Epithelial Rests of Malassez Form a Wnt-Dependent Mechanoresponsive Network

ERMs in the Supracrestal Region Are Larger Than ERMs along the Root

The PDL and JE are anatomically distinct structures, but based on their shared attachments to a moving tooth, their responses to hyper-loading would likely be coordinated to provide appropriate tissue-level responses. We hypothesized that the 3-dimensional ERM meshwork was a candidate tissue for this task. ERMs were observed in the PDL space (Fig. 1G, H) and the supracrestal connective tissue, coronal to the alveolar bone crest (Fig. 1G, I), under normal masticatory loading conditions. The size of these anatomically distinct ERMs was quantified (see Methods and Materials), and those in the supracrestal region were consistently, significantly larger than those along the root (green bars, Fig. 1J).

Hyper-loading Triggers the Selective Expansion of Supracrestal ERMs

Larger ERMs were located in the supracrestal region, which experienced higher distortional strains compared to the PDL space (Fig. 1). To test whether the size or number of ERMs was influenced by strain, an in vivo hyper-loading model was employed (see Methods and Materials; Xu et al. 2019). The periodontium of teeth subjected to hyper-loading for 3, 7, 14, and 28 d was compared against baseline data from periodontia subjected to normal masticatory loading (Fig. 1J).

Within 1 d of hyper-loading, more ERMs were detected (yellow arrows, Fig. 1K; quantified in Fig. 1L). The total number of ERMs increased between days 1 and 3, as well as again between days 3 and 7 of hyper-loading (Fig. 1L). After 7 d of hyper-loading, the number of ERMs remained relatively constant and always significantly more than observed around normal loading controls (Fig. 1L).

Changes in ERM size were also observed. Under hyper-loading conditions, supracrestal ERMs were larger than root ERMs (compare Fig. 1M, N), and over the duration of hyper-loading, this difference in size became more pronounced (Fig. 1O). Thus, ERM number and size exhibited a time-dependent responsiveness to masticatory hyper-loading.

Hyper-loading by Itself Does Not Trigger Periodontal Inflammation

Factors other than mechanical force could influence ERM number and/or size. For example, hyper-loading could cause an inflammatory response (Leone et al. 2026), which indirectly affected the number and size of ERMs. To evaluate this possibility, the distribution and number of neutrophils and macrophages in the hyper-loading group were compared to controls subjected to normal masticatory loading. No apparent differences between hyper-loaded and control groups in terms of neutrophil and macrophage number and distribution in either the PDL space or the JE were observed (Appendix Fig. 1).

Hypo-loading Reduces the Size of Supracrestal ERMs

Hyper-loading correlated with an increase in the number and size of ERMs; we next tested whether hypo-loading had the opposite effect. After 4 wk of hypo-loading (see Methods and Materials; Zhang et al. 2019), ERMs along the root surface (Fig. 1P, Q) and ERMs in the supracrestal region (Fig. 1P, R) were significantly smaller compared to both controls and the hyper-loaded group (Fig. 1S). These data further supported the conclusion that the ERM network was responsive to changes in masticatory forces.

ERMs Are Mechanoresponsive

Do ERMs possess the molecular machinery to detect a physical force and convert it into a biochemical signal? Human ERMs expressed the mechanically gated ion channel Piezo1 (Fig. 2A); single-cell RNA sequencing (scRNA-seq) data confirmed expression of Piezo1 in human ERMs and in JEs (Fig. 2B). The mechanosensitive adhesion protein β4 integrin (Fig. 2C, D), the mechanosensitive linker protein plectin (Fig. 2E, F), and the mechanoresponsive ion channel TRPV4 (Fig. 2G, H) were also detected in human ERMs and in the JE.

OPEN IN VIEWER

Figure 2.

