Contrast enhancement in 1p/19q-codeleted anaplastic oligodendrogliomas is associated with 9p loss, genomic instability, and angiogenic gene expression

Abstract

Background

The aim of this study was to correlate MRI features and molecular characteristics in anaplastic oligodendrogliomas (AOs).

Methods

The MRI characteristics of 50 AO patients enrolled in the French national network for high-grade oligodendroglial tumors were analyzed. The genomic profiles and IDH mutational statuses were assessed using high-resolution single-nucleotide polymorphism arrays and direct sequencing, respectively. The gene expression profiles of 25 1p/19q-codeleted AOs were studied on Affymetrix expression arrays.

Results

Most of the cases were frontal lobe contrast-enhanced tumors (52%), but the radiological presentations of these cases were heterogeneous, ranging from low-grade glioma-like aspects (26%) to glioblastoma-like aspects (22%). The 1p/19q codeletion (n = 39) was associated with locations in the frontal lobe (P = .001), with heterogeneous intratumoral signal intensities (P = .003) and with no or nonmeasurable contrast enhancements (P = .01). The IDH wild-type AOs (n = 7) more frequently displayed ringlike contrast enhancements (P = .03) and were more frequently located outside of the frontal lobe (P = .01). However, no specific imaging pattern could be identified for the 1p/19q-codeleted AO or the IDH-mutated AO. Within the 1p/19q-codeleted AO, the contrast enhancement was associated with larger tumor volumes (P = .001), chromosome 9p loss and CDKN2A loss (P = .006), genomic instability (P = .03), and angiogenesis-related gene expression (P < .001), particularly for vascular endothelial growth factor A and angiopoietin 2.

Conclusion

In AOs, the 1p/19q codeletion and the IDH mutation are associated with preferential (but not with specific) imaging characteristics. Within 1p/19q-codeleted AO, imaging heterogeneity is related to additional molecular alterations, especially chromosome 9p loss, which is associated with contrast enhancement and larger tumor volume.

Anaplastic oligodendrogliomas (AOs) account for approximately 10% of gliomas.¹ Despite the homogeneous histological appearances of the AOs, the survival times of AO patients range from a few years to more than 15 years. This clinical heterogeneity has been related to molecular heterogeneity.^2,3 Three main molecular subgroups of AO can be distinguished.⁴ Anaplastic oligodendrogliomas with 1p/19q codeletion (virtually all mutated by the isocitrate dehydrogenase [IDH] gene) display the best prognosis. The IDH-mutated AOs, without 1p/19q codeletions, have an intermediate prognosis. Finally, non-1p/19q-codeleted and non-IDH-mutated AOs have a poor prognosis. In addition to the clinical and molecular heterogeneities, the radiological presentations of AO also vary. In AO, the 1p/19q codeletion has been shown to be associated with distinct radiological characteristics, particularly frontal lobe location, blurred tumor borders, and intratumoral signal heterogeneity.^5–9 In glioblastomas, several studies have correlated imaging, genomic, and gene expression features. These studies identified gene expression modules associated with contrast enhancement, edema, and necrosis and showed that IDH mutation, amplification of the epidermal growth factor receptor gene (EGFR), and glioblastoma transcriptomic subgroups were associated with preferential radiological features.^10–13 The aims of the present study were to describe the radiological characteristics of 50 AO patients enrolled in the French network for AO and to correlate these characteristics with tumor molecular profiles (ie, copy number and gene expression profiles) and IDH mutational status.

Materials and Methods

The POLA Network

The scarcity of AO requires collaborative multicenter networks to investigate large cohorts of AO patients. To improve the clinical, biological, and translational research focused on AO patients, in 2009 the French Institut National du Cancer supported the creation of a national network named “Prise en charge des OLigodendrogliomes Anaplasiques” (POLA). This network prospectively collects samples and data from patients with a diagnosis of high-grade oligodendroglial tumor made in the main academic centers of the country.¹⁴

Patients

Fifty newly diagnosed AO patients in the POLA network for whom initial MRI results were available have been included in this study. The diagnosis of AO was confirmed by a central pathological review (D.F-B., A.J., K.M., E.U.C.) using criteria from the World Health Organization.¹⁵ All of the patients provided written consent for the clinical data collection and the genetic analysis, according to national and POLA policies.

