TMEM2: A New Dimension in Hyaluronan Biology

Abstract

Hyaluronan (HA) is one of the most abundant components of extracellular matrices. HA is a huge polysaccharide — a single linear HA polymer often exceeds 25,000 disaccharide units in length (~10⁷ Da) and occupies the volume of a 300 nm diameter sphere. These unique biochemical and biophysical properties are accompanied by extremely rapid turnover of HA, which emphasizes the importance of not only its biosynthesis but also degradation in regulating the homeostasis and biological functions of HA. Further supporting the specific importance of HA degradation, a large body of evidence demonstrates that biological functions of HA are dependent on its size and degree of fragmentation. While considerable research has revealed the roles of the HYAL family hyaluronidases in HA catabolism and biology, the discovery of TMEM2 as a functional cell surface hyaluronidase, coupled with increasing data demonstrating its remarkable biological functions, have added a new dimension of research to the field of HA biology.

1 |. Introduction — Mechanisms of HA Degradation at the Cellular and Organismal Levels

A general model for systemic HA turnover, as we know it today, is based largely on classical biochemical and metabolic studies performed during 1980s and 1990s, highlighted by contributions from Torvard Laurent and colleagues at the University of Uppsala [1-8]. Results from these studies collectively suggested the systemic HA degradation proceeds according to the following scheme: (i) HA is released from the peripheral extracellular matrix (ECM) either enzymatically (by hyaluronidases) or non-enzymatically (e.g., by oxidants) and enters the lymphatic system; (ii) Most of the HA in the lymphatic system is partially degraded in lymph nodes into intermediate-sized fragments, which then drain into the general circulation; (iii) HA in the blood stream is taken up and further degraded predominantly by the liver, and finally excreted into urine [4]; and (iv) HA degradation in the liver is primarily mediated by sinusoidal endothelial cells [9], while the degradation in lymph nodes is thought to be carried out by lymphatic endothelial cells [10]. Remarkably, these pre-molecular biology studies revealed an extremely rapid systemic HA turnover. In skin, containing half of the HA of the body, HA turns over in 1–2 days. The half-life for HA injected into the blood stream is 2–5 min and one third of total body HA turns over daily [2,4,11,12].

Our understanding of the mechanisms of HA degradation at the cellular level remains less complete. It has been well established that the final step of HA degradation (i.e., complete depolymerization into monosaccharides) occurs in lysosomes via the concerted actions of lysosomal hyaluronidase, β-d-glucuronidase, and β-N-acetyl-d-hexosaminidase [13]. On the other hand, earlier steps in the HA degradation process are less clearly defined. Since high-molecular weight (HMW) HA molecules are synthesized by hyaluronan synthases at the plasma membrane and extruded directly into the extracellular space, HA molecules to be broken down in lysosomes must be internalized at some point during the degradation process. While multiple cell surface proteins, including CD44 [14] and Stabilin-2 (also known as hyaluronan receptor for endocytosis or HARE) [15], have been suggested to play a role at the internalization stage, molecular mechanisms by which HA is endocytosed still remain unclear. Another unresolved issue concerning HA degradation at the cellular level has been the identity of the hyaluronidase(s) that mediates the initial fragmentation of extracellular HA. These two issues are intricately linked; both problems need to be addressed to fully understand the molecular mechanism of cellular HA degradation.

2 |. HA Degradation on the Cell Surface

The involvement of hyaluronidase activity on the extracellular side of the plasma membrane in cellular HA catabolism was first proposed by Robert Stern and his colleagues [16]. Stern's model postulates that the catabolism of extracellular HA is initiated by the action of a lipid raft-associated hyaluronidase on the cell surface, which he speculated to be HYAL2. HA fragments thus generated are internalized and transported ultimately to lysosomes, where a lysosomal hyaluronidase, presumably HYAL1, further degrades the fragments into oligosaccharides. This model is mainly based on the finding in 2001 that HYAL2 is expressed as a GPI-anchored protein in transfected cells [17]. However, there have been several observations inconsistent with the notion that HYAL2 is a "full-time" cell surface hyaluronidase. Harada et al. [18] reported that stable transfection of HEK293 cells with HYAL2 alone does not endow the cells with the capability of degrading HA added to culture media; cotransfection with CD44 is required for the acquisition of cell surface HA-degrading activity. Other data also suggest that the cell surface is not a constitutive site of HYAL2 expression and action. For example, endogenous HYAL has been identified in intracellular compartments in some studies [19,20], and like lysosomal enzymes, HYAL2 favors acidic conditions for optimum activity [21,22] (Table 1). These results suggest that the actual mechanism of cell surface HA degradation is not as simple as the Stern model postulates.