Epithelial rests of Malassez (ERMs) express mechanosensitive and mechanoresponsive proteins. Human ERMs express (A) PIEZO1, which was confirmed via (B) single-cell RNA sequencing (scRNA-seq) analysis. Expression of PIEZO1 in ERMs (green) was compared to its expression in the junctional epithelium (JE) (blue). (C) β4 INTEGRIN immunostaining in a representative human ERM. (D) scRNA-seq analysis confirmed expression in ERMs (green) and JE (blue). (E) PLECTIN immunostaining in a human ERM, confirmed by (F) scRNA-seq analysis. (G) Human ERMs express TRPV4, confirmed by (H) scRNA-seq analysis. Murine ERMs (arrows) express Piezo1 under (I) normal loading and (J) hyper-loading conditions. (K) Quantification of Piezo1. Murine ERMs (arrows) also express (L, M) β4 integrin, (N, O) plectin, (P, Q) TRPV4, (R, S) YAP, and (T, U) phosphorylated FAK under normal and hyper-loading conditions. (V) Phalloidin staining identifies F-actin in an ERM that (W) colocalizes with phospho-FAK, indicating (X) that ERMs are actively responding to mechanical stimuli under conditions of mechano-loading. Abbreviation: c, cementum; d, dentin. Asterisks as indicated in Figure 1.

Murine ERMs also expressed Piezo1 (Fig. 2I), and its expression was significantly elevated in load-activated ERMs (Fig. 2J, K). The mechanosensitive proteins β4 integrin (Fig. 2L, M) and plectin (Fig. 2N, O), as well as TRPV4 (Fig. 2P, Q) and YAP (Fig. 2R, S), were all expressed in ERMs under both normal and hyper-loading conditions, with expression being consistently higher in ERMs induced by hyper-loading.

Of particular note was the pattern of phosphorylated focal adhesion kinase (phospho-FAK) expression in ERMs under normal and hyper-loading conditions (Fig. 2T, U). Phalloidin staining identified F-actin⁺ microfilaments (Fig. 2V), and colocalization with phospho-FAK (Fig. 2V, W, X) provided strong evidence that ERMs were involved in transducing mechanical signals into biochemical signals.

ERMs Initiate Matrix Remodeling in Response to Hyper-loading

Thus far, our data demonstrated that ERMs possess the molecular machinery to sense and respond to mechanical forces; we postulated these forces were transmitted to ERMs via surrounding connective tissue fibers. Transmission electron microcopy illustrated the proximity of human ERMs to collagen fibers (Fig. 3A) and to oxytalan fibers (Fig. 3B). Histological staining for oxytalan fibers illustrated their close association with ERMs (Fig. 3C). βIII tubulin⁺ neurons encircled ERMs (red arrow, Fig. 3D). Expression of matrix metalloproteinase 13 (MMP13) in ERMs (Fig. 3E) suggested their local extracellular matrix was prone to remodeling.

OPEN IN VIEWER

Figure 3.

Remodeling of the extracellular matrix around epithelial rests of Malassez (ERMs) is stimulated by hyper-loading. Representative transmission electron microscopic images showing (A) collagen fibers and (B) oxytalan fibers encircling ERMs. (C) Human ERMs stained with oxytalan to illustrate circumferential oxytalan fibers. Immunostaining for (D) βIII tubulin; red arrow indicates a βIII tubulin⁺ neuron and (E) matrix metalloproteinase 13 (MMP13). Representative images from murine tissues (F) stained with picrosirius red and visualized under polarized light to visualize collagen fibers and (G, H) stained with Gomori’s aldehyde fuchsin to visualize oxytalan fibers. Immunostaining for (I) βIII tubulin (red arrow indicates a βIII tubulin⁺ neuron) and (J) MMP13. Under hyper-loading conditions, (K) collagen-stained fibers around ERMs appear fragmented. (L) Collagen density around ERMs as a function of the duration of hyper-loading. Under hyper-loading conditions, (M) oxytalan-stained fibers around ERMs appear truncated. (N) MMP13 expression in ERMs is significantly increased under hyper-loading conditions. (O) Quantification of MMP13 under normal loading and after 28 days of hyper-loading. Abbreviation: c, cementum; col, collagen; d, dentin; erm, epithelial rests of Malassez; je, junctional epithelium; oxy, oxytalan; PDL, periodontal ligament. Asterisks as indicated in Figure 1.

