Author contributions: Paula Ordonez provided the first version of the manuscript and compiled Figure 1. Nick Di Girolamo provided the outline for the review and was responsible for drafting the manuscript and for performing the histological studies that contributed to Figures 2 and 3.
Disclosure of potential conflicts of interest is found at the end of this article.
First published online in STEM CELLS EXPRESS November 30, 2011.
The cornea contains a reservoir of self-regenerating epithelial cells that are essential for maintaining its transparency and good vision. The study of stem cells in this functionally important organ has grown over the past four decades, partly due to the ease with which this tissue is visualized, its accessibility with minimally invasive instruments, and the fact that its stem cells are segregated within a transitional zone between two functionally diverse epithelia. While human, animal, and ex vivo models have been instrumental in progressing the corneal stem cell field, there is still much to be discovered about this exquisitely sensitive window for sight. This review will provide an overview of the human cornea, where its stem cells reside and how components of the microenvironment including extracellular matrix proteins and their integrin receptors are thought to govern corneal stem cell homeostasis. STEM CELLS 2012; 30:100–107.
The World Health Organization estimates that 45 million people suffer from bilateral blindness worldwide, 10 million of whom are affected by corneal disease. The prevalence of corneal disease varies globally; however, the burden placed upon an individual, community, and economy is uniformly immense . Corneal disorders precipitated by trauma, ulceration, scarring, and other complications encompass a range of disease entities that negatively impact the integrity of the cornea, resulting in partial or complete blindness .
Of the corneal disorders, limbal stem cell deficiency (LSCD) is perhaps the most severe and difficult to treat. This disease is characterized by the inability to replenish the corneal epithelium, resulting in chronic inflammation, vascularization, pain, and blindness . Chronic corneal epithelial defects develop because corneal stem cells (SCs) and/or their niche is damaged through chemical or thermal burns, severe ocular infections, autoimmunity, exposure to cytotoxic agents, long-term contact lens wear, extensive surgeries, and genetic anomalies . In such patients, a corneal graft is ineffective at restoring ocular health and vision. Instead, more specific and targeted therapies are required that involve harvesting autologous  or allogeneic  SCs from the healthy contralateral eye, a living relative or cadaver, expanding them in vitro, and transplanting them using a variety of substrates and delivery devices [2, 6]. Progress in establishing the most effective therapy for these patients has been slow and difficult to gauge due to multiple causes and severity of disease, length of patient follow-up, type of graft, and culture conditions used [2, 7]. If corneal SCs can be better identified, isolated, and maintained ex vivo, graft quality can be improved for more promising patient outcomes.
The Human Cornea
The anterior surface of the eye is defined by the cornea that is separated from the surrounding conjunctiva by a transition zone termed the limbus (Fig. 1A, 1B) . During development, the human cornea is one of the last ocular structures to form. Interactions between the lens vesicle and overlying surface ectoderm as well as involvement of Wnt signaling pathways  and the oculogenic transcription factor Pax6  are critical for epithelial development. Corneal stromal keratocytes (fibroblasts) and endothelial cells are derived from the neural crest.
The adult cornea is a lamellar-structured tissue composed of a stratified epithelium, a collagenous substantia propria (stroma) sparsely populated with keratocytes, and an inner monolayer of endothelium, with each cell type separated by a specialized membrane, anteriorly by Bowman's layer and posteriorly by Descemets membrane (Fig. 1C). The corneal epithelium is further segregated into three layers: basal, wing, and squames (Figs. 1C, 2A, 2B) . Basal cells secrete matrix molecules that can be incorporated into the basement membrane (BM) and stroma. Wing cells and squames differentiate from basal cells. Squames form lateral tight junctions that protect against external environment, while wing cells participate in wound healing . Constant cyclic shedding and replacement or “self-renewal” maintains corneal integrity and ability to fulfill its refractive and protective roles.
Epithelial cell migration, proliferation, and replication during self-renewal can be explained by the XYZ hypothesis, which proposes that epithelial loss from the surface (Z) equals the proliferation of basal epithelial cells (X) plus the number of cells moving in a centripetal manner from the periphery (Y) . Over the past four decades, in vivo animal studies have consistently demonstrated cells migrating circumferentially and centripetally toward the central cornea from the limbus during re-epithelialization . Furthermore, patients with severe corneal injuries have been successfully treated with limbal tissue grafts [2, 5, 7, 14], compelling evidence that the limbus is the repository for corneal epithelial SCs.
