Microarray Studies Reveal Macrophage-like Function of Stromal Keratocytes in the Cornea

Abstract

purpose. To elucidate biological processes underlying the keratocyte, fibroblast, and myofibroblast phenotypes of corneal stromal cells, the gene expression patterns of these primary cultures from mouse cornea were compared with those of the adult and 10-day postnatal mouse cornea.

methods. Murine Genome_U74Av2 arrays (Affymetrix Inc., Santa Clara, CA) were used to elucidate gene expression patterns of adult and postnatal day-10 corneal stroma, keratocytes, fibroblasts, and myofibroblasts.

results. Mobilization of stromal cells by culturing led to a wound-healing cascade in which specific extracellular matrix and cornea-transparency–related genes were turned off, and a repertoire of macrophage genes were switched on. Thus, novel transparency-related crystallins detected in the corneal gene expression patterns were downregulated in culture, whereas macrophage genes, mannose receptor type-1, Cd68, serum amyloid-A3, chemokine ligands (Ccl2, Ccl7, Ccl9), lipocalin-2, and matrix metalloproteinase-3 and -12 of innate immunity were induced in primary keratocyte cultures. Fibroblasts expressed the growth-related genes lymphocyte antigen 6 complex locus-A and preprokephalin-1, and myofibroblasts expressed annexin-A8, WNT1-inducible signaling pathway protein-1, arginosuccinate synthetase-1, and procollagen XI of late-stage wound healing.

conclusions. The emergent biological process suggests a dual role for resident stromal keratocytes in the avascular cornea: one of cornea maintenance, which involves synthesis of proteins related to the extracellular matrix and corneal transparency, and a second of barrier protection macrophage functions, which are switched on during corneal infection and injury.

The cornea is a unique connective tissue that combines transparency, refractive power for correct vision, tensile strength, and protection against infections. ^{1 2} From a biological standpoint, it does so with remarkable finesse. Its multilayered protective epithelium; a transparent, refractive stroma of highly organized collagen fibrils; and a single layered endothelium, which regulates hydration, are suited for these functions. For undisrupted vision, the cornea minimizes inflammation and maintains an immune-privileged status. ³ In addition to the epithelium and the tear film, which are most important to the barrier function of the cornea, the avascular stroma provides additional barrier protection. With minimal protein turnover in the adult cornea, the stromal extracellular matrix (ECM) genes are transcriptionally active during corneal maturation and healing. Exploring wound healing, investigators in recent studies have reported global gene expression changes in the adult rat and mouse cornea after wounding where the entire cornea was used to obtain average gene expression patterns of all cell types. ^{4 5}  

A key to understanding healing after injury is the keratocyte, the major cell type of the corneal stroma and the focus of this study. On injury, the keratocytes migrate to the wound and reportedly transdifferentiate into fibroblasts. ⁶ At later stages of healing and scarring, keratocytes have well-developed F-actin stress fibers, express α-smooth muscle actin, and become myofibroblastic. The latter is implicated in ECM regeneration and fibrosis in all healing tissues. ^{7 8 9} These three cellular phenotypes are often reproduced in culture to study specific aspects of the resting and wound healing cornea. Primary cultures of keratocytes in a serum-poor condition preserve the stellate morphology and still synthesize stromal keratan sulfate proteoglycans. Exposed to serum, these cells become fibroblastic; treated with TGF-β1, the stromal cells differentiate into myofibroblasts. ^{7 10} Beyond the morphologic differences and expression of a few genes, little is known about gene expression changes that distinguish these cellular phenotypes. ⁷  

In the present study, we investigated gene expression patterns of primary keratocytes, fibroblasts, and myofibroblasts compared with that of the adult and postnatal 10-day-old (P10) corneal stroma. Our results show that mobilization of stromal cells by culturing induced a wound–infection response. Prolonged treatment of stromal fibroblasts with TGF-β1 induced anti-inflammatory and profibrogenic genes in the myofibroblast phenotype. Overall, gene expression patterns of the cultured stromal cells resembled that of the immature cornea more closely than the adult. 