Similar relationships were noted between murine ERMs and picrosirius red–stained collagen fibers (Fig. 3F) and purple-stained oxytalan fibers, which are typically oriented perpendicular to collagen fibers (Fig. 3G, H). βIII tubulin⁺ neurons were in close proximity to murine ERMs (red arrow, Fig. 3I). Murine ERMs were also immunopositive for MMP13 (Fig. 3J).

Hyper-loading in the mouse model resulted in apparent collagen fiber micro-ruptures (Fig. 3K, Appendix Fig. 2); collagen content analyses indicated that continued hyper-loading was detrimental to fiber density (Fig. 3L). Fragmented oxytalan fibers were also evident around load-activated ERMs (Fig. 3M, Appendix Fig. 2). Load-induced damage to the extracellular matrix correlated with significantly increased expression of MMP13 in and around murine ERMs (Fig. 3N, O).

ERMs Release Wnt Signals That Influence the Mechanoresponsiveness of the PDL

To coordinate mechanosensitive responses in the PDL, supracrestal connective tissue, and JE, we speculated that load-activated ERMs released signaling molecules that could locally influence cellular behavior. The following lines of evidence implicated Wnt/β-catenin signaling in this process.

The first line of evidence came from lineage-tracing the fates of Wnt-responsive cells in ERMs. Tamoxifen was delivered to the Wnt lineage tracer strain Axin2Cre^ERT2;R26R^mTmG/+. Three days later, small Keratin5⁺ ERMs, some of which costained with green fluorescent protein (GFP), were detected (Fig. 4A–A′′). Very few PCNA⁺ cells were detectable in the PDL (Appendix Fig. 3), indicating that under normal loading conditions, Wnt-responsive ERMs were generally quiescent.

The second line of evidence came from fate-mapping Wnt-responsive cells under hyper-loading conditions. Tamoxifen was delivered to Axin2Cre^ERT2;R26R^mTmG/+ mice, and immediately thereafter, hyper-loading was initiated. After 3 d, Keratin5⁺ ERMs were significantly larger (arrows, Fig. 4B; quantified in Fig. 1), and the number of Wnt-responsive cells increased (Fig. 4B′, B′′), which corresponded with an increase in PCNA⁺ cells in the PDL (Appendix Fig. 3). Most Wnt-responsive cells were positioned adjacent to the cementum, and some colocalized with the Keratin5⁺ ERMs (Fig. 4B–B′′).

OPEN IN VIEWER

Figure 4.

Supracrestal epithelial rests of Malassez (ERMs) adjacent to the junctional epithelium (JE) show the most significant enlargement in response to hyper-loading. Axin2Cre^ERT2;R26R^mTmG/+ mice received tamoxifen and 3 days later immunostaining for (A) Keratin5 and (A′) GFP identified (A′′) Wnt-responsive ERMs under normal loading conditions. Axin2Cre^ERT2;R26R^mTmG/+ mice under conditions of hyper-loading for 3 days, followed by immunostaining for (B) Keratin5 and (B′) GFP to identify (B′′) Wnt-responsive ERMs. Under hyper-loading conditions for (C, C′, C′′) 7 days and (D, D′, D′′) 14 days. Seven days after tamoxifen delivery, comparison of β-catenin immunostaining in ERMs of (E) control and (F) Wls mutant mice. Expression of Piezo1 in ERMs of (G) control and (H) Wls mutant mice; quantified in (I). Expression of YAP in ERMs of (J) control and (K) Wls mutant mice; quantified in (L). Expression of β4 integrin in ERMs of (M) control and (N) Wls mutant mice; quantified in (O); and in the JE of (P) control and (Q) Wls mutant mice; quantified in (R). Supracrestal ERMs in (S) control and (T) Wls mutant mice; size quantified in (U). Picrosirius red-stained collagen in the periodontal ligament (PDL) space of (V) control and (W) Wls mutant mice. Dotted white lines indicate root surfaces. Abbreviations: c, cementum; d, dentin. Asterisks as indicated in Figure 1.

