Does endometrial morular metaplasia represent odontogenic differentiation?

Morular metaplasia (MorM) is a common type of altered differentiation found in endometrioid lesions [1]. The nature of MorM has not yet been defined: several authors support that it is an immature form of squamous differentiation (SqD), while others consider it a completely different entity [1‐8]. Morphologically, MorM has been defined by the typical syncytial appearance with bland nuclei and profuse cytoplasm. The lack of overt squamous features, such as prominent cell membranes and keratinization, has been used to distinguish MorM from conventional SqD [8]. However, we and others highlighted that MorM not uncommonly displays overt squamous features and a characteristic ghost cell keratinization [9‐11]. Immunophenotypically, MorM differs from conventional SqD. Squamous cell markers, such as p40, p63, and cytokeratin (CK) 5/6, can be expressed in MorM, although less commonly and intensely than conventional SqD. Instead, the expression of β-catenin (nuclear, due to CTNNB1 mutation), CD10, and CDX2 appears as a hallmark of MorM. Unlike SqD, MorM typically shows decreased CK7 expression compared to the background endometrium; CK8/18 expression, typically strong in both MorM and conventional SqD, may be lost in squamous cells originating from MorM [5, 6, 8, 9, 12‐15].

Based on the presence of nuclear β-catenin accumulation and the ghost cell keratinization, it has been suggested that MorM might represent a differentiation towards hair [10]. In fact, tumors originating from the hair matrix, such as pilomatrixoma and pilomatrix carcinoma, often show nuclear β-catenin expression (due to CTNNB1 mutation) and ghost cell keratinization (due to the presence of hard keratin). However, the described features are also present in tumors exhibiting odontogenic differentiation, such as calcifying odontogenic cyst and adamantinomatous craniopharyngioma (ACP) [16, 17]. In these tumors, the presence of ghost cells and hard keratin seems to represent an irregular deposition of tooth enamel matrix rather than aberrant hair formation [18, 19]. In spite of these similarities, a systematic comparison between MorM and hard keratin-expressing tumors has never been performed. Furthermore, to our knowledge, hard keratins have never been assessed in MorM. The aim of this study was to assess whether MorM may represent a differentiation towards hair or tooth, by comparing MorM to hard keratin-expressing tumors.

This study showed a wide morphological and immunophenotypical overlap between MorM and the WS of ACP; no morphological or immunophenotypical analogies were found between MorM and hair matrix tumors.

The nature of MorM has long since remained a mystery. On the one hand, several features have led to consider MorM as a putative immature form of SqD, e.g., the profuse eosinophilic cytoplasm, the expression of Bcl2 intermediate between glands and squamous areas, the focal presence of tonofilaments on electron microscopy [3, 8, 12]. On the other hand, the unique immunophenotype of MorM (nuclear β-catenin+, CD10+, CDX2+) contrasts with such hypothesis [1, 9, 15]. In addition, some results reported in the literature that might have been misleading in this field. It was reported that MorM showed evidence of neuroectodermal differentiation [7], although such findings have been considered inconclusive due to the inconsistency of the results [21]. Furthermore, several studies separated MorM from conventional SqD based on the absence of overt squamous features and keratinization [5‐8, 12, 13]. Due to this view, many cases of conventional SqD with positivity for the MorM markers (nuclear β-catenin, CD10, CDX2) have been reported [5, 12]. As discussed in our previous study, we believe that squamous/keratinizing features are not a criterion to differentiate between MorM and conventional SqD, since it can be observed in both entities. Instead, MorM seems to differ from conventional SqD based on the presence of the typical MorM cellularity with the typical MorM immunophenotype, the variable tendency to maintain a “morular” shape, and the ghost cell keratinization [9]. Therefore, we believe that part of the previously described cases of SqD with MorM immunophenotype, or of mixed MorM/SqD, may represent in fact MorM with squamous/keratinizing features. The evidence of squamous/keratinizing features within MorM appears in agreement with previous observations [10, 11].

The characteristic features of MorM, in particular the nuclear β-catenin accumulation and the ghost cell keratinization, suggest its similarity with hard keratin-producing tumors. Such tumors mainly include hair matrix tumors (i.e., pilomatrixoma and pilomatrix carcinoma) and odontogenic tumors (i.e., calcifying odontogenic cyst and ACP). Analogously to MorM, these tumors harbor CTNNB1 mutation (which underlies the nuclear β-catenin accumulation); the ghost cell is related to the presence of hard keratin [16, 17]. The analogy between MorM and hard keratin-producing tumors has previously been postulated by Tanaka, who supported that MorM could represent a differentiation towards hair [10]. However, contrary to it was previously suggested [17], hard keratin in odontogenic tumors seems to represent a component of the tooth enamel and not of hair [18, 19]. In fact, the presence of enamel-related proteins has been shown in the ghost cell of ACP [18]. Nonetheless, to the best of our knowledge, MorM has never been systematically compared to hard keratin-producing tumors.