Materials and Methods

Sample Preparation

According to protocols approved by The Johns Hopkins University and in compliance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research, 6-week- (adult) and postnatal 10-day-old (P10) mice (Charles River Laboratories, Wilmington, MA) were used to obtain corneas. Corneas were dissected from the limbus, the epithelium and the endothelium were scraped off from the stroma before extraction of stromal total RNA (TRIzol; Invitrogen-Gibco, Grand Island, NY). It is conceivable that scraping of the epithelium can alter gene expression patterns in the stroma. However, the corneas were treated with extraction reagent to isolate the RNA within minutes of scraping, which should prevent further changes in gene expression. Independent total RNA samples were prepared from three pools of adult stroma (3.5 μg RNA per pool of 50 corneas) and two pools of P10 stroma (5 μg of RNA per pool of 40 corneas) for microarray hybridizations. 

Stromal cells were isolated from adult intact corneas and cultured in DMEM containing 1% platelet-poor horse serum, as described before. ¹⁰ The epithelium and endothelium were not scraped off (as they were for isolating RNA from the intact corneal stroma), as these culture conditions do not support epithelial cell survival. After 48 hours the cells were placed in (1) DMEM with 1% platelet-poor horse serum (keratocytes in serum-poor), (2) DMEM containing 10% FBS (fibroblasts in serum-rich), and (3) DMEM containing 10% FBS and 10 ng/mL TGF-β1 (myofibroblasts) ¹⁰ for 4 days. Three independent RNA samples were prepared from primary cultures of each cell type established from separate pools of mouse corneas, to obtain three gene expression patterns per cell type. 

Microarray

Biotinylated cRNA was prepared from total RNA according to recommended protocols. Murine genome_U74Av2 arrays (9824 unique transcripts; Affymetrix Inc., Santa Clara, CA) were used. The error in expression signals between stromal RNA preparations is low (correlation coefficient > 0.99), but is higher between RNA preparations of cell cultures. The lowest correlation coefficient, between two fibroblast profiles, is 0.92 and the highest, between two keratocyte profiles (serum-poor) is 0.99. Variable growth rates in serum-rich medium and consequent differences in cell density at harvest time may be factors contributing to gene expression variability. 

Immunofluorescence

Cells plated on coverslips were fixed in 4% paraformaldehyde. Cells and paraffin-embedded sections (5–6 μm) from adult corneas were immunostained with antibodies (Santa Cruz Biotechnology, Santa Cruz, CA) as indicated in Figure 3 . Actin filaments were visualized by staining with Alexa Fluor 568 Phalloidin (Molecular Probes, Eugene, OR). 

Data Analysis

Scanned images were analyzed by the GeneChip Analysis Software, ver. 5.0 (Affymetrix Inc.). Background intensities ranged between 41 ± 1.04 and 54 ± 0.59 arbitrary units. The model-fitting DNA chip-analyzer software (dChip; Affymetrix Inc.) and an invariant-set normalization method were used for data normalization and to calculate a model-based expression value for each transcript. ^{11 12}  

We used hierarchical clustering to visualize genes with similar expression levels across sample types and grouped sample types based on similar gene expression. As the distance between two genes we chose 1 – r, where r is the Pearson correlation coefficient between the expression values of two genes across samples, standardized so that the mean is 0 and the standard deviation is 1. We used agglomerative clustering with centroid-linkage as implemented in the chip-analyzer program. ¹²  

Gene annotation was based on the NetAffx (Affymetrix Inc.) and Gene Ontology (GO) (August 20, 2003, www.geneontology.org). LocusLink, UniGene, (www.ncbi.nih.gov/LocusLink/), Online Mendelian Inheritance in Man (OMIM; www.ncbi.nlm.nih.gov/omim/), and ENZYME, Swiss-Prot and TrEMBL, Swiss Institute of Bioinformatics (www.expasy.com) were used to procure information on gene functions. The primary data are available at the Gene Expression Omnibus Database (www.ncbi.nlm.nih.gov/geo/). (LocusLink, UniGene, OMIM, and the Gene Expression Omnibus Database are provided in the public domain by the National Center for Biotechnology Information, Bethesda, MD.) 

