The soil and the seed: the relationship between Descemet’s membrane and the corneal endothelium

Abstract

Descemet’s membrane (DM), the basement membrane of the corneal endothelium, is formed from the extracellular matrix (ECM) secreted by corneal endothelial cells. The ECM supports the growth and function of the corneal endothelial cells. Changes to DM are central to the diagnosis of the most common corneal endothelial disease, Fuchs endothelial corneal dystrophy (FECD). Changes in DM are also noted in systemic diseases such as diabetes mellitus. In FECD, the DM progressively accumulates guttae, “drop-like deposits” that disrupt the corneal endothelial cell monolayer. While the pathophysiologic changes to corneal endothelial cells in the course of FECD have been well described and reviewed, the changes to DM have received limited attention. The reciprocity of influence between the corneal endothelial cells and DM demands full attention to the latter in our search for novel treatment and preventive strategies. In this review, we discuss what is known about the formation and composition of DM and how it changes in FECD and other conditions. We review characteristics of guttae and the interplay between corneal endothelial cells and guttae, particularly as it might apply to future cell-based and genetic therapies for FECD.

1. Introduction

The corneal endothelium, a monolayer of cells on the posterior surface of the cornea tasked with maintaining control of corneal hydration, begins secreting extracellular matrix (ECM) during prenatal development. The ECM organizes to form the basement membrane of the corneal endothelium, Descemet’s membrane (DM), which increases in thickness throughout life. When disease affects the corneal endothelium, in addition to loss of the non-proliferative corneal endothelial cells (CEnCs), the CEnCs secrete a disordered ECM that is reflected in the overall structure and composition of DM. Guttae, drop-like deposits in DM, are the hallmark clinical diagnostic finding for the most common disease of the corneal endothelium, Fuchs endothelial corneal dystrophy (FECD). While CECs secrete the components of ECM that form the guttae, the altered ECM environment reciprocally affects CEnC function. In this review, we discuss the development of DM, changes with disease, the interplay between DM and CEnCs, and the implications for future therapeutic strategies for corneal endothelial diseases.

2. Normal development of Descemet’s membrane

The development of DM has two main phases —prenatal and postnatal. The prenatal layer is unchanged with aging while the postnatal layer grows continuously. The time of onset of corneal endothelial disease can be determined by the level of DM affected.

2.1. Prenatal development

In human prenatal development, the corneal endothelial cell (CEnC) monolayer is populated at week 8 (Wulle, 1972). These cells then secrete thin lamellae of basement membrane that become stacked until 30-40 lamellae are present at the time of birth, forming a layer approximately 3 μm thick that remains unchanged throughout life (Johnson et al., 1982; Murphy et al., 1984). The major collagen components of the prenatal DM are collagen type IV and collagen type VIII (Ljubimov et al., 1995; Sawada et al., 1990). Collagen type VIII in particular consists of α1 and α2 chains, which form homotrimers that combine to form tetrahedrons (Hansen and Karsdal, 2016). The tetrahedrons assemble into hexagonal lattices (Hansen and Karsdal, 2016; Stephan et al., 2004), which when stacked upon each other form characteristic bands (Levy et al., 1996). These bands are connected to each other by thin rods measuring roughly 100nm, an arrangement that is characteristic of the prenatal DM and is called “wide-spaced collagen” (Levy et al., 1995; Levy et al., 1996). Due to the banded appearance, the prenatal structure of the DM is termed the anterior banded zone (ABZ) (Johnson et al., 1982).

2.2. Postnatal development

Postnatally, endothelial cell secretion of collagen type VIII decreases, while collagen IV secretion continues (Ali et al., 2016; Kabosova et al., 2007). The DM thickens and develops a nonlamellar, homogeneous appearance without the banded pattern that is seen in the prenatal ABZ (Figure 1A) (Murphy et al., 1984). This layer is termed the posterior nonbanded zone (PNBZ) and increases in thickness with age at approximately 0.1 μm/year (Johnson et al., 1982; Murphy et al., 1984). While predominantly homogeneous in appearance, the PNBZ of healthy individuals occasionally contains focal lacunae of irregularities consisting of wide-spaced collagen similar to that seen in the ABZ (Figure 1B), or fibrillar collagen seen in disease (Johnson et al., 1982; Murphy et al., 1984). Occurrence of these irregularities in isolated lacunae within DM in healthy eyes suggests the occurrence of sporadic disruptions in ECM gene expression in CEnCs. Similar, but much more extensive deposits are seen within the DM of patients with guttae (Murphy et al., 1984).

