Characteristics of gliomas in patients with somatic IDH mosaicism

Introduction

Mutations in the IDH1 or IDH2 genes are found in the majority of adult diffuse grade II and grade III gliomas and are considered as the earliest oncogenic event in these tumors [50]. These mutations result in the abnormal production of 2-hydroxyglutarate (2-HG) which is structurally similar to alpha-ketoglutarate. 2-HG competitively inhibits multiple alpha-ketoglutarate enzymes leading to histone and DNA hypermethylation, altered cell differentiation, activation of enzymes implicated in HIF degradation and PDGFRA overexpression [14, 26, 30, 47]. In addition to gliomas, IDH mutations are particularly frequent in cartilaginous tumors [2, 50]. Ollier disease and Maffucci syndrome are two rare non-hereditary enchondromatosis syndromes characterized by the development of multiple benign cartilaginous tumors (enchondromas). Enchondromas appear during childhood and may progress to chondrosarcomas in up to 30 % of cases [3, 36]. In Maffucci syndrome, enchondromas are associated with soft tissue hemangiomas. The analysis of IDH mutations in multiple cartilaginous tumors and non-neoplastic tissues from enchondromatosis patients led to the conclusion that these pathologies are due to early post-zygotic acquisition of IDH mutations, resulting in somatic mosaic mutations of IDH1 or IDH2 [3, 36]. In addition to skeletal tumors, enchondromatosis patients may develop other neoplasm including juvenile granulosa tumors, cholangiocarcinomas, pituitary adenomas, acute myeloid leukemia and gliomas. The aim of the present study was to determine whether gliomas in enchondromatosis patients might also result from somatic IDH mosaicism and whether their characteristics are similar to those of sporadic IDH-mutated gliomas. For this purpose, we analyzed the characteristics of 6 newly diagnosed and 32 previously reported cases of enchondromatosis patients who developed gliomas, and compared them to those of a consecutive series of 159 patients with sporadic IDH-mutated gliomas.

Materials and methods

We retrospectively reviewed the medical and radiological records of 6 patients with Ollier disease (n = 5) or Maffucci syndrome (n = 1) who were referred to our neuro-oncology departments (Pitié-Salpêtière Hospital in Paris, Hospices Civils of Lyon, CHU of Bordeaux and Rennes) for the diagnosis of glioma, as well as 159 consecutive patients who were diagnosed with an IDH1 or IDH2 mutated sporadic glioma between 2010 and 2014 (Hospices Civils of Lyon). Multicentric gliomas were defined as multiple gliomas without connecting T2/FLAIR signal abnormality and distinguished from multifocal gliomas (i.e. multiple gliomas with connecting T2/FLAIR signal abnormality) [1]. IDH mutations and 1p/19q co-deletion status were determined based on DNA extracted from blood and FFPE tumor using a standard protocol (Qiagen, QIAmp DNA mini Kit). IDH1 codon 132 and IDH2 codon 172 were sequenced using the Sanger method with the following primers: IDH1-Forward: TGTGTTGAGATGGACGCCTATTTG; IDH1-Reverse: TGCCACCAACGACCAAGTC; IDH2-Forward: GCCCGGTCTGCCACAAAGTC and IDH2-Reverse: TTGGCAGACTCCAGAGCCCA, as previously reported [21]. The 1p/19q co-deletion was determined based on the loss of heterozygosity technique (LOH) using microsatellite polymorphism markers as previously described [18]. In 3 patients, 1.5 μg of DNA extracted from frozen tumor tissue was outsourced to Integragen Company for the determination of the genomic profile based on Illumina SNP arrays [22]. In enchondromatosis gliomas, immunohistochemistry was performed on 4 μm thick sections of formalin-fixed paraffin embedded blocks with a Ventana Benchmark XT Device. The following antibodies were used after antigen retrieval to assess the expression of ATRX (anti-ATRX, Sigma, polyclonal, dilution 1/400), IDH1R132H (anti-IDH1R132H, Dianova, clone H09, dilution 1/50) and TP53 (clone DO.7, Dako, dilution 1/200). Previously reported cases of patients with Ollier disease or Maffucci syndrome who developed gliomas were identified through PubMed searches from January 1970 until September 2015 using the terms “Ollier disease”, “Maffucci syndrome”, “glioma” and “brain tumor”. We retrieved all relevant articles and checked additional references quoted in these articles. Categorical comparisons were performed using Fisher’s exact test and a t-test was used for quantitative variables. The threshold for statistical significance was p = 0.05.

Results

The characteristics of our 6 patients are presented in Table 1 and those of the 32 previously published patients are shown in Additional file 1: Table S1 [5–8, 11, 12, 15, 16, 19, 20, 23–25, 31–35, 37–40, 43–46, 48, 49]. The characteristics of all of the patients are summarized in Table 2.