After 7 d of hyper-loading, the number and size of ERMs continued to significantly increase (arrows, Fig. 4C–C′′; quantified in Fig. 1); load-activated ERMs maintained their Wnt-responsive status (Fig. 4C–C′′). GFP⁺ Wnt-responsive cells remained associated with the enlarged Keratin5⁺ ERMs throughout longer periods of hyper-loading (Fig. 4D–D′′). Together, these data strongly suggested that Wnt signaling was involved in ERM responses to hyper-loading.

To directly test whether Wnt signals emanated from ERMs, a K14Cre driver was crossed with a Wntless^fl/fl strain. GFP immunostaining in K14Cre^ERT2;R26R^mTmG/+ mice confirmed tissue specificity of the Cre to ERMs (Appendix Fig. 4) and the JE, and then K14Cre^ERT2;Wls^fl/fl mice were generated. The chaperone protein Wntless (Wls) is essential for Wnt protein secretion (Banziger et al. 2006), and immunostaining for β-catenin confirmed that canonical Wnt signaling was inhibited in ERMs of K14Cre^ERT2;Wls^fl/fl mice (compare Fig. 4E, F).

A third line of evidence that Wnt/β-catenin was involved in ERM signaling came from molecular analyses. Seven days after tamoxifen delivery, expression of Piezo1 (Fig. 4G–I) and YAP (Fig. 4J–L) was lost in Wls mutant ERMs. β4 integrin expression was absent in Wls mutant ERMs and in Wls mutant JEs (compare Fig. 4M, N and 4O, P; quantified in Fig. 4Q, R). These data demonstrated that Wnt/β-catenin signaling was required for ERM mechanosensitivity, as well as JE attachment.

Our fourth line of evidence supporting an essential role for Wnt/β-catenin signaling in ERM mechanotransduction came from analyses of Wls mutant mice after initiating hyper-loading. Normally, supracrestal ERMs enlarge in response to hyper-loading; in Wls mutants, these supracrestal ERMs did not expand (compare Fig. 4S, T; quantified in Fig. 4U). In addition to defects in the JE and supracrestal region, PDL fibers deteriorated. Picrosirius red–stained collagen fibers in the PDL of control mice exhibited their characteristically dense, linear, elongated morphology (Fig. 4V). In Wls mutants, PDL fibers were sparse, fragmented, and truncated (Fig. 4W).

Functionless Relics or an Essential Mechanosensory Element in the Periodontium?

The vestige of (Hertwig’s) epithelial root sheath, ERMs emerge during tooth development and endure for the lifetime of most mammals (Reitan 1961). Embryonic tissues are equipped to remove redundant cells (Abud 2004); consequently, this persistence of an ERM network argues against them being simply “developmental jetsam” (Cohen 1976). We started this study by asking, Why does this ERM network persist? And what are the circumstances that provoke ERM reactivation?

The ERM Network Is Ideally Positioned to Detect and Respond to Mechanical Input

Our hypothesis was that ERMs function as a mechanotransducing network, which is supported by 4 lines of evidence. First, ERMs are embedded in a mechanoresponsive extracellular matrix (ECM). ERMs in supracrestal connective tissues are encircled by collagen and oxytalan fibers, and because the JE and supracrestal fibers are firmly attached to the tooth surface, these tissues, along with supracrestal ERMs, experience stresses and strains created by masticatory loading (Fig. 1). The same relationship holds true in the PDL space: because of its attachment to the tooth surface, the PDL and root ERMs experience stresses and strains in response to masticatory forces (Fig. 1). Thus, the ERM network appears to be ideally positioned to detect and respond to changes in mechanical loading of the dentition.

ERMs Are Responsive to Changes in Masticatory Load

Hyper-loading of the dentition triggers an increase in the size and number of ERMs; conversely, hypo-loading results in a significant reduction in ERM size (Fig. 1). This change in size is most significant in the supracrestal region, where distortional strains are twice as high as in the PDL (Fig. 1). These experiments demonstrate a correlation between mechanical loading and ERM biology.