Figure 1.

Structure of normal and diseased Descemet’s membrane. A. Normal post-natal Descemet’s membrane and B. variant of normal Descemet’s membrane. C-F. Disruptions to Descemet’s membrane in Fuchs endothelial corneal dystrophy (FECD). (a) anterior banded zone with wide-spaced collagen, (b) posterior nonbanded zone, (c) corneal endothelial cells, (d) lacunae of fibrillar collagen, (e) lacunae of wide-spaced collagen, (f) posterior banded zone (PBZ), (g) guttae in the PBZ disrupting dysfunctional corneal endothelial cells, and (h) posterior fibrillar zone (PFZ).

3. Descemet’s membrane changes in disease

In our review of the published literature on the structural changes to DM with disease, we encountered multiple confusing terminologies for the abnormal posterior layers. We have summarized these descriptions with a unifying terminology in this section that is also illustrated in Figure 1 and is based upon the observations of Waring (Waring, 1982) and Levy (Levy et al., 1996).

Corneal endothelial disease initiated in utero is often reflected by changes to the ABZ of DM whereas disease beginning later in life is characterized by changes to the posterior layers of DM with normal ABZ. DM in late-onset FECD has an absent or decreased thickness to the PNBZ (Waring, 1982). The thickness of the PNBZ may provide an estimate of the age of onset of endothelial disease. The increased thickness of DM in FECD is accounted for by a thick posterior collagenous layer (PCL) that lies posterior to the PNBZ (Waring, 1982). The abnormal PCL that is found in FECD is also found in several other diseases, including iridocorneal endothelial syndrome, posterior polymorphous dystrophy, congenital hereditary endothelial dystrophy, and aphakic corneal edema, although the exact structure and composition of the PCL varies between diseases (Levy et al., 1996; Waring, 1982). The material of the PCL includes wide-spaced banded collagen as well as nonbanded collagenous fibrils (Murphy et al., 1984), that resembles the material found in the ABZ (wide-spaced banded collagen) and in isolated lacunae of the PNBZ (wide-spaced banded collagen and fibrillar collagen) in healthy individuals (Johnson et al., 1982; Murphy et al., 1984).

The PCL is subdivided into a posterior banded zone (PBZ) and a posterior fibrillar zone (PFZ) (Figure 1C and 1D) (Bourne et al., 1982; Iwamoto and Devoe, 1971; Waring, 1982). In the PBZ, wide-spaced collagenous material predominates and forms a characteristic banding pattern that can be observed with electron microscopy (Waring, 1982). The PFZ, named for the fine collagenous fibrils that predominate in this layer, lacks the banding pattern seen in the PBZ (Waring, 1982). In FECD, the PBZ is always present and harbors guttae (Figure 1C) (Waring, 1982). In contrast, in other endothelial diseases, the PBZ may also be present but lacks characteristic guttae. The PFZ is not always present in FECD, but when present, lies posterior to the PBZ and can “bury” the guttae present in the PBZ (Figure 1D) (Iwamoto and Devoe, 1971; Xia et al., 2016). The PFZ may be thin or absent in FECD corneas with minimal edema (Bourne et al., 1982), but is significantly thicker when marked corneal edema is present, suggesting that this layer forms later in the course of FECD (Hribek et al., 2021).

3.1. Descemet’s membrane changes in systemic disease: diabetes mellitus

Changes to DM occur not only corneal endothelial diseases, but also in systemic diseases such as diabetes mellitus, a disease with well-known ocular manifestations. Most common is diabetic retinopathy, a disease with vascular endothelial basement membrane thickening and accumulation of advanced glycation end products, culminating in loss of microvascular endothelial cells (Beltramo and Porta, 2013). A similar pattern of disturbances due to diabetes occur in DM and CEnCs. Advanced glycation end products in DM, specifically the glycation of fibronectin and laminin, attenuate the attachment and spreading of cultured CEnCs (Kaji et al., 2001). These observations support clinical observations where diabetes increases CEnC pleomorphism and variability of cell area, and decreases CEnC densities (Ljubimov, 2017).