Table 1.

Comparison of biochemical and biological properties between TMEM2, the HYALs, and HYBID.

Hyaluronidase	TMEM2	HYAL1	HYAL2	SPAM1/PH-20	HYBID	References
Protein molecular mass (human)	154,374 Da	48,368 Da	53,860 Da	57,848 Da	152,998 Da	GeneCards
Primary subcellular localization	Cell surface (transmembrane)	Lysosome	Lysosome Cell surface (GPI)	Acrosome of sperm cells	Secretory	[17,19,20,23,24,28]
Optimum pH of the enzyme	pH 6–7	pH 4.0	pH 3.7	pH 4.0 (53 kDa isoform) pH 7.0 (64 kDa isoform)	N/A (no intrinsic catalytic activity)	[11,21,22,28,114]
Substrate specificity of the enzyme	HA	HA CS	HA CS	HA CS	N/A ¶	[23,28,115,116]
Phenotype of KO mice – Embryonic lethality	Die by E10.5 (null) Die by E13.5 (NCC-specific CKO)	Born alive (null)	Born alive (null)	Born alive (null)	Born alive (null)	[51] Murao et al. §
Phenotype of KO mice – Overt developmental phenotype	Gastrulation defects (null) Craniofacial and cardiovascular defects (neural crest CKO)	No	Short nose (in a C57BL/6 background) Expanded heart valves; heart hypertrophy (in a mixed background)	No	No	[36,47,51] Murao et al. §
Phenotype of KO mice – Blood HA accumulation	~40-fold within 3 weeks of induced KO	2–5 fold in adult null mice	10–20 fold in adult null mice	N.D.	N.D.	[35,38,47]
Phenotype of KO mice – Adult phenotype	Not studied yet	Osteoarthritis	Cardiopulmonary dysfunction (mixed background)	Reduction in age-related cartilage degeneration	Memory impairment; Shorter long bones	[36,37,45,46,117]

^¶HEK293 cells transfected with HYBID selectively degrade HA, but not chondroitin sulfate, dermatan sulfate, or heparan sulfate [23].

^§Murao, A. Irie, F., Tobisawa, Y., and Yamaguchi, Y.: manuscript in preparation. See text for more information.

N.D., not determined; HA, hyaluronan; CS, chondroitin sulfate; KO, knockout; CKO, conditional knockout; NCC, neural crest cell; E, embryonic day.

Other than HYAL2, the secretory protein HYBID (hyaluronan-binding protein involved in hyaluronan depolymerizarion; also known as KIAA1199; official gene symbol CEMIP) has been reported as a candidate that mediates extracellular/cell surface HA degradation [23]. According to this study, HYBID-mediated HA degradation requires the presence of cells with functional clathrin-coated pit pathway — neither recombinant HYBID nor conditioned media from HYBID-transfected HEK293 cells exhibit HA-degrading activity [23,24]. This indicates that HYBID itself is not a hyaluronidase but participates in HA degradation in an unknown manner, perhaps as an HA-binding protein. Interestingly, mutations in the CEMIP gene have been implicated as the cause of non-syndromic deafness [25].

3 |. TMEM2 is a Hyaluronidase that Functions on the Cell Surface

TMEM2 (transmembrane protein 2; official gene symbol CEMIP2) was originally identified as an open reading frame that encodes a large transmembrane protein. Its extracellular domain has a 48% overall amino acid homology with HYBID. The biological significance of this putative protein was first demonstrated by the discovery that zebrafish mutations at the tmem2 locus (wickham and frozen ventricle) induce developmental heart defects characterized by abnormal heart looping and endocardial cushion development [26,27]. Subsequently, Yamamoto et al. [28] demonstrated that TMEM2 is a hyaluronidase that functions on the cell surface. Specifically, this study showed that: (i) TMEM2 is expressed as a type II transmembrane protein; (ii) Transfection of TMEM2 confers HEK293 cells with the ability to cleave HMW HA into fragments up to 5–10 kDa; (iii) Unlike HYAL family hyaluronidases, which degrade not only HA but also chondroitin sulfate and prefer an acidic pH for the optimum activity, the enzymatic activity of TMEM2 is specific for HA with a pH optimum of 6–7; and (iv) TMEM2-transfected cells degrade substrate-immobilized HA in a contact-dependent manner [28]. Together, these initial data have suggested that TMEM2 is a strong candidate for the physiological cell surface hyaluronidase. Data subsequently obtained further demonstrate that TMEM2 has biochemical and biological properties that are distinct from those of the HYAL family hyaluronidases and HYBID (Table 1), presumably reflecting TMEM2's unique functional significance as a "professional” cell surface hyaluronidase.

