Imaging Correlates of Molecular Signatures in Oligodendrogliomas

Oligodendrogliomas were the first of the malignant gliomas shown to be responsive to chemotherapy and the first for which specific molecular predictors of chemotherapeutic response and survival were identified (1, 2, 3, 4). In anaplastic oligodendrogliomas, response to chemotherapy, duration of response to chemotherapy, progression-free survival after radiotherapy, and overall survival have been associated with allelic loss of chromosomal arm 1p (1, 2, 3, 4, 5, 6, 7, 8). Moreover, loss of 1p is highly associated with loss of 19q, together constituting the earliest known molecular change in 50–70% of such neoplasms (9). Combined loss of 1p and 19q appears to define a treatment-sensitive malignant glioma, one which may sometimes be curable with current therapies (7). Indeed, diagnostic molecular testing may soon guide the management of patients with low- and high-grade oligodendrogliomas.

Like genetics, imaging is also revolutionizing neuro-oncology. Today, the detection, surgical management, radiation planning, chemotherapeutic assessment, and follow-up evaluation of patients with brain tumor are highly dependent on magnetic resonance imaging (MRI). Tissue sampling is needed for precise diagnosis, but the type of brain tumor can often be correctly inferred from the clinical context and MRI. Moreover, in adults with cerebral gliomas, enhancement with gadolinium is usually indicative of a high-grade anaplastic tumor, with ring-enhancement being characteristic of the most malignant of these, glioblastoma. To further explore the potential of MRI to provide diagnostic information, we tested the hypothesis that 1p and 19q allelic loss in an oligodendroglioma might be associated with specific imaging characteristics.

After institutional review board approval, cases for study were culled from a database of patients with oligodendrogliomas. The eligibility criteria included the following: (a) a newly diagnosed, untreated oligodendroglioma; (b) a preoperative MRI scan; and (c) data on 1p and 19q status.

MRI scans were assessed by a neuroradiologist (D. H. L.) and two neurosurgeons (J. F. M., E. K.) blinded to the genetic alterations in the tumors. Only T₁-weighted, T₁ post-gadolinium, and T₂-weighted images were consistently available and reviewed. The following image characteristics were evaluated qualitatively: (a) sharp versus indistinct tumor border; (b) homogeneous versus heterogeneous tumor signal intensity (i.e., hypointense on T₁ and hyperintense on T₂ versus mixed intensity signals on T₁ and T₂); (c) paramagnetic susceptibility effect, present versus absent [i.e., shortening on T₁ and T₂ (10, 11)]; (d) contrast enhancement, present versus absent; and (e) percent cortical involvement by tumor (i.e., maximum cortical extent divided by maximum tumor diameter). The images were scored by consensus. Pathology slides and reports on each case were re-reviewed by a neuropathologist (D. N. L.) to ascertain whether calcification or hemorrhage had been present in the tumor at the time of initial diagnosis.

Allelic status was assessed by loss of heterozygosity assays in constitutional/tumor DNA pairs using microsatellite markers on 1p36 (D1S2734, D1S199, D1S508) and 19q13 (D19S219, D19S112, D19S412, D19S596; Ref. 3). Tumor DNA was extracted from microdissected, formalin-fixed, paraffin-embedded sections and constitutional DNA from blood leukocytes or paraffin sections of adjacent uninvolved brain (12).

Fisher’s exact test was used to assess the significance of associations between 1p loss and 19q loss and imaging features (with % cortical involvement dichotomized at its median). The overall type 1 error of 0.05 was controlled using a bootstrap resampling method with 10⁵ samples (13), as implemented in SAS, PROC MULTTEST (SAS Institute Inc., Cary, NC). The Wilcoxon rank-sum test was used to assess the association between measured cortical involvement and genetic features. The adjusted P values are reported.

Forty sets of scans were evaluated on 21 men and 19 women, ages 24–82 years (median, 40 years). The histopathological, molecular genetic, and imaging features are summarized in Table 1. There were 12 WHO grade II tumors, 13 grade II-III tumors, and 15 grade III oligodendrogliomas. Oligoastrocytomas were excluded from this study. Loss of chromosomal arm 1p was detected in 21 tumors, loss of 19q was observed in 23 (three were non-informative), and combined loss of 1p and 19q was noted in 18. Both alleles, 1p and 19q, were intact in 13. The tumor border was indistinct on T₁-weighted images in 28 cases and indistinct on T₂ images in 23. In no instance was the border sharp on T₁ and indistinct on T₂. The tumor signal intensity was heterogeneous on T₁ and T₂ in 28 cases, and susceptibility effect was present in 15. Twenty tumors displayed evidence of contrast enhancement. Calcification was present in the tumor sections or described in the pathology reports in 12. There was no evidence of hemosiderin deposition in any case.

