Identification of different subgroups of gliomas based on the immune profile
Samples from 28 patients diagnosed with glioma (Table 1) were dissociated and analyzed by flow cytometry (Individualized data in Additional file 1). Tumors were classified based on the histology (GBM vs LGG) and based on the presence of IDH mutations. As expected, the majority of LGG were IDHmut (7/9), whereas only 3 out of 19 GBMs carried these mutations. The gating strategy described in Additional file 4 was used to characterize different immune populations. In agreement with the literature [14], there was a significant increase in the number of CD45 + cells in IDHwt GBMs compared to IDHmut GBMs or to LGGs (Fig. 1A). This increase was observed in both the lymphoid (Fig. 1B) and the myeloid (Fig. 1C) components. Among the IDHwt GBMs we identified a subgroup of tumors with a low content of CD45 + cells (less than 15% of the cellular suspension) (herein called GBMwt_lo) (Fig. 1D), with a very similar percentage of immune cells to the one measured in IDHmut tumors (either LGGs or GBMs) and very different from the other group of IDHwt GBMs (herein called GBMwt_hi), where CD45 + cells account for almost 50% of the tumor content (Fig. 1E). Similar differences between the defined glioma subtypes were obtained when we measured independently lymphoid (Fig. 1F) or myeloid (Fig. 1G) cells.
Table 1
Characteristics of the study population.
N
|
28
|
Age (years)
Median
Range
|
52 years
30–82 years
|
Gender
Female
Male
|
N = 12; 43%
N = 16; 57%
|
Grade of resection
Complete
Partial
|
N = 23; 82%
N = 5; 18%
|
KPS after surgery
100
90
80
≤ 70
|
N = 10; 36%
N = 9; 32%
N = 4; 14%
N = 5; 18%
|
Histological Diagnosis
Astrocitoma
Oligodendroglioma
|
N = 25; 89%
N = 3; 11%
|
Tumor grade
II
III
IV
|
N = 3; 11%
N = 6; 22%
N = 19; 67%
|
IDH 1/2
Mutated
w/t
|
N = 10; 36%
N = 18; 64%
|
ATRX
Mutated
w/t
|
N = 8; 28%
N = 21; 72%
|
MGMT
Methylated
Unmethylated
Unknown
|
N = 21; 72%
N = 2; 7%
N = 5; 21%
|
TERT promoter
Mutated
Wild Type
Unknown
|
N = 19; 62%
N = 10; 34%
N = 1; 4%
|
1st line therapy
Stupp (RT + temozolomide)
Temozolomide
RT + PCV
None
|
N = 21; 75%
N = 1; 4%
N = 2; 7%
N = 4; 14%
|
We performed an analysis of the genetic profile of the tumors. As expected, epidermal growth factor receptor (EGFR) alterations were enriched in IDH1wt GBMs, whereas TP53 mutations were common among the IDH1mut gliomas (Additional file 5A). However, we did not detect clear differences between the genetic profile of GBMwt_lo and GBMwt_hi tumors. Regarding the clinical data, patients carrying IDH mutations survived longer than the wild-type counterparts. However, there was no significant differences in the clinical behavior of patients from both GBMwt immune subgroups (Additional file 5B).
Characterization of the myeloid and the lymphoid compartments in the different subgroups of gliomas
In order to gain further insight into the composition of the immune infiltrate in the glioma subgroups, we dissected out the myeloid component in the tumor suspension. We combined the LGG and GBM IDH1mut (herein called IDHmut) for the subsequent comparisons with the other two groups of IDHwt GBMs. We observed that the percentage of neutrophils (Fig. 2A), myeloid derived suppressor cells (MDSCs) (Fig. 2B) and macrophages (Fig. 2C) was increased in GBMwt _hi compared to both, IDHmut and GBMwt_lo gliomas. We also analyzed the presence of CD206, a typical marker of alternatively activated (M2) macrophages. We found a higher proportion of CD11b + CD206 + cells in GBMwt_lo and GBMwt_hi, compared to IDHmut gliomas (Fig. 2D). Our panel was not designed to recognize specifically resident microglia, but we found no differences in the transcription of P2RY12, which is highly expressed on microglia, between the three subgroups (Additional file 6A). Moreover, ionized calcium binding adaptor molecule 1 (IBA1) positive cells were detected in high proportion in all the tumors analyzed (Additional file 6B). The number of microglial cells (Additional file 6C), as well as the total amount of IBA1 protein (Additional file 6D-E), was very homogenous among the different gliomas. These results suggest that the main differences in the immune compartment of the distinct glioma subgroups are due to the entrance of cells from the blood. Notably, the ratio of myeloid to lymphoid cells was lower in GBMwt_hi compared to the other two subgroups (Fig. 2E), suggesting that T cells infiltrate this subgroup of gliomas in particular. In agreement with that, we observed that the proportion of T cells (CD3+) (Fig. 2F), in particular the CD4 + subset (Fig. 2G), was higher in GBMwt_hi tumors than in the other two subclasses. However, there was an increase in the percentage of CD3 + cells in GBMwt_lo