Until recently, there has been no direct evidence demonstrating that the TMEM2 protein itself has intrinsic HA-degrading activity. The fact that TMEM2 is a bona fide hyaluronidase possessing intrinsic catalytic activity has now been demonstrated using highly purified, recombinant soluble TMEM2 [29]. Purified TMEM2 protein exhibits HA-degrading activity irrespective of the species of origin (human and mouse) and the position of epitope tag insertion (N- and C-terminal tags). These studies also demonstrated that the HA-degrading activity of recombinant TMEM2 is comparable to that of recombinant HYAL2, and that recombinant TMEM2 degrades not only fluorescence-labeled (i.e., chemically modified) HA but also unlabeled native HMW HA [29].

Prior to the demonstration of intrinsic enzymatic activity of recombinant TMEM2, there have been conflicting data as to whether TMEM2 possesses intrinsic catalytic activity. Hohenester and colleagues [30] reported that recombinant human TMEM2 they produced did not show HA-degrading activity in in vitro assays. Sato et al. [31] reported that they could not detect HA-degrading activity of human TMEM2 using a cell-based assay with transfected 293T cells, although they could show HA-degrading activity with mouse TMEM2 in the same assay. A series of control experiments have now resolved the causes of these inconsistencies [29]. Briefly, these inconsistencies appear to be due to the quality of fluorescence-labeled HA products used and the presence of unidentified inhibitor(s) in cell-based assays (see Narita et al. [29] for details). In fact, by optimizing assay conditions, Hohenester's group has now confirmed that their recombinant human TMEM2 possesses HA-degrading activity (Hohenester, E., author comment published on June 9, 2024; see the updated version of Ref. [30]; the relevant experimental data are available from Dr. Hohenester). Additionally, in collaboration with our laboratory, Yamada's group at Meijo University has analyzed the mode of TMEM2-mediated HA cleavage and showed that TMEM2 is an endo-β-N-acetylglucosaminidase cleaving the β-1,4-glycosidic linkage between N-acetylglucosamine and glucuronic acid (Dr. Shuhei Yamada, personal communication).

While the available data indicate that the cell surface is the physiological site of TMEM2 action, it is an open question whether TMEM2 is released from cells and acts in the extracellular space under certain non-physiological conditions. Hogan et al. [32] reported that TMEM2 is present in exosome-like vesicles from urine, and that this level is 2.1-fold higher in individuals with autosomal dominant polycystic kidney disease with PKD1 mutations. This observation suggests that exosome-mediated shedding of TMEM2 may also occur under other pathological conditions, such as cancer. It is also interesting whether exosome-associated TMEM2 can function as a hyaluronidase at remote sites.

4 |. Structure of TMEM2

A crystal structure of the ectodomain of TMEM2 has recently been determined by Hohenester's group [30], which matches remarkably well with the model predicted by AlphaFold (AlphaFold DB: Q9UHN6). According to these models, the TMEM2 ectodomain is composed of a main body of the protein, which is a 70 Å-long right-handed parallel β-helix, and two small domains [30] (SD1 and SD2) (Figure 1). The β-helix is a structural motif often found in bacterial polysaccharide-degrading enzymes [33]. The dominant feature of the TMEM2 surface is a large groove that spans the width of the main body. Interestingly, the crystal structure of bee venom hyaluronidase, an ortholog of the mammalian HYAL family hyaluronidases, is also characterized by the presence of a single pronounced groove, although its shape is not very similar to the large groove of TMEM2. In bee venom hyaluronidase, the groove serves as the substrate-binding site and also contains the catalytic residues [34]. This suggests that the large groove of TMEM2 may serve similar functions, although there are currently no data supporting or refuting this assumption.

Figure 1 |. Structure of TMEM2.

(A) The primary domain structure of TMEM2. (B,C) The space-filling and ribbon diagrams of the 3D structure of the TMEM2 extracellular domain predicted by AlphaFold [118,119] (AlphaFold DB: Q9UHN6). In A-C, the same coloring scheme is used to represent individual domains. Red rectangles in B indicate the large groove in the main body.

