Recent Advances in Stem Cells of Corneal Epithelia

Abstract

The cornea, which refracts light toward the retina and acts as a protective barrier for the eye, is critical for our visual system. The corneal epithelium, continuously renewed by corneal epithelial stem cells (CESCs), is required to maintain corneal transparency and thus is critical for vision. Recent experiments in corneal epithelial lineage tracing, live imaging, and single-cell RNA sequencing have deepened our understanding of CESCs and uncovered their unexpected properties. In this review, we focus on recent progress in elucidating these essential properties of CESCs, including their location and how they maintain their populations and their heterogeneity. Additionally, we discuss the limitations of current studies and highlight critical issues that urgently need to be addressed in the field of CESCs. These new insights into CESCs will evolve into game-changing therapeutics for corneal diseases by harnessing CESCs in regenerative medicine.

As one of the most important components of our visual system, the cornea provides most of the refractive power for the eye to refract light toward the retina for normal vision. Structurally, the cornea is composed of three layers: the epithelial cell layer, stromal layer, and endothelial cell layer (from anterior to posterior). The corneal epithelium, located at the anterior surface of the cornea, plays a critical role in maintaining corneal transparency and provides a protective barrier for the eye. The corneal epithelium continuously turns over and is replenished by corneal epithelial stem cells (CESCs), namely, limbal epithelial stem cells (LESCs), named for their anatomical location at the limbus (the boundary between the cornea and conjunctiva). LESCs proliferate, migrate and differentiate into mature corneal epithelial cells (CECs) to maintain the corneal epithelium. Any damage to LESCs or their specialized microenvironment, the limbal niche, results in LESCs dysfunction and limbal stem cell deficiency (LSCD) and, thus, serious visual impairment.^1,2 LESCs transplantation has become an effective therapy for LSCD^2–4; however, its long-term outcomes need further improvements. Recent advances in corneal epithelial researches, including lineage tracing, live imaging, and single-cell RNA sequencing (scRNA-seq), have revealed unexpected properties of CESCs and deepened our understanding of their essential properties. These studies have addressed several key questions: whether CESCs scatter throughout the cornea (but not exclusively at the limbus), how LESCs maintain their population, and whether LESCs are heterogenous. These new advances will provide novel ideas for treating corneal diseases, especially transplantation of stem cells to treat LSCD. 

Where Are CESCs? LESCs vs. CESCs Models

Although the LESCs model of CESCs has been accepted widely and LESCs transplantation has been used for LSCD treatment, a key question remains: whether CESCs also scatter throughout the cornea but not exclusively at the limbus (CESCs model). In this section, we describe the differences between the LESCs model and CESCs model and present the evidence that supports each model. Then, we discuss how recent corneal epithelial lineage tracing and live imaging resolve this question and support the LESCs model, as well as how to explain the evidence supporting the CESCs model. 

The LESCs Model

The longstanding concept that CESCs reside mainly in the limbus, a narrow transitional zone between the cornea and conjunctiva, was established in the late 1980s and early 1990s. LESCs are located in the basal epithelial layer of limbus, and renew corneal epithelium through their proliferation, differentiation and centripetal migration. Key evidences that support the LESCs model include (1) slow-cycling cells, a critical attribute of stem cells and can be detected experimentally as label retaining cells (LRCs), are found exclusively in the basal layer of limbal epithelium but not corneal epithelium^5,6; (2) limbal basal epithelial cells are relatively undifferentiated, as evidenced by the absence of key corneal epithelial differentiation markers (CK3/12)⁷; (3) central corneal epithelial wounds preferentially stimulate cell proliferation of limbal epithelium,⁵ and limbal epithelial cells show higher growth capacity and colony-forming efficiency in vitro compared with peripheral and central CECs^8–10; (4) limbal and peripheral corneal epithelium migrate centripetally and replace central corneal epithelium^11–13; (5) abnormal integrity and wound healing capacity of corneal epithelium after partial or total limbal ablation^14–16; and (6) restoration of corneal transparency and sight in severely damaged ocular surface after transplantation of limbal tissue or cultured cells containing LESCs.^{2–4,17–23} Subsequently, corneal epithelial progenitor cells or transit amplifying cells (TACs), which replicate relatively rapidly and are capable of extensive cell population expansion, are found in the basal layer of peripheral (early TACs, which replicate at least twice during homeostasis and have additional potential of cell divisions) and central (late TACs, which usually divide only once) cornea.⁶ In this scheme of LESCs model (Fig. 1A), LESCs reside at the limbus where they divide slowly to self-renew and produce daughter TACs, which divide more quickly and migrate centripetally from the basal layer of limbal epithelium to the basal layer of corneal epithelium. After a final cell division, TACs leave the basal layer, differentiate, and move rapidly through the suprabasal layer before being shed from the surface as terminally differentiated cells. LESCs become more active and produce more TACs during wound healing, although small wounds may be healed by TACs without the activation of LESCs. 

Figure 1.

 

Where are CESCs? LESCs vs. CESCs model. (A) Schematic representation of the LESCs model and CESCs model. The LESCs model proposes that rare, immortal, and slow-cycling stem cells reside only in the basal epithelial layer of the limbus (LESCs). These LESCs give rise to abundant, short-lived, and fast-cycling progenitors (or TACs) that proliferate, migrate centripetally, and differentiate to renew the corneal epithelium. The corneal epithelial basal TACs also move upward through the suprabasal layer and are eventually shed from the surface as terminally differentiated cells. In contrast, CESCs model proposes that stem cells reside not only in the limbus, but also within the cornea. LESCs are only activated for wound repair, whereas CESCs, scattered throughout the corneal epithelium, self-renew and produce centrifugal-moving TACs to maintain the corneal epithelium. Key evidence supporting the LESCs model and CESCs model is presented. (B) The Confetti construct and its four possible recombination outcomes when the reporter is expressed in a single allele. (C) Predicted clonal growth patterns in the LESCs model and CESCs model, and the outcomes of lineage tracing based on the Confetti cassette driven by CK14-Cre^ERT2. These results strongly suggest that stem cells of the corneal epithelium mainly, perhaps exclusively, reside at the limbus. The experimental photograph is re-used with permission from Stem Cell Reports (2017).³⁸

