The clinicopathological features of EPN-ZFTA in previous studies were compared with those in the present study. In our study, cysts were observed in 8/9 cases, which was higher than that (12/19 cases) in a previous study on EPN-ZFTA [4]. Although male or pediatric predominancy, which was similar to previous studies, was also observed, it is important to note that this predominancy remains controversial [19, 20]. The frequency of EPN-ZFTA among ST-EPN was 33% (10/30) in the present study, which was similar to a previous one reporting 40% (17/42) [20].

The prognosis of EPN-ZFTA has not yet been established. According to Branger et al. [4], 5-years PFS and OS rates were 67.5 and 72.2%, respectively. Pajtler et al. [21] reported 5-years PFS and OS rates of 29 and 75%, respectively, and concluded that patients with EPN-RELA (EPN-ZFTA) had a dismal prognosis. Although the prognosis of EPN-ZFTA was better in the present study, the number of cases examined was small. Although the treatment, such as GTR/STR and radiotherapy, may underlie the favorable prognosis noted in the present study, it is difficult to explain why the prognosis of patients differed between the present study and previous ones because the details of treatment thereof were not described. In addition, previous studies did not search for CDKN2A gene abnormalities, which may have been associated with the worse prognosis of patients. Although a relationship was previously reported between Ki-67 LI ≥ 7% and the poor prognosis of patients with EPN [22], other studies suggested that the WHO grading and MIB-1 LI were not associated with a poor prognosis [8, 21]. Therefore, the prognostic value of proliferative indices, such as MIB-1 LI and the WHO grading, remains still unestablished. In the present study, all cases classified as WHO grade 2 relapsed, whereas 50% (3/6 cases) classified as WHO grade 3 did not. Moreover, all cases with low MIB-1 LI (< 8.6%) relapsed, whereas 42% (3/7 cases) with high MIB-1 LI (≥ 18.8%) did not. Therefore, the present results highlight the difficulty of using MIB-1 LI and/or the WHO grading as prognostic factors for future recurrence. The median OS of GBM and PFA-EPN was worse than that of EPN-ZFTA, and no deaths or recurrences were observed for SC-EPN in the present study. In general, it was reported that a better prognosis was observed for SC-EPN compared to EPN-ZFTA and PFA-EPN [23]; however, it should be noted that the number of cases is too small in the present study, especially for SC-EPN and PFA-EPN. The prognosis of GBM is, as is sufficiently expected, miserable compared to SC-EPN, PFA-EPN and EPN-ZFTA [24].