Diabetes also results in ultrastructural and biomechanical changes to DM. Diabetic rats and humans have an increased number of wide-spaced collagen fibrils (similar to those seen in the PBZ in FECD), possibly due to excessive nonenzymatic glycosylation that has been described in other basement membranes (Akimoto et al., 2008; Rehany et al., 2000). Although these changes are not clinically observable in patients, biomechanical changes have been observed in corneas from donors with diabetes. Donor graft preparation for Descemet’s membrane endothelial keratoplasty (DMEK) involves separation of the DM-CEnC complex from the remaining stroma. Compared to corneas from donors without diabetes, corneas from donors with diabetes have a higher risk for tears in DM that render the tissue unusable for transplantation (Greiner et al., 2014; Vianna et al., 2015). This may be explained by the stronger mechanical adhesion strength of DM to the stroma in corneas from donors with advanced diabetes compared to those with mild or no diabetes (Schwarz et al., 2016). While studies of the DM-CEnC complex from donors with and without diabetes have demonstrated significant differences in the proteome relating to mitochondrial function and intercellular junctions, the relationship between compositional changes and biomechanical properties of DM remains unknown (Skeie et al., 2018).

3.2. Changes in composition of Descemet’s membrane in endothelial disease

A proteomics study of DM has identified several proteins that are altered in specimens from FECD patients compared to specimens from patients with pseudophakic bullous keratopathy (Poulsen et al., 2014). Although comparing endothelial disease tissue to healthy donor corneal tissue would better describe the true pathology of DM in corneal endothelial diseases, there have been no such studies to date. FECD and pseudophakic bullous keratopathy are both characterized by low endothelial cell density, however, only FECD has guttae in DM. Thus, the data presented by Poulsen et al. may reflect changes specific to guttae and FECD. Altered proteins in FECD, when compared to pseudophakic bullous keratopathy, include upregulation of agrin, apolipoprotein D, clusterin, transforming growth factor beta induced protein (TGFBIp), collagen VI, decorin, keratocan, and matrilin-3, and downregulation of collagens I, II, III, IV, and V, fibrillin-1, tenascin, and various types of keratin (Poulsen et al., 2014). Several of these proteins with altered expression levels (TGFBIp, collagen IV, collagen I, collagen II, clusterin, and fibrillin-1) were among the most abundant proteins of the DM/endothelial layer in FECD (Poulsen et al., 2014). Collagen VIII, which is upregulated in COL8a2 mutation associated early-onset FECD (Levy et al., 1996), was not shown to be upregulated among late-onset FECD patients (Poulsen et al., 2014). Several of the altered proteins are potential key players in organization of the ECM/DM. Some studies show TGFBIp and clusterin colocalize in areas of DM with guttae (Jurkunas et al., 2009; Poulsen et al., 2014), whereas other studies did not find clusterin and TGFBIp colocalization or even increased expression (Goyer et al., 2018). Matrilin-3 (upregulated) mediates interactions between collagen fibrils and other ECM components, and mutations in matrilins are known to disrupt ECM assembly (Nicolae et al., 2007). Furthermore, decorin (upregulated), shows strong interactions with collagen VI (upregulated) (Bidanset et al., 1992). Type XII collagen is one of the most abundant proteins in DM, and although its expression levels are unchanged in FECD, there is significant upregulation of decorin and TGFBIp which interact with type XII collagen (Font et al., 1996; Runager et al., 2013).

3.2.1. FECD and Transcription Factor 4 (TCF4) trinucleotide repeat-associated changes in gene expression of ECM components

In the Caucasian population, one of the most common mutations found in association with FECD is a CTG trinucleotide repeat expansion in the TCF4 gene (Fautsch et al., 2021; Wieben et al., 2012). The expanded CUG repeat RNA transcripts of TCF4 accumulate in nuclear foci in CEnCs from FECD patients (Du et al., 2015; Hu et al., 2018; Mootha et al., 2015). This leads to sequestration of RNA splicing factors, functionally depleting them, thus resulting in global mis-splicing of RNA (Wieben et al., 2017). The majority of the mis-splicing events occur in genes coding for proteins that are involved cell cytoskeleton, cell adhesion, and ECM organization (Wieben et al., 2017). Altered splicing of genes encoding ECM components might be responsible for the disrupted ECM of DM in FECD. And changes in ECM are translated to functional changes in the cells through cell adhesion and cytoskeletal proteins thus providing a means for dysregulated ECM to further impact cell function (Romani et al., 2021).

