Both spinal ependymomas (SPE) and myxopapillary ependymomas (MPE) exhibit a characteristic methylation profile [6]. However, for a significant portion of SPE, many pathologists have reported conflicting results between morphologic diagnosis and the DNA methylation-based classification [9, 11]. Indeed, a study by the German Glioma Network reported that nearly one-third of the histologically diagnosed spinal ependymomas were assigned by methylation to the MPE class [14]. In the current study, we address this topic and focus on SPE cases exhibiting a methylation profile of MPE.

We performed immunohistochemical, AI-assisted morphological, as well as methylation analyses on a cohort of 100 MPE and SPE. In addition, mass spectrometry-based proteomic analysis was conducted on 54 of these tumors, including 16 MPE, 23 SPE and 15 discrepant cases. A detailed description of methods is provided in supplementary materials. Data refinement included filtering for precursor peptides and proteins with q-values > 0.01. Next, low-quality samples with less than 5000 identified proteins were excluded. Additionally, proteins with a high incidence of missing values (i.e., more than 50% in both MPE and SPE) were excluded from the analysis. This approach removed 10 tumor samples, with the final cohort constituting 16 MPE, 17 SPE, and 11 discrepant cases. In these samples, we identified by MS/MS an average of 69,658 (± 12,648) peptides and 8029 (± 758) proteins per sample at a false discovery rate (FDR) lower than 1% at both peptide and protein level.

Pearson correlation matrix showed higher similarity between discrepant cases and MPEs, compared to discrepant cases and SPEs (Fig. 1). In agreement, principal component analysis (PCA) revealed weakly predominant clustering of the discrepant cases with MPEs (Fig. 1). In addition, proteomic analysis identified HOXB13 as the most differentially expressed protein between MPE and SPE. HOXB13 was indeed highly prevalent in all MPEs (16 out of 16 samples) and not identified in any of the SPEs (0 out of 17 samples) (Supplementary Fig. 1). Immunohistochemical staining in a cohort of 100 tumor samples demonstrated 100% sensitivity and 100% specificity (Supplementary Figs. 2 and 3). Differently from previous reports, we did not observe weak or focal staining of HOXB13 in the tumor samples [2, 8, 11, 13]. This can be attributed to the usage of a different antibody compared to previous research (Supplementary Materials). HOXB13 showed strong and diffuse positive staining in all discrepant cases, thus perfectly aligning with proteomic data.

Pearson correlation matrix (left panel) and principal component analysis (right panel) of proteomes of 44 tumor samples

Next, we investigated the morphological features of the cohort. The extracellular proteoglycan versican (VCAN) was overrepresented in the myxoid matrix of MPE (Supplementary Fig. 1). VCAN immunohistochemistry showed excellent alignment with Alcian blue stain; however, VCAN immunohistochemistry demonstrated a more prominent and reliable staining. Employing digital pathology tools and VCAN immunohistochemistry, we demonstrated that discrepant cases frequently contain myxoid foci (Supplementary Figs. 4–6 and Supplementary Table 1).

CNV summary plots of MPEs and SPEs were compatible with previous reports [7]. Noteworthy, discrepant cases appeared to exhibit a lower rate of chromosomal aberrations in comparison to MPEs and SPEs. Overall, the summary CNV plot of discrepant cases was more similar to MPE than SPE (Supplementary Fig. 7).

The morphologic and methylation-based classification difference between MPE and SPE has been the subject of several studies [9, 11, 14]. Also, the observation of strong expression of HOXB13 in a fraction of ependymomas with a focus on MPE has been previously reported [2, 3, 5, 8, 11, 13]; thus, our observations corroborate previous reports. As a unique feature, the present study focuses on the conflicting morphological and methylation-based classification of MPE and SPE, while proposing a systematic approach to resolve this discrepancy.

Previous studies observed a differential expression of HOXB13 in morphologically diagnosed MPE [2, 13]. A subsequent study demonstrated strong HOXB13 expression predominantly in MPE, but not in other spinal ependymomas or astrocytomas [8]. After the advent of methylation-based classification, a considerable fraction of morphologically unequivocal SPE produced an mcMPE profile, as reported in several studies [9, 10, 14].

Altogether, the different layers of information demonstrated that discrepant cases show a pattern more similar to MPE than SPE. Our results speak in favor of extending the spectrum of tumors diagnosed as MPE. The diagnosis of an MPE should include the discrepant cases with an SPE-like morphology, but with nuclear expression of HOXB13, and a methylation profile of myxopapillary ependymoma. Adoption of such a pipeline in clinical practice would require modifications of the WHO definition and the essential and desirable diagnostic criteria for these tumors.

It should be noted that cauda equina neuroendocrine tumor, a tumor within the extended differential diagnosis of MPE and SPE, also exhibits nuclear expression of HOXB13 [1, 4, 12].

We have not observed a classical MPE without strong nuclear HOXB13 expression. We therefore propose classifying tumors that resemble morphologically SPE, but exhibit an mcMPE profile or have nuclear expression of HOXB13 as MPE irrespective of their morphological appearance. This would imply considerable changes to the criteria provided by WHO for the diagnosis of MPE. A suggestion is presented in Fig. 2.

Proposal for diagnostic criteria of myxopapillary ependymoma