The CESCs Model

Despite the widespread acceptance of the LESCs model, there are credible reports that challenge the conventional LESCs model. Key evidences that contradict the LESCs model are as follows. First, LESCs do not participate in steady-state corneal epithelial renewal and small corneal epithelial wound healing. Through detecting cell proliferation across the whole cornea after central corneal epithelial wounds^24–26 and examining central corneal epithelial wound healing ability of corneas that are either separated from the limbus^27,28 or deficient in LESCs,^29–31 studies indicate that the healing of small central corneal wounds is independent of limbal epithelium and LESCs. In addition, limbal transplants from transgenic mice implanted into the limbus of recipient athymic mice do not participate in the steady-state corneal epithelial renewal.³⁰ These studies suggest that corneal epithelium has self-renewal and, therefore, regenerative capacity. Second, there is centrifugal migration of CECs. When central corneal transplants are implanted into the central cornea of recipient athymic mice, CECs from these transplants are able to migrate centrifugally toward the limbus.³⁰ Third, CECs have the ability to form colony. CECs from various species, including pig, sheep, rabbit, rat, and human, contain many keratinocyte colony-forming cells that can be serially passaged in vitro.^10,30 Fourth, corneal transparency is maintained after LESCs ablation. Cornea can remain transparent and undergo normal reepithelialization after cauterizing the limbal circumference to destroy the limbal epithelium (including LESCs),^30,31 and a cohort of patients with LSCD have been reported to exhibit persistent islands of healthy central corneal epithelium.³² These findings indicate the possibility that stem cells exist within the corneal epithelium itself. Collectively, the CESCs model proposes that there are stem cells in the limbus (LESCs), but they are only activated to repair wounds. CESCs, scattered throughout the corneal epithelium, divide to self-renew and produce centrifugal-moving daughter TACs, and are in charge of their own limited area of corneal epithelium, thereby ensuring corneal epithelial homeostasis^30,33 (Fig. 1A). 

Outcomes of Corneal Epithelial Lineage Tracing and Live Imaging

The advancement of experimental technologies, especially lineage tracing and scRNA-seq, provides new approaches to identify the location of CESCs. Lineage tracing is a technique to monitor the fate of a cell and its progeny within tissues. To achieve this goal, a stable reporter is introduced into a cell so that, when it divides, the reporter is passed to its daughter cells faithfully. If introduced into a stem cell, the reporter will be inherited permanently by all daughter cells and the marked stem cell can be tracked as an expanding clone indefinitely in a lineage-restricted manner. Conversely, if the reporter is introduced into TACs or differentiated cells, these cells eventually disappear from the trace owing to tissue turnover. This process allows direct testing of the LESCs or CESCs model without disturbing normal homeostasis, bypassing disadvantages of complicate artificial culture and transplantation. 

The introduction of a stable reporter into a cell is based mainly on the Cre-loxP system, which allows for temporal and spatial control of the reporter expression. Temporal control is achieved by fusing Cre recombinase to a mutated human estrogen receptor (CreERT), which is maintained in an inactive state in the cytosol by heat shock proteins. Upon administration of the synthetic ligand tamoxifen (TAM) or its active metabolite 4-hydroxy-tamoxifen, CreERT is released from its chaperone proteins and translocates to the nucleus, where Cre recombinase acts freely on loxP sites to initiate recombination, resulting in the expression of the reporter. Spatial control is achieved in a cell-specific manner, where the expression of CreERT is driven by an appropriate promoter. In some cases, the unbiased and stochastic lineage tracing system is used, where the expression of CreERT is controlled by ubiquitous CAGG promoter, allowing the labeling of all cell types. The stable reporter can be either a LacZ reporter or a fluorescent protein reporter. The LacZ reporter expresses β-galactosidase, and these labeled cells and their progeny can be stained using X-Gal as a substrate. The fluorescent protein reporters can be visualized directly under fluorescence microscopy, enabling real-time monitoring and intravital imaging. The fluorescent protein reporters can be one color or multiple colors. For example, the Brainbow 2.1 or Confetti cassette allows stochastic expression of one of four fluorescent proteins (mCFP, nGFP, cYFP, and cRFP) in a cell when only one copy of this cassette is expressed³⁴ (Fig. 1B). 

Up to now, specific/ubiquitous induction and LacZ/fluorescent protein reporter have been used in several studies to test the LESCs or CESCs model. We focus on describing the different predictions of clonal growth for these two models and the outcome of corneal epithelial lineage tracing using the Confetti cassette driven by CK14-Cre^ERT2. For other cases, please refer to additional reports.^35–37 After the administration of tamoxifen, limbal and corneal basal epithelial cells will stochastically express one of four fluorescent proteins (mCFP, nGFP, cYFP, and cRFP). After long-term lineage tracing, the pattern of the traced cells is used to distinguish the LESCs and CESCs model. Under the LESCs model, the cornea forms a pattern like multicolored pinwheel, with all CECs originating from the limbus. In contrast, the CESCs model predicts a pattern like a multicolored painting disc, with distinct clonal regions originating from both limbus and cornea (Fig. 1C). 

Experimental results show that short radial stripes appear at the limbus, peripheral and central cornea in the short term. However, after a long-term lineage tracing, these short stripes are replaced by longer and more persistent stripes that originate from the limbus^38,39 (Fig. 1C). Thus, these findings suggest that CESCs either do not exist or do not significantly contribute to corneal epithelial maintenance during normal homeostasis. Instead, stem cells of corneal epithelium reside mainly, perhaps exclusively, at the limbus (LESCs). 

LESCs Leakage Hypothesis

Alternatively, a compromise between the LESCs and CESCs models has been proposed, which hypothesizes a leakage of stem cells from the limbus (LESCs) into the cornea, where they become CESCs.⁴⁰ The stem cell leakage model is based on observations using the CK14-Cre^ERT2-Confetti lineage tracing mice, where the distribution of stem cells (small clusters of labeled cells) peaks in the peripheral zone near the limbus and in the central cornea. This model proposes that LESCs might occasionally undergo symmetric cell divisions, producing two daughter LESCs. Subsequently, additional LESCs are evicted into the cornea, becoming CESCs. Then, centripetal migration leads to the accumulation of CESCs in the central cornea. 

However, this clear two-peak distribution of stem cells in the limbus and central cornea is only observed in 10-week-old mice. In older mice, only limbus-derived stem cells are detected.³⁸ As established, CK14-expressing stem cells/progenitors are widely distributed across cornea before 10 weeks of age. Therefore, the accumulation of so-called CESCs in the central cornea might simply result from the centripetal migration of these CK14-expressing stem cells/progenitors toward the central cornea, rather than from the leakage of LESCs. Moreover, these small clusters of labeled cells only persist 2 weeks, which are more likely to be short-lived TACs rather than true stem cells. 