compared to IDHmut gliomas (Fig. 2F). Furthermore, there was no difference between the percentages of CD8 + cells between the two subgroups of GBMs (Fig. 2H), which was higher in both compared to IDHmut tumors. As a result, the CD4/CD8 ratio was lower in the GBMwt_lo compared to GBMwt_hi tumors (Fig. 2I). This ratio has been linked to the appropriate lymphocyte function in other types of cancer [15]. Moreover, the proportion of PD1 + cells, which labels T cell exhaustion, was higher in GBMwt_hi compared to GBMwt_lo gliomas (Fig. 2J), whereas the amount of regulatory T cells (Tregs) was similar in the two groups (Fig. 2K). By contrast IDHmut gliomas presented fewer exhausted T cells (Fig. 2J) and Tregs (Fig. 2K). Taken together, these findings highlight the important dissimilarities in the immune profile of IDHmut vs IDHwt gliomas and suggest that the GBMwt_lo subgroup resembles IDHmut gliomas in their percentage of myeloid cells. Besides, we found differences in the lymphocyte content and function between the two subgroups of GBMs.
Enrichment of programmed death ligand 1 (PD-L1) expression in the immune cells of highly infiltrated gliomas
Our flow cytometry analysis in gliomas revealed two levels of expression of PD-L1 (herein defined as PD-L1_lo and PD-L1_hi) (Fig. 3A), both in tumor (CD45-) and in immune (CD45+) cells. The expression profile was similar in GBMwt_lo and IDHmut tumors and very different from the GBMwt_hi gliomas (Fig. 3B), which showed a strong increase in the percentage of CD45/PDL1 double positive cells. The increment was significant in both the PD-L1_hi (Fig. 3C) and the PD-L1_lo (Fig. 3D) populations. Notably, there were no differences in the amount of tumor cells expressing high levels of PD-L1 among the different subgroups of gliomas (Fig. 3E). Altogether, our data suggest that there is a subgroup of IDHwt GBMs that contain a high immune infiltrate, enriched in myeloid cells and with a strong immunosuppressive profile: high content of MDSCs, CD206 + myeloid and PD-L1 + cells.
The immune stratification of the tumors correlates with different vascular phenotypes
It has been proposed that the three different transcriptomic subtypes of gliomas (Proneural, PN, Classical, CL and Mesenchymal, MES) are associated with a different immune microenvironment7. When we analyzed our cohort of gliomas using qRT-PCR we noticed that, as expected, PN and MES transcripts were enriched (Additional file 7A-C) and diminished (Additional file 7D-F), respectively, in IDHmut gliomas compared to their wild-type counterparts. However, there were no differences in the expression of PN (Additional file 7A-C) or MES (Additional file 7D-F) markers between the two subgroups of IDHwt GBMs. Therefore, neither the genetic (Additional file 5A) nor the expression profiles seem to explain the existence of two distinct immune patterns in the aggressive IDHwt gliomas.
In order to find disparities between the two subgroup of IDHwt GBMs that could correlate with their distinct immune landscapes we performed a macroscopic analysis of the tumors. Preoperative magnetic resonance imaging (MRI) revealed clear differences between IDHmut and IDHwt gliomas, especially in the contrast enhanced sequences (Fig. 4A) [16]. However, the T1 + C and T2 images of GBMwt_lo tumors were very similar to the ones obtained in the immune-high GBM subgroup (Fig. 4A). By contrast, the macroscopic analysis of different vascular features revealed that the blood vasculature score (unbiased annotations from neurosurgeons) and the number of vessels with a large lumen (herein called dilated blood vessels (BVs)) were higher in the GBMwt_hi tumors (Fig. 4B and Additional file 8). The IHC labelling of endothelial cells confirmed that the number of dilated vessels (Fig. 4C), as well as the CD34 density (Fig. 4D), correlated with the percentage of CD45 + cells measured by flow cytometry. To obtain an independent confirmation of these results, we performed a qRT-PCR analysis. We found a strong correlation between the CD45 content and the expression of the endothelial marker CD34 (Fig. 4E), as well as with the expression of CD248 (Fig. 4F) which is highly expressed by tumor-pericytes in gliomas [17]. Moreover, the transcription of CD34 (Fig. 4G), EMCN (another marker of endothelial tumor cells) (Fig. 4H) and CD248 (Fig. 4I) were increased in the GBMwt_hi group compared to the rest of gliomas. Notably, only the expression of CD248, was increased in GBMwt_lo tumors compared to IDHmut gliomas (Fig. 4I), which correlated with the higher CD248 score measured by IHC (Fig. 4B and Additional file 8), suggesting a direct correlation between the absence of IDH mutations and the increase in tumor pericytes, as we have recently described [11].