Radiological Study

Two neuroradiologists (N.M.D., M.L.) and 2 neurologists (G.R.B., C.D.), all of whom were blinded to molecular status, retrospectively analyzed the preoperative conventional brain MRIs (1.5 or 3 T). All MRIs were obtained within 4 weeks of the histological diagnosis (biopsy or surgery) and included the following images: T1-weighted, T1-weighted postgadolinium, and T2-weighted or T2 fluid attenuated inversion recovery (FLAIR). The following characteristics were assessed (Fig. 1): (i) tumor location (frontal, temporal, insula, parietal, occipital, basal ganglia, corpus callosum); (ii) unilobar or multilobar involvement; (iii) contrast enhancement as absent, blurry (nonmeasurable), or measurable (nodular or ringlike); (iv) shape of the tumor borders (sharp, blurred) on T2/FLAIR sequences; (v) signal intensity (homogeneous vs heterogeneous), where heterogeneous intratumoral signal intensity was defined as the coexistence of hyposignal and hypersignal zones; (vi) intratumoral cysts (>1 cm); and (vii) tumor volume, as calculated by manual segmentation (OsiriX v3.8.1 32-bit software).

Fig. 1.

A representative example of the MRI features assessed in the 50 AOs. Upper panel: tumor topography—F: frontal, T: temporal, P: parietal, O: occipital, CC: corpus callosum. Middle panel: Contrast enhancement types (left), tumor borders in T2/FLAIR sequences (right). Lower panel: Lobar extension—unilobar or multilobar, intratumoral signal, cyst (diameter >1 cm).

DNA and RNA Extraction

The iPrep ChargeSwitch Forensic Kit was used to extract DNA from frozen tumor samples, and the RNeasy Lipid Tissue Mini Kit (Qiagen) was used to extract total RNA. Both the RNA and the DNA were assessed for integrity and quantity, following stringent quality control criteria established by the program protocols of the Cartes d'Identité des Tumeurs (http://cit.ligue-cancer.net). A 1-µg volume from each DNA sample was outsourced to the Integragen Company (Paris, France) for single-nucleotide polymorphism (SNP) array experiments.¹⁴ A 1-µg volume from each RNA sample was used to perform the gene expression analysis.

SNP Array and Gene Expression Array Procedures

As mentioned above, the SNP array experiments were outsourced to Integragen. Two types of platforms were used: HumanCNV370-Quad and Human610-Quad from Illumina.¹⁴ Because the molecular abnormalities were included in the medical management of the patients (ie, non-1p/19q-codeleted patients were included in the European Organisation for Research and Treatment of Cancer 26053-22054 trial if they were eligible), the tumor DNA was run prospectively to obtain its genomic profile within 10 days of the tumor resection. The gene expression profiles of 25 tumors with sufficient RNA available were studied on gene expression arrays. These arrays were performed using the IGBMC (Institut de Génétique et de Biologie Moléculaire et Cellulaire) microarray platform. Total RNA was amplified, labeled, and hybridized to the Affymetrix Human Genome U133 plus2 GeneChip following the manufacturer's protocol. The microarrays were scanned using an Affymetrix GeneChip Scanner 3000, and the raw intensities were quantified from the subsequent images using GCOS 1.4 software (Affymetrix). The data were normalized using the Robust Multi-array Average method implemented in the R package affy.¹⁶

1p/19q Codeletion

Tumors were considered as codeleted if there was an entire loss of 1p and an entire loss of 19q with centromeric breakpoints.¹⁴

IDH1 and IDH2 Mutational Status

IDH1 codon 132 and IDH2 codon 172 were sequenced using the Sanger method with the following primers as previously reported:¹⁷ IDH1 forward: TGTGTTGAGATGGACGCCTATTTG, and IDH1 reverse: TGCCACCAACGACCAAGTC; and IDH2 forward: GCCCGGTCTGCCACAAAGTC, and IDH2 reverse: TTGGCAGACTCCAGAGCCCA.