3.2.2. DM composition changes related to endothelial-mesenchymal transition

Endothelial mesenchymal transition (EMT), the differentiation of CEnCs to a myofibroblast-like phenotype, is observed in FECD (Ong Tone et al., 2021). CEnCs from patients with late-onset FECD show upregulation of EMT inducing genes zinc finger E-box binding homeobox 1 (ZEB1) and snail family transcriptional repressor 1 (SNAI1) (Okumura et al., 2015). Upregulation of these genes is associated with excessive production of the ECM proteins type I collagen and fibronectin (Okumura et al., 2015). The increase in fibronectin precedes the accumulation of other ECM components and it is possible that fibronectin accumulation may play an early role in PBZ deposition and guttae formation (Goyer et al., 2018).

3.2.3. UV stress-associated changes in Descemet’s membrane

Ultraviolet light is an important environmental factor in FECD pathogenesis. The central cornea receives higher doses of UV radiation than the peripheral cornea and this coincides with the central location of guttae (Doutch et al., 2012). With FECD disease progression, the distribution of guttae is more likely to increase horizontally rather than vertically (Rosenblum et al., 1980), potentially due to the increased exposure to UV-light through the palpebral fissure.

In rabbits (Palazzo et al., 2020) and mice (Liu et al., 2020), exposure to UV light leads to thickening of the DM. In addition, the UV-A light induced model of FECD in mice also results in guttae formation (Liu et al., 2020). Constant exposure to UV light induces biological damage in exposed tissues, consisting of DNA damage, increase of reactive oxygen species (ROS), and decrease of antioxidant protective enzymes. Signs of all of these disruptions have been reported in FECD (Ong Tone et al., 2021). However, the changes to composition of DM due to UV light exposure are not known. By contrast, data is known for corneal stroma, where UV light exposure leads to alterations in gene expression of ECM remodeling components: collagens, proteoglycans, matrix metalloproteinases (MMPs), and tissue inhibitors of metalloproteinases (TIMPs) (Gendron and Rochette, 2015).

3.3. Structural changes to Descemet’s membrane: Guttae

The classic disruption to DM in FECD is the progressive accumulation of central corneal guttae, believed to arise from aberrant deposition of ECM by the corneal endothelium. Guttae (plural noun; singular = gutta; adjectives = guttata and guttate) (Eghrari and Gottsch, 2010) appear as excrescences molded within the PBZ of the PCL. These true guttae are easily distinguished clinically from pseudoguttae (caused by transient disruptions to endothelial cell function and resolve spontaneously without involvement of DM), and Hassall-Henle bodies (non-progressive, age-related guttae located in the peripheral cornea) (Moshirfar et al., 2019). Guttae are likely a common manifestation of genetic and metabolic derangements of CEnCs resulting in changes to ECM composition. FECD is a genetically heterogeneous disease, and the considerable variation in guttae morphology and PCL structure and composition might be explained by multiple pathophysiologic pathways.

3.3.1. Distinct guttae morphologies

Although only one term, guttae, is used to describe the excrescences on DM that are observed clinically, there are several distinct entities histologically. Guttae vary in their structure and distribution between the rare early-onset FECD and the common late-onset FECD (Gottsch et al., 2005; Jackson et al., 1999). Corneal guttae in patients with COL8A2 mutations associated with early onset-FECD are small, rounded, and associated with the endothelial cell center. This contrasts with late-onset FECD, in which guttae are larger, sharply peaked, and positioned at edges of endothelial cells (Gottsch et al., 2005).

Furthermore, there are different guttae morphologies even among patients with late-onset FECD. Xia et al classified FECD into three subtypes with different guttae morphologies and varying PCL structure (Xia et al., 2016). In type I, the PCL consists of a PBZ with abnormally deposited wide-spaced collagen. The PBZ is continuous with the guttae, which protrude posteriorly into the endothelial cell layer (Figure 1C). Type II FECD DM is similar to type I, but also has a PFZ which buries the guttae of the underlying PBZ (Figure 1D). Type III FECD DM has an additional layer of guttae within the PFZ protruding into the endothelial cell layer (Figure 1E and 1F). These guttae appear to have the same composition as the PBZ despite being embedded within the PFZ. On TEM they appear as “islands”, distinct from the PFZ that surrounds them (Figure 1E and 1F). It is unclear whether this “island structure” is a distinct entity or a result of the cross-sectional histology of mushroom shaped guttae. It is possible to observe all types of guttae in the same specimen.

