Expression of Integrin αvβ3 in Gliomas Correlates with Tumor Grade and Is not Restricted to Tumor Vasculature

Abstract In malignant gliomas, the integrin adhesion receptors seem to play a key role for invasive growth and angiogenesis. However, there is still a controversy about the expression and the distribution of αvβ3 integrin caused by malignancy. The aim of our study was to assess the extent and pattern of αvβ3 integrin expression within primary glioblastomas (GBMs) compared with low‐grade gliomas (LGGs). Tumor samples were immunostained for the detection of αvβ3 integrin and quantified by an imaging software. The expression of αvβ3 was found to be significantly higher in GBMs than in LGGs, whereby focal strong reactivity was restricted to GBMs only. Subsequent analysis revealed that not only endothelial cells but also, to a large extent, glial tumor cells contribute to the overall amount of αvβ3 integrin in the tumors. To further analyze the integrin subunits, Western blots from histologic sections were performed, which demonstrated a significant difference in the expression of the β3 integrin subunit between GBMs and LGGs. The presented data lead to new insights in the pattern of αvβ3 integrin in gliomas and are of relevance for the inhibition of αvβ3 integrin with specific RGD peptides and interfering drugs to reduce angiogenesis and tumor growth.


INTRODUCTION
Gliomas are the most common brain-derived neoplasms with glioblastoma (GBM) multiforme being the most frequent and also the most malignant subtype. Marked tumor cell proliferation, rapid invasion into the surrounding brain tissue and intense microvascular angiogenesis are the biologic hallmarks of malignant gliomas (25,26,34). They depend on the complex interactions of tumor cells with the extracellular matrix (ECM) and neighboring endothelial cells as well as various other cell types (14,40). These mechanisms include signaling via integrins, which are involved in adhesive properties of proliferating cells and play fundamental roles in the regulation of migration and invasion of tumor cells as well as of neoangiogenesis (23,36). The integrin family consists of 24 different heterodimeric receptors of two transmembrane subunits of which the 18a and 8b subunits are known at present (5,23,41). In contrast to many other cell-surface receptors, integrins are not only able to transduce signals to the interior of the cells ("outside-in" signaling) but can also transmit information about the activational state of the cells to their microenvironment ("inside-out" signaling) (13,41).
The integrins a vb3 and avb5 have been demonstrated to be necessary for tumor-induced angiogenesis in a variety of tumors, particularly malignant gliomas via basic fibroblast growth factor (bFGF) and tumor necrosis factor a (TNF-a) (9,10,15). Indeed, integrin avb3 has been found to be located in small blood vessels in GBMs, where it is thought to promote the extensive tumor progression, while in the tissue of the normal brain, it is barely detectable (16). Interestingly, immunohistochemical studies of autopsy material from patients with brain tumors revealed that the expression of integrin avb3 might not be restricted only to proliferating endothelial cells (32). Integrin avb3 was also detected and colocalized with matrix metalloproteinase 2 (MMP-2) at the surface of invasive tumor cells (6,11). Yet, the level of avb3 integrin expression and its cellular distribution caused by malignancy has not been expounded up to now. This has become even more important as avb3 integrin antagonists have reached phase I/IIa clinical trials in patients with malignant gliomas, where they are used in combination with temozolomide as an antiangiogenic therapy. The current study was undertaken to elucidate the expression level and pattern of the integrin avb3 in high-and low-grade gliomas (LGGs) by analysis of snap-frozen tumor samples.

Patients and tissue samples
The brain tissues were obtained from 20 primary GBMs, corresponding to the World Health Organization (WHO) grade IV, and five diffuse astrocytomas, corresponding to WHO grade II. Written informed consent was obtained from all patients for the scientific use of tumor tissue. The histologic diagnosis was made on paraffin sections according to the WHO criteria (25). For the immunohistochemical analysis of avb3 integrin expression (GBM: n = 12; LGG: n = 4; histologically suitable quality) and Western blot analysis (GBM: n = 20; LGG: n = 5), corresponding tissue samples were snap frozen in liquid nitrogen and processed as described below. Additionally, samples of five different solid peripheral tumors (malignant melanoma, neurofibroma, skin cancer metastases, malignant fibrillary histiocytoma and sarcoma; n = 1 each) were processed (kindly provided by the Department of Pathology, Technical University of Munich, Munich, Germany).