We have recently described that Tau, a protein associated with neurodegenerative diseases, is also expressed in glioma cells, particularly in the less aggressive tumors, where it obstructs glioma progression by blocking the formation of novel and aberrant tumor vessels. These effects were associated with a limited capacity of the gliomas cells to contribute to the pool of pericytes in Tau-high tumors, which results in a reduced amount of dilated BVs and a less efficient fueling of tumor growth [11]. We measured the amount of Tau in our cohort of gliomas and we observed that it accumulates in IDHmut gliomas (Fig. 5A-B). This result was not surprising given that the MAPT/Tau gene is epigenetically induced by the presence of mutant IDH proteins [11]. However, we also found an enrichment of Tau in GBMwt_lo compared to GBMwt_hi tumors (Fig. 5A-B), suggesting a possible correlation between this protein and the immune landscapes of gliomas. To test this hypothesis, we performed an in silico analysis of the TCGA database, which revealed a strong inverse correlation between the amount of MAPT/Tau transcription and overall survival or the expression of vascular- (CD34 and CD248) (Fig. 5C and Additional file 9A-B) and immune- (CD3, CD4, CD11b and CD68) (Fig. 5C and Additional file 9C-F) related genes. Notably, the transcript levels of MAPT/Tau and CD248 were inversely and directly correlated, respectively, with several of the signatures associated with different immune cell populations (Fig. 5C) and with the inflammatory- and cytokine-related pathways (Fig. 5D). We also analyzed which genes were downregulated in Tau-high gliomas and we found that many of them were linked to the immune response (Additional file 9G).
To get further insight we overexpressed Tau in GL261 cells, a well-known mouse glioma model. Tau overexpression reduced tumor growth (Additional file 9H) and increased survival (Fig. 5E) of mice bearing orthotopic tumors, which is in agreement with the increased survival of patients with low MAPT/Tau expression (Additional file 9I). The analysis of the tumors revealed a decrease in the amount of infiltrating CD3 lymphocytes, in parallel with a reduction in the number of dilated BVs in the Tau-overexpressing gliomas (Fig. 5F-G). The transcriptomic analysis of the tumor tissues confirmed the inhibition of the expression of vascular- (Fig. 5H) and immune- (Fig. 5I and Additional file 9J) related genes in GL261-Tau tumors, compared to their control counterparts. These findings suggest that Tau modulates both, the immune phenotype and the vascular features of gliomas.
We have previously shown that the Tau expression is induced by IDH mutations and repressed by wild-type IDH1 [11], which is upregulated in primary GBMs and promotes aggressive growth and therapy resistance [18]. Moreover, it has been shown that the expression of wild-type IDH1 reshapes the methylome and also affects gene expression [19]. In agreement with these data, we found that total IDH1 expression was upregulated in those GBMs with a higher immune content (Additional file 10A). This result suggests an explanation for the downregulation of Tau expression in the GBMwt_hi subgroup, which could be responsible, at least in part, for the increase in the vascular abnormalities and with the immune-enriched TME observed in these gliomas. However, we cannot discard that epigenetic changes induced by a higher amount of wild-type IDH could be affecting the expression of other immunomodulatory molecules as well. One such gene could be HLA-A, whose expression can be modulated by epigenetic mechanisms [20]. As a matter of fact, we observed a decrease in the amount of HLA-A transcription in the G-CIMP gliomas (Additional file 10B), a phenotype associated with the presence of IDH mutations. When we analyzed our cohort, we observed that HLA-A transcription was elevated in the GBMwt_hi subgroup, in comparison with the rest of gliomas (Additional file 10C). Moreover, we found a strong correlation between the expression of HLA-A and IDH1wt in the TCGA dataset (Additional file 10D), similar to the one observed between MAPT/Tau and IDH1. Taken together, our results suggest that the balance between mutant and wild-type IDH function in gliomas is controlling the expression of Tau, and probably other proteins, to shape the vascular and the immune niche of gliomas.