3.3.2. Potential mechanisms of guttae formation

Guttae formation through secretion of proteinaceous material by endothelial cells is supported by the observation that guttae appear to grow over time (Laing et al., 1981; Son et al., 2014). Given that guttae appear to have the same composition as the PBZ, it is possible that ECM secretion is accelerated in the vicinity of specific cells where guttae form (Iwamoto and Devoe, 1971).

Another proposed mechanism of guttae formation involves the fusion of membrane bound vacuoles containing aberrant ECM with the basal endothelial cell membrane (Son et al., 2014). Son et al observed that in mice with the collagen VIII α2 Q455K mutation, the dilated rough endoplasmic reticulum became closely approximated with the basal endothelial cell membrane (Son et al., 2014). This may allow for direct attachment of the proteinaceous material within the rough endoplasmic reticulum to the DM. Guttae have been reported to sometimes have an “anvil” or “mushroom-like” appearance, and it is possible the “stalk” of these structures forms by fusion of the membranes of the intracellular vacuoles with the basal cell membrane (Son et al., 2014). Other studies have also noted dilated rough endoplasmic reticulum and large vacuoles in FECD specimens, and have investigated the role of dysfunctional autophagy in FECD pathogenesis. One function of autophagy is to degrade protein aggregates in response to endoplasmic reticulum stress, and is designed to protect cells from accumulation of misfolded proteins (Meng et al., 2013). In the collagen VIII α2 Q455K mouse model, lithium administration was found to increase autophagy in CEnCs (Kim et al., 2013). Lithium treated Q455K mice had fewer guttae, higher CEnC density, and increased CEnC survival against endoplasmic reticulum and oxidative stress when compared to controls (Kim et al., 2013). Rough endoplasmic reticulum stress and altered autophagy likely play a role in guttae formation, but the mechanism of vacuole fusion with the basal endothelial cell membrane does not explain how guttae grow over time, since the growth of guttae would cease with death of the overlying endothelial cell.

An alternative mechanism for guttae formation is globally increased CEnC secretion of ECM proteins with guttae “molded” into ECM through the altered expression and balance of MMPs and TIMPs. ECM architecture is normally regulated by MMPs (proteases responsible for degradation of the ECM) and TIMPs (which serve numerous functions, one of which is inhibition of MMPs) (Iyer et al., 2012; Xu et al., 2021). Focal downregulation of MMPs or upregulation of TIMPs could result in focal ECM accumulation, resulting in the formation of a gutta. MMPs and TIMPs show altered expression patterns in FECD. There are discrepancies in their expression patterns, however, suggesting heterogenicity in the disease. While one group reported a downregulation of MMP10, MMP14, and TIMP1 in FECD patients (De Roo et al., 2017), another reported upregulation (Weller et al., 2014). A more recent study found downregulation of MMP2 and MMP10 in FECD (Xu et al., 2021). MMP2 is capable of degrading denatured collagen, as well as fibronectin, laminin, and type IV collagen (Cabral-Pacheco et al., 2020). MMP10 is also capable of degrading fibronectin, laminin, and type IV collagen (Cabral-Pacheco et al., 2020). These are all important components of guttae and downregulation of the proteases responsible for their degradation might explain the accumulation of these ECM components.

3.3.3. Guttae in systemic disease

Furthermore, guttae have been reported in association with other diseases such as Marfan syndrome, myotonic dystrophy, and certain mitochondrial disorders, suggesting that there are multiple etiologies for guttae formation. In a case series of 41 patients with Marfan syndrome, 17 patients were found to have guttae (Setälä et al., 1988). Guttae were more likely to be present in Marfan patients who had lens subluxation than in those who did not have lens subluxation. Marfan syndrome is caused by mutations leading to altered or reduced fibrillin-1 (Hayward and Brock, 1997; Sakai et al., 2016), a protein which is also downregulated in FECD (Poulsen et al., 2014). Changes in fibrillin-1 expression affect its interactions with other ECM components and its roles in mechanical support, ECM formation, cell-matrix interactions, and cell behavior (Schrenk et al., 2018). However, its precise role in FECD has not been studied.