Paraffin sections
For the histologic tumor grading, tissues from all patients were fixed for at least 24 h in phosphate-buffered formalin (4%, pH 7.3), dehydrated in graded ethanol followed by xylol and embedded in hot paraffin. Sections (6 mm) were cut from a cold paraffin block and mounted on slides. After drying and deparaffination in graded alcohol, routinely Hematoxilin & Eosin (H&E), Elastica van Gieson and Gomorri stains were performed. Additional immunohistochemical investigations included the proliferation index by staining proliferation marker MIB-1 and glial fibrillary acid protein (GFAP) according to standard procedures (antibodies were obtained from DakoCytomation, Hamburg, Germany).
All immunohistochemical staining procedures and antibody concentrations were adjusted to avoid overstaining (saturation), understaining (beyond detection limit) or non-specific background artifacts. In detail, for staining a vb3 integrin, series of dilutions of the secondary antibody (previous incubation without primary antibody) were performed on positive control (xenotransplanted human M21 melanoma cells expressing avb3, see below) to determine the highest concentration where no background artifacts or non-specific binding occurred (= maximum sensitivity of secondary detection). Then, based on this setting, series of the primary antibody concentration were tested on the positive control. As staining intensity demonstrates a saturation effect, the concentration developing a moderate staining intensity (about half of the maximum intensity of a blank sample) was further used. At this concentration, no non-specific background staining was detectable.

Murine tumor models
Xenotransplanted human melanoma cells (M21) expressing avb3 integrin (12,19) served as a positive immunohistochemical control. Cell culture conditions for M21 were set up as described elsewhere (20). The experimental protocol involving animals was approved by the Committee of Veterinarian Medicine of the State of Bavaria; the handling of animals was performed according to the standards set by the Committee of Veterinarian Medicine.
To obtain the tumor tissue, the mice were injected 1.5 ¥ 10 6 M21 subcutaneously, which led to the formation of tumors with a diameter of approximately 8 mm within 8 weeks. Four weeks after inoculation, the tumors were dissected, cut and frozen for further processing (20).

Analysis of immunohistochemical staining
The immunohistochemical detection of avb3 integrin was determined by a special imaging software, which was calibrated for this purpose. Within a selected area, it measured (i) the mean intensity of the desired immunohistochemical staining and (ii) the immunohistochemically positive fraction at a desired immunohistochemical staining intensity (software written by the author B. Krebs) (33).
In detail, all sections stained in parallel were scanned through an upright microscope (Axiovert 200M and Axiocam MRc5, Carl Zeiss AG, Jena, Germany), keeping the settings including light intensity constant. In the digitized pictures (983 ¥ 737 mm), the area of tissue was selected by excluding possible empty areas of glass light only. Consequently, the measurement was related to tissue only. Then, the selected area was scanned for immunohistochemical detection. Thereby, the software mathematically extracted the component of the desired immunohistochemical staining per pixel (Fast Red maximum emission: 580 nm) and determined its intensity. From this data, the mean intensity of the immunohistochemical staining was calculated. For calibrating the intensity measurement, a sample stained in parallel of xenotransplanted human M21 melanoma cells expressing avb3 was used. This sample acted as a positive control (12,19) and was set to 100%. The tissue that was counterstained only served as a negative control (set to 0%). The procedure resulted in (i) the relative level of avb3 integrin expression in the respective sample (= mean immunohistochemical staining intensity). In addition, to evaluate the distribution pattern of the immunohistochemical staining intensity within the selected area, the fraction of immunohistochemically positive area at a desired intensity level was calculated. In this study, three different levels of staining intensity defined as mild, moderate or strong (about 25%, 50% and 75% of positive control) were investigated. This resulted in (ii) a quantification based on the desired levels of staining intensity (= fraction at different intensity levels).