THE LIMBUS AND OTHER SC NICHES
The concept that a SC niche provides a unique and appropriate microenvironment to support self-renewal and multipotential activity was first proposed in the late 1970s . Niches are three-dimensional SC-sheltering, highly organized interactive structural units which commonly occur at tissue intersections or transition zones (e.g., corneo-limbal, esophago-gastric, endo-ectocervical) . Molecular crosstalk from surrounding cells and soluble signals from the immediate vasculature or from extracellular matrix (ECM)-sequestered mediators unique to the microenvironment are thought to provide the differential cues that dictate SC homeostatic or activation programs . The physical and chemical signals between cells, matrix glycoproteins, and the three-dimensional spaces they form allow molecular interactions that are critical for regulating SC function. Identification and characterization of tissue niches has revealed a constellation of components; however, the mechanisms that govern how niches are established and maintained to support SC functions are just beginning to be understood . Moreover, new tools for labeling SC in situ have facilitated the localization and characterization of SC niches in mammalian tissues [18, 19].
The hair follicle has emerged as one of the most extensively studied models for examining adult SC. Multipotent hair follicle SCs (HFSCs) responsible for regenerating skin, hair, and sebaceous glands reside in a region of the outer root sheath termed the hair follicle bulge [20, 21]. Early studies that took advantage of the slow-cycling nature of bulge cells permitted their identification and isolation as label-retaining progenitors [20, 21]. These cells formed large colonies in vitro and intact hair follicles when transplanted into hairless mice . The discovery of molecular markers to better identify HFSCs has significantly advanced our understanding of the biology and physiology of these cells including their adhesion to ECM proteins and the molecules involved in cell-cycle control.
Like the HFSC niche, the mechanisms involved in controlling adult hematopoietic SCs (HSCs) remain largely unknown. HSCs are a subset of bone marrow-derived cells capable of self-renewal or multipotential differentiation. Quiescent HSCs traverse along the inner surface of bone lined with osteoblasts . When HSCs mature, they lose contact with surrounding stromal cells and begin to proliferate. Numerous experiments involving osteoblast ablated mice, and mice genetically altered to increase osteoblast number, have led to one theory of regulation which suggests that HSCs retain their quiescent characteristics due to their ability to adhere to osteoblasts through N-cadherin-mediated adheren junctions [22, 23]. Another study has demonstrated that HSCs express calcium sensing receptors and HSCs that lacked this receptor failed to localize to the endosteal niche and did not function normally after transplantation . This highlights the importance of the ionic mineral content of the bone and its matrix in retention of HSCs within the niche [22, 23]. Furthermore, osteoblasts are known to express angiopoietin-1 which interacts with a tyrosine kinase receptor on quiescent HSCs, thereby enhancing their adhesive capability and offering an optimal environment for hematopoiesis .
A unique feature of the cornea that distinguishes it from other SC-harboring organs is its transparency (Fig. 1). Furthermore, due to its superficial anatomical location, the limbus is the only SC niche that is readily visualized using noninvasive slit-lamp and in vivo confocal microscopy. The prevailing notion is that unipotent corneal SCs within the basal layer of the limbus (hereinafter referred to as limbal SCs [LSCs]) maintain the corneal epithelium during normal cell turnover and following injury [9, 11, 12]. Clues as to the location of cells with regenerative capacity originate from studies that demonstrated the migration route of pigmented cells from the limbal zone . Afterward, landmark studies by Cotsarelis et al.  provided convincing evidence of label-retaining, slow-cycling, stem-like cells at the limbus, which progressively lost their radio-label as they migrated toward the central cornea. More recent studies using chimeras have demonstrated radial striping pattern of transient amplifying cells (TACs) streaming from the limbus toward the central cornea [13, 26].
LSCs are located within the basal interpalisade epithelial papillae of the Palisades of Vogt (Figs. 1C, 2C, 2D) and are often visualized in small clusters (Fig. 2D, inset) [27, 28]. Groups of similar cells have also been noted in the limbal transition zone. Transmission electron microscopy has revealed that these clusters are composed of several small primitive appearing cells that are in direct contact with the BM, have heterochromatin-rich nuclei and a sparse cytoplasm with few mitochondria, ribosomes, and hemidesmosomes. The remaining cells within these aggregates are slightly larger with more prominent organelles and are thought to represent early TACs . Furthermore, label-tracing studies in animals have identified slow-cycling cells predominantly in the limbus . The Palisades of Vogt are a series of fibrovascular ridges, radially orientated, and more densely located at the superior and inferior limbal borders as assessed by biomicroscopy on human subjects . The stromal matrix and BM of the ridges are closely associated with LSCs, enabling them to communicate through cell-to-cell, cell-to-ECM, or paracrine signals . Furthermore, the depth of the epithelial–stromal papillae may shelter SCs from physical shearing stressors, while the palisades provide a larger surface area to accommodate more SCs within a confined zone. Pigmented caps on LSC and the presence of interspersed melanocytes are thought to protect against ultraviolet radiation and oxidative DNA damage [28, 32].