While SD1 and SD2 are located at opposite ends of the linear protein sequence, they are spatially proximal to each other in the 3D structure (Figure 1). Both SD1 and SD2 are lectin-like domains [30], but whether these domains actually bind HA or any other glycans have not been tested. SD1 contains an arginine residue (Arg-265), which positionally corresponds to Arg-187 of HYBID. This arginine residue in HYBID is the site of mutations identified in familial non-syndromic deafness [25,28], and mutagenesis of this residue to cysteine, as seen in the cases of non-syndromic deafness, results in a reduction, but not total abrogation, of the HA-degrading activity of HYBID-transfected cells [23]. Mutagenesis of Arg-265 and a few other neighboring residues in TMEM2 similarly impairs the HA-degrading activity [28]. This suggests that SD1 is somehow involved in the enzymatic function of TMEM2; however, the mechanism by which SD1 affects TMEM2 activity is not yet known.

5 |. TMEM2 in Systemic HA Catabolism

The physiological contribution of individual hyaluronidase species to systemic HA catabolism can be assessed by analyzing the accumulation of HA in bodily fluids and organs in mice knocked out for the respective hyaluronidases (Table 1, Phenotype of KO mice – Blood HA accumulation). Constitutive Hyal1 knockout mice have been shown to exhibit a modest 2-fold increase in serum HA [35]. On the other hand, constitutive Hyal2 knockout mice at the age of 4–7 months exhibit a >20-fold increase in serum HA [35,36]. HA levels in constitutive HYBID knockout mice have been examined only in hippocampal extracts, which exhibit a modest 2.5-fold increase [37]. These results indicate that, among these proteins, HYAL2 makes a substantially more significant contribution to systemic HA catabolism than HYAL1 and HYBID. It should be noted that, since all these studies were performed with adult constitutive knockout mice, the observed HA accumulation is a long-term consequence of embryonic loss of the respective hyaluronidases and therefore may involve secondary effects on the expression of other hyaluronidases and overall cell metabolism. To exclude these potential secondary effects, it is desirable to assess acute changes in the level of HA upon the induction of gene deletion. Toward this goal, Tobisawa et al. [38] employed a tamoxifen-inducible Tmem2 knockout model (CAG-Cre^ER;Tmem2^flox/flox) to determine the effect of global TMEM2 inactivation on HA accumulation shortly after the induction of gene inactivation. Remarkably, these mice exhibit a ~40-fold increase in plasma HA (~7500 ng/ml) within 3 weeks of tamoxifen-induced Tmem2 inactivation. This study also demonstrated that ongoing HA degradation in the liver and the lymphatic system is significantly impaired upon Tmem2 inactivation, and that TMEM2 is strongly expressed both in liver and lymphatic endothelial cells in wild-type mice [38]. These observations strongly suggest that TMEM2 is a major player in systemic HA catabolism. Thus far, comparable studies using inducible global knockout models have not yet been reported for other hyaluronidases. Comparison of HA accumulation in similar models for other hyaluronidases should help to understand the relative importance of individual hyaluronidase molecules in vivo.

6 |. TMEM2 in Embryonic Development

As mentioned above, studies in zebrafish first demonstrated that TMEM2 plays an essential role in embryonic development [26,27]. Since then, zebrafish TMEM2 has also been shown to play roles in other aspects of development, including muscle morphogenesis [39], angiogenesis [40], and atrioventricular canal development [41]. In these studies, excess HA deposition has been demonstrated in each of the relevant tissues [27,40,41], supporting the idea that zebrafish TMEM2 functions as a hyaluronidase.

HA has long been thought to play important roles in mammalian embryonic development. Studies using constitutive and conditional hyaluronan synthase-2 (Has2) knockout mice provided the first direct evidence for this speculation, revealing striking embryonic phenotypes in the cardiovascular [42], skeletal [43], and craniofacial tissues [44]. On the other hand, our understanding of the role of HA degradation in mammalian development is quite limited. This is at least partly due to the lack of remarkable embryonic phenotypes in knockout mice of previously known hyaluronidase genes; Hyal1, Hyal2, Spam1 (PH-20), and Cemip (HYBID) knockout mice are all born alive with undetectable or relatively modest developmental defects [36,45-47] (Table 1). One interpretation of these results could be that, unlike the synthetic arm of HA, the catabolic arm does not play any significant role in embryonic development. However, the rapid and dynamic changes in the intensity and spatial pattern of HA distribution that occur during early and middle gestational stages [48-50] suggest the involvement of HA degradation during these periods. Recent studies using Tmem2 mutant mice indicate that TMEM2 is the hyaluronidase that plays the key role in HA degradation during mammalian embryonic development.

