vascular proliferation, a hallmark of GBM. We assayed 24 hr conditioned medium from the three GBM-M2 cell lines including U251-M1A and U251-M1B compared to their parental lines on a platform that queries expression of 89 proteins (Multi-Analyte Profile; Rules-Based Medicine, Austin, TX) http://www.rulesbasedmedicine.com. Figure 2 shows a heat map with fold changes described in the supplementary table [Additional File 4], revealing four cytokines and growth factors in all three GBM-M2 lines, GM-CSF, IL-6, BDNF, and IL-8 that were highly elevated in GBM-M2 cells (DBM2, U87-M2 and U251-M2) compared to their parental cell lines (DBTRG-05MG, U87 and U251). In addition, GM-CSF, IL-6 and IL-8 are all reported to be associated with poor prognosis in patient GBM 1618. In addition, monocyte chemotactic protein-1 (MCP-1), which is elevated in patients with GBM 26, is also highly elevated in U87 and U251 sub-lines. It is striking that GBM-M2 ELM selection of three separate cell lines markedly enhanced the expression of the same interleukins and cytokines that are of prognostic significance in GBM tumors. These results encouraged us to analyze the growth and histopathologic characteristics of this animal model for intracranial tumor growth.

Metastatic DBM2 cells grow orthotopically in mouse brain with a diffuse tumor boundary (Figure 3A, a–c) and finger-like protrusions (Figure 3A, c) indicative of infiltrative growth. Insufficient intracranial growth of parental DBTRG-05MG cells led to compare DBM2 intracranial growth with the orthotopic growth of parental U251 xenograft tumors. In contrast to DBM2 tumors, U251 tumors maintained a distinct border with the brain parenchyma with little localized invasion (Figure 3A, d–f). Analysis of tissue sections from DBM2 tumors for human c-MET and uPAR expression pinpointed the location of invasive glioblastoma cells in the brain parenchyma and at the same time examined an important mechanism for cellular invasion (Figure 3B). c-MET oncoprotein signaling promotes the activation of urokinase and its receptor (uPAR) 27 and both are associated with GBM invasion in patient tumors 24272829. Adjacent to the main tumor xenograft, we observed human c-MET and uPAR staining of cells invading the normal brain parenchyma (Figure 3B) showing that DBM2 cells are highly invasive.

Certain pathological features are associated with aggressive behavior of many cancer types, including GBM 1530. DBM2 orthotopic tumors show many of these features. They are markedly pleomorphic and possess regions of central necrosis lined by a row of crowded tumor cells (Figure 3Ca, b arrows). Further, the orthotopic tumors exhibit extensive vascular hyperplasia (Figure 3Ce), vascular invasion (Figure 3Cd) as well as invasion of vessel walls (Figure 3Cc arrow), thrombus formation (Figure 3Cd). Glomeruloid body-like abnormal vasculature formation was observed upon staining with CD31 antibody (Figure 3Cf). Together, the invasive and aggressive growth behavior and cytokine profile of ELM selected xenografts strongly resemble human disease and validate this animal model for testing of drugs for inhibition of intracranial tumor growth.

As DBM2 orthotopic tumors grow, we observed that the opening created for tumor cell inoculation increases in size, allowing both intra and extracranial tumor growth [Additional File 1]. This opening allows high-resolution intravital imaging of DBM2 tumor growth [Additional File 1;B]. Ultrasound imaging revealed poorly distinct tumor margins, consistent with invasive growth. Further, ultrasound measurements demonstrated that the increase of tumor volume was accompanied by a proportional increase of the skull erosion at the DBM2 cell inoculation site [Additional File 2]. This was confirmed by CT technology (data not shown). We compared the dimensions of the skull erosion obtained by ultrasound [Additional File 1;A,c], the distance between the arrows) to measurements with conventional calipers [Additional File 1;A,d] at the time of necropsy and observed good correlation between the two approaches (γ = 0.87, n = 10). Beneath the skull erosion, tumor volume was determined from the ultrasound images [Additional File 2;C]. Moreover, we found a high correlation (γ = 0.95, n = 96), [Additional File 2;D] between tumor volume and the size of the skull opening measured by ultrasound. Thus, the skull opening provides a simple way to monitor tumor growth during therapeutic intervention.

We found that, with Doppler and contrast injection ultrasound, both the amount of blood flow and the direction of the flow in the orthotopic DBM2 tumor can easily be visualized. Under the Doppler mode [Additional File 1;B, a], we see strong energy signals that accumulate in the skin, indicating the existence of "macro" blood vessels with high blood flow in these tissues. However, the tumor mass is mostly dark, indicating that the tumor vasculature does not emit a Doppler signal. To enhance the visualizing of tumor blood vessels, we injected a contrast reagent through the tail vein before ultrasound measurement. Following injection, we saw a rich vascular network extending from the bone-tumor margins along the intracranial boundary of the tumor [Additional File 1;B,b]. Strikingly, almost all the tumor provided a contrast signal, indicating that the DBM2 orthotopic tumors have micro-blood vessels with a lower flow rate than abundant large, mature blood vessels. This makes the DBM2 intracranial glioblastoma model particularly useful as a preclinical model to evaluate novel therapeutic interventions against vascular flow and formation. Given the resemblance of this animal model to patient GBM we proceeded with the evaluation of the 17AAG for inhibition of intracranial tumor growth.

17AAG is an HSP90 inhibitor that is in clinical phase I trials targeting different types of cancers, but its use has not been reported against glioblastoma 192031. With the SQ model, 17AAG at 60 mg/kg gave significant growth inhibition after 4 weeks of dosing (Figure 4A, P < 0.05 at day of 32). When the orthotopic model was used, however, results with the 60 mg/kg-day group growth rate was significantly lower than that of mice in the non-treated DBM2 control group (Figure 4B, P < 0.05 at day 21). Moreover, administration of 17AAG at 60 mg/kg-day significantly prolong the survival of mice bearing DBM2 intracranial tumors in dose-dependent manner (Figure 4C, p < 0.05).

We also tested if 17AAG can inhibit DBM2 ELM metastasis, for the purpose of determining whether the drug would inhibit this invasion dependent metastasis assay. Our results show that, at 60 mg/kg-day, 17AAG can significantly block DBM2 metastasis formation in lungs and other organs (Table 2, P < 0.05). Moreover, the harvested lungs from the 60 mg/kg-day group demonstrated significantly less tumor burden than those from the 20 mg/kg-day and control groups (Table 2, P < 0.05). We conclude that 17AAG inhibits intracranial DBM2 tumor growth at the same dose (60 mg/day) as tumor growth and metastasis formation in the SQ and ELM models. This strongly encourages testing of a novel application for 17AAG in patients with GBM.