Statistical Analyses

The SNP array analysis was performed as previously described.¹⁴ The association of chromosome arm imbalances (either loss or gain) with radiological variables was estimated using either Fisher's exact test (for factors) or Student's t-test (for quantitative variables). The gene expression array analysis was performed as previously described.¹⁶ The published centroid-based classifier of Verhaak et al¹⁸ was used to classify our samples according to their system.

Results

Radiological Spectrum of Anaplastic Oligodendrogliomas

The clinical and molecular characteristics of the 50 patients are summarized in Table 1, and their radiological characteristics are summarized in Table 2. Most of the cases were located in the frontal lobe (n = 35, 70%), were contrast enhanced (n = 37, 74%), and had heterogeneous intratumoral signal intensities (n = 42, 84% on T1; n = 37, 74% on T2 or FLAIR). However, as shown in Fig. 2, the radiological presentation was highly variable, ranging from a non-contrast-enhanced tumor (n = 13, 26%), suggestive of low-grade glioma, to a ringlike enhanced tumor (n = 11, 22%), suggestive of glioblastoma.

Table 1.

The clinical and relevant molecular characteristics of the 50 patients

Median age (range)	48 y (24–78)
Sex (M/F)	n = 31/n = 19
Symptoms at diagnosis
Seizures	n = 26 (52%)
Headache	n = 11 (22%)
Neurological deficit	n = 13 (26%)
Cognitive impairment	n = 12 (24%)
Time from symptom onset to diagnosis
Median (range)	2.7 mo (0.2–172)
Time from MRI to surgery
Median (range)	10 d (1–32)
Surgery
Total resection	n = 13 (26%)
Subtotal	n = 17 (34%)
Partial resection	n = 10 (20%)
Biopsy	n = 10 (20%)
Molecular alterations
IDH mutation	n = 43 (86%)
IDH1	38
IDH2	5
1p/19q codeletion	n = 39 (78%)
9p loss	n = 16 (32%)
10q loss	n = 5 (10%)
EGFR amplification	n = 2 (4%)

Table 2.

The MRI features of the 50 patients according to 1p/19q codeletion and IDH mutation

	Total, n (%)	1p/19q Codeletion, n (%)			IDH Mutation, n (%)
		Yes	No	P	Yes	No	P
Topography
Frontal	35 (70)	32 (82)	3 (27)	.001	33 (77)	2 (28)	.02
Temporal	8 (16)	5 (13)	3 (27)	NS	7 (16)	1 (14)	NS
Parietal	3 (6)	1 (2)	2 (18)	NS	1 (2)	2 (28)	.05
Occipital	3 (6)	1 (2)	2 (18)	NS	2 (5)	1 (14)	NS
Insula	18 (36)	13 (33)	5 (45)	NS	15 (35)	3 (43)	NS
Corpus callosum	15 (30)	13 (33)	2 (18)	NS	14 (32)	1 (14)	NS
Extension
Unilobar	32 (64)	25 (64)	7 (64)	NS	28 (65)	4 (57)	NS
Multilobar	18 (36)	14 (36)	4 (36)	NS	15 (35)	3 (43)	NS
Contrast enhancement
No	13 (26)	12 (31)	1 (10)	NS	12 (28)	1 (15)	NS
Yes	37 (74)	27 (69)	10 (90)	NS	31 (72)	6 (85)	NS
Blurry	9 (18)	9 (23)	0 (0)	NS	9 (21)	0 (0)	NS
Nodular	17 (34)	11 (28)	6 (54)	NS	15 (35)	2 (28)	NS
Ringlike	11 (22)	7 (18)	4 (36)	NS	7 (16)	4 (57)	.03
Tumoral borders
Sharp	19 (38)	15 (39)	4 (36)	NS	17 (41)	2 (28)	NS
Indistinct	30 (62)	23 (60)	7 (63)	NS	25 (59)	5 (72)	NS
Intratumoral signal
Heterogeneous T1	42 (84)	35 (89)	7 (63)	P = .06	33 (77)	5 (45)	NS
Heterogeneous T2 or FLAIR	37 (74)	33 (84)	4 (36)	P = .003	34 (79)	3 (43)	P = .06
Other
Intratumoral cyst	14 (28)	13 (33)	1 (9)	NS	14 (32)	0 (0)	NS
Median volume (cm3)	108	120	70	P = .06	111	100	NS