Results

Overall Gene Expression Patterns of Corneal Stromal Cells in Culture

Gene expression patterns were obtained from three independently isolated total RNA preparations of keratocyte, fibroblast, myofibroblast primary cultures and the adult stroma, and two such preparations from P10 stroma. The postnatal cornea was included to have a complete representation of stromal genes. A significant proportion of stromal transcripts may be underrepresented in the adult cornea, where the stroma is quiescent, with little turnover of stromal proteins. 

We filtered the gene expression data from 14 arrays of the cornea and cell culture samples by mean expression index ≥100 arbitrary units and a “present call” in ≥50% of sample replicates, to obtain 5002 significantly expressed transcripts, of which 3608 are known genes. These were analyzed further by hierarchical clustering of genes with similar levels of expression along the vertical axis and samples with similar expression patterns along the horizontal axis (Fig. 1) . The cell types clearly clustered separately from the corneal profiles, substantiating a greater degree of similarity between the cornea samples of two ages than corneal cells in culture. We assessed similarities in gene expression of 15 functional groups between each of the cell types and the cornea of two ages (Table 1) . For most functional clusters, the cellular phenotypes resembled P10 more closely than adult, as indicated by the percent similarities (Table 1) . The cell cultures are most different from the cornea in angiogenesis. A complex process prevents vasculature of the cornea that seems to be deregulated in cultured cells. Also notable is the greater similarity between myofibroblasts and the cornea in immune and inflammation response (Table 1 , asterisks), evidently a reflection of TGF-β responsive downregulation of specific genes (discussed later). 

Using the adult corneal gene expression pattern as a baseline, we next compared expression of the 5002 genes/expressed sequence tags (ESTs) in P10 and cell cultures. The analysis yielded 865 differentially expressed genes/ESTs (705 known genes), with a median false discovery rate of 0.0% between each comparison (Fig. 2) . The following general trends in the gene expression patterns were evident: (1) Corneal housekeeping functions that include transparency, oxidative stress protection, and detoxification genes were downregulated in culture. (2) Of the genes expressed in the cornea, a small subset was retained in primary keratocyte cultures. (3) Acute phase, wound healing, cell adhesion, ECM, and ECM-remodeling genes were upregulated in culture. (4) Macrophage-like gene expression was induced in keratocytes. (5) Late-healing and profibrogenic TGF-β responsive changes occurred in myofibroblasts. 

Genes Related to Corneal Housekeeping Functions, Transparency, Oxidative Stress Protection, and Detoxification Downregulated in Culture

Several corneal maintenance genes were downregulated in the primary cultures. Crystallins, other heat shock proteins, glutathione S-transferase, transketolase, aldolase, and aldehyde dehydrogenase are enzymes or stress proteins. Studied extensively in the lens, these have multiple roles, ranging from detoxification to protection against oxidative, inflammatory, and ultraviolet light stresses, to regulating refractive, transparent properties. ^{13 14} These may have similar functions in the cornea. Several known and novel crystallins (Aim1, Cryaa, Cryba1, Crybb2, Crygb, Crygd), ² glutathione S-transferase α-4, omega-1 (Gsta4, Gsto1), and transketolase (Tkt), expressed in the cornea of both ages, were downregulated in culture (Fig. 2) . A recent serial analysis of gene expression (SAGE) reported Gsta4 and Gsto1 as abundantly expressed in corneas, and by in situ hybridization further localized the transcripts in the epithelium. ¹⁵ We found glutathione S-transferase expression in de-epithelialized stroma, suggesting that there may be some residual epithelial RNA in our preparations. However, we also detected Gsta4 and Gsto1 expression in the cultured keratocytes, where epithelial contamination is very unlikely (epithelial cells are lost in culture). 