Neural crest cells (NCC), which give rise to a variety of tissues, including cardiovascular and craniofacial tissues, arise from the edge of the neural tube and migrate long distances to the target sites. The NCC migratory route surrounding the neural tube contains high levels of HA [49]. Using NCC-targeted Tmem2 conditional knockout (Wnt1-Cre;Tmem2^flox/flox) mice, Inubushi et al. [51] demonstrated the essential role for TMEM2 in NCC migration and the subsequent development of NCC-derived structures. Wnt1-Cre;Tmem2^flox/flox embryos develop severe craniofacial and cardiovascular defects and result in embryonic death prior to 13.5 days of gestation. While NCC generation from the neuroectoderm in the neural tube is not affected in these embryos, the emigration of NCCs from the neural tube and their subsequent migration along the dorsolateral migratory route are both impaired [51]. This suggests that NCCs employ TMEM2 for their efficient migration. Additionally, this study revealed an unexpected effect of TMEM2 inactivation on the survival of NCCs. In Wnt1-Cre;Tmem2^flox/flox embryos, significantly increased numbers of apoptotic cells are observed in the nasal processes and branchial arches, which are major NCC target tissues [51]. This susceptibility of Tmem2-deficient NCCs to apoptosis might be a reflection of the reported protective effect of TMEM2 on ER stress [52] (see TMEM2 in cell stress response below).

Neural crest development is not the only developmental process in which TMEM2 plays an essential role. Our recent study has shown that Tmem2^−/− embryos deviate from normal development during gastrulation and totally degenerate by 10.5 days of gestation. These embryos exhibit conspicuous morphological defects in the ectoderm and mesoderm (Murao et al., manuscript in preparation), suggesting that TMEM2 has a quite fundamental function in embryonic tissue morphogenesis.

7 |. TMEM2 in Cell Adhesion and Migration

HA is a major factor in defining the biophysical and biological properties of the pericellular and extracellular matrices and has been thought to modulate cellular adhesive and migratory functions. Previous studies have nevertheless yielded diverse sets of data that fail to provide a unanimous mechanistic view of the function of HA and hyaluronidases in cell adhesion and migration. For example, HA can be both adhesive and anti-adhesive depending on the cellular and experimental contexts [53]. On the one hand, HA supports adhesion of cells, mostly leukocytes and lymphocytes, that express CD44 [54-56]. On the other hand, the presence of high levels of HA in the pericellular and extracellular matrices is inhibitory to cell adhesion [57-62]. This anti-adhesive effect is thought to be primarily due to steric exclusion by a gel-like HA matrix, which could interfere with direct engagement of cell surface adhesion receptors, such as integrins, with their ligands. In this latter context, it can be envisioned that highly migratory cells, such as invasive tumor cells, may employ the cell surface HA degradation machinery to remodel anti-adhesive matrices surrounding them, thereby promoting adhesive interaction with the matrix and migration. The impaired migration in Tmem2-deficient NCCs in Tmem2 conditional knockout embryos (see above) appears to be consistent with this hypothetical model.

Irie et al. [63] demonstrated that a variety of tumor cells exhibit the ability to remove substrate-immobilized HA in a tightly localized pattern that corresponds to the distribution of focal adhesions (FAs) and stress fibers. In U2OS cells, this FA-localized HA degradation is mediated by TMEM2, but not by other hyaluronidases. TMEM2 is present in FAs and binds at least two integrins (α₅β₁ and α_Lβ₂) via direct interaction between their respective extracellular domains. Functionally, TMEM2 knockdown attenuates U2OS cell adhesion and migration on mixed 2D substrates consisting of type I collagen and HA [63]. These results support the aforementioned idea that tumor cells may utilize TMEM2-mediated HA degradation as a means of enhancing integrin-ligand engagement and consequent FA development (Figure 2). Also, these observations imply that cell surface TMEM2 may act as an ECM-remodeling enzyme in the mold of MT1-MMP (membrane type 1 matrix metalloproteinase; also known as MMP14). MT1-MMP promotes tumor cell migration and invasion via its actions on ECM proteins at points of cell-substrate contact [64]. In this vein, it is intriguing that there is increasing data showing association of TMEM2 overexpression with aggressive subtypes of cancers (see TMEM2 and cancer below).