Fig. 2.

Representative MRIs of the 50 patients classified according to their 1p/19q codeletion and IDH mutation status. The MRIs of 1p/19q-codeleted AOs were further classified according to contrast enhancement. Upper panel: 1p/19q-codeleted and IDH-mutated AOs without (A) and with contrast enhancement (B). Middle panel: AOs without 1p/19q codeletion and without IDH mutation. Lower panel: Non-1p/19q-codeleted AOs with IDH mutation.

Fig. 3.

Copy number alterations in 1p/19q-codeleted AOs. (Top) AO without contrast enhancement. (Bottom) AO with contrast enhancement. Genomic gain and genomic loss are indicated in red and green, respectively.

1p/19q Codeletion and IDH Mutation Are Associated With Distinct Radiological Features

Thirty-nine AOs (78%) were 1p/19q codeleted and IDH mutated, 4 AOs (8%) were non–1p/19q codeleted and IDH mutated, and 7 AOs (14%) had none of these alterations. Their radiological characteristics, categorized by 1p/19q codeletion and IDH mutation, are summarized in Table 2 and illustrated in Fig. 2.

The 1p/19q codeletion was associated with a frontal lobe location (32/39 vs 3/11, P = .001), heterogeneous intratumoral intensity on T2 or FLAIR (33/39 vs 4/11, P = .003), and an absent or nonmeasurable contrast enhancement (21/39 vs. 1/10, P = .01). Intratumoral cysts also occurred more frequently in the 1p/19q-codeleted tumors (13/39 vs 1/11, P = .1). The IDH mutation was also associated with a frontal lobe location (33/43 vs 2/7, P = .01; Table 2). The IDH wild-type AOs were more frequently parietal (2/7 vs 1/43, P = .05) and more frequently displayed ringlike enhancement than IDH-mutated AOs (4/7 vs 7/43, P = .03).

Although 1p/19q codeletion and IDH mutation were associated with distinct radiological features, no radiological pattern was specific to these molecular alterations. As shown in Fig. 2, the radiological presentations of the 1p/19q-codeleted AOs remained variable, ranging from a low-grade-like aspect without contrast enhancement (12/39, 30%) to a glioblastoma-like aspect with ring contrast enhancement (7/39, 18%).

Contrast Enhancement Is Associated With Chromosome 9p Loss in 1p/19q-Codeleted Anaplastic Oligodendrogliomas

To assess whether the radiological heterogeneity within 1p/19q-codeleted AOs was related to their underlying molecular heterogeneity, the genomic profiles of non-contrast-enhanced and contrast-enhanced 1p/19q-codeleted AOs were compared. The contrast-enhanced AOs had a larger mean tumor volume than the non-contrast-enhanced 1p/19q-codeleted AOs (145 vs 61 cm³, P = .001), were more frequently multilobar (13/27 vs 1/12, P = .02), and tended to infiltrate the corpus callosum more frequently (11/27 vs 2/12, P = .2). In addition, contrast enhancement was associated at the histological level with microvascular proliferation (26/27 vs 6/12, P = .002), necrosis (10/27 vs 0/12, P = .02), and a higher proliferation index (mean labeling of KI-67 1, 20% vs 15%, P = .04).