In the present study, we did not find significant analogies between MorM and hair matrix tumors, since the latter ones showed diffuse nuclear β-catenin accumulation in the absence of the typical MorM cellularity. By assessing ACP, we noticed that the WS showed strong morphological analogies with MorM. Indeed, the rounded shape, the syncytial appearance, the bland round/ovoidal-to-spindled nuclei, and the profuse cytoplasm were identically observed in MorM and WS. The squamous/keratinizing features were also similar, with the presence of round/ovoidal clear cell with prominent membranes, overtly squamous cells in a “morular” arrangement and ghost cell keratinization. We noticed that, while the ghost cells of pilomatrixoma always were densely stratified, those of MorM and ACP often appeared as multiple single, round ghost cells within areas with syncytial cellularity. Although ACP and MorM also show layers of stratified ghost cells, Rumayor et al. showed that these were dyscohesive on electron microscopy, while the ghost cells of pilomatrixoma appeared tightly cohesive [22]; such ultrastructural finding is in agreement with our observation. The immunophenotype of MorM and WS was widely superimposable, with consistent β-catenin (nuclear), CD10, and CDX2 positivity that contrasted with the negative background. Interestingly, CDX2 was previously reported to be negative in ACP [23]. Instead, we found a CDX2 positivity limited to the WS, which could be missed at low magnification. The proliferation marker ki67 was low/absent in both MorM and WS. In endometrioid carcinomas, the low proliferation index of MorM contrasted with the highly proliferating glandular component. In ACP, the proliferation index was also low in the stellate reticulum around WS; however, one of us previously reported that ki67 expression remained low in WS even when the surrounding cells showed increased proliferation [20]. The CK pattern was also similar between MorM and WS, despite being less consistent than β-catenin, CD10, and CDX2; by contrast, hair matrix tumors showed weak/absent expression of CK8/18, which was the most strongly expressed CK in both MorM and WS. The only evident difference was p63, which was positive in MorM only in the presence of overt squamous features, while it was consistently positive in WS. However, such difference might be attributed to the difference between endometrioid carcinoma and ACP, since p63 is negative in the former and positive in the latter. In both endometrioid carcinoma and ACP, positivity for β-catenin (nuclear), CD10, and CDX2 was observed around several ghost cell-keratinizing areas, consistently with their derivation from MorM/WS [9]. In addition to these findings, most MorM cases in our series showed focal/multifocal positivity for hard keratin. Such positivity, although not comparable to that of ACP (which showed diffuse positivity in the ghost cells), was completely absent in all but one SqD case, which showed weak and focal staining. The described findings suggest that MorM of endometrioid carcinomas might be biologically similar to the WS of ACP. Interestingly, a recent study demonstrated that WS are analogous to structures termed “enamel knots,” which play a crucial role in the tooth development [19]. On the account of these findings, we suggest that MorM, just like WS, might be analogous to enamel knots. This would mean that MorM is a form of odontogenic differentiation and the keratinizing process found in MorM is an aberrant mimic of the normal tooth development. Remarkably, while MorM resembles enamel knots, the equivalents of other odontogenic components (such as stellate reticulum) are not found in endometrioid lesions. The reason why endometrium would differentiate into a specific odontogenic structure might lie in the different gene expression among different odontogenic components. In fact, enamel knots differ from the other odontogenic components by nuclear β-catenin accumulation and increased expression of several proteins (such as p21 and ectodysplasin A receptor) [19]. It is possible that endometrioid proliferations share the expression of key molecules with enamel knots; in such scenario, a specific molecular event (e.g., CTNNB1 mutation) might trigger the development of MorM. This would explain why MorM is considerably more common in endometrium than in other tissues. Further studies are necessary in this field.

MorM shows nuclear β-catenin expression and not rarely exhibits ghost cell keratinization, as observed in hair matrix and odontogenic tumors. However, we did not find striking morphological or immunophenotypical analogies between MorM and hair matrix tumors. On the other hand, MorM showed considerable morphological and immunophenotypical overlap with the WS of ACP. This may suggest that MorM mimics odontogenic differentiation, in particular the enamel knot of the normal tooth development. Further studies are warranted in this field.