The advent of powerful imaging techniques has led to unique but controversial discoveries in defining the human limbal architecture. Several novel structures have been identified, which are thought to be part of the limbal complex. Using light and transmission electron microscopy, Dua et al.  identified “limbal epithelial crypts” as solid cords of cells arising from the underside of the interpalisade epithelial rete ridges of the Palisades of Vogt which varied in size, extended deep into the limbal stroma, and were observed as evenly distributed clusters around the limbus. The hypothesis that this novel structure acts as a SC repository was supported by the expression of well-accepted LSC markers . Shortt et al.  used confocal and scanning electron microscopy to identify limbal crypts, which were deemed different to those observed by Dua's group. These well-circumscribed structures were located adjacent to the Palisades of Vogt, laterally, and surrounded by a highly cellular and vascularized stroma. The same investigators also identified “focal stromal projections” that extended upward into the corneal limbal epithelium, were surrounded by smaller, tightly packed basal cells, and contained a central blood vessel. Areas of limbus abundant in crypts and stromal projections also contain cells expressing LSC markers . We too have noted crypt-like structures (Fig. 2E) and stromal-like projections (Fig. 2F) composed of tightly packed small basal-like cells that stain for p75 on limbal epithelial progenitors (Fig. 2G)  within the limbus. However, it should be noted that donor-to-donor variation (including age), tissue orientation, and depth of sections may contribute to uncertainties regarding the precise architecture of the limbal complex. Further comprehensive and independent studies are warranted to determine whether these structures are indeed part of the limbal SC niche complex as well as establishing their precise role.
Limbal Epithelial SC Function and Morphology
LSCs are slow-cycling  undifferentiated cells that are proposed to undergo asymmetric division giving rise to corneal epithelial cells . Likewise, cells that retain DNA label  and undergo asymmetric division to promote stratification and differentiation  have been identified in the mammalian skin. Upon dividing, one daughter cell remains within the niche to replenish the SC population, while the other detaches from its BM and becomes a TAC committed to differentiate (Fig. 3D). Although TACs divide at an exponential rate, this activity is limited as they senesce into superficial terminally differentiated cells (squames) after migrating centripetally from the peripheral limbus (Fig. 1C) [8, 26]. LSCs characteristically possess a large nuclear to cytoplasm content, are highly clonogenic and proliferate long-term under ideal culture conditions [20, 40]. Suprabasal limbal epithelial cells are generally larger and are associated with increased differentiation and decreased colony-forming efficiency and growth capacity . Immunohistochemical analysis consistently reveals the presence of the differentiation marker CK-3 in the epithelium of the central cornea and in suprabasal cells of the limbus but not in the basal limbal epithelium, suggesting basal cells are more primitive [28, 42].
Limbal Epithelial SC Markers
The ability to identify SCs in different organs has been hampered by unreliable and nonspecific markers. The discovery of robust markers would enhance isolation and expansion strategies, thereby providing more effective cell-based therapies. Several molecules have been suggested as potential markers; however, to meet the selection criteria, they must be expressed in a restricted manner and functional studies conducted to validate their inclusion.
The ATP-binding cassette protein (ABC) G2, is a cell surface transporter protein  thought to protect LSCs from oxidative stress induced by toxins , a factor that also supports progenitors by transporting small regulatory molecules required for their proliferation, differentiation, and apoptosis . Side population cells are visualized by flow cytometry and commonly include SCs that are able to efflux the fluorescent dye Hoechst 33342 . ABCG2 is responsible for effluxing this dye from side population cells , and the presence of these cells within the limbus correlates with ABCG2 expression [43, 44].
The transcription factor p63 is important for epithelial development  and has been used to distinguish SCs from their TAC progeny in the cornea . Only the ΔNp63α isoform has been shown to be highly expressed in the limbus during resting state, whereas in the cornea, it is undetectable . p63+/ABCG2+ cells display morphological and functional features of LSCs . N-Cadherin and cytokeratin-15 are two other promising candidate LSC antigens, which colocalize to clusters of progenitor-like cells in the limbus . Overall, it is unclear as to whether any of these markers are specific enough to distinguish between stem and early TACs. Perhaps, clues to identifying a specific marker(s) for LSCs may become more obvious after determining the precise composition of the limbal BM to which these cells adhere.