As shown in Fig. 3, loss of chromosome 9p and the cyclin-dependent kinase inhibitor 2A gene (CDKN2A) were more frequent in contrast-enhanced 1p/19q-codeleted AOs compared with non-contrast-enhanced 1p/19q-codeleted AOs (12/27 vs 0/12, respectively, P = .006). The contrast-enhanced tumors also exhibited more complex genomic profiles compared with the non-contrast-enhanced tumors. Indeed, the mean number of chromosome alterations and the number of cases displaying more than 3 chromosome arm abnormalities (excluding 1p/19q codeletion) were higher in contrast-enhanced tumors than in non-contrast-enhanced tumors (3.9 vs 1.9, P = .03 and 8/27 vs 0/12, P = .04, respectively). In addition, there was a positive correlation between the tumor volume and the number of chromosome arm alterations (r = 0.33, P = .04). Consistent with the association between contrast enhancement and tumor volume, chromosome 9p and CDKN2A loss in 1p/19q-codeleted AO was also associated with larger tumor volume (median 163 vs 100 cm³, P = .02), multilobar involvement (9/12 vs 5/27, P = .001), and corpus callosum infiltration (9/12 vs 4/27, P = .005).

Contrast Enhancement Is Associated With the Expression of Angiogenesis-related Genes in 1p/19q-Codeleted Anaplastic Oligodendrogliomas

To identify genes for which expression was associated with contrast enhancement in the 1p/19q-codeleted AOs, the gene expression profiles of contrast-enhanced (n = 18) and non-contrast-enhanced tumors (n = 7) were compared using Affymetrix expression arrays. Fifty-nine genes were differentially expressed with a fold change ≥2 and a P value <.005 (Supplementary Table 1). The list of upregulated genes in the contrast-enhanced tumors (n = 33) was significantly enriched in genes involved in angiogenesis (P < 10⁻⁴), basement membrane (P < 10⁻⁴), and cell adhesion (P < 10⁻⁴). Angiogenesis-related genes consisted of both angiogenic and anti-angiogenic genes (Table 3). In addition, the expression of several angiogenic genes, including angiopoietin 2 (ANGPT2) and vascular endothelial growth factor A (VEGFA), were positively correlated with tumor volume (Table 3). Consistent with VEGFA upregulation, gene set enrichment analysis demonstrated significant enrichments of VEGFA targets (P < .001) and hypoxia-related genes (P < .001) in the contrast-enhanced tumors (Supplementary Table 2).¹⁹ The contrast-enhanced tumors were also enriched in genes upregulated in glioblastomas compared with low-grade glioma vessels (P = .02)²⁰ and in genes positively correlated with contrast enhancement in glioblastomas (P = .1)¹¹ (Table 3, Supplementary Table 2). Finally, the 25 tumors were classified according to the system of Verhaak et al¹⁸ to assess whether the type of contrast enhancement was associated with AO molecular subclasses. Sixteen samples were assigned to the proneural subtype, 6 samples were assigned to the mesenchymal subtype, and 3 samples were assigned to the neural subtype. The AOs with nodular and ringlike contrast enhancements were more frequently classified as mesenchymal than were AOs with blurry or no contrast enhancement (5/11 vs 1/14, P = .05).

Table 3.