Limitations of Traditional Methods and Explanations for Evidence of the CESCs Model

Because stem cells of corneal epithelium reside at the limbus (LESCs), an open question arises: how do we interpret these experimental results that support the CESCs model? One key factor to consider is that the traditional methods used to identify the existence of stem cells, such as colony formation, cell/tissue grafting, and LRCs assay, are technically limited. These methods are often under stress conditions but not normal homeostasis, or indirectly indicate the function of stem cells in vivo. For instance, colony formation and cell grafting assays involve the digestion of cell-surface proteins that join cells together, stimulation of detachment-induced apoptosis owing to cell suspension, and deprivation of original niche signals, which are replaced by feeder cells or a limited number of extracellular matrix proteins. The LESCs culture in vitro itself can affect function and behavior of cells as the lack of limbal niche. The tissue grafting assay involves the injury of tissue and the change of niche signals because of the emergence of injury response. In the LRCs assay, LRCs can originate not only from slow-cycling stem cells, but also from some differentiated cells. Moreover, an increasing number of studies on adult stem cells in various organs and tissues, especially epithelial stem cells, have shown that adult stem cells are not necessarily slow-cycling or quiescent (see Section Populational Asymmetry of Adult Stem Cells (Stochastic Model)). 

Of note, studies supporting the CESCs model involve the following experimental results, which are limited or can be explained. (1) Genetically labeled limbal transplants do not contribute to corneal epithelial homeostasis; in contrast, central corneal transplants are able to move centrifugally toward the limbus.³⁰ The corneal epithelium can self-repair small wounds successfully after a single injury. However, after repetitive injuries, the corneal epithelium repairs more slowly and fails to heal the defect completely, while LESCs are the main source of regenerative epithelial cells.²⁸ The LESCs model accepts that the central corneal epithelium has limited self-renewal capacity owing to the existence of TACs across the corneal epithelium, and these TACs contribute to small wound healing without involving LESCs, indicating limited and poor regenerative capacity. In one study using two-photon live imaging, vertical paths of terminally differentiated trajectories from the basal layer to the most superficial layer are described for both the limbus and cornea during homeostasis. Although terminal differentiation at the limbus follows a direct upward trajectory, the trajectory at the cornea follows an arced path in a centrifugal direction.⁴¹ However, this centrifugally arced trajectory does not indicate that corneal basal cells are stem cells, because these cells themselves originate from the limbus. Another factor that might lead to mistakes is the use of athymic mice as recipients. One study shows that immunodeficient mice exhibit abnormalities in the quiescence of LESCs, epithelial thickness, and wound healing.³⁹ The absence of immune cells in the cornea of athymic mice might change the cellular function and behavior of both recipient and transplanted limbus/cornea. (2) Surgical or thermal depletion of the entire limbal epithelium (including LESCs) has no detrimental impact on the cornea, which can remain transparent for several months without resulting in LSCD. This result can be explained by the dedifferentiation of CECs into LESCs, which can maintain normal tissue dynamics of the corneal epithelium.^42–44 (3) Central CECs display the capacity of proliferation and wound healing after transplantation, as well as colony formation in vitro.³⁰ The limitations of colony formation in vitro are described elsewhere in this article. Additionally, colonies are classified artificially into three distinct types based on their area, shape, and components: holoclone, meroclone, and paraclone, which are believed to result from stem cells, progenitor cell (TACs), and differentiated cells, respectively. However, a continuum of colony sizes is observed in the colony formation assay. Collectively, adult stem cells should be defined as biologically functional cells that sustain the generation of relevant progeny through proliferation and differentiation over a long period or their lifetime. However, all these evidences supporting the existence of CESCs fail to demonstrate this biological function, and thus these so-called CESCs cannot be considered as true stem cells. 

Limitations of Cre-loxP–Based Lineage Tracing

Despite it is powerful for looking for stem cells, the Cre-loxP system has several limitations. (1) The use of transgenic mice is a potential limitation. No matter the Cre recombinase is randomly integrated into the genome or precisely inserted into the downstream of a specific promoter, the transgenic method might disrupt endogenous genes or alter their expression. (2) The Cre recombinase and its off-target cleavage of genomic loxP sites might cause genomic instability and cytotoxic effects,⁴⁵ and the off-target labeling will lead to mistaken results and conclusions. (3) Tamoxifen and its active metabolite 4-hydroxy-tamoxifen might be cytotoxic,⁴⁶ potentially affecting the function and behavior of stem cells. Another factor that should be considered is the permeability of 4-hydroxy-tamoxifen across the cornea. Although the limbus is rich with blood vessels, the cornea is avascular. The exposure of 4-hydroxy-tamoxifen might be uneven across limbal and corneal epithelium, which results in heterogeneous labeling of the reporter. (4) The choice of promoter used to drive the expression of Cre recombinase can affect the result of lineage tracing. Most current lineage tracing studies rely on specific promoters (e.g., CK14 or other LESCs markers). However, the disturbance of endogenous genes of LESCs markers might alter the function and behavior of stem cells. Moreover, the use of a single marker to trace stem cells might introduce bias, which focuses on the fate of marker-positive stem cells rather than the entire stem cells population. The use of ubiquitous promoters to drive Cre recombinase expression could be a better choice. In the future, the development of methods with low off-target labeling, low toxicity, and Cre-independent approaches for tracing stem cells will help to avoid the limitations of the present lineage tracing techniques. 

How to Maintain the LESCs Pool? Cellular vs. Populational Asymmetry

A key question in the field of adult stem cells is how stem cells maintain their identity and population. A widely accepted model is the hierarchical model, where adult stem cells undergo asymmetric division to maintain their identity and population. However, symmetric cell divisions are observed in adult stem cells frequently, especially in epithelial stem cells.⁴⁷ This elicits another way to maintain stem cells pool: the populational asymmetry (stochastic model). In this model, the stem cells pool is maintained through populational, rather than single-cell, asymmetric division. In this section, we discuss the differences between these two models, expound on recent results of corneal epithelial lineage tracing that support the populational asymmetry of LESCs, and discuss several key questions about the stochastic model of stem cells. 