Paired box gene 6 (Pax6) expression, high in P10 and adult cornea, was low in culture. Based on its role in activating certain vertebrate lens crystallins, ¹⁶ we speculate that it may have a similar crystallin-regulating role in the cornea. Pax6 regulates ocular development and is known to be expressed strongly in the epithelium. It is also present at lower levels in the stroma, where it is essential for normal stromal development, as shown in wild-type/pax6 ^−/− chimeric animals. ¹⁷ In the primary cultures, Pax6 was downregulated, but paired mesoderm homeobox 1 and 2 (Prrx1 and Prrx2) were induced, raising the question as to whether these are involved in negative regulation of Pax6 in culture. Keratoepithelin (or TGF-β induced, Tgfbi) was dramatically downregulated in culture. Tgfbi regulates ECM-cell adhesion and its mutations cause several human corneal dystrophies. ¹⁸  

Corneal Gene Expression Pattern Retained by Primary Keratocyte Cultures

Stromal keratocytes cultured in serum-poor medium, rather than fibroblasts in serum-rich medium, have long been considered to resemble the in situ keratocytes of the cornea. ^{7 10} Therefore, we looked for corneal stroma–like gene expression in our cultured keratocytes. The corneal ECM proteoglycan keratocan, ¹⁹ and the eicosanoid metabolic enzyme PGD2 synthase (Ptgds, regulates prostaglandins in inhibiting platelet aggregation and release of cAMP for cellular signaling) were earlier shown to be present in the cornea and cultured keratocyte. Both of these keratocyte markers were present in our keratocyte profiles (Fig. 2) . We further confirmed these in cultured keratocytes and the cornea by immunofluorescence (Fig. 3) . Consistent with prior findings that keratocytes from transparent cornea, but not fibroblasts from the opaque ocular sclera, express aldehyde dehydrogenase (Aldh1a3) and transketolase (Tkt), ²⁰ our keratocyte profiles were positive for aldolase (Aldo3) and transketolase transcripts. Pax6 and its target ephrin B2 (Efnb2) was expressed in keratocytes (Fig. 2) . By immunofluorescence, we detected ephrin B2 in keratocytes and in the corneal stroma (Fig. 3) . Ephrins signal in neural crest cell migration and synaptic plasticity. ²¹ Therefore, ephrin B2 may function in corneal innervations or in maintaining the typical cytoskeletal structure of corneal keratocytes. Other noteworthy overlaps in gene expression between cultured keratocytes and the corneal stroma that bolster functional similarity between in situ keratocytes and primary keratocyte cultures include phospholipase A2 group VII (Pla2g7, inactivates proinflammatory platelet activating factor in the cornea), agrin and glutamate-ammonia ligase (Glul). Agrin interacts with acetylcholine receptors at neuromuscular junctions; functional implications of its expression in the cornea and keratocytes are unknown. Glutamate-ammonia ligase (Glul) eliminates free ammonia by converting neurotoxic glutamate to nontoxic glutamine and functions in corneal detoxification. 

Upregulation of Acute Phase, Wound-Healing, Angiogenesis, Cell Adhesion, ECM, and ECM-Remodeling Genes in Cultured Stromal Cells

Stromal cells in culture upregulated many wound-healing–related genes (Fig. 2) . Acute phase and inflammatory responses included overexpression of serum amyloid A3 (Saa3), IL-1 receptor antagonist (Il1rn), IL-1 receptor 1 (Il1r1), IL-1 receptor-like 1 (Il1rl1), IL-6 signal transducer (Il6st), endoglin (Eng), kit ligand (Kitl), colony-stimulating factor 1 (Csf1), and TNF receptor superfamily members (Tnfrsf1a, Tnfrsf11b). Several of these were also reported in Pseudomonas aeruginosa–infected mouse corneas. ²² Additional immune response genes are discussed in the context of macrophage-like properties of cultured keratocytes. Coregulated Wnt signaling pathway genes (Wnt5a, Fzd2, and Wisp1) and increased tenascin may be indicative of additional healing-related functions. Increased thrombospondins (Thbs1, Thbs2) and vascular endothelial growth factor A (Vegfa) suggest angiogenesis-related activities. 

Homeodomain-containing genes Prrx1, Prrx2, Pitx2, Irx3, and Shox2, of early embryonic development, were differentially upregulated in culture and may have tissue regenerative functions during corneal wound healing (Fig. 2) . In contrast, the somatic growth regulatory genes glypican 3 (Gpc3) and cyclin-dependent kinase inhibitor 1C (Cdkn1c or p57), preferentially elevated at P10, were not induced in cell culture. Gpc3 and Cdkn1c mutations are linked to overgrowth syndromes, ^{23 24} and may have somatic growth regulatory roles in the postnatal cornea. 