Guttae are also found in patients with myotonic dystrophy type 1 (Gattey et al., 2014; Mootha et al., 2017; Winkler et al., 2018). Similar to the FECD association with a trinucleotide repeat in the TCF4 gene, myotonic dystrophy is associated with a CTG trinucleotide repeat in the noncoding region of the DMPK gene (Gattey et al., 2014), and RNA toxicity and altered RNA splicing play a role in its pathogenesis (Gattey et al., 2014; Winkler et al., 2018).

Guttae have also been reported in patients with macular dystrophy associated with the mitochondrial point mutation A3243G, seen with the mitochondrial disorders maternally inherited diabetes and deafness (MIDD) and mitochondrial encephalopathy with lactic acidosis and stroke-like episodes (MELAS) (Bakhoum et al., 2018). In these patients, the corneal endothelium and DM have features similar to FECD, namely polymegathism and guttae formation (Bakhoum et al., 2018). The A3243G mitochondrial mutation alters energy production (decreased adenosine triphosphate), increases lactic acid, and increases reactive oxygen species (Lin et al., 2019). These are also findings in FECD pathophysiology (Jurkunas et al., 2010; Vallabh et al., 2017), implicating common mechanisms in guttae formation.

4. Interactions between corneal endothelial cells and guttae

4.1. In vitro studies

In both early- and late-onset FECD, the changes to the biochemical and biomechanical properties of DM promote phenotypic changes in CEnCs and CEnC loss (Kocaba et al., 2018; Leonard et al., 2019). In mouse models of early-onset FECD, it is clear that the early physiologic changes in CEnCs lead to the deposition of wide-spaced collagens in DM. The wide-spaced collagens result in reduced elastic modulus and a biomechanically softer DM in FECD compared to normal DM (Leonard et al., 2019). The softer DM in FECD is present in both early- and late-onset FECD (Leonard et al., 2019; Xia et al., 2016). It has been proposed that the softer DM leads to altered mechanotransduction signaling pathways in CEnCs leading to continued synthesis of abnormal wide-spaced collagen (Leonard et al., 2019).

The structural disruptions to DM from guttae in FECD also direct non-diseased CEnCs to a phenotype seen in FECD. This has been seen with both native DM samples from FECD patients, and with tissue culture surfaces micro-engineered with guttae-type buttons (Rizwan et al., 2016). With both native and engineered guttae surfaces, formation of an intact CEnC monolayer is dependent upon the size of guttae. Small diameter (< 20 μm), short (< 10 μm) guttae do not inhibit monolayer formation but larger and taller guttae do inhibit monolayer formation (Kocaba et al., 2018; Rizwan et al., 2016). In addition, large (>30 μm) guttae induced EMT, and promoted senescence, and apoptosis (Kocaba et al., 2018). However, guttae alone did not induce other cellular perturbations previously observed in CEnCs in FECD, such as deficient antioxidant capacity, the unfolded protein response, DNA damage, and mitochondrial dysfunction. Initial disruptions to CEnCs in disease likely lead to secretion of an aberrant ECM with guttae formation further contributing to endothelial decompensation (Kocaba et al., 2018).

Furthermore, the composition and elasticity of ECM affects the growth and morphological characteristics of CEnCs, independent of the architectural changes from guttae. It is well known that the growth of CEnCs in vitro is dependent upon the coating of the cell culture surface. CEnCs have been expanded with varying degrees of success on collagen type I and type IV (Choi et al., 2013), laminin (Choi et al., 2013; Yamaguchi et al., 2011), FNC coating mix (Athena Enzyme Systems, Baltimore, MD) (Zhu and Joyce, 2004), and fibronectin (Blake et al., 1997; Choi et al., 2013). In addition, endothelial cell density, cell size, cell shape, and expression of α-smooth muscle actin (indicative of EMT) were altered when CEnCs were grown on matrices of various elastic moduli and matrix coatings (uncoated, fibronectin, collagen type I, laminin, collagen type IV, or combined laminin and collagen type IV) (Palchesko et al., 2015). The optimal CEnC phenotype was observed when CEnCs were cultured on collagen type IV-coated substrate with elastic modulus of 50kPa (Palchesko et al., 2015), similar to the elastic modulus of native DM (Last et al., 2009), highlighting the importance of the composition and biomechanics of the extracellular matrix on CEnC phenotype and physiology.