Cellular Asymmetry of Adult Stem Cells (Hierarchical Model)

The traditional hierarchical model of stem cells proposes that stem cells sit at the apex of a hierarchy, and they are rare and slow-cycling (divide infrequently or be quiescent) to minimize the risk of mutations associated with DNA replication.⁴⁸ To maintain stem cells pool and their progeny, individual stem cells undergo invariant asymmetric division, yielding two cells that adopt different fates: one becomes a new stem cell, while the other becomes a progenitor cell. A classical example of adult stem cells that conform to this hierarchical model is hematopoietic stem cells (HSCs). HSCs are multipotent; in other words, a single HSC can generate all blood lineages. HSCs divide asymmetrically, producing two cells that adopt different fates: one becomes a new HSC and the other becomes a lineage-restricted progenitor (e.g., the common lymphoid progenitor). As these daughter cells continue to divide, they migrate down the hierarchy and become progressively more lineage-restricted (e.g., common lymphoid progenitors differentiate into T/B cells). This process occurs through a well-orchestrated series of discrete steps, which are unidirectional and ultimately yields all mature blood cell types.^48–50 

Populational Asymmetry of Adult Stem Cells (Stochastic Model)

In fast-renewing tissues (e.g., intestinal epithelium), proposed markers of stem cells often label a broad population of frequently dividing cells.⁵¹ The scRNA-seq also captures abundant and fast-cycling stem cells. Additionally, these stem cells do not strictly divide asymmetrically. Instead, both asymmetric and symmetric cell divisions are observed. These findings support an alternative hypothesis, the stochastic model of stem cells, where asymmetry is achieved at the population level rather than at the single-cell level.⁴⁷ This model posits that stem cells are equipotent and abundant. The daughters of each stem cell have an equal probability to become a new stem cell or differentiate into a progenitor cell (or TAC). These equipotent daughters compete neutrally with each other for the niche of stem cells, and their fates are unpredictable. However, on average, each stem cell division is asymmetric, producing a daughter stem cell and a daughter progenitor cell (or TAC). This model implies a postmitotic period, during which the fates of stem cells daughters are undetermined and unknown. The neutral competition (e.g., for the limited niche of stem cells) among equipotent stem cells at the population level results in a stochastic pattern of behavior known as neutral drift dynamics, where the tissue is maintained through an ever-diminishing number of clones with ever-increasing size.⁴⁷ This stochastic model and neutral drift dynamics of equipotent and abundant stem cells have been described in several tissues, such as intestinal stem cells at the crypt,^51–53 spermatogonial stem cells in the testes along seminiferous tubules,^54–56 and oral, epidermal, and esophageal epithelial stem/progenitor cells.^57–59 

Stochastic Competition and Neutral Drift of LESCs

The traditional LESCs model follows similar principles to the hierarchical model, with LESCs positioned at the apex of the hierarchy. LESCs divide asymmetrically and slowly to self-renew, producing a daughter LESC and a daughter TAC. TACs then divide and differentiate rapidly and move centripetally from the limbus to the central cornea. Although LRCs assays indicate the existence of rare and slow-cycling stem cells at the limbus, previously identified LESCs markers typically label a broad population of limbal epithelial basal cells. Moreover, both asymmetrical and symmetrical divisions are observed across the corneal epithelium.^60–62 Although lineage tracing using LacZ as a reporter shows dynamic changes of radial limbus-derived stripes, this dynamic could also be attributed to the merging of radial stripes or the loss or quiescence of LESCs,^36,37 thus rendering the question of how LESCs maintain their identity and pool unresolved. 

The use of a multicolor Confetti cassette has made it possible to distinguish between the hierarchical and stochastic model of LESCs. If LESCs are rare cells with hard-wired asymmetrical division, the size of LESCs-derived stripes would remain constant. In contrast, if LESCs follow the stochastic model, the size of LESCs-derived stripes would not stabilize, but would change dynamically over time (Fig. 2A). Long-term lineage tracing reveals that the heterogeneity in clonal size distribution (width of LESCs-derived stripes) increases over time, with some stripes becoming wider and larger, while others shrinking or disappearing (Fig. 2A). Additionally, clonal size distribution scales with their averages.^38,39,63 These mathematical analyses suggest that the clonal dynamics of LESCs cannot be explained by the traditional hierarchical model, but instead aligns more closely with the stochastic model of stem cells. In this model, abundant LESCs constantly compete neutrally with each other for the niche of stem cells and exhibit the neutral drift behavior. 

Figure 2.

 

How to maintain LESCs pool? Cellular vs. populational asymmetry. (A) Schematic representation of the predictions of cellular and populational asymmetry of LESCs, and the outcomes of lineage tracing based on the Confetti cassette driven by CK14-Cre^ERT2. Cellular asymmetry proposes that the division of a single LESC is invariant asymmetry, which yields two daughter cells that adopt different fates (one being new LESC, the other being TAC) to maintain stem cells pool and their progeny. In contrast, populational asymmetry proposes that LESCs are equipotent and abundant, and daughter cells of LESCs division have on average the same probability to become a new LESC or a TAC. At the population level, stem cell division is asymmetry, producing a daughter LESC and a daughter TAC. All LESCs neutrally compete with each other for the stem cell niche and their fates are unpredictable. The stochastic competition between equal LESCs at the population level will lead to neutral drift dynamics. Experimental photographs are re-used with permission from Stem Cell Reports (2017).³⁸ (B) An experimental strategy to directly test the stochastic competition and neutral drift model of LESCs. Initially, at the first tracing after the administration of tamoxifen (TAM), the loser stripe (e.g., *yellow stripe) is replaced by adjacent winner stripe (e.g., red stripe neighboring the *yellow stripe). At the second tracing after the administration of TAM, the remaining pair of loxP sites with opposite orientation can be inverted by Cre recombinase, accompanying with the change of fluorescent protein. These LESCs are relabeled and retraced to form new stripes. If the previously losing stripe can subsequently become a new winner stripe, this strongly indicates that LESCs are equipotent and there is no hierarchy between these LESCs.

