2.1. The MGMT Promoter Is Variably Methylated in Various Non-Neoplastic CNS Diseases
We investigated autopsy and biopsy samples of healthy and diseased brains in terms of MGMT promoter methylation. Our analysis included brains without evidence of any pathological changes (“healthy controls”), as well as brains with infectious non-demyelinating diseases, i.e., bacterial, mycotic, inflammatory, viral (induced by HSV or HIV), or parasitic (toxoplasmosis), those with inflammatory-demyelinating diseases, i.e., multiple sclerosis (MS) and progressive multifocal leukoencephalopathy (PML), as well as those with non-neoplastic and non-inflammatory CNS conditions that damage the myelin sheaths, i.e., central pontine myelinolysis (CPM), extrapontine myelinolysis (EPM) and Wallerian degeneration. The degree of MGMT promoter methylation was compared with that in healthy controls (Table 1
, Figure 1
) as well as in hypermethylated gliomas (Supplementary Table S1
). The mean percentage of methylated cytosine residues, measured at five CpG sites in each sample, was 5.6% in healthy controls (SD 0.88). The highest methylation rate in our controls was 6.6%. A sample was defined as “hypermethylated” if the mean methylation rate of all five CpG sites exceeded 7.7%, which was equivalent to the third quartile of the methylation rate in all non-neoplastic samples (Table 1
, Figure 1
The methylation rate of the MGMT promotor varied significantly across all samples, including those that were non-methylated as well as those with prominent hypermethylation. We identified MGMT promoter hypermethylation not only in samples from patients with MS and PML, but also in those with metabolic and degenerative diseases, such as EPM, CPM and Wallerian degeneration (Table 1
, Figure 1
). Isolated samples did exhibit methylation rates similar to those in moderately to highly hypermethylated gliomas (Supplementary Table S1, Supplementary Figure S1
). The hypermethylated samples could not be assigned to distinct pathological conditions, and, in addition, diseases associated with hypermethylation also included non-methylated samples.
Interestingly, no samples from the group of infectious non-demyelinating diseases had MGMT promoter methylation rates higher than those in healthy controls, which stands in contrast to the findings in extra-CNS samples. In fact, all samples with hypermethylation were found in the groups with inflammatory demyelinating diseases or non-inflammatory metabolic and degenerative diseases accompanied by damaged myelin sheaths (Figure 1
The extent of hypermethylation showed no significant correlations with the age or sex of the patients (age: Pearsons r = −0.03; p
= 0.82; sex: Pearsons r = 0.15; p
= 0.18) or with the total number of apoptotic cells measured by immunohistochemical staining for caspase-3 (Pearsons r = 0.23, p
= 0.24). These results are shown in Supplementary Table S2
. In order to rule out any influence of cause of death on the extent of methylation, we categorized our autopsy samples into five groups concerning their cause of death, namely (1) multiorgan failure (n
= 7), (2) sepsis (n
= 4), (3) cardiac reasons (n
= 5), (4) pulmonary reasons (n
= 14) and (5) others (n
= 13). The means of methylation of those groups were compared to each other via ANOVA afterwards. We did not find any significant differences between the groups (p
> 0.05). Therefore, the (reason of) death itself does not seem to affect MGMT methylation (Supplementary Figure S2
It is worth mentioning that MGMT promoter methylation rates differed significantly between samples obtained from MS patients by biopsy or autopsy (biopsy: mean 4.13%, SD 0.8 vs. autopsy: mean 8.72%, SD 9.8; p
< 0.01). All deceased MS patients whose autopsy samples we analyzed had suffered from long-standing chronic MS, whereas most patients who had been diagnosed with MS by brain biopsy had a short history of symptoms and disease course (i.e., a few months maximum; the only exception was MS sample 8). In each case, the biopsy had been performed because a tumor had been suspected. Therefore, it seems plausible that MGMT promoter methylation is associated with long-standing and chronic MS disease rather than with more acute and active forms. Of note, the autopsy samples were from MS patients significantly older than those who had had a brain biopsy (autopsy: mean 56.1 years, SD 11.5 vs. biopsy: mean 35.6 years, SD 15.4; p
< 0.01). However, since there was no association between age and MGMT promoter methylation across all analyzed samples (Supplementary Table S2
) it is much more likely that the activity status of the disease is responsible for the differences in MGMT methylation.
Since 50% of the hypermethylated samples (ten out of 20 samples) were linked to non-inflammatory diseases, such as CPM, EPM and Wallerian degeneration, an association of MGMT promoter methylation to inflammatory infiltrate per se and, in particular, to specific inflammatory cells (e.g., lymphocytes, granulocytes, plasma cells) was not expected.