List of the angiogenesis-related genes upregulated in contrast-enhanced (n = 18) vs non-contrast-enhanced (n = 7) 1p/19q-codeleted AOs (fold change ≥ 2, P < .005)

Gene Symbol	Fold Change	Gene Ontology	Upregulated in GBM-associated Blood Vessels	Role	Pearson Correlation Coefficient with tumor volume, r
COL1A1	11.3	Blood vessel development	Yes42		0.45 (P = .03)
COL3A1	8.1	Cell adhesion	Yes20		0.6 (P = .002)
COL1A2	4.6	Blood vessel development	Yes20		0.6 (P = .003)
ANGPT2	4.3	Angiogenesis	Yes20	Angiogenic41	0.57 (P = .004)
CD93	3.8	Cell adhesion	Yes20	Angiogenic43	0.41 (P = .04)
COL4A1	3.8	Angiogenesis	Yes42		0.47 (P = .02)
COL4A2	3.3	ECM organization	Yes20	Anti-angiogenic44	0.42 (P = .04)
TIMP1	3.2	ECM organization		Anti-angiogenic45	0.56 (P = .004)
COL6A3	3	Cell adhesion	Yes20		0.42 (P = .04)
HMOX1	2.7	Angiogenesis		Angiogenic46	NS
ENPEP	2.6	Angiogenesis	Yes20	Angiogenic47	NS
VEGFA	2.6	Angiogenesis		Angiogenic48	0.45 (P = .02)
NID2	2.4	Cell adhesion	Yes20		0.41 (P = .04)
ELTD1	2.4	Neuropeptide signaling	Yes20		NS
FN1	2.3	Cell adhesion	Yes20	Angiogenic49	NS
VWF	2.3	Blood coagulation		Anti-angiogenic50	0.44 (P = .03)
COL15A1	2.2	Angiogenesis		Anti-angiogenic51	NS
MYOF	2.2	VEGFR signaling pathway		Angiogenic52	NS
IGFBP4	2.1	Cell proliferation	Yes20	Anti-angiogenic53	NS
MYOIB	2.1	Regulation of cell shape	Yes20		NS
TFPI	2	Blood coagulation		Anti-angiogenic54	NS
SPRY1	2	EGFR signaling	Yes20	Anti-angiogenic55	NS

Abbreviations: GBM, glioblastoma multiforme; COL1A1, collagen, type I, alpha 1; COL3A1, collagen, type III, alpha 1; COL1A2, collagen, type I, alpha 2; CD93, CD93 molecule; COL4A1, collagen, type IV, alpha 1; COL4A2, collagen, type IV, alpha 2; ECM, extracellular matrix; TIMP1, tissue inhibitor of metalloproteinase 1; COL6A3, collagen, type VI, alpha 3; HMOX1, hemeoxygenase (decycling) 1; ENPEP, aminopeptidase A; NID2, nidogen 2; ELTD1, EGF, latrophilin and 7 transmembrane domain containing 1; FN1, fibronectin 1; VWF, von Willebrand factor; COL15A1, collagen, type XV, alpha 1; MYOF, myoferlin; IGFBP4, insulin-like growth factor binding protein 4; MYO1B, myosin 1B; TFPI, tissue factor pathway inhibitor; SPRY1, sprouty homolog 1, antagonist of fibroblast growth factor signaling (Drosophila).

Discussion

This study demonstrates that AO molecular subtypes are associated with different radiological characteristics and that, as previously shown with glioblastomas,^10–13 correlating AO imaging and molecular features can be used to approach molecular alterations implicated in tumor growth.

The 1p/19q codeletion was previously shown to be associated with frontal lobe location, intratumoral signal heterogeneity, and blurred tumor borders.^5–9 Except for the tumor border characteristics, our study confirmed these findings. The IDH mutation was also associated with distinct radiological features, including a frontal lobe location and a lower rate of ringlike contrast enhancement. These findings are very similar to those reported in recent series of glioblastomas, in which the IDH-mutated tumors were predominantly frontal and had a lower rate of ringlike contrast enhancement.^10,12,13 As IDH mutation precedes 1p/19q codeletion, it is likely that the preferential frontal lobe location of the 1p/19q-codeleted AOs is explained by the fact that nearly all 1p/19q-codeleted AOs are IDH mutated.^4,21 The predominant frontal lobe location of the IDH-mutated gliomas may be explained by a preferentially tumorigenic effect of this mutation in specific forebrain neural progenitors.^13,22