Figure 2 |. A model for the role of TMEM2 in integrin-mediated cell adhesion and migration.

The graphical model shown here proposes a mechanism by which TMEM2 is involved in the initial assembly of nascent FAs and their maturation. The central tenet of the model is that TMEM2 removes HA at cell-matrix adhesion sites, thereby promoting integrin-ligand engagement and FA formation/maturation. (A) High levels of HA in the ECM are inhibitory to the direct engagement of integrins to their ECM ligands. (B) In the presence of TMEM2, HA in the ECM is locally removed. This create a microenvironment permissible to the direct integrin-ECM interaction, leading to the activation of integrins and the assembly of nascent FAs. (C) The physical association between TMEM2 and integrins [51] helps promote the formation of additional FAs and their maturation via further removal of HA in the vicinity of the integrin-ECM engagement. (D) This in turn induces in the assembly of fully matured FAs that are connected to the actin cytoskeleton, integrin-mediated signaling, and downstream cellular responses. This cascade of events is thought to occur not only in stationary cells but also in the leading edge of migrating cells. Note that it remains to be determined when the association between TMEM2 and integrins occurs in this process. It is possible that a fraction of TMEM2 constitutively associates with integrins or the association occurs at a certain stage in this process.

The functional significance of TMEM2 in cell-surface molecular interactions may not be limited to integrin-ECM interactions. It has been shown that thick pericellular HA coats, anchored to the cell surface via CD44, serve as barriers that obstruct physical contact of phagocytes to their targets and consequent phagocytic receptor engagement [65]. Thus, it is conceivable that, by modulating the property of the glycocalyx, TMEM2 may also have functional influence on receptor engagement events that are dependent on direct cell-cell and cell-particle interactions.

It should be noted that mechanisms by which a hyaluronidase modulates cell migration and invasion may not be limited to the degradation of HA in the matrix. Short HA fragments (<35 kD) and oligomers (4–10 mer) added to cultured cells have been shown in some cases to stimulate cell migration [66,67]. It is thus possible that TMEM2 modulates cell migration via the generation of short, bioactive HA species. However, at least in the context of the aforementioned study of cell migration on substrate-bound HA [63], it seems rather unlikely that small quantities of HA fragments released from the substrates during the assay would achieve the concentrations needed to induce cell migration in the HA fragment/oligomer studies (e.g., 10 μg/ml [66] and >0.02% [67]). In any event, further studies will be needed to fully elucidate the mechanisms by which TMEM2 promotes cell migration.

8 |. TMEM2 in Cell Stress Response

In a truly unexpected development, Andrew Dillin's group at the University of California, Berkeley, reported in 2019 that TMEM2 is a mediator of ER stress tolerance [52]. They performed a whole-genome CRISPR knockout screen in human immortalized fibroblasts for genes that, when inactivated, sensitized cells to ER stress. This screen identified TMEM2 as one of the most significant genes in this regard. The ability of TMEM2 to promote ER stress resistance is dependent on its hyaluronidase activity and is mediated by CD44 and the MAPK pathway, rather than by the canonical unfolded protein response of the endoplasmic reticulum (UPR^ER). More remarkably, this study further showed that ectopic overexpression of human TMEM2 in C. elegans protected animals from ER stress and increased their longevity [52]. This study therefore revealed a previously unanticipated connection between HA metabolism and ER stress response, and further links these processes to organismal survival and longevity. However, there are several unaddressed issues with this study. First, it is generally believed that C. elegans does not produce HA and does not possess HA synthase or hyaluronidase genes. Thus, the effects of TMEM2 overexpression on C. elegans are somewhat incongruent with their observations with human fibroblasts; namely, how can ectopically expressed human TMEM2 in C. elegans affect survival/longevity if there is no HA in C. elegans? (More recently, the same group claims that TMEM2 cleaves non-sulfated chondroitin present in C. elegans. See below.) Another unresolved issue is the possible involvement of anoikis, a form of programmed cell death that is induced by loss of adhesion [68], in the observed phenotype. Given the fact that loss of TMEM2 leads to impaired FA formation and cell adhesion in HA-rich environments [51] (see above), it is possible that Tmem2-deficiency renders cells susceptible to anoikis-mediated cell death. The observed effects on fibroblasts and C. elegans may therefore be due to, or involve, the ability of TMEM2 to promote FA formation and cell adhesion, and thereby prevent anoikis.