Western blot
Parallel to the immunohistochemical stainings, Western blotting was performed from the frozen sections of the brain tumor biopsies as described elsewhere (27). In brief, frozen brain sections were cut in 6-mm sections and mounted on glass slides (same as for immunohistochemical stainings). Afterwards, the sections were overlaid with section lysis buffer [2% sodium dodecyl sulfate (SDS), 0.05 M dithiothreiotol, 10% glycerol, 1 mm EDTA, 16 mm Tris-HCl, pH 7.2] using a total volume of 20 mL per 10 mm 2 of tissue. The solubilized tissue was collected, boiled for 10 minutes and subjected to sodium dodecyl sulfate polyacrylamide gel electrophoresis. The very small amount of tissue used in this study was not suitable for standard methods of protein measurement (eg, Bradford assay). Therefore, a protein loading control was determined separately (b-actin, see beyond). Proteins were electrophoretically separated (New Page Gel 12%, Invitrogen, Karlsruhe, Germany) with 100 V over 24 minutes in running buffer [0.1% SDS, 1 mm EDTA, 15 mm 3-(N-morpholino) propansulphonic acid (MOPS), 50 mmol Tris-HCl, pH 7.2]. After that, the proteins were transferred to a polyvinyldifluoride (PVDF) membrane (0.45 mm, Immobilon, Millipore, Eschborn, Germany) by a semidry blotting technique with freshly prepared blotting buffer (20% methanol, 192 mmol glycin, 25 mmol Tris-base, pH 8.3). Next, the PVDF membranes were blocked in blocking buffer (0.05% w/v casein, 0.5% v/v Tween in phosphate buffer saline, pH 7.4) over 1 h at room temperature. The primary antibodies against integrin subunits (polyclonal rabbit anti-integrin a v subunit AB1930 and polyclonal rabbit anti-integrin b3 subunit AB1932; Chemicon, Temecula, CA, USA) were diluted in the respective blocking buffer (1:1000) and incubated overnight at 4°C. After a washing step, bound antibodies were labeled using a secondary antibody coupled to alkaline phosphatase (1:2000 polyclonal goat anti-mouse and anti-rabbit immunoglobulin/AP, DakoCytomation, Hamburg, Germany) over 45 minutes. Visualization followed a second washing step using chromogen nitroblue tetrazolium (NBT) salt and 5-bromo-4-chloro-3-indulylphosphate (BCIP) according to standard protocols (28). The wet blots were scanned using a highresolution flatbed scanner before reprobing. For the detection of the endothelial component, the blots were reprobed with anti-CD31 (CD31); for the assessment of the amount of tissue, Western blots were reprobed with anti-b-actin (Actin I-19, Santa Cruz Biotechnology, Heidelberg, Germany) as the ubiquitous component of the cytoskeleton (21). Densitometric measurements were performed with the software TotalLab V 2.01 (Nonlinear Dynamics, Newcastle upon Tyne, UK) by using series of dilution for calibration of the concentration and a protein ladder (Invitrogen, Karlsruhe, Germany) for the calibration of the molecular weights.

PCR
For the analysis of the mRNA expression of integrin subunits av and b3, equal amounts of tumor tissue from GBMs (n = 5) and LGG (n = 3) patients were homogenized and suspended in TRIZOL ® reagent (Invitrogen, Karlsruhe, Germany). RNA was isolated according to the manufacturer's instructions (Invitrogen, Karlsruhe, Germany). The first-strand cDNA synthesis was done by MMLV-RT (New England Biolabs, Frankfurt, Germany).

Statistics
Staining intensities and densities of immunopositive structures as well as densitometric measurements of the Western blots were tested for normal distribution and groups were compared by Student's t-test for unpaired values and unequal variances. All values were given with standard deviation.