The diagnosis of Ollier disease (n = 28) or Maffucci syndrome (n = 10) was made during childhood based on the occurrence of multiple enchondromas predominating on the articulations of the knees, hands and toes. The median age at glioma diagnosis was 25.6 years. At this time, six patients (19 %) had a previous history of chondrosarcoma. Clinical presentation consisted of a variable association of seizures (38 %), progressive focal deficits (41 %) and intracranial hypertension (21 %). Radiological presentation was suggestive of diffuse glioma in all of the patients (Fig. 1). Tumors were most frequently located in the frontal lobe (54 %) and in the brainstem (21 %). Contrast enhancement was present in 27 % of cases. Twelve patients (32 %) had multicentric gliomas that were synchronous in eleven patients and metachronous in one patient. Multicentric gliomas were confined to the hemispheres in 8 patients and involved both the hemispheres and the brainstem in 4 patients. Two patients had gliomatosis cerebri. Ten patients (36 %) had another intracranial tumor at the time of glioma diagnosis (skull base enchondroma n = 7, pituitary adenoma n = 2, skull base chondrosarcoma n = 1).

The glioma was histologically proven in 33 patients. Among the 5 patients without histological diagnosis, 4 patients had a typical aspect of brainstem glioma consisting of an infiltrative intra-axial T2/FLAIR hypersignal without contrast-enhancement, and 1 patient had an aspect suggestive of hemispheric low-grade glioma that was managed with initial follow-up (patient 2, Fig. 1). Histology consisted of diffuse low-grade glioma in 24 patients (73 %), anaplastic glioma in 7 patients (21 %) and glioblastoma in 2 patients (6 %). An astrocytic (n = 22, 66 %) or oligo-astrocytic phenotype (n = 8, 25 %) was more frequently observed than a pure oligodendroglial phenotype (n = 3, 9 %). IDH mutations were assessed in 8 cases (based on sequencing in 6 patients and on immunohistochemistry only in 2 patients) and detected in the tumors of 7 patients (IDH1 R132H n = 5, IDH2 R172S n = 2). In the negative case, no expression of IDH1 R132H was detected on immunohistochemistry, but other IDH mutations were not assessed [38]. In two patients (patient 5 and a previously reported patient [34]) the presence of an IDH mutation could be analyzed in both the glioma and a skeletal tumor and the same mutation (IDH1 R132H in patient 5 and IDH2 R172S in [34]) was identified in these tumors. Since none of our patients underwent a biopsy of more than one glioma, the analysis of IDH mutations in different gliomas from the same patient could not be performed. No 1p/19q co-deletion was identified in the 6 cases in which it was assessed (oligo-astrocytoma n = 4, oligodendroglioma n = 1, glioblastoma n = 1). TP53 expression was studied in 6 patients and found to be expressed in 5 patients. In an additional patient, sequencing demonstrated missenses TP53 mutations [34]. ATRX expression was studied in 5 patients. A homogeneous loss of expression was observed in 4 patients and a heterogeneous loss of expression in 1 patient (patient 3). Genomic profiles of the tumors were obtained for 2 patients; one patient had an isolated LOH of the 17p region covering the TP53 locus (patient 3), and the other patient had a LOH of chromosome 9 associated with partial losses of chromosomes 14q and 15q and a partial gain of chromosome 11q (patient 5). In patient 3, double staining with anti-ATRX and anti-IDH1 R132H antibodies identified: (i) areas with maintained ATRX expression in tumor cells, and (ii) areas with diffuse loss of ATRX expression in tumors cells. In these last areas, some very rare IDH1 R132H and ATRX positive cells with a normal morphology were observed (Fig. 1). Treatment consisted of surgical resection (46 %), radiotherapy (80 %) and or chemotherapy (42 %). The median survival after glioma diagnosis was 5 years.

Independent of histology and 1p/19q co-deletion status, gliomas in enchondromatosis patients were diagnosed at an earlier age than sporadic IDH mutated gliomas (25.6 years versus 44 years, p < 0.001, Fig. 2). In addition, they were more frequently multicentric (32 % versus 1 %, p < 0.001), more frequently involved the brainstem (21 % versus 1 %, p < 0.001) and were not associated with a 1p/19q co-deletion (0/6 versus 59/123, p = 0.03).

Discussion

The high rate of multicentric gliomas, the identification of IDH mutations in all of the gliomas in which it could be fully assessed and the identical IDH mutations found in both the glioma and the skeletal tumor of two patients strongly suggest that, in most cases, similar to cartilaginous tumors, the development of gliomas in enchondromatosis patients results from somatic IDH mosaicism. The very rare IDH1 R132H and ATRX positive cells with a normal morphology found in patient 3 may further support this hypothesis. Consistent with the demonstration of occasional IDH mutated cells in the normal bone, blood and bone marrow of some enchondromatosis patients, we speculate that these cells may correspond to normal glial cells with IDH mutations [3, 36]. However, this observation must be taken with caution since morphology is subjective and ATRX expression can be heterogeneous [41].