4.2. In vivo Observations

The importance of DM composition and architecture to proper CEnC function draws attention to emerging therapies for corneal endothelial diseases where the substrate for the CEnCs is bare corneal stroma or diseased DM. In Descemet’s stripping without endothelial keratoplasty (DWEK, also referred to as Descemet’s stripping only, DSO; Figure 2A), performed for FECD involving the central cornea but with a clear peripheral cornea, the diseased central DM and CEnCs are removed by surgical stripping, leaving bare corneal stroma on the posterior cornea (Garcerant et al., 2019). Re-endothelialization of the cornea is proposed to occur by central migration of peripheral corneal endothelial cells (Joyce, 2003), possibly with cell proliferation (Garcerant et al., 2019). In corneal endothelial cell injection therapies (Figure 2B), diseased corneal endothelial cells are dislodged from DM with a silicone needle and non-diseased cultured CEnCs and a rho kinase inhibitor (ROCKi) are injected into the anterior chamber to allow cells to repopulate the exposed DM during prone positioning (Kinoshita et al., 2018; Ueno et al., 2022). Based upon our in vitro knowledge of the interactions between CEnCs and DM, we consider the implications of the CEnC environments for optimization of outcomes in both of these conditions.

Figure 2.

Emerging therapies for Fuchs endothelial corneal dystrophy. Diseased cells have red borders. A. Descemet’s stripping without endothelial keratoplasty (DWEK) with repopulation of cornea with migration +/− proliferation of peripheral FECD cells. B. Corneal endothelial cell injection (CEnC) with injected non-diseased CEnCs in yellow. C. Possible future therapy with simultaneous DWEK and CEnC injection. Shades of blue represent potential newly deposited ECM which will likely vary due to different cell (seed) and substrate (soil) combinations under each condition (A. FECD cells and corneal stroma, B. normal CEnCs and guttae, C. normal CEnCs and corneal stroma).

The success of DWEK depends upon removal of the diseased DM and CEnCs, followed by endothelialization of the bared posterior cornea without addition of donor cells or tissues. The mechanism of endothelial repopulation is likely due to cell migration from the peripheral cornea with potential cell proliferation (Arbelaez et al., 2014; Macsai and Shiloach, 2019; Moloney et al., 2017). In two case series and one case report that measured preoperative and postoperative central and peripheral ECD, there was a statistically significant 10-40% decrease in peripheral ECD after DWEK (Macsai and Shiloach, 2019; Moloney et al., 2017; Ploysangam and Patel, 2019). This suggests that endothelial cells migrate from the peripheral cornea to populate the region of bare stroma. Bilobed nuclei have also been reported by confocal microscopy after DWEK, potentially representing mitoses and proliferative activity (Moloney et al., 2017). Creating a defect in the endothelial layer possibly relieves the contact inhibition that plays a role in endothelial cell cycle arrest (Joyce, 2003) and may allow for endothelial cell proliferation. ROCKi eye drops have been used to aid cell proliferation following DWEK surgery. ROCKi relieves the block from G1 to S cell cycle progression in CEnCs Okumura et al., 2014). Topical application of ROCKi eye drops appears to significantly increase final endothelial cell density following DWEK, and aids the healing process in DWEK as evidenced by increased rates of corneal clearing when ROCKi are used (Moloney et al., 2017; Schlötzer-Schrehardt et al., 2021). The enhanced corneal clearing following DWEK with ROCKi may also be secondary to the effects of ROCKi aiding cell migration over the bare stroma (Ho et al., 2022), decreasing EMT, or upregulating genes involved in corneal endothelial barrier function (Schlötzer-Schrehardt et al., 2021), in addition to enhancing cell proliferation.

The importance of the interactions between the CEnCs and bare stroma is evident in observations from DWEK surgeries. Two methods have been used in DWEK to remove the central diseased DM. One method involves scoring the DM and the other involves tearing the DM in a Descemetorhexis fashion (Yuan and Pineda, 2021). The favorable outcomes (increased rates of corneal endothelial healing) of the Descemetorhexis technique have been attributed to the decreased stromal trauma and clean DM edge compared to the scoring method (Davies et al., 2018; Garcerant et al., 2019; Macsai and Shiloach, 2019). Posterior stromal scarring at the edge of the DM stripping have been observed with the scoring technique and may inhibit corneal endothelial cell migration (Davies et al., 2018; Iovieno et al., 2017; Moloney et al., 2017).