Two types of patterns of avb3 and CD31 expression
Immunohistochemical staining of avb3 integrin was investigated in sections of GBMs (n = 12), diffuse low-grade astrocytomas (n = 4) as well as non-CNS tumors (n = 5). The expression of avb3 integrin was found in the microvessels as well as in the glial tumor cells. The majority of avb3 expression in glial tumors was located in samples from GBMs, which are characterized by extensive proliferation of pleomorphic glial cells accompanied by microvascular proliferates with branching glomeruloid vessels. A sample of an HE stain is given in Figure 1A. Figure 1C,D demonstrates the immunohistochemical stainings of an area with vital tumor (left) and a hypercellular zone (mid) surrounding the necrotic areas (right). The staining for avb3 integrin, shown in Figure 1C, demonstrates a high endothelium-associated expression in this hypercellular zone. This expression is almost identical to the staining of endothelial marker CD31, as demonstrated in Figure 1D (microvascular association). Nevertheless, numerous tumor cells, especially enlarged giant pleomorphic astrocytes, show a strong immunoreactivity for the avb3 integrin, too. In other samples or areas with dense tumor growth, the expression of avb3 integrin seems to be associated to glial tumor cells as well. A typical sample is demonstrated by an overlay of fluorescent stainings of avb3 integrin and CD31 in Figure 1B. The small branched proliferating microvessels marked by CD31 in red are only partially colocalized with the expression of the avb3 integrin in green, which is predominantly located in the glial tumor cells (glial association). The nuclei are counterstained in blue. In contrast to the GBMs, a low expression of avb3 integrin is found in WHO grade II diffuse astrocytomas, where the staining was more diffuse and only barely associated with the microvessels ( Figure 2B).

avb3 Protein expression is associated with malignancy
The different samples were simultaneously stained for avb3 integrin (Figure 2A,B) and analyzed via an imaging software ( Figure 2C,D). The measurements were calibrated to xenotransplanted human melanoma cells (M21) expressing avb3. These acted as a positive control and were set to 100% integrin avb3 expression. The overall mean staining intensity of avb3 integrin in GBMs reached 76.53 Ϯ 12.08%, which was significantly higher than in LGGs with 58.43 Ϯ 2.30% (P < 0.05, Figure 3A). In comparison, the samples of the non-CNS tumors reached 85.18 Ϯ 11.95%, (P < 0.05).
To investigate the expression pattern in more detail, the fraction of immunohistochemically positive structures (density) were calculated for mild, moderate and strong positive detection of avb3 integrin (demonstrated by different colors in Figure 2C,D). In the samples of GBMs, 54.62 Ϯ 23.52% of all tissue were found at least mildly stained, whereas the samples of LGGs reached only 22.20 Ϯ 14.99% (P < 0.05, Student's t-test; Figure 3B). The moderate staining of avb3 integrin was nearly exclusively associated with malignancy, with 9.07 Ϯ 9.25% in GBMs and 0.06 Ϯ 0.07% in low-grade astrocytomas (P < 0.05, Student's t-test). The strong staining intensities were restricted to GBMs only (2.83 Ϯ 4.08%; Figure 3B). Therefore, the average avb3 integrin expression in GBMs exceeds avb3 integrin expression in LGGs by far. However, different GBMs show a very heterogeneous expression of avb3 integrin, which ranges from slightly to very strong integrin expression, similar to high expressing non-CNS tumors.

Glioma cells substantially contribute to avb3 integrin expression
To access the glial cell specific avb3 integrin expression within the tumor, small areas were selected and analyzed where no endothelial structures were present. These areas showed once more a significantly higher mean staining intensity in GBMs (69.47 Ϯ 9.49%) than in LGGs (56.41 Ϯ 6.83%; P < 0.05). The density of immunohistochemically positive tumor cells was measured in GBMs (mild: 46.31 Ϯ 22.23%, moderate: 2.02 Ϯ 2.88%, strong: 0.23 Ϯ 0.49%) and LGGs (mild: 12.31 Ϯ 5.46%, moderate: 0.01 Ϯ 0.01%, strong: 0%), which were significantly different (P < 0.05), although the smaller areas might increase the sampling error. The latter findings indicate that the tumor cells contribute substantially to the avb3 integrin expression.
In conclusion, GBMs demonstrated a higher expression of avb3 integrin than low-grade astrocytomas that was not only caused by a high focal reactivity in proliferating microvessels but also by stronger expression in glial tumor cells. The pattern of avb3 integrin expression depended also on the subtype of tumor. In malignant gliomas, more than three quarters of the overall integrin expression (about 85%) is derived from glial tumor cells.