Differences Between Limbal and Corneal Extracellular Matrices
Characterizing the precise molecular composition of the limbal niche has been a major limitation in the field. While in vitro models and clinical studies of corneal reconstruction with fibrin and human amniotic membrane have been used widely for decades [3, 7, 28], these bioscaffolds only provide a short-term solution until they are effectively remodeled on the eye. Hence, it is possible that unsuccessful cell therapies for patients with LSCD are attributable to inappropriate molecular grafts which bear little resemblance to the actual SC niche. Researchers are therefore focusing on developing appropriate biomaterials to maintain SC function long term in vivo.
It is widely accepted that interactions between SCs and their microenvironment are regulated and mediated through intrinsic and extrinsic factors [9, 17, 51]. During corneal development, a diverse signaling program may be in place as the niche models and remodels structurally, functionally, and biochemically. For example, it is not until just prior to eyelid opening in humans that SC activity accumulates within the rudimentary limbal niche, and it is not until the postnatal period that the LSC niche is properly formed . During adulthood, stromal keratocytes are thought to contribute signals, which support limbal epithelial cell homeostasis through the cytokines and growth factors they secrete . Limbal fibroblast conditioned media contains factors that are therapeutically beneficial in animal models of LSCD . Furthermore, these cells regulate limbal epithelial cell proliferation, differentiation, and phenotype  by synthesizing ECM proteins that are deposited within the substantia propria and basal lamina. Components of BMs vary according to tissue location and this compositional variation may be responsible for differences in cellular behavior . On the ocular surface, the stromal matrix and BM of the corneal, limbal, and conjunctival epithelia differ significantly [27, 31], but the precise compositional make-up is yet to be fully elucidated. Elegant tissue recombination studies using rabbit limbal and central corneal epithelial sheets placed on either limbal or corneal stroma demonstrated that the limbal stroma maintained “stemness,” while the corneal stroma promoted cell differentiation, proliferation, and apoptosis . In other tissue recombination studies, adult central corneal epithelium responded to specific signals derived from embryonic dermis to undergo transdifferentiation into epidermal cells . Conversely, a corneal limbal microenvironment facilitated the reprogramming of HFSCs into corneal epithelium . These studies highlight the importance of signals that emanate from stromal elements in dictating cell fate.
Collagens are the major structural component of BMs. Collagen types VII, XVI , and IV  are more abundant in the limbal compared to the corneal BM. Differences in collagen type IV (Col-IV) distribution is thought to be attributed to developmental cues and differentiation status of corneal cells. For example, the human fetal corneal BM displays uninterrupted Col-IV immunoreactivity, contrasting remarkably with the absence of this factor from its adult counterpart . Similar regional changes in the expression of tenascin-C have been noted in the developing human cornea, widely observed in preterm, less so in neonatal and restricted to the limbal BM in adults . This consolidated expression of specific BM components during development is analogous to SC segregation into the rudimentary limbal niche with increasing gestational age in humans . Interestingly, tenascin-C colocalizes to ABCG2+/p63+/CK-19+ cell clusters within the limbus , suggesting a role in supporting SC homeostasis. In addition, limbal  and cutaneous  epithelial cells that rapidly adhered to Col-IV possessed SC-like characteristics. Other proteins more abundant in the limbal BM include laminin (LN)-α1, α2, β1 chains, agrin, and vitronectin (VN) [27, 31] (Fig. 3A, 3B, 3D). Given these compositional differences, receptors for their ligands, including integrins, could be used to identify LSCs.
Corneal Integrins: Potential Markers of SCs
Integrins are a superfamily of multifunctional cell-surface receptors that comprise heterodimers of α and β subunits. They possess large extracellular domains (some which form a docking site for Arginine-Glysine-Apartate [RGD] peptide motifs), a transmembrane region, and an intracellular domain that binds cytoskeletal components (Fig. 3C). Currently, 24 α and 9 β subunits have been identified, forming an array of heterodimeric combinations that localize to distinct cellular layers. Integrins regulate organogenesis, cell adhesion, proliferation, differentiation, migration, and death  by transducing signals between the extracellular and intracellular compartments. An added level of complexity exists as integrins interact with growth factor receptors to activate intracellular signaling cascades . Integrins that mediate cell attachment to proteins on BMs can identify SCs as they anchor cells in specific zones within a tissue. Loss of integrin expression is thought to trigger departure from the niche, initiating a differentiation program (Fig. 3D).