Enhancing the outcomes of DWEK will benefit many individuals with an indication for this technique. Key benefits of DWEK are the lack of need for donor corneal tissue and correspondingly, an absence of risk of graft rejection. At this time, key hurdles in DWEK surgery are expanding eligibility to individuals with advanced FECD, improving rates for clearance of corneal edema, improving cell density, and promoting long term improvement in outcomes. In the future, DWEK may be an ideal procedure to pair with gene therapy targeted at causes of FECD. In DWEK, without gene therapy, the CEnCs repopulating the area of DM stripping will still harbor the disease mutations and in younger patients, disease recurrence remains a risk.

In studies reported to date on cell injection therapy for endothelial diseases, the diseased endothelial cells are removed, but the diseased DM is not removed. Injected CEnCs repopulate on the diseased DM. This is in contrast to DWEK, where the diseased DM is removed and diseased cells repopulate the cornea. Results of cell injection therapy studies show functional and anatomic restoration of the corneal endothelium, with improved visual acuity in over 80% of patients (Kinoshita et al., 2018; Numa et al., 2021). The biological quality of injected endothelial cells also influences clinical outcome (Ueno et al., 2022). Compared to patients receiving cell injection consisting of a lower proportion of mature CEnCs, patients who received cell injection with a high proportion of mature, well-differentiated CEnCs achieved higher mean endothelial cell density, had a faster reduction in central corneal thickness, and had less reduction in endothelial cell density over time (Ueno et al., 2022).

A key concern with cell injection therapy is the presence of the diseased DM. In FECD patients, the guttae remain present at 3-5 years post cell injection therapy (Numa et al., 2021; Ueno et al., 2022). In light of prior in vitro experiments (Kocaba et al., 2018; Rizwan et al., 2016), the guttae microenvironment is detrimental to cell health, and the altered composition of the PCL of DM in disease presents a stress to corneal endothelial function. Nevertheless, 5 years after cell injection therapy, these patients with FECD appear to maintain sufficient endothelial cell density and function to maintain corneal clarity (Numa et al., 2021). A hybrid strategy of Descemetorhexis with cell injection may also be a good option because it allows for removal of the diseased DM (Figure 2C). Two patients in the cell injection study (Numa et al., 2021) had stripping of DM because of unintentional damage of DM during endothelial cell removal. Those patients had functional restoration of the endothelial monolayer following cell injection, thus suggesting that cell injection therapy may be successfully combined with DM stripping procedures. All of these studies, including the in vitro data on the effects of DM composition and architecture on CEnC function, illustrate the complex interactions between the cells and their soil (Kocaba et al., 2018; Leonard et al., 2019; Rizwan et al., 2016).

5. Future considerations

DM formation begins in utero and continues growth throughout life. Early manifestations of corneal endothelial diseases, such as in FECD, reveal disruption to the formation of DM before phenotypic changes in CEnCs are evident. Therapeutics targeting these early changes in ECM, through drugs or gene therapies, may suffice to limit subsequent CEnC dysfunction and death. Later in endothelial disease, changes in DM composition and architecture affect the function of CEnCs leading to cell dysfunction and death. Design of future therapies for late-stage disease of the corneal endothelium should ideally address disease of both the CEnCs and the DM.

Abbreviations:

FECD

Fuch’s endothelial corneal dystrophy

DM

Descemet’s membrane

ECM

extracellular matrix

CEnC(s)

corneal endothelial cell(s)

ABZ

anterior banded zone

PNBZ

posterior nonbanded zone

PCL

posterior collagenous layer

PBZ

posterior banded zone

PFZ

posterior fibrillar zone

EMT

endothelial-mesenchymal transition

UV

ultraviolet

DMEK

Descemet’s membrane endothelial keratoplasty

ROS

reactive oxygen species

MMP

matrix metalloproteinase

TIMP

tissue inhibitor of metalloproteinase

ROCKi

rho-kinase inhibitor

DWEK

Descemet stripping without endothelial keratoplasty

DSO

Descemet stripping only