More recently, the Dillin group reported that TMEM2-mediated ECM remodeling induces altered mitochondrial function in both human cells and C. elegans [69]. This study was motivated by the fact that the effect of TMEM2 on ER stress response is not mediated by the canonical UPR^ER pathway [52]. Accordingly, the authors examined other types of cell stress responses and found that TMEM2 regulates mitochondrial homeostasis in human fibroblasts. Remarkably, manipulation of TMEM2 expression alters not only the metabolic function but also the morphological dynamics of mitochondria [69]. With regard to the question why TMEM2 overexpression has effects on C. elegans, which does not produce HA, the authors argue that fragments of unsulfated chondroitin, not those of HA, are the mediator of this mitochondrial response. This claim is based only on the decreased Alcian blue staining of TMEM2 overexpressing animals. Biochemical experiments to directly test the activity of TMEM2 on unsulfated chondroitin would be required to fully substantiate the claim. It has previously been shown that TMEM2 does not degrade a variety of chondroitin sulfate species (chondroitin sulfate A, chondroitin sulfate C, chondroitin sulfate D, and dermatan sulfate) [28], but the reactivity with unsulfated chondroitin has not been tested.

9 |. TMEM2 and Cancer

There is a large body of published data showing the effects of HA expression and degradation on tumorigenesis and tumor progression [70,71]. However, as with the effects of HA on cell migration and invasion, these data are often conflicting in terms of the stimulatory versus inhibitory effects of HA on tumor progression [72]. This is thought to be partly due to the multifaceted effects of HA, which depend on its size and physical state. In some cases, however, the use of non-physiological experimental conditions, such as exogenous addition of large amounts of HA and non-physiological overexpression of HA synthase and hyaluronidase genes, may have contributed to spurious effects. Based on a wealth of descriptive and experimental reports, the following general observations seem to be more or less applicable to a variety of cancers. They may therefore be useful in appreciating the role of HA degradation in tumor progression. First, a number of clinicopathological studies have demonstrated that HA levels in tumor stroma correlate positively with the aggressiveness and poor prognosis of human cancers [70,73-75]. Such a correlation is observed in the majority of adenocarcinomas, and HA accumulation is especially prominent in adenocarcinomas with desmoplastic reaction, such as pancreatic ductal adenocarcinoma [76-78]. Second, in adenocarcinomas, HA is generally produced by stromal cells, rather than by neoplastic epithelial cells [72,75,76,79]. Third, expression levels of hyaluronidases are also positively correlated with the aggressiveness and poor prognosis of adenocarcinomas [80-83].

Initial evidence for the association of TMEM2 with aggressive cancer was reported in 2016, prior to the discovery that TMEM2 is a hyaluronidase. Searching by RNA-seq for transcriptional targets of SOX4, a driver of breast cancer invasion and metastasis, Lee et al. [84] identified TMEM2 as a gene whose expression correlates with reduced overall survival in high-risk grade 3 breast cancers. The authors further showed that knockdown of TMEM2 in highly metastatic MDA-LM2 breast cancer cells significantly impaired in vitro invasion and reduced the formation of metastatic colonies in the lung in mouse xenograft experiments. More recently, further correlations of TMEM2 overexpression with poor prognosis have been reported for triple-negative breast cancer [85], pancreatic ductal adenocarcinoma [86,87], glioblastoma [88-90], and laryngeal cancer [91]. It is interesting that the overexpression of HYBID has also been shown to associate with poor prognosis in several cancers [92-96]. Kudo et al. [97] reported that concurrent overexpression of TMEM2 and HYBID strongly predicts shorter survival in patients with pancreatic ductal adenocarcinoma. These results suggest that pharmacological intervention targeting TMEM2 alone or together with HYBID can be a novel therapeutic paradigm for aggressive cancers.

10 |. TMEM2 and Non-neoplastic Disorders

Recent association studies implicate TMEM2 not only in cancers but also in several non-neoplastic disorders. One of the most compelling possibilities is the involvement of TMEM2 in fibrosis. Excessive accumulation of HA in tissues and biological fluids has long been implicated in fibrosis [98-100], and elevation of serum HA levels has been used as a clinical marker for liver fibrosis/cirrhosis [101]. These observations suggest that impairment of HA degradation may underlie the pathogenesis of fibrosis. Clinical and experimental studies demonstrate associations of TMEM2 (and downregulation of TMEM2 expression in the case of transcriptomic studies) with fibrotic diseases [102-105]. As observed in the induced global TMEM2 knockout study [38], reduction in TMEM2 expression results in the rapid accumulation of HA not only in plasma but also in a variety of organs, including the liver. However, it remains to be explored whether TMEM2 knockout induces not only HA accumulation but also the quintessential features of fibrosis, including the induction of epithelial-mesenchymal and endothelial-mesenchymal transition (EMT/EndoMT), generation of myofibroblasts, and accumulation of type I collagen. Long-term follow-up studies of the induced global Tmem2 knockout model are expected to address this issue.