An Experimental Strategy to Test Stochastic Competition and Neutral Drift

Unfortunately, the evidence supporting the stochastic competition and neutral drift model of LESCs is currently based on mathematical analysis; there is no direct experimental evidence indicating that LESCs are equipotent and can be replaced by each other. These phenomena could also be explained by a hierarchical model with populational heterogeneity of stem cells, just like spermatogonial stem cells.⁵⁶ Here, we propose an experimental strategy to test directly their equipotentiality and stochastic competition of LESCs (Fig. 2B). As known, the Confetti cassette retains a pair of loxP sites with opposite orientation after the first cleavage by Cre recombinase, and this pair of loxP sites can be further inverted by Cre recombinase, resulting in changes in the expression of fluorescent protein.⁶⁴ So, after an initial pulse of tamoxifen, the cornea develops into a multicolored pinwheel pattern, with some stripes being replaced by their neighbors. We can designate these stripes that become wider to be winners, whereas those that are replaced to be losers. Then, after a second pulse of tamoxifen, all zones are relabeled, and previously labeled stripes can change their fluorescent protein tags. If these newly labeled stripes in previous losers’ zones can become winners and replace previous winners’ zones (Fig. 2B), it would provide strong evidence indicating that there is no hierarchy between LESCs, and instead LESCs compete with and replace each other stochastically. 

Key Questions About Stochastic Competition and Neutral Drift

The stochastic competition and neutral drift of stem cells require the complete stochasticity and equipotency of stem cells. However, in fact, biased drift often occurs owing to both extrinsic and intrinsic factors. For example, the location of intestinal stem cells within the niche of crypt affects their competitiveness. Although intestinal stem cells are endowed intrinsically equipotent via the transfer of stem cells between different locations, intestinal stem cells residing at the bottom of the crypt experience a survival advantage over those at the periphery, which have a higher likelihood to be lost and replaced.⁵³ Furthermore, intestinal stem cells are shown to follow an asymmetric cell division-dominant neutral drift model, which reduces the equality of stem cells.⁶⁵ Additionally, stem cells that acquire oncogenic mutations can skew neutral drift dynamics and outcompete wild-type stem cells, and the accelerated proliferation is the main driving force behind this biased drift behavior.^66,67 

A key question that the stochastic model of stem cells must address is how these active, fast-dividing stem cells avoid the occurrence of cancer, given the unavoidable random mutations that arise during DNA replication. Quiescence, a prolonged yet reversible cell-cycle exit, has been used widely as a defining characteristic of adult stem cells. Quiescence minimizes endogenous stress caused by cellular respiration and division of stem cells, decreasing the accumulation of DNA damage and preventing stem cells from being cancerous. The lifetime risk of many cancers is correlated strongly with the total number of divisions of normal self-renewing stem cells that maintain tissue homeostasis.⁶⁸ However, the assumption that the division rate of stem cells necessarily correlates with mutational accumulation has been challenged. Whole-genome sequencing of organoid cultures derived from primary multipotent cells of the small intestine, colon, and liver biopsies allows for the quantification of mutation rates in these tissues. Mutations accumulate steadily over time at a rate of approximately 40 novel mutations per year, despite significant variation in cancer incidences among these tissues.⁶⁹ This finding suggests that the division rate of stem cells is not the main factor contributing to mutational accumulation and cancer incidence. In bone marrow HSCs, a typical example of quiescent stem cells, quiescence promotes mutagenesis in response to ionizing irradiation. This is because nonhomologous end-joining–mediated DNA repair in quiescent HSCs is more error-prone than high-fidelity homology-dependent repair in proliferating HSCs.⁷⁰ A similar reliance of quiescent adult stem cells on nonhomologous end-joining for DNA damage repair is also observed in hair follicle bulge stem cells.⁷¹ This finding suggests that quiescence of stem cells is a double-edged sword: it protects stem cells against endogenous stress, but also renders them more vulnerable to mutagenesis after DNA damage. In the epidermis, fast-dividing epidermal stem cells with DNA double-strand breaks are cleared by selective differentiation and delamination through the DNA damage response–p53-Notch/p21 axis. Moreover, the selective elimination of stem cells with DNA double-strand breaks is coupled with symmetric cell divisions of surrounding intact stem cells.⁷² This finding suggests that neutral competition and drift of stem cells at the population level can safeguard genomic integrity. Additionally, although a super competitor mutation might outcompete more abundant wild-type stem cells and lead to the expansion of mutated stem cells, stem cells carrying harmful mutations are also likely to be replaced by abundant neighboring healthy stem cells. Together, fast cell division and neutral competition among abundant stem cells may serve as a protective mechanism that decreases the risk of cancer, especially in fast-renewing tissues. 

The ocular surface squamous neoplasia (OSSN) comprises of a spectrum of tumors that affect the ocular surface.⁷³ OSSN usually occurs within the interpalpebral fissure (the space between the open upper and lower eyelids), particularly at the nasal limbus, which receives the greatest intensity of ultraviolet radiation and contains the most concentrated LESCs.^74,75 A similar pattern is observed in pterygia (a benign lesion), where most lesions are located at the nasal limbus.⁷⁶ This finding implies that the origin of OSSN and pterygia may be LESCs, and exposures to ultraviolet radiation and viruses are the main pathogenic causes of OSSN. Although OSSN often occurs at the limbus, no study has explored whether the proliferation or quiescence of LESCs promotes mutagenesis in OSSN. Interestingly, two compartmentalized subpopulations of LESCs (quiescent LESCs [qLESCs] and active LESCs [aLESCs]) have been identified at the limbus in mice (see Section Heterogeneity of LESCs in Mice), making the limbus an ideal tissue to examine the relationship between the rate of cellular division (fast division vs. quiescence) of adult stem cells and mutagenesis. 

Heterogeneity of LESCs

With the development and application of scRNA-seq, more and more cell populations that previously were thought to be one type of cells are now revealed to be heterogeneous. This trend is also emerging in the field of adult stem cells. However, the initial discovery of the heterogeneity of LESCs results from corneal epithelial lineage tracing in mice. Subsequently, scRNA-seq has also uncovered the heterogeneity of LESCs in humans. In this section, we review the findings of heterogeneity of LESCs in mice, the potential microniches that regulate LESCs subpopulations, and the interconversion between LESCs subpopulations. We then discuss the heterogeneity of LESCs in humans and potential challenges associated with using scRNA-seq to identify LESCs subpopulations. 

Heterogeneity of LESCs in Mice

Long-term lineage tracing studies suggest that LESCs conform to the stochastic model of stem cells, implying that LESCs are abundant, equipotent, and fast-cycling. However, LRCs assays indicate the existence of slow-cycling stem cells at the limbus.^5,77 These findings encourage the consideration that the limbus may contain two distinct depots of LESCs: qLESCs and aLESCs.⁶⁴ The first evidence of these two depots of LESCs was reported in 2015, when both LESCs-derived stripes across the entire cornea and scarce limbal clones with minimal expansion or migration were observed in mice.⁷⁸ However, this study does not resolve two critical questions regarding these two depots of LESCs. How are these two depots of LESCs distributed across the limbus? And, what are the functional differences between these two depots of LESCs? 