The stromal keratocytes in particular, upregulate basement membrane (Col4a1, Col4a2, Col15, Col18a1, laminin α2 and -4) and cell adhesive (Cdh5, discussed later) genes. All cellular phenotypes also overexpressed fibronectin and its integrin receptor (Itga5) and fibulin 2 (Fbln2, stabilizes fibronectin fibrils) necessary for making a fibronectin-rich wound matrix. Stress fiber–related cytoskeletal changes include upregulation of rho-associated coiled-coil forming kinase 2 (Rock2), cofilin, profilin-1 and -2, tubulins and tropomyosins (Tpm1, Tpm4), and focal adhesion vinculin and talin. ECM-remodeling matrix metalloproteinase gene Mmp12 and Mmp3 were overexpressed in keratocytes and fibroblasts, as reported earlier in corneal fibroblasts treated with interleukin 1α. ²⁵ Upregulation of interstitial ECM collagen assembly genes, Col1a2, Col5a1, and lysyl oxidase (collagen cross-linking) in all three cell types, may represent a stroma-rebuilding and healing effort that in vivo is likely to persist to late stages of cornea repair. In addition, genes for major collagen-binding corneal ECM proteoglycans, lumican, decorin, and related, noncorneal proteoglycans, fibromodulin, and biglycan were all elevated in our primary cultures. ^{26 27}  

Induction of Macrophage-like Gene Expression in Cultured Keratocytes

An overwhelming cellular response, evident in primary keratocytes in serum-poor medium, was increased expression of tissue-resident macrophage genes, ²⁸ mostly silent in fibroblasts in serum-rich medium (Figs. 2 4) . These keratocytes expressed mannose receptor type 1 (Mrc1), chemokine ligands (Ccl2, Ccl7, Ccl9, and Cxcl12), IL-1 receptor 1, Ptgds (eicosanoid metabolism), lipocalin 2, serum amyloid A3 and Cd68 (expressed by monocytes and macrophages), cathepsins, (Ctsb and Ctsd, involved in antigen degradation/presentation), complement component pathway genes (C1qa, C1qr1, C3, C4), and moesin (involved in T-cell antigen-presenting-cell interactions). There was no significant expression of major histocompatibility complex (MHC) class II genes of monocyte lineage in our keratocyte profile. However, class I MHC genes (H2-d1, H2-q1, H2-q7) were elevated in cultured keratocytes, consistent with recent studies that support an integral role for MHC class I–mediated antigen presentation in stress and acute phase response. ²⁹  

Late Healing and Profibrogenic TGF-β–Responsive Changes in Myofibroblasts

Myofibroblasts play a key role in the healing and scarring of stromal wounds in corneas and are implicated in ECM regeneration and fibrosis of healing tissues. ^{8 9} We generated corneal myofibroblasts by maintaining primary stromal cells in medium with 10% fetal calf serum (as for fibroblasts) and 10 ng/mL TGF-β1. ¹⁰ Genes responding to TGF-β consisted of 15 upregulated and 18 downregulated genes indicative of late-healing events (Table 2 ; Figs. 2 4 ). Connective tissue growth factor (Ctgf) and WNT1-inducible signaling pathway protein 1 (Wisp1, a Ctgf family member of the WNT pathway), elevated in the myofibroblast phenotype, may function in cell growth and migration. Argininosuccinate synthase (Ass1) was increased more than sixfold and is necessary for arginine synthesis (from citrulline, urea cycle). The sole substrate for nitric oxide synthase, it may be involved in cellular defense. Cellular retinoic acid binding protein II (Crabp2, a retinoic acid–inducible, embryonic development–related gene), upregulated in myofibroblasts only, may regulate cellular differentiation. Specific, potentially fibrosis-related changes included overexpression of fibulin 2 (stabilization of elastic fiber suprastructures), Col11a1, encoding a minor cartilaginous fibrillar collagen ³⁰ also detected in donor human cornea ³¹ and Col3a1 (Figs. 2 4) . Major downregulations in the myofibroblast phenotype were platelet-derived growth factor receptor α (Pdgfra), IL-6 signal transducer (Il6st), Mmp3, Mmp12, Ccl2, Ccl7, Ccl9, which relate to TGF-β1–mediated healing and anti-inflammatory and profibrogenic processes (Figs. 2 4) . The TGF-β1–responsive changes described herein are quite different from those reported in an earlier study of mouse embryonic fibroblasts exposed to TGF-β1 for 0 to 4 hours. ³² We used a prolonged (4-day) TGF-β1 treatment to induce a myofibroblastic phenotype frequently present in late stages of healing. 