IDH mutations are considered to be the earliest oncogenic events in the majority of lower grade gliomas [10]. In animal models, IDH mutations have been shown to be sufficient to induce enchondromas and chondrosarcomas [17, 29]. To our knowledge, this demonstration has not been previously reported for gliomas, but the occurrence of gliomas in enchondromatosis patients, in whom IDH mutations are thought to occur as an early post-zygotic event suggests that IDH mutations can initiate gliomagenesis [3, 36]. In addition, the median age at glioma diagnosis in these patients (25 years) suggests that IDH driven gliomagenesis is a very slow process. In contrast to enchondromas, however, IDH mutation alone is probably not sufficient to induce gliomas. Additional alterations, such as ATRX and TP53 mutations, are necessary [10]. Consistently, Moriya et al. identified a TP53 mutation in the glioma but not in an enchondroma of their patient, though both lesions shared the same IDH2 mutation [34]. This could explain why enchondromas appear much earlier than gliomas in enchondromatosis patients.

It is unknown why only a small percentage of enchondromatosis patients (approximately 5 %) develop gliomas [45]. It could be related to a variable distribution and proportion of mutated cells within the brain, to the fact that only a small percentage of IDH mutated glial cells acquire additional alterations, or to the type of IDH mutation present. In cartilaginous tumors of enchondromatosis patients the IDH1 R132C mutation is more frequent than the IDH1 R132H mutation (70 and 15 %, respectively) [3, 36], while the IDH1 R132H mutation is the most frequent mutation (90 %) in sporadic IDH mutated gliomas [50]. Enchondromatosis patients with an IDH1 R132H mutation could have a higher risk of developing gliomas than patients with an IDH1 R132C mutation. In our series, an IDH1 R132H mutation was present in 5 out of the 7 enchondromatosis patients who developed diffuse glioma.

Not surprisingly, like sporadic IDH mutated gliomas, enchondromatosis gliomas were frequently located in the frontal lobe and were more frequently diffuse low-grade or anaplastic gliomas than glioblastomas. However, they differed from sporadic IDH-mutated gliomas in several aspects. First, they were diagnosed at an earlier age. This difference could be explained by the fact that in sporadic IDH-mutated gliomas, the IDH mutation is acquired later than in enchondromatosis patients. Second, enchondromatosis gliomas, compared to sporadic IDH mutated gliomas, more frequently involved the brainstem, which may also be related to an earlier origin of enchondromatosis gliomas (i.e., during the first years of life, which is a period associated with the development of infratentorial gliomas). Since both H3-K27M mutations (which are present in most cases of children brainstem gliomas) and IDH mutations alter histone methylation, brainstem gliomas may therefore require histone modification of precursor cells at an early stage of development [9, 28, 30]. This particularity could help explaining why early acquisition of IDH mutations mimics the regional preference of H3-K27M mutated gliomas and why gliomas in that location are rare in adults. At last, none of the 6 enchondromatosis gliomas that could be tested were associated with a 1p/19q co-deletion. This observation needs confirmation in a larger series, however, again it could be related to a different timing of oncogenesis because the 1p/19q co-deletion is virtually absent in pediatric gliomas [42]. Consistently, in the present series, as in previously reported series, patients with 1p/19q co-deleted gliomas were older at diagnosis than those with IDH mutated non 1p/19q co-deleted gliomas [13].

Finally, multicentric gliomas were much more frequent in enchondromatosis than in sporadic IDH-mutated gliomas. Multicentric gliomas (i.e., with no apparent continuity between tumors) account for 2 to 5 % of sporadic gliomas [1]. Their oncogenesis is debated. In some cases, evidence has been provided that the multiple tumors actually originated from the same clone [1]. However, the multicentric presentation of gliomas reported in patients with inherited glioma predisposition and the high incidence of systemic cancers reported in patients with multicentric gliomas also suggest that multicentric gliomas may originate from different clones in patients who have inherited genetic alterations predisposing them to gliomas [4, 27]. Our study shows that somatic mosaicism might be another mechanism leading to the development of multicentric gliomas. It remains to be determined whether somatic IDH mosaicism might be responsible for the occurrence of multicentric gliomas in patients without enchondromatosis.

Conclusions

In addition to its retrospective design and its small sample-size, limitations of the present study include the limited number of patients in whom molecular characterization was possible, the absence of comprehensive molecular analysis and the absence of IDH mutation assessment in several gliomas from the same patient. Nevertheless, the analysis of glioma characteristics in patients with enchondromatosis sheds new light on IDH driven gliomagenesis. It provides evidence that the IDH mutation can initiate gliomagenesis and that the timing of IDH mutation acquisition might influence the location and molecular characteristics of gliomas.