Two back-to-back studies report the coexistence of two separate and well-defined subcompartments within the mouse limbus, termed as the outer and inner limbus, where qLESCs and aLESCs are located respectively.^39,41 The inner limbus (approximately 240 µm) hosts a population of aLESCs with a division rate of approximately 4.2 days. These aLESCs predominantly undergo symmetric divisions and are required to sustain the population of progenitors (or TACs) that constantly support the homeostasis of the corneal epithelium. The outer limbus (approximately 100 µm) hosts a population of relatively qLESCs with a slower division rate of approximately 8.1 days. These qLESCs minimally, if at all, contribute to the homeostasis of the corneal epithelium, instead mainly function as a boundary between conjunctival and corneal epithelia to prevent conjunctival cells from invading the cornea during homeostasis (Fig. 3A). Both aLESCs and qLESCs participate in corneal epithelial wound healing, even though qLESCs play a minimal role in the homeostasis of the corneal epithelium.³⁹ The inner aLESCs are defined as stem cells rather than early progenitors (or early TACs), because they generate large (comprising thousands of cells), long-lived (4 months) stripes. This new LESCs model, featuring two distinct subpopulations of stem cells, provides an ideal framework for understanding how functionally diverse stem cell populations within compartmentalized organization support the maintenance and regeneration of an adult organ (Fig. 3B). 

Figure 3.

 

Heterogeneity of LESCs. (A) The coexistence of qLESCs and aLESCs at the limbus of mice. qLESCs are located at the outer limbus, while aLESCs at the inner limbus. (B) Behavior of qLESCs and aLESCs during homeostasis and wound healing, as revealed by lineage tracing. During homeostasis, both qLESCs and aLESCs follow the rule of stochastic competition and neutral drift dynamics. aLESCs maintain the homeostasis of the corneal epithelium, while qLESCs maintain a boundary between the conjunctival and corneal epithelium. Notably, qLESCs also contribute minimally to the homeostasis of the corneal epithelium. After injury, both qLESCs and aLESCs participate in the wound healing of the corneal epithelium.

Intriguingly, the serial HSCs transplantation assay, which is considered as the gold standard for functional definition of HSCs, also reveals the co-existence of long-term HSCs and short-term HSCs. Long-term HSCs, which are predominantly quiescent, contribute minimally to the steady-state native hematopoiesis. In fact, the depletion of the majority of long-term HSCs does not perturb the generation of blood lineages during homeostasis. This finding indicates that short-term HSCs, which are actively dividing population, sustain the bulk of steady-state hematopoiesis without the involvement of long-term HSCs.^79–81 In contrast, a unique class of long-term HSCs is an important source of megakaryocyte/platelet–restricted progenitors during homeostasis or following transplantation.^82,83 Similarly, qLESCs contribute minimally to corneal epithelial replenishment during homeostasis, but readily enter mitosis in response to corneal injury. This finding suggests that different subpopulations of stem cells work cooperatively and are all indispensable for fulfilling their complete biological functions during homeostasis. Moreover, quiescent stem cells may serve as a reservoir to deal with injury and stress. 

Boundary Formation and Microniches of LESCs Subpopulations

Boundary formation often separates groups of cells with distinct functions or fates during development. It plays a vital role in segmenting tissues and maintaining cellular compartments, thereby supporting diverse physiological functions. In adult tissues, boundary formation is also important in maintaining tissue compartmentalization, which can be disrupted in diseases such as cancer.⁸⁴ Two basic types of boundaries can be defined. The first is a nonlineage boundary, where cell identity is plastic, and cells can move across the boundary and adapt their identity to match that of their local neighbors. The second is a lineage boundary (or compartment boundary), where cell identity is inherited, and a sharp boundary is maintained between cell groups with distinct identities through cell sorting.⁸⁵ 

It is clear that a sharp epithelial boundary exists between the conjunctiva and limbus (a lineage boundary), and between the limbus and cornea (a nonlineage boundary).⁸⁶ One key question is how the compartmentalized organization between the cornea, inner limbus, outer limbus, and conjunctiva is achieved, where epithelial cell groups must be able to distinguish themselves from their neighbors. Cell signaling is a crucial factor in maintaining boundaries. Two major cell-cell signaling networks, Eph and Notch, can sense their cellular microenvironment to distinguish like from nonlike cells and are involved in boundary formation.⁸⁴ Both of these signaling networks distinguish cells by their nonadhesive characteristics and asymmetrical distribution of ligands and receptors in neighboring cells. In the cornea, the ephrin–A1 ligand is concentrated in the limbal epithelium and extends into the limbal–corneal epithelial junction, whereas the EphA2 receptor is localized mainly in the more differentiated corneal epithelium.⁸⁶ This asymmetrical distribution of ligand and receptor enables ephrin–A1/EphA2 signaling complexes to play a key role in limbal–corneal epithelial compartmentalization.⁸⁶ Notch signaling is vital for the differentiation and proliferation of LESCs,^87–89 and it will be interesting to examine its role in boundary formation across the ocular surface. 

Another important factor to maintain boundary is the deposition of the extracellular matrix. Pronounced heterogeneity of basement membrane and extracellular matrix components is observed at the limbus, cornea–limbus transition zone and cornea.⁹⁰ Interestingly, collagen fibers at the limbus are circumferential, whereas those in the cornea are centripetal, extending toward the central cornea.^41,91 This finding indicates that the clonal growth direction of qLESCs or aLESCs is aligned with the extension of collagen fibers. Furthermore, the circumferential collagen fibers are enriched at the outer limbus,⁴¹ where certain immune cells also localize, and T cells play a critical role in maintaining the quiescence of LESCs.³⁹ 

Thus, it is highly likely that the narrow limbus develops at least two distinct microniches that participate in the regulation of qLESCs and aLESCs and the formation of boundary between the outer and inner limbus. In the hair follicle, precise control over morphogenesis and regeneration is achieved by compartmentalizing stem cells into microniches along epithelial–mesenchymal interfaces, where some stem cells specify lineages immediately, whereas others retain potency.⁹² In the future, it will be essential to explore the components of these microniches of LESCs subpopulations. 