b3 Integrin subunit expression shows an essential difference between GBMs and LGGs in Western blot analysis
To investigate the immunohistochemical expression of integrin avb3 expression in more detail, we performed Western blot analysis from frozen brain tumor samples as described in the methods. Each lane was loaded with an equal tissue volume of 30 nL (5 mm 2 of tissue). The Western blots were incubated with subunit specific antibody for b3 chain and av chains. Thereby, the integrin subunits  Figure 4A shows the detection of integrin subunits from five different GBM patients (left five lanes; the one lane on the right demonstrates CD31 and b-actin). The group of GBMs demonstrate a heterogenous pattern of b3 chains, whereas the portions of av light chains were strongly present in all samples. The expression in low-grade astrocytomas was lower, especially the b3 chains, which were barely detectable. The densitometric analysis ( Figure 4B) of the samples, calibrated by the dilution series of a reference probe, revealed an overall density of b3 chain at 96 kDa and av light chains at 25 and 27 kDa of 102.8 Ϯ 43.0 in GBMs (n = 20) and 69.1 Ϯ 30.9 in low-grade astrocytomas (n = 5), which corroborates the immunohistochemical measurements (but is not statistically significant).
Further on, microvascularization was assessed by Western blot analysis using the endothelial cell marker CD31, which was detected at 80 kDa ( Figure 4A, right lane). Densitometric analysis demonstrated that samples of GBMs had significantly higher staining intensities (25.2 Ϯ 12.9) than LGGs (4.9 Ϯ 2.2, P < 0.05). The ratio of the overall integrin avb3 detection (sum GBMs 102.8 Ϯ 43.0, sum LGGs 69.1 Ϯ 30.9) to the amount of CD31 demonstrated quite different values of 4.1 in GBMs and 14.1 in LGGs, which indicates that the detection of CD31 positive endothelial cells overcomes the higher expression of avb3 integrin in GBMs (Table 1). Consequently, the numerous proliferating endothelial cells do not contribute at all to the avb3 integrin expression. Densitometric analysis revealed significant differences for the b3 subunit between malignant gliomas and low-grade gliomas (LGGs) (P < 0.05; Student's t-test; indicated with an asterisk). The av subunit (about 137 kDa) and its degradation products (av light chains, 25/27 kDa) showed no significant differences between high-and low-grade gliomas. Table 1. Densitometric analysis of integrin subunits av and b3 and microvascularization in glioblastomas (GBMs), low-grade gliomas (LGGs) and non-CNS tumors (non-CNS). Overall density of integrin subunits av (25/27 kDa) and b3 (96 kDa) as well as the endothelial cells (CD31; 80 kDa) of the tumors was assessed by the Western blot analysis and calibrated by a dilution series of a reference probe. In GBMs, the b3 subunit at 96 kDa showed a significantly elevated expression compared with LGGs (P<0.05; Student's t-test Integrin subunits av mRNA and b3 mRNA both found in GBMs and low-grade astrocytomas As immunohistochemical stainings and Western blot analysis revealed an avb3 integrin expression in malignant GBMs and in a low level in low-grade astrocytomas, we investigated the expression on mRNA level. Cryoconserved tissue of low-and high-grade gliomas was subjected to RT-PCR as described in the materials. Figure 5 demonstrates the PCR product after 24 to 36 recombination cycles for the subunits av and b3 in a sample of a malignant GBM and a low-grade astrocytoma. The mRNA of subunit av was strongly amplified in both high-and low-grade gliomas. The mRNA of the b3 subunit was also present in both tumor entities but amplified on a lower level. Therefore, on the mRNA level, both malignant GBM and low-grade astrocytoma were able to express active avb3 integrin, as suggested by the previous immunohistochemistry and Western blotting, which corroborates the previous results of immunohistochemistry and Western blotting.