Although 1p/19q codeletion and IDH mutation were associated with distinct imaging characteristics, we were not able to identify a specific radiological pattern associated with these molecular characteristics. Importantly, the ringlike contrast enhancement (which was more frequent in IDH wild-type AOs) was also observed in nearly 20% of 1p/19q-codeleted and IDH-mutated AOs. Other MRI techniques might be more powerful for identifying features specifically associated with 1p/19q codeletion or IDH mutation. MRI texture analysis has been shown to predict 1p/19q codeletion with high sensitivity and specificity in low-grade gliomas,²³ and several studies have suggested that diffusion,²⁴ perfusion,^25–27 and MR spectroscopy²⁸ may also help to noninvasively identify 1p/19q codeletion. However, a recent study has demonstrated that the use of multimodal MRI only marginally improves the accuracy of conventional MRI for the identification of 1p/19q codeletion.²⁹ In contrast, MR spectroscopy seems to be a particularly promising technique for identifying IDH-mutated gliomas, as it can detect the intratumoral production of 2-hydroxyglutarate that specifically results from this mutation.^30,31

In the 1p/19q-codeleted AOs, contrast enhancement was associated with a larger tumor volume and distinct histological, genomic, and gene expression features. Although all of the cases demonstrated endothelial hyperplasia, contrast enhancement was associated with microvascular proliferation, necrosis, and a higher proliferation index. At the genomic level, contrast enhancement was associated with chromosome 9p loss, CDKN2A loss, and chromosome instability. These findings are consistent with previous studies showing that chromosome 9p loss and CDKN2A loss are implicated in tumor progression³² and the development of microvascular proliferation in 1p/19q-codeleted oligodendrogliomas.³³ Both CDKN2A and p14ARF (the alternative CDKN2A gene product) have been shown to play anti-angiogenic roles.^34–39 Specifically, CDKN2A has been demonstrated to downregulate VEGF expression in gliomas³⁵ and to inhibit colon tumor angiogenesis.⁴⁰ The murine homolog of p14ARF, P19(ARF), has been shown to inhibit angiogenesis through the translational control of VEGFA mRNA.³⁴ At the gene expression level, contrast enhancement in 1p/19q-codeleted AOs was associated with the expression of angiogenesis-related genes; the expression of several of these genes, including VEGFA and ANGPT2, was positively correlated with tumor volume. Interestingly, most of the angiogenesis-related genes upregulated in the contrast-enhanced 1p/19q-codeleted AOs were previously shown to be upregulated in glioblastoma-associated vessels, suggesting that despite completely different genetic backgrounds, 1p/19q-codeleted AOs and glioblastomas have similar angiogenic profiles.²⁰ Therefore, anti-angiogenic therapies developed for glioblastomas may also be active in 1p/19q-codeleted AOs. Recent studies have demonstrated that targeting ANGPT2 could greatly enhance the efficacy of anti-VEGF treatments.⁴¹ As ANGPT2 was the most overexpressed angiogenic gene in the contrast-enhanced 1p/19q-codeleted AOs, it might be a particularly interesting target in these tumors.

This study has several limitations. Besides its retrospective nature, it was based on the analysis of only conventional MRI characteristics. In addition, due to the prospective nature of the POLA network, the follow-up was too limited to perform correlations with the outcomes. Nevertheless, this study provides additional evidence of the relationship between AO imaging and molecular heterogeneity and shows that correlating pathological, radiological, and molecular data is an interesting strategy for identifying molecular alterations associated with tumor progression.

Supplementary Material

Supplementary material is available at Neuro-Oncology Journal online (http://neuro-oncology.oxfordjournals.org/).

Funding

This work was funded by the French Institut National du Cancer and part of the national program Cartes d'Identité des Tumeurs (http://cit.ligue-cancer.net/), funded and developed by the Ligue Nationale Contre le Cancer.