Gemistocytic astrocytoma represents a small subgroup accounting for approximately 5% of IDH mutant diffuse astrocytic gliomas. The WHO classification defines gemistocytic astrocytoma by the presence of “a conspicuous, though variable, fraction of gemistocytic neoplastic astrocytes” [7]. These cells should account for at least 20% of tumor cells. The term “gemistocytic” was coined by Nissl in 1904 for cells with homogeneous, faintly eosinophilic cytoplasm with short branching processes. It originates from the German word “gemästet” (filled, stuffed, swollen), sometimes also referred back to the Greek “gemistos” with similar meaning.

The neoplastic nature of gemistocytic cells in astrocytoma could be clearly demonstrated by binding of an isocitrate dehydrogenase 1 (IDH1) R132H mutant protein specific antibody [1]. However, so far no molecular drivers characteristic for gemistocytic differentiation have been identified. Previous molecular analyses reported higher frequencies of TP53 and PTEN mutations [16], lower frequency of IDH mutations, and alterations in RRAS and ERCC1 [9]. Several studies found shorter progression-free survival in gemistocytic astrocytoma patients as compared to fibrillary astrocytoma patients [5, 10,11,12]. However, there appears to be a contradiction between accelerated progression and lower proliferative activity of gemistocytes compared to other tumor cells in the same sample, or in general in diffuse glioma [3,4,5, 17].

To identify characteristic genetic alterations associated with gemistocytic morphology, we performed high-throughput, high-resolution, genetic and epigenetic analyses on a set of 24 gemistocytic astrocytomas. In all samples, gemistocytic cells constituted more than 20% of tumor cells (on average 59%). The control group consisted of 47 IDH mutant diffuse astrocytomas WHO grade II, 104 IDH mutant anaplastic astrocytomas WHO grade III, and 293 IDH wild-type glioblastomas WHO grade IV. The Illumina Infinium HumanMethylation450 BeadChip (450 k) array data was used for methylation profiling and to calculate a genome-wide copy number profile (CNP) as previously described [14]. Targeted resequencing was performed as reported previously [13].

A distinctive feature of tumors with gemistocytic histology was a recurrent focal copy number gain in the telomeric region of chromosomal arm 12p. The size of the involved regions varied between cases. However, cyclin D2 (CCND2) was constantly affected in all cases (Fig. 1a). An integrated analysis of copy number variation (copy number algorithm by Stichel et al. in preparation) in all gemistocytic astrocytomas in the cohort also indicated 12p as the most commonly altered locus (Fig. 1b). In particular, focal gain of CCND2 and adjacent regions was seen in 8 of 9 (89%) gemistocytic IDH mutant astrocytomas WHO grade II and in 13 of 15 (87%) IDH mutant anaplastic astrocytomas with distinct gemistocytic morphology (Table 1). In contrast, this alteration was observed in only 5 of 47 (11%) IDH mutant fibrillary astrocytomas WHO grade II and in only 19 of 104 (18%) IDH mutant anaplastic astrocytomas lacking gemistocytic morphology. The alterations detected by analysis of copy number plots based on 450 k analysis were confirmed by fluorescence in situ hybridization (FISH) in a subset of 18 cases (11 gemistocytic, 7 fibrillary), yielding concordance in 17 of 18 (94%) cases (Fig. 1c). The single non-concordant case was an anaplastic gemistocytic astrocytoma without indications of chromosome 12 gain by 450 k, but low-level gain detected by a FISH probe directed against 12p12 encompassing CCND2. This difference was highly significant within grade II and grade III gliomas, respectively (each p < 0.0001, Fisher’s exact test). In addition, the event of 12p/CCND2 gain was significantly associated with gemistocytic histology over the entire diffuse astrocytic glioma cohort (p < 0.0001, Table 1). Unsupervised clustering of methylome profiles did not separate gemistocytic from fibrillary astrocytoma (data not shown).

Fig. 1
figure 1

Representative copy number profile of a gemistocytic astrocytoma (a). Integrated copy number analysis across all gemistocytic astrocytomas (b). Fluorescence in situ hybridization of a case with amplification (left) and low-level gain of CCND2 (middle, higher magnification right, c)

Full size image
Table 1 Subtypes of diffuse glioma and 12p status
Full size table

To assess the mutational landscape of gemistocytic astrocytoma, 17 cases for which enough DNA was available were subjected to panel sequencing. All cases harbored IDH1R132 mutations (16/17 IDH1R132H, 1/17 IDH1R132G) and TP53 mutations. ATRX was mutant in 10 of 17 cases. Other genes recurrently affected by single-nucleotide variants or small insertions/deletions were ALK (2/17), CSF1R (2/17), FGFR1 (2/17), GSE1 (2/17), MSH6 (2/17), NF1 (2/17), and SMO (2/17). However, since no germline control was available and none of these variants included known hot-spot mutations, no conclusions can be drawn regarding the somatic status of these variants. These findings are in line with studies describing high rates of TP53 mutations in gemistocytic astrocytomas, but contrast reports on high frequencies of PTEN mutations [16] of which none was found in the present set. Recent reports suggested aberrations of copy number and methylation of ERCC1 and RRAS as a possible marker for gemistocytic astrocytoma [9]. However, we could not detect these aberrations in our dataset (mean beta values for ERCC1 promoter sites 0.09 and 0.1, for RRAS promoter sites 0.08 and 0.09, in gemistocytic and control samples, respectively).

The function of the CCND2 gene product cyclin D2 in cell-cycle regulation might also explain previous cytological observations in gemistocytic cells. Disruption of the regular cell-cycle by higher abundance of cyclin D2 might prevent the transition from S to G2 phase, resulting in lower mitotic activity despite higher pleomorphism and higher number of multinucleated cells. Also, activating CCND2 mutations are an alternative for PI3K/AKT/mTOR pathway activation in altering neural cell differentiation [8]. In turn, the activation of the mTOR pathway has previously been implicated in the increase of cell size in subependymal giant cell astrocytoma and balloon cells in focal cortical dysplasia [2, 6]. Thus, it is attractive to speculate that gain of CCND2 might determine gemistocytic cytology along similar mechanisms. However, further studies are warranted to explore the exact functional consequences of CCND2 gain. Moreover, the recent approval of inhibitors of the CDK4/6 axis [15], both interacting with CCND2, also opens a possible additional therapeutic approach for this glioma subtype.