Interconversion Between LESCs Subpopulations

Given that the phenotypes of qLESCs and aLESCs are regulated by their respective microniches, the next open question arises: can qLESCs and aLESCs interconvert? Of note, quiescence is a relative concept rather than an absolute measure. This is because the exact cell-cycle time is difficult to determine, and the division rates of different stem cells can vary significantly. The division rates of aLESCs (approximately 4.2 days) and qLESCs (approximately 8.1 days) are average values, and thus do not exclude the possibility that stem cells with slower or faster division rates also exist. As known, T cells serve as niche cells for qLESCs and play a critical role in maintaining quiescence, and the deficiency of T cells results in increased cell proliferation at the outer limbus.³⁹ Moreover, qLESCs inevitably proliferate and convert into aLESCs to deal with extensive corneal epithelial wound. Conversely, these aLESCs will inevitably convert into qLESCs once the wound healing is completed. Furthermore, it is highly likely that aLESCs can convert into qLESCs when exposed to the microniche of qLESCs. This presumption could be tested through the precise ablation of qLESCs and surrounding conjunctival epithelium using femtosecond laser. After such ablation, aLESCs could migrate into the outer limbus, providing an opportunity to observe potential interconversion between LESCs subpopulations. 

Furthermore, quantitative analysis of lineage-traced LESCs has revealed that approximately 3% of radial stripes across the cornea possess a base or foot at the outer limbus.³⁹ These rare radial stripes may result from two possibilities: (1) qLESCs give rise to aLESCs, thereby contributing to corneal epithelial replenishment, or (2) the clone of qLESCs and the radial stripe of aLESCs are labeled by the same fluorescent protein purely by chance. The scRNA-seq analysis of the LESCs/CECs using pseudotime algorithm suggests that outer qLESCs can give rise to inner aLESCs.³⁹ Therefore, it is highly likely that qLESCs are the origin of aLESCs in these 3% of radial stripes. 

Of particular importance is the fact that both aLESCs and qLESCs follow stochastic dynamics, yet they exhibit different doubling times.³⁹ This implies that LESCs follow both the stochastic and hierarchical models, although the degree of hierarchy is minimal. Consequently, it becomes clear that LESCs cannot be described by a simple model, owing to their heterogeneity and the distinct functions and behaviors of aLESCs and qLESCs. 

Heterogeneity of LESCs in Humans

The heterogeneity of LESCs is likely to be more diverse and complex in humans, owing to the complicated structure of limbus in humans.⁹³ With the development and application of scRNA-seq in human cornea, an increasing number of markers and cell states of LESCs have been identified.^94–101 The expanded toolbox of available markers of LESCs includes Tspan7, SOX17, GPHA2, TP63, CCL20, and others. In one study, four subtypes of limbal stem/progenitor cells were identified by unsupervised subclustering⁹⁵: (1) the TP63⁺ subcluster, characterized by the classical LESCs marker TP63; (2) the CCL20⁺ subcluster, with high expression of the chemokine CCL20, which can induce cell migration and proliferation; (3) the GPHA2⁺ subcluster, a novel marker of qLESCs in mice³⁹ and humans⁹⁷; and (4) the Krt6B⁺ subcluster, which rapidly divides and inhibits migration of mitotic cell populations from the basal layer. TP63⁺ and CCL20⁺ cells exhibit high stemness, whereas GPHA2⁺ and Krt6B⁺ cells show a greater degree of differentiation.⁹⁵ The biological processes of differentially expressed genes across these four subtypes appear to be distinct: hemidesmosome assembly and cell population proliferation are enriched in TP63⁺ cells; RNA metabolic process and RNA splicing are enriched in CCL20⁺ cells; epidermis development and epithelial cell differentiation are enriched in GPHA2⁺ cells; and cornification is enriched in Krt6B⁺ cells.⁹⁵ Notably, it remains challenging to determine whether these cell states are biologically distinct or whether they result from technical differences, such as tissue dissection or scRNA-seq data analysis. Furthermore, discrepancies often arise between different scRNA-seq datasets⁹⁴ and between different experimental approaches used to characterize LESCs, such as scRNA-seq and immunohistochemistry. Therefore, the heterogeneity of LESCs and the specificity of these LESCs markers in humans should be validated rigorously, especially for the identification and isolation of LESCs for clinical application. 

Although the heterogeneity of human LESCs during homeostasis requires further validation, the existence of two populations of human LESCs (aLESCs and qLESCs) in culture is well-established, and their states/phenotypes can be controlled by the surrounding extracellular matrix. For instance, human LESCs encapsulated in gelatin methacrylate–based scaffolds exhibit an active state, whereas those encapsulated in the hyaluronic acid glycidyl methacrylate–based scaffolds are in a quiescent state.¹⁰² Similarly, the presence of heavy chain–hyaluronic acid/pentraxin 3, purified from amniotic membrane, maintains human LESCs in a quiescent state.¹⁰³ Mechanistically, the suppression of canonical Wnt signaling, coupled with the activation of noncanonical Wnt signaling (planar cell polarity), may mediate the maintenance of quiescence in human LESCs.^102,103 

New Markers of LESCs and Their Functions

ScRNA-seq and immunostaining have identified markers for aLESCs as Krt15-GFP⁺/ATF3⁺/Mt1-2⁺, and for qLESCs as Krt15⁺/GPHA2⁺/IFITM3⁺/CD63⁺ in mice. Kyoto Encyclopedia of Genes and Genomes enrichment analysis reveals that the differences between aLESCs and qLESCs involve pathways such as regulation of cell proliferation, regulation of cell motility, response to lipid, and response to oxygen-containing compounds.³⁹ Notably, the Krt15-GFP does not overlap with endogenous Krt15 expression, consistent with a previous report in hair follicle stem cells.¹⁰⁴ Krt15-GFP specifically marks aLESCs, whereas Krt15 is used to mark qLESCs, even if Krt15 is also expressed in conjunctival epithelial cells.³⁹ 

Activating transcription factor 3 (ATF3), a member of the AP-1 superfamily of transcription factors, has been found to maintain the quiescence of stem cell–enriched limbal basal cells in mice.¹⁰⁵ Similarly, ATF3 preserves skeletal muscle stem cells (also known as satellite cells) population by actively suppressing precocious activation.¹⁰⁶ ATF3 also plays an important role in preventing stress-induced exhaustion of HSCs. Specifically, the deletion of ATF3 dramatically impairs the long-term reconstitution capability of long-term HSCs following a series of bone marrow transplantations.¹⁰⁷ Notably, satellite cells and HSCs are classically considered as quiescent adult stem cells. Therefore, the role of ATF3 in LESCs needs further study. 