DISCUSSION
Integrins are adhesion receptors that mediate cell-cell as well as cell-ECM interactions associated with tumor growth and angiogenesis in malignant gliomas (5,23,31,39). However, there is still a controversy about the extent and the distribution of avb3 integrin expression caused by malignancy and the cell type of origin in gliomas. In literature, 30% of glioma cells were described positive for avb3 integrin in anaplastic astrocytomas and GBMs, whereas LGGs did not express detectable amounts of this integrin (16)(17)(18)38). In situ studies for the two different integrin subunits av and b3 revealed that the expression of integrin avb3 was focused more on small blood vessels of GBMs (16,38). One other group detected individual glial tumor cells of GBMs to be positive for the subunits av and b3 (32). Only recently, Bello et al reported a more prominent expression of integrin avb3 on glioma cells but without providing a detailed expression pattern (5). Yet, this is a matter of major interest for therapeutic delivery of integrin avb3 interfering agents to reduce tumor angiogenesis and growth (7,9,29,38). The efficacy of this approach is clearly dependent on the intratumoral extent and distribution of avb3 integrin, as it has to be supposed that mainly the endothelial cell bound avb3 integrin portion is reached by intravenous application. For this reasons, the extent, distribution and pattern of integrin avb3 expression as well as its correlation to tumor grade were investigated. The computed immunohistochemistry and calibrated staining analysis provided high accuracy and the possibility to bypass the very crucial point of visual quantification of immunohistochemical staining results. In GBMs, a high expression of avb3 integrin was found not only in the endothelial cells of proliferating microvessels but also in tumor cells. The great variability of the expression might, in part, explain why the studies that have been carried out so far showed controversial results according to the amount of avb3 integrin and its cellular and regional distribution within the tumor. While previous reports focused more on the impact of integrin avb3 on angiogenesis (5,16), we demonstrated that glial tumor cells themselves contribute significantly, accounting for approximately 85% of the overall expression of avb3 integrin. In contradiction to former studies, we were also able to detect integrin avb3 expression even in LGGs (18), but on a significantly lower level and predominantly on tumor cells, as there is a lower vascularization.
Interestingly, our miniaturized Western blot analysis revealed a significant difference only for the expression of the b3 subunit between GBMs and LGGs, whereas the av subunit and its degradation products (12) showed no significant difference between these tumor entities. This might be caused by the fact that the integrin av subunit is also needed with the b5 subunit to form the avb5 integrin, which is also abundant in highly vascularized malignant gliomas (5). Moreover, there are other b subunits (b1, b6, b8) that heterodimerize with the av subunit, whereas the b3 subunit in brain tumors only partners with av and with aIIb in thrombocytes (5,18,23,24,32). Therefore, the endothelial expression of avb3 integrin seems to be characteristic for malignant microvessel transformation, as it has already been described that the b3 subunit expression in GBMs is most prominent in endothelial and perivascular cells associated with tumor angiogenesis (5,16,18). Moreover, integrin avb3 has already been detected in colocalization with MMP-2, where it was found to correlate with the dedifferentiation and invasive behavior of invasive endothelial cells as well as tumor cells (5,8,9,11,37). As MMP-2 is an important metalloproteinase for the invasion and dissemination of tumor cells, its functional relationship to the avb3 integrin expression might play a key role in tumor progression (30). However, recent studies on genetically altered mouse models raise the question whether the integrin avb3 is truly proangiogenic and necessary for tumor angiogenesis, as mice lacking either one of these subunits are viable and fertile and show extensive vascularization (1,22,35). Despite these still unanswered questions, antiangiogenic therapy with integrin avb3 antagonists has reached clinical phase I/IIa trials for the patients with malignant gliomas. Therefore, minute detection and quantification of avb3 integrin expression could be a prerequisite for the selection of patients suitable for this kind of additional therapy to analyze the potential relationship between avb3 expression and antitumor activity. In the context of a modern personalized cancer therapy, the benefits should be calculated for each individual patient, especially, as GBMs show a very heterogenous expression of avb3 integrin, which ranges from only mild to very strong. This might also include the noninvasive detection of avb3 expression in gliomas through molecular imaging methods (2-4, 19, 20, 30). Thus, future studies will have to confirm that either tumor angiogenesis or higher malignancy in gliomas in fact correlate with the amount of integrin avb3 expression. This could also lead to new diagnostic and therapeutic approaches for integrininterfering agents.