The present study indicates the utility of MTAP and p16 IHC as surrogate markers for CDKN2A HD [12]. MTAP is the key enzyme in the methionine salvage pathway [10] and is normally expressed in various tissues; however, its expression was previously shown to be lost in some tumors [25, 26]. The p16 protein plays an important role in blocking the G1 to S phase transition via the inhibition of CDK4 and CDK6 [12, 27] and is encoded by CDKN2A (9p21). The gene coding MTAP is located on 9p21, which is only 165 kb telomeric to CDKN2A. In a previous study, IHC revealed that GBM cases with CDKN2A HD did not express p16 [9]. Satomi et al. [12] showed the potential of the deficiency of MTAP expression by IHC as a predictive surrogate marker for CDKN2A HD in IDH-mutant astrocytomas. In the present study, we examined MTAP and p16 IHC in EPN-ZFTA with or without CDKN2A HD, and found that neither MTAP nor p16 was expressed in the two cases with CDKN2A HD. The three cases that were positive for MTAP or p16 harbored the CDKN2A hemizygous deletion. The relationship between the CDKN2A hemizygous deletion and the expression of MTAP or p16 has not yet been reported for ependymal tumors. The frequency of CDKN2A HD in ST-EPN was previously reported to be 9.6% (3/31 cases) [28]. In the present study, the frequency of CDKN2A HD in EPN-ZFTA was higher at 20% (2/10 cases). Since non-ZFTA fusion cases were included in the previous study, the frequency of CDKN2A HD may have been lowered by the inclusion of non-ZFTA fusion cases. CDKN2A HD was previously, in fact, detected in EPN-ZFTA [10]; Junger et al. [10] detected CDKN2A HD in 9 out of 54 (16.7%) EPN-ZFTA cases, which is consistent with the present results. On the other hand, extracranial metastasis is rare in ependymal tumors. In a previous study that searched 258 cases of EPN, extracranial metastasis was only observed in 5 cases, which was equivalent to 2% [29]. CDKN2A HD was also detected in a case of extracranial metastasis including lung metastasis [30]. In the present study, one case (case 1) had lung metastasis and also harbored CDKN2A HD. The other case with CDKN2A HD (case 2) developed recurrence at a relatively early stage after surgery, despite the fact that there was no MVP or necrosis, being histologically classified into WHO grade 2. In conjunction with an animal study on adult mice showing a significantly worse prognosis in those with CDKN2A HD [31], these results indicate a relationship between CDKN2A HD and the poor prognosis of EPN-ZFTA. Regardless of the morphological features such as MVP, necrosis, and the WHO grading, EPN-ZFTA with CDKN2A HD showed a poor prognosis represented by death or early recurrence in the present study; furthermore, characteristic histological features indicating the presence of CDKN2A HD were not identified in these EPN-ZFTA cases. These findings suggest the necessity of searching for CDKN2A HD regardless of the histological features or the WHO grading in EPN-ZFTA. Based on previous findings and the present results, we propose a practical flowchart for the stratification of EPN-ZFTA (Supplemental Fig. 4), which is applicable to the routine diagnostic practice of EPN-ZFTA. Cases expressing MTAP and p16 may be regarded as non-CDKN2A HD; MLPA to detect CDKN2A HD is thus unnecessary. Cases with the loss of both MTAP and p16 expression may be regarded as CDKN2A HD; although MLPA to detect CDKN2A HD is not necessary, it is better to perform it to confirm IHC results. On the other hand, MLPA is essential for the following cases: (1) negative for MTAP or p16, and (2) showing inadequate immunostaining for MTAP (for example, when physiologically positive cells are also negative) or p16 (when nuclear staining is uneven and, thus, assessments with a low magnification are difficult).

A previous study confirmed that PD-L1 IHC, such as SP263, was expressed on tumor cell membranes, but not on immune cells [32]. Among L1CAM-positive ST-EPN cases, 40% (6/15 cases) were positive for SP263 and 20% (3/15 cases) for E1L3N. The latter antibody, E1L3N, which was used in the present PD-L1 IHC study, yields positive immunoproducts on tumor cell membranes [13]. Another study clearly identified PD-L1 in supratentorial EPN-RELA using various techniques, including Western blotting, flow cytometry, and IHC, and all ten cases were positive for SP263 [33]. The reason for the absence of PD-L1 immunopositivity in the present study remains unclear; however, it is unlikely to be due to insufficient immunohistochemical techniques because we detected the expression of PD-L1 in positive control tissue sections.

B7-H3, also known as CD276, is a transmembrane protein from the B7 family that was initially reported in 2001 [34]. It is regarded as a third group of immune checkpoints [35]. Although the ligand for B7-H3 has not yet been identified, it is known to be expressed on T-cells, B-cells, and dendritic cells [34, 36]. The expression of B7-H3 was previously shown to be higher in the tumor vascular endothelium than in the normal endothelium and was associated with tumor proliferation [37, 38]. B7-H3 expression was detected in cases of medulloblastoma and GBM and correlated with a worse prognosis [14, 39, 40]. Although the expression of B7-H3 has not yet been reported in EPN-ZFTA, we herein confirmed its expression in EPN-ZFTA, indicating possible involvement of the immune checkpoint molecules in EPN-ZFTA, and found an inverse correlation between its expression levels and the number of infiltrating CD8-positive lymphocytes. CD8-positive lymphocytes are cytotoxic T-cells associated with a favorable prognosis in gliomas [13]. As CD8-positive lymphocytes were suppressed by immunosuppressive function of B7-H3 as was reported previously [41], the inverse correlation between CD8-positive lymphocytes and B7-H3 would indicate the usefulness of B7-H3 as a target of immune checkpoint chemotherapy for EPN-ZFTA via B7-H3 pathway. Similar findings have been reported for GBM [14]. In GBM, the expression of MYC regulates cell differentiation and high expression levels were associated with high tumorigenicity [40]. Another study confirmed that the knockdown of B7-H3 regulated the differentiation of GBM by modulating MYC expression [42]. Therefore, B7-H3, which modulates MYC expression, has potential as a target of immune checkpoint chemotherapy in EPN-ZFTA. Regarding lymphocyte markers such as CD3, CD8, and CD20 in the present study, the case of EPN-ZFTA without CD3-positive and CD8-positive lymphocytes (case 4) harbored relatively a favorable prognosis. On the other hand, two out of the three cases with CD20-positive lymphocytes also harbored relatively a favorable prognosis (cases 6, 8). While a probable better prognosis without CD3-positive and CD8-positive lymphocytes as in case 4 is inconsistent with a previous report [13], that with CD20-positive lymphocytes as in cases 6 and 8 is consistent with a previous report [13]. After all, the story behind B7-H3 and/or the presence or absence of lymphocytes in the immune microenvironment of EPN-ZFTA is likely not so straightforward, but given the present results, futher investigations into the prognostic values of B7-H3 and/or lymphocytes are warranted for EPN-ZFTA.