However, since injured axonal networks display the fundamental commonality of all hypermethylated samples, we suspected a link between axonal damage and MGMT promoter methylation. SMI31 is a marker for phosphorylated neurofilaments which display the integrity of axonal networks [19
]. We therefore stained all samples against SMI31 and correlated the density of SMI31-positive phosphorylated neurofilaments to MGMT promoter methylation rate. Nonetheless, we were unable to verify our hypothesis in this investigation (Pearsons r = 0.05; p
= 0.78). However, this negative result might be due to relatively small differences in methylation levels compared with healthy controls, as well as a heterogeneity of hypermethylation and non-methylation within one disease entity.
Since the degree of hypermethylation was relatively low, we next wanted to investigate whether the slight differences might be reflected in lowered MGMT mRNA and protein expression levels. To begin with, we attempted to quantify MGMT mRNA levels in formalin fixed paraffin embedded (FFPE) samples from healthy controls and those with hypermethylation using qPCR. Suitable results were only obtained for a few samples (healthy control (control 1–4); CPM/EPM: CPM 2, Wallerian degeneration: WAL 2, WAL 3), and a reliable statistical assessment was not possible due to the small sample size. Nonetheless, MGMT mRNA levels tended to be lower in hypermethylated samples compared with healthy controls (Wallerian degeneration: 1.36-fold lower levels, CPM/EPM: four-fold lower levels) (Supplementary Figure S3
In a next step, all samples were stained immunohistochemically against MGMT for further analysis of MGMT protein expression levels. For the quantification of MGMT expression in glial cells, we automatically measured the numbers of MGMT-positive glial cells per mm2
. Of note, small differences in methylation rates of the MGMT promoter resulted in significantly lower MGMT protein expression levels (Figure 2
). Healthy controls (Figure 2
C) showed the constitutive expression of MGMT in almost all glial cells, i.e., astrocytes and particularly in oligodendrocytes, as well as in neurons, while the staining intensity of glial cells was significantly reduced in all samples with hypermethylated MGMT promoter (Figure 2
< 0.01, Wallerian degeneration not significant due to small sample size). Indeed, the hypermethylated samples showed only scattered MGMT-expressing glial cells in the areas of the lesions (Figure 2
D–H), while the neurons showed stable staining intensities across all diagnostic groups (Figure 2
C–H, insets; for quantification see Supplementary Figure S4
), serving as an internal positive control and thereby ruling out staining artifacts. The number of MGMT-positive glial cells was found to be independent of the extent of myelination in MS lesions (chronic inactive plaque (CIAP) vs. remyelinated shadow plaques) (Figure 2
D,E) (no significant difference between the groups).
2.2. The Protein Expression of Demethylase TET1 Is Associated with MGMT Promoter Methylation
After realizing that MGMT promoter hypermethylation is a frequent phenomenon in non-neoplastic CNS diseases, the question arose regarding the underlying mechanism. Enzymes of the TET protein family are fundamental players in stabilizing DNA methylation patterns by regulating DNA demethylation. TET1 and TET2 are important members of this enzyme family, which are expressed in brain tissue [20
]. Bearing in mind that DNA-hypermethylation is often a result of an imbalance of DNA methylation and demethylation which can be based on reduced TET1 protein activity as shown in gliomas, we investigated the protein levels of TET1 and TET2 via immunohistochemistry. Since MGMT expression differences have only been demonstrated in glial cells but not in neurons (see Supplementary Figure S4
) and since diseases with disintegration of the myelin sheaths are located in the white but not the grey matter, we analyzed and in particular quantified TET expression in glial cells only.
Healthy controls showed a broad spectrum of nuclear TET1-staining intensity with numerous highly positive glial cells (i.e., high TET1-expressing cells) as well as few faintly positive glial cells (i.e., low TET1-expressing cells) (Figure 3
A,C). Hypermethylated samples from various conditions, however, showed numerous glial cells with only weak or even completely absent TET1 positivity and only small numbers, if any, of highly positive glial cells (Figure 3
A, D–H). Since neurons showed stable staining intensities across all diagnostic groups (Figure 3
D–H, insets), they served as an internal positive control and hence staining artifacts were ruled out. Therefore, we were able to demonstrate an overall reduction in TET1 protein expression levels in the different kinds of conditions showing MGMT promoter hypermethylation (Figure 3
A–H). Notably, TET1 expression levels were independent of the degree of myelination in MS lesions (CIAP vs. shadow plaque) (Figure 3
D,E), similar to the staining results of MGMT (Figure 2
It is noteworthy that the percentage of high TET1-expressing glial cells correlated negatively with the MGMT promoter methylation rate (Figure 3
B, lower panel), while conversely, that of low TET1-expressing glial cells (Figure 3
B, upper panel) was positively correlated with it.
In contrast to TET1, there were no significant correlations between TET2 expression and MGMT promoter methylation (Supplementary Figure S5