A significant association was observed between tumor genotype and tumor border on T₁ images. Oligodendrogliomas with 1p and 19q loss were significantly more likely to have an indistinct border, whereas a sharp border between tumor and adjacent brain was characteristic of those with intact 1p and 19q alleles (Fig. 1). Of 26 with an indistinct border and full genetic data, 17 had 1p and 19q loss, and of 12 with a sharp border only one had 1p and 19q loss (P = 0.005). The association of tumor genotype with an MRI characteristic likely to be indicative of invasiveness is consistent with an earlier study in which all bi-hemispheric oligodendrogliomas harbored deletions of 1p and 19q (14). The biological basis of this association is unclear, although one potential explanation posits that 1p and 19q allelic loss confers an invasive phenotype. Indeed, in a recent microarray analysis, oligodendrogliomas with 1p and 19q loss had a gene expression profile similar to that of normal brain (15), which might occur if normal brain tissue was trapped within an invasive tumor. Alternatively, susceptibility to transformation by 1p and 19q loss might occur preferentially in cells that are intrinsically motile. In this scenario, the cell of origin determines the invasive character of the tumor, the latter reflected in the MRI. In either case, the fact that oligodendrogliomas with 1p and 19q loss may be especially invasive could have therapeutic implications. Successful treatment for oligodendrogliomas harboring 1p and 19q allelic loss might require drugs that safely penetrate the entire central nervous system.

An association was also observed between genotype and tumor signal intensity on T₁ and T₂ images. Oligodendrogliomas with 1p and 19q loss were more likely to display mixed signal intensity, whereas a uniform signal on T₁ and T₂ images was more characteristic of those with intact alleles (Fig. 1). Of 26 tumors with mixed signal intensity, 16 had 1p and 19q loss, and of 12 with homogeneous intensity only two had 1p and 19q loss (P = 0.047).

In addition, there was a significant association between tumor genotype and paramagnetic susceptibility effect. Oligodendrogliomas with 1p and 19q loss were more likely to display susceptibility change, whereas the absence of susceptibility was characteristic of tumors with intact alleles (Fig. 1). Of 15 tumors with susceptibility present, 13 had 1p and 19q loss, and of 23 without susceptibility 6 had 1p and 19q loss (P = 0.007). Calcification and tumor-associated hemorrhage are histopathological features of oligodendroglioma (16). Because both features might contribute to signal heterogeneity and susceptibility effect, we re-reviewed the pathology slides and reports on all cases. Hemosiderin was not detected. There was, however, a significant association between intratumoral calcification and the presence of paramagnetic effects (P = 0.03). There was also a significant association between calcification and 1p and 19q loss. Of 12 tumors with calcification, 10 had 1p and 19q loss, and of 25 with no apparent calcification 7 had 1p and 19q loss (P = 0.007).

In the computed tomography era of brain tumor diagnosis, calcification in a supratentorial intra-axial tumor was highly suggestive of an oligodendroglioma. Calcification was a common radiographic finding in low-grade oligodendrogliomas and in anaplastic tumors that had evolved from a lower grade lesion, a natural history profile associated with 1p and 19q loss (17). Indeed, in our series of cases diagnosed by computed tomography, 17 of 19 calcified tumors had 1p and 19q loss, one had 1p loss (19q, non-informative), and one had intact alleles.⁵ It is tempting to speculate that calcification is a biological effect downstream of 1p and 19q loss, that calcification is reflected by paramagnetic effects on MRI, and therefore, that paramagnetic susceptibility changes constitute a marker of 1p and 19q loss in oligodendrogliomas. However, because calcium is a non-paramagnetic element, the relationship between intratumoral calcification and susceptibility effects is likely indirect. Perhaps, in addition to often being calcified, oligodendrogliomas with 1p and 19q loss contain paramagnetic elements. Certainly such findings encourage additional study.

The association of 1p and 19q loss with an indistinct border, mixed signal intensity, and susceptibility was also true for 1p loss, with or without 19q loss (Table 2). There was no statistical association between 19q loss and any imaging parameter. A similar pattern was seen in an earlier study in which radiographic response to chemotherapy was tightly linked to 1p and 19q loss and also to 1p loss but not to 19q loss (3). Loss of 19q, a common finding in high-grade gliomas, appears to be associated with a more favorable prognosis (18). However, 19q loss per se is less characteristic of oligodendroglial histopathology than is 1p loss.

Clinical MRI, as performed here, will not replace tissue-based diagnosis. However, the possibility that molecular alterations in gliomas might be detected by quantitative imaging, novel sequences (19), or biochemical surrogates (20) is strongly encouraged by our findings. Furthermore, the association of 1p and 19q loss with specific imaging and histopathological features may provide clues to the function of genes on chromosomes 1p and 19q that appear critical to the genesis of oligodendrogliomas. A note of caution is warranted. The aforementioned associations require confirmation in a larger prospective study that includes a more comprehensive set of MRI sequences and quantitative image analysis.