Hyper-loading is not the sole stimulus that leads to ERM activation. ERM “reactivity,” especially in the supracrestal region, has been observed after severing supracrestal fibers in a gingivectomy procedure (Ramfjord et al. 1966), after orthodontic tooth movement (Gilhuus-Moe and Kvam 1972; Talic et al. 2003), after ligature-induced periodontitis (Manisagian et al. 2018), and in a variety of genetic models of periodontal disease (Sojod et al. 2017; Wang et al. 2018). A shared feature of all these conditions is that they disrupt the mechanical integrity of the periodontium. Thus, our proposed function for the ERM as a mechanosensory network aligns with ERM activation in these other models of injury and/or disease.

ERMs Possess the Molecular Machinery for Mechanotransduction

The third line of evidence supporting our hypothesis that ERMs function as a mechanotransducing network is the fact that ERMs express multiple proteins with well-described mechanosensitive and mechanoresponsive functions (Fig. 2). ERM-specific expression of mechano-sensitive and mechano-responsive proteins significantly increases in response to hyper-loading (Fig. 2). The fact that hyper-loading led to increased expression of the mechanosensory proteins further suggests that ERMs adjust this molecular machinery to adapt to a changing mechanical environment, perhaps via a positive feedback loop. Other epithelial tissues harbor such a positive feedback loop involving Piezo1 and Wnt/β-catenin signaling (He et al. 2024).

ERMs Function as an Essential Wnt-Dependent Mechanotransducive Network

Several studies confirm that Wnt signaling is directly regulated by mechanical cues (Przybyla et al. 2016; Zhou et al. 2020; Li et al. 2021; Oak et al. 2025), and our fourth line of evidence supporting a mechanoresponsive function for ERMs came from studies involving manipulation of Wnt/β-catenin signaling in ERMs. Fate-mapping experiments demonstrated that ERMs and, significantly, load-induced ERMs are Wnt-responsive (Fig. 4).

The essential role of Wnt signaling in ERM mechanotransduction was demonstrated by genetically deleting Wls from K14-expressing cells. This strategy resulted in loss of mechanosensitive protein expression in ERMs (Fig. 4). ERMs typically expand in response to hyper-loading (Fig. 1), but Wnt-deficient supracrestal ERMs failed to enlarge, demonstrating that they were no longer sensitive to mechanical input (Fig. 4). In the Wls mutant, PDL fibers were thin, short, and sparse, and the JE began to lose its attachment (Fig. 4). Together, these data support our conclusion that the ERM network functions as a Wnt-dependent mechanosensor.

Limitations of the Current Study

In this study, we integrated analyses conducted in human tissues with analyses of murine tissues, and one might legitimately ask whether tissues from 2 species should be compared. There are multiple variables to consider: for example, it was not possible to ascertain if patients from which ERMs were collected had normal masticatory function or not, whereas a mouse model permitted control over variables such as masticatory loading. There is also the issue of age: in humans, ERM numbers change with age (Simpson 1965), and in our study, murine age was kept constant. Consequently, we are comparing not only different species but also tissues from different life stages. Further, human samples come from a genetically diverse population, whereas our murine population is far more genetically homogeneous.

Despite these and other differences, there are 2 major justifications for undertaking such comparisons. First, in both species, ERMs share a developmental origin (e.g., Hertwig’s Epithelial Root Sheath [ERS]). This shared origin suggests a conserved biological identity and, potentially, conserved functions. Second, the signaling pathways controlling tooth development and periodontal homeostasis (e.g., Wnt, Notch, Shh, BMP) are highly conserved across mammals; consequently, it is reasonable to presume that the mechanisms controlling ERM quiescence/activation are substantially the same in both species. Finally, human tissues allow us to make observations about ERMs, from which we can make hypotheses that are testable within a mouse model. If our experimental data support a particular hypothesis, we may then be able to formulate therapeutic strategies based on those animal models that have potential value in treating human disease.