Metallothioneins (Mts) are a family of low-molecular-weight, cysteine-rich, metal-binding proteins that have been implicated in a range of roles, including toxic metal detoxification, protection against oxidative stress, and regulation of zinc and copper homeostasis.¹⁰⁸ In particular, knock-out of Mt1/2 impairs hepatic proliferation and regeneration following partial hepatectomy and oxidative stress injury induced by thioacetamide.^109,110 In satellite cells, mild upregulation of Bmi1 expression leads to a remarkable improvement of muscle function in a mouse model of Duchenne muscular dystrophy, which is mediated by Mt1-driven modulation of resistance to oxidative stress.¹¹¹ 

During the induced differentiation of LESCs in vitro, both GPHA2 and IFITM3 are down-regulated.^39,97 Knock-down of GPHA2 and IFITM3 in LESCs using small interfering RNA decreases Krt15 expression, increases the expression of corneal epithelial differentiation markers CK3/12, and significantly reduces colony-forming efficiency.^39,97 These findings indicate that both GPHA2 and IFITM3 support the stemness of LESCs in vitro. However, although GPHA2 and CD63 become undetectable in immune-deficient mice, IFITM3 remains detectable, suggesting different regulation of these qLESCs markers by immune cells.³⁹ 

GPHA2 is known to form a heterodimer with GPHB5, which activates the thyroid-stimulating hormone receptor pathway.¹¹² However, neither transgenic overexpression of GPHA2 nor deletion of GPHB5 has been shown to produce an overt phenotype in mice. In the pituitary, GPHA2 is expressed in quiescent pituitary stem cells and is regulated by the Notch2 signaling pathway.¹¹³ Given that Notch signaling is crucial for the differentiation and proliferation of LESCs,^87–89 it would be intriguing to investigate whether GPHA2 also serves as a downstream target of Notch signaling in LESCs. 

IFITM3 localizes in the endosomal–lysosomal system and broadly inhibits the entry of diverse pathogenic viruses by preventing virus–cell membrane fusion.¹¹⁴ This antiviral activity of IFITM3 is mediated by its abilities to bind cholesterol and sort local lipid, thereby increasing the concentration of lipids that disfavor viral fusion at the hemifusion site.^115,116 Strangely, LESCs also express TMPRSS2 and ACE2, two proteins that facilitate SARS-CoV-2 viral entry, making the limbus as a portal for viral entry.^117,118 Interestingly, the differences between aLESCs and qLESCs include response to lipid, with several lipid metabolism-related genes (e.g., ApoE, LCN2, and CCDC3) enriched in qLESCs.³⁹ Moreover, ACE2 is shown to regulate LESCs proliferation via the TGFA/EGFR/LCN2 pathway.¹¹⁹ Thus, it will be an important field to explore the role of lipid metabolism in the regulation of LESCs. 

CD63 is a member of the tetraspanins, a large superfamily of cell surface–associated membrane proteins characterized by four transmembrane domains. CD63 is present abundantly on the cell surface, as well as in late endosomes and lysosomes. In late endosomes, CD63 can be secreted as exosomes through the fusion of endosomes with the plasma membrane. This complex localization pattern suggests that the intracellular trafficking and distribution of CD63 must be regulated tightly.¹²⁰ In the liver, CD63 serves as a novel marker for liver stem cells. Specifically, CD63⁺CD56⁺ cells are identified as activated liver stem cells, whereas CD63⁺CD56⁻ cells are quiescent liver stem cells.¹²¹ However, the function of CD63 in adult stem cells, including LESCs, remains unidentified. Given the cholesterol-sorting ability of CD63,¹²² its potential role in lipid trafficking should be considered, especially in light of the emerging importance of lipid metabolism in the regulation of LESCs. 

Tspan7⁺ and SOX17⁺ cells are distributed sparsely and exclusively in the basal epithelial layer of limbus, but not in the suprabasal, superficial layers of limbal epithelium or the full layers of the peripheral and central corneal epithelium. Knock-down of Tspan7 or SOX17 using specific small interfering RNAs suppresses the proliferation of primary human limbal epithelial cells, suggesting their roles in the regulation of cell proliferation in LESCs.⁹⁶ Tspan7, a member of the tetraspanin family, encodes a cell surface glycoprotein. Increased expression of Tspan7 is observed in numerous cancers, where it promotes the migration and proliferation of cancer cells via epithelial-to-mesenchymal transition.^123,124 Moreover, Tspan7 is involved in the formation of migrasome.¹²⁵ However, the function of SOX17, a member of the SoxF family, requires further investigation. 

Of note, most of these new LESCs markers are identified by their exclusive expression in the limbal epithelial basal layer, where LESCs are located. However, discrepancies often arise between different studies, highlighting the challenges in consistently identifying and validating these markers. Thus, the reliability of these newly identified proteins as LESCs markers, as well as their functions in LESCs, require further examinations. 

Summary and Conclusions

Over the last decade, the development and application of lineage tracing and scRNA-seq have advanced our understanding of CESCs significantly, including where they are and how they maintain their populations and their heterogeneity. In the current model of CESCs, these cells reside in the basal epithelial layer of the limbus, namely, LESCs, which consists of two distinct subpopulations: aLESCs and qLESCs. aLESCs rapidly undergo cell divisions, neutrally compete with each other for the niche, and are required to sustain the population of progenitors (or TACs) that move centripetally to continuously support the homeostasis of corneal epithelium. In contrast, qLESCs divide slowly and contribute minimally to corneal epithelial replenishment, instead they mainly maintain a boundary between the conjunctival and corneal epithelium to prevent the invasion of conjunctival cells into cornea during homeostasis. Additionally, qLESCs also serve as a reservoir to deal with corneal epithelial injury and stress. This two-depot model of LESCs offers an unique example of how functionally diverse stem cell populations play markedly different roles during normal homeostasis and wound healing, thereby supporting both the maintenance and regeneration of tissue. These recent advances not only deepen our understanding of CESCs, but also pave the way for the application of LESCs in the treatment of corneal diseases. 

Acknowledgments

Supported by the Science and Technology Innovation Project of Army Medical University (2020XQN14), the National Key Research and Development Program (2024YFA1108700) and the National Natural Science Foundation of China (81900832). 

Author Contributions: Y. Li wrote the manuscript, and Y. Liu reviewed the manuscript. 

Disclosure: Y. Li, None; Y. Liu, None