Macrophages infiltrating cancer tissues are called tumor-associated macrophages (TAMs). Iba-1 is a pan-histiocytic marker for both M1 and M2 macrophages, and CD204 is an M2 macrophage marker. While M1 TAMs are induced by lipopolysaccharide and interferon-γ stimulation, regulating acute inflammation, M2 TAMs exert anti-inflammatory effects, play roles in tissue remodeling and angiogenesis, and contribute to tumor growth [43,44,45,46]. Although TAMs were often historically described in the context of M1/M2 activation, single-cell RNA sequencing has revealed the complexity of macrophage responses, moving us beyond the linear M1/M2 activation paradigm in GBM [47]. Previous studies revealed that the infiltration of M2 macrophages was associated with the higher malignancy and worse prognosis of astrocytic tumors, including GBM [15, 48]. The number of TAMs in gliomas may be associated with the glioma type and grade; in a study that examined astrocytic tumors and oligodendrogliomas, the number of microglia was significantly higher in astrocytomas than in oligodendrogliomas, regardless of whether they were grade II or III [49]. Limited information is currently available on TAM in EPN. Nam, et al. [13] showed a high ratio of CD163/CD68 + cells (cut-off value of 1.007), indicating that M2 macrophages were associated with a poor prognosis in EPN patients. In previous studies, the prognosis in SC-EPN was better than PFA-EPN and EPN-ZFTA [22], and the prognosis of GBM was the worst among SC-EPN, PFA-EPN, and EPN-ZFTA [24], while the expression of Iba-1 and CD204 was the highest in GBM and the lowest in SC-EPN in the present study. The quantitative results of TAM among EPN tumor subtypes and GBM in the present study indicates the possibility for the prognostic value of TAM; this is because the possible correlation between poor prognostic tumor types and TAM expression was observed (Fig. 2). To confirm further the prognostic value of TAM in EPN-ZFTA, more detailed prognostic analyses in association with the amount of TAM expression would be an interesting issue. Moreover, a positive correlation between B7-H3 expression and M2 macrophage expression was reported in a previous study [50]. Therefore, a more detailed understanding of the relationship between B7-H3 expression and TAM subsets in the immune microenvironment of EPN-ZFTA will surely contribute to the development of new therapies for EPN-ZFTA.

As has been discussed above, the prognostic value of lymphocytes or TAMs is probable in EPN-ZFTA; in this regard, the association of lymphocytes or TAMs with CDKN2A HD, the latter of which is likely a potential prognostic determinant of EPN-ZFTA according to the present study, is an interesting issue. But the pathological behavior of T-lymphocytes/B-lymphocytes or TAMs was seemingly not different between EPN-ZFTA with CDKN2A HD and those without it. Furthermore, the small number of EPN-ZFTA with CDKN2A HD, which was just 2 in the present study, made it all the more difficult to evaluate the differences in the qualitative and quantitative behavior of T-lymphocytes/B-lymphocytes or TAMs among EPN-ZFTA cases with or without CDKN2A HD. Thus, the exact relationship between the immune microenvironment (lymphocytes or TAMs) in EPN-ZFTA and CDKN2A genetic abnormality still remains to be established. But since the issue of the immune microenvironment and CDKN2A abnormality in EPN-ZFTA, and its association with the prognosis of the tumor, are surely interesting, it awaits further studies in the future.