Integrins are differentially expressed across the ocular surface epithelium [66, 68] as they engage with BM proteins including fibronectin (FN), VN, LN, and Col-IV, which themselves are deposited in a restricted manner (Fig. 3D) [27, 31]. The most abundant corneal epithelial integrins include α2, −3, −5, −6, −v and β1, −4, and −5. Integrins α2, −3, −v and β1 and −5 are localized at sites of cell-to-cell contact within the rat corneal epithelium, while α5 and −6 and β4 are found in basal epithelial cells . Just as BM proteins, Col-IV  and tenascin-C  become restricted to the limbal zone during corneal development, so too does α9-integrin , and because tenascin-C binds α9 , it was proposed that this integrin could be used to identify a population of mouse corneal epithelial progenitors. However, this was not the case, as DNA label-retaining cells did not include α9-positive cells; instead, slow-cycling cells were represented by high levels of β1 and β4 integrins . Recently, α6-bright/CD71-dim cells isolated from human corneo-scleral rims were enriched for a minor population of small clonogenic cells that stained with LSC markers . This combination of cell-surface molecules was also used to successfully isolate human keratinocyte  and murine esophageal  SCs. Likewise, mouse spermatogonial SCs were distinguished by excessive α6 and β1 . In vitro and in vivo experiments conducted on human prostate epithelium also provide significant supporting evidence for integrins as possible SC markers and as factors that support SC function. Rapid adherence to Col-I was used to isolate α2β1 bright cells that demonstrated higher colony forming ability in comparison to nonadherent α2β1 dim counterparts and formed a fully differentiated prostate epithelium when grafted into nude mice .
A major point of contention in the ocular field has been deciphering which (if any) of the current markers decorate quiescent LSCs with pin-point accuracy. Perhaps, researchers in the general SC field have approached this problem narrow mindedly by focusing solely on specific SC types (e.g., corneal, esophageal, bladder, skin) to derive candidate positive and negative markers. Less than a decade ago a rather novel approach was used by three independent laboratories, each comparing global gene-expression profiles of three distinct SC types including embryonic, neuronal, and hematopoietic to their differentiated progenies [77–79]. After intersecting the list of genes, the data revealed a single common gene, namely α6-integrin . Interestingly, the same integrin was used to isolate limbal , cutaneous , and esophageal  SCs. These investigations have sparked ongoing debate as to whether there exists a universal stemness marker or transciptome. Nonetheless, the identification of a common gene expressed by three diverse SC types highlights the relevance of integrin molecules as critical niche adhesion factors and potential SC markers.
The advent of tissue-specific knockout mice has highlighted the relevance and multiple roles of integrins. Deletion of β1 in the epidermis is associated with perhaps the most profound phenotype, including defective hair follicle development, BM disruptions, reduced cell proliferation and differentiation, and impaired wound healing which may be due to a depleted or disorganized SC compartment [39, 80]. However, other studies in mice suggest the presence of compensatory mechanisms that result in normal tissue phenotype after selective integrin ablation . The impact of selective integrin deletion in the cornea has not yet been assessed, although mice lacking αvβ5 integrin progressively develop blindness due to retinal dysfunction . Furthermore, selective deletion of αv-integrins affects basal epithelial cells of the eyelid skin and conjunctiva by inducing tumorigenesis . These data support the notion that integrin–ECM coupling is a mechanism that keeps cell differentiation, proliferation, and transformation in check.
Although knowledge of the structural and compositional make-up of the limbal niche has grown exponentially during the past decade, there is still much to be discovered about this unique SC repository. Determining the precise proteomic profile of the BM that spans the ocular surface may inform researchers as to the best conditions to effectively expand LSC in vitro, while preserving their phenotype and differentiation status. Additional mechanisms that govern stemness under homeostatic conditions may arise from growth factors that are sequestered on selective glycoproteins (tenascin-C, LN, and VN) found in certain BM regions along the cornea that are able to influence cell phenotype and behavior through intracellular signaling . Moreover, single or most likely combinations of extracellular matrices could be used as a surrogate BM in culture models to induce transdifferentiation of SCs from related or diverse epithelia (conjunctival, nasal, mucosal, and cutaneous) into cells of corneo-limbal lineage for clinical use. Currently, no definitive marker exists to illuminate LSCs or to distinguish them from their early TAC progeny. However, integrins may prove ideal markers as their expression on cell membranes provides a tethering point for isolating, enriching, and assessing cells and in combination with appropriate ECM factors, improvements in culture conditions, and current cell-based therapies may be achieved for patients with LSCD. Finally, SCs of many self-renewing epithelia share molecular properties; hence, information gained from ocular studies may inform researchers developing treatment strategies for other organ systems.
N.D.G. is supported by a Career Development Award (no. 455358) from the National Health and Medial Research Council of Australia. Other grants supporting this study are University of NSW GoldStar Award and Australian Stem Cell Centre Strategic Development Award.