Alzheimer's disease (AD) is another human disease in which the association of TMEM2 is implicated. Mathys et al. [106] analyzed single-nucleus transcriptomes from the prefrontal cortices of individuals with varying degrees of AD pathology, and determined cell type-specific expression signatures in these samples. This study identified TMEM2 as one of the top three genes most significantly downregulated in microglia in AD brain. This result has since been confirmed by a more stringent statistical analysis of the same dataset [107]. Another single-nucleus transcriptomic study in aged brain identified the association of TMEM2 expression with a microglial cluster defined by the expression of genes involved in senescence [108]. There are additional genomic and transcriptomic studies reporting the association of TMEM2 with a subpopulation of AD cases [109] and age-related cognitive decline [110]. These data seem to suggest that TMEM2 expression, presumably in microglia, may have functional relevance for the cognitive function and the pathogenesis of AD. However, there are currently no experimental data either to support or refute this possibility.

In addition to the studies described above, there are solitary reports showing TMEM2 association with other human diseases, including impaired lung function [111], emphysema [112], and psoriatic arthritis [113]. However, the significance of these associations is uncertain at present, and will require independent confirmation.

11 |. Concluding Remarks

The past five years have seen significant progress in our knowledge and understanding of TMEM2. Studies using Tmem2 knockout mouse models have demonstrated the functional significance of TMEM2 in systemic HA catabolism [38] and in mammalian embryonic development [51]. These results are generally in line with the phenotypes expected for knockout mice lacking a hyaluronidase that physiologically acts on extracellular HA. It is anticipated that further research using mouse models will continue to unveil roles of TMEM2 in additional aspects of development, tissue homeostasis in adult organs, and human disease.

Studies at the cellular level, on the other hand, have yielded some rather surprising results. While the functional involvement of TMEM2 in cell adhesion, migration, and FA formation [63] has been somewhat anticipated, its effects on cellular stress response, survival, and even the longevity of C. elegans [52,69] were unexpected. Investigation of the in vivo relevance of these observations in higher animals will be an important research focus in the next several years.

Unlike the progress in the function-focused research, there is still much to learn about the regulation of TMEM2 and its HA-degrading activity. Since ECM-remodeling enzymes are potentially damaging to tissues, they are usually under tight regulatory control. For example, MT1-MMP, the prototypical ECM-remodeling enzyme, is regulated at multiple levels. Proteolytic processing, intracellular trafficking, phosphorylation, and transcriptional/epigenetic regulation have been shown to play roles in temporal (short- and long-term) as well as spatial control of its activity [64]. It is possible that TMEM2 is similarly regulated at multiple levels. Also, little information is available regarding the potential functional coordination of TMEM2 with cell surface and extracellular HA binding proteins, including CD44, RHAMM, Stabilin-2/HARE, TSG-6, and HYBID. The profound metabolic and developmental phenotypes observed in Tmem2 knockout mouse models suggest the possibility that the co-presence of these HA binding proteins in tissues may have significant enhancing effects on TMEM2 activity in vivo, perhaps by sequestering HA in the vicinity of TMEM2. Compound mutants of TMEM2 and these HA binding proteins could provide useful insights into this possibility.

Finally, since its enzymatic active site is localized in the extracellular domain, TMEM2 is surmised to be a readily accessible target not only for small molecules but also for biologics, such as neutralizing antibodies. Given the suggested role of TMEM2 in tumor cell invasion and metastasis, inhibitors for TMEM2 may have therapeutic potential against TMEM2-overexpressing aggressive cancers. Despite the recent advent of novel anti-cancer therapies, invasion and metastasis still remain the most life-threatening aspects of cancer. New drugs targeting novel classes of molecules or pathways underlying invasion and metastasis are urgently needed. Inhibitors to TMEM2 have the potential to fill this gap in cancer therapeutics.

Abbreviations:

AD

Alzheimer's disease

CKO

conditional knockout

ECM

extracellular matrix

FA

focal adhesion

HA

hyaluronan

HMW

high molecular weight

KO

knockout

NCC

neural crest cell

UPR

unfolded protein response