Distinguishing Features of Cultured Keratocytes, Fibroblasts, and Myofibroblasts

To identify distinguishing markers for each cell type, we compared the keratocyte, fibroblast, and myofibroblast gene expression directly with each other and identified 130 uniquely expressed genes (Fig. 4) . Several genes differentially expressed in keratocytes were discussed earlier. Fatty acid synthesis genes sterol-C4-methyl oxidase-like (Sc4mol), farnesyl diphosphate synthetase (Fdps), and prostaglandin D2 synthase (Ptgds), uniquely upregulated in keratocytes, may relate to their response to serum-poor conditions. ^{33 34} Cadherin 5 (Cdh5), of endothelial cell junctions, implicated in Cdh5/β-catenin–mediated endothelial cell survival, was expressed in keratocytes. By immunostaining we confirmed its presence in cultured keratocytes and in epithelial desmosome and hemidesmosome areas of the cornea (Fig. 3) . Keratocytes in situ exist in an interconnected network, and expression of Cdh5 in cultured keratocytes may be a survival response to the loss of cell–cell adhesion. Fibroblasts expressed Ly6a (lymphocyte antigen 6 complex locus A), which influences antigen-induced differentiation of CD4⁺ naive T cells and the growth-related genes Penk1 (preproenkephalin 1), Mest (mesoderm specific transcript), and Shox2 (short stature homeobox 2). By immunofluorescence, we confirmed Ly6a in the cytoplasm of cultured fibroblasts (Fig. 3) . Annexin A8, a vascular anticoagulant, was expressed in myofibroblasts and the cornea (Figs. 2 4) , and, by immunofluorescence, we detected it on myofibroblasts (Fig. 3) . 

Discussion

The stromal keratocytes play a key role in establishing the ECM-rich refractive stroma. During stromal repair after injury or infection they reportedly differentiate into fibroblasts and myofibroblasts. ^{6 7 10 35} To elucidate biological processes that define these phenotypic changes in stromal keratocytes, we compared gene expression patterns of primary cultures of mouse keratocytes, fibroblasts, and myofibroblasts with those of the corneal stroma. Our study revealed an overwhelming upregulation of injury-response–related genes in cultured stromal cells. The stromal cells seem to perceive epithelial cell death during culturing and loss of stromal ECM environment as corneal “trauma” or “injury.” The differential gene expression patterns of stromal cells in culture have unraveled a multistep process, that is likely to represent specific aspects of normal corneal repair from injury or infection (Fig. 5) . Major responses to stromal cell culturing or trauma are downregulation of several corneal housekeeping transparency-related genes, reversion to early developmental programs, and an induction of acute phase response genes. Also, differential expression of angiogenic genes in the primary cultures suggest a breakdown of normal antiangiogenic conditions of the resting cornea. 

Blood serum is not a natural presence in the cornea, and in that respect, the serum-poor culture conditions may be closer to the stromal environment. Consequently, genes expressed in cultured keratocytes, unlike fibroblasts or myofibroblasts, may reflect the earliest changes that occur in the cornea in response to wounding. The cultured keratocytes displayed invasive, cell-adhesive characteristics, expressing more basement membrane–related ECM genes, such as procollagen XV (Col15, adhesion of interstitial connective tissue to basement membrane), procollagen XVIII (Col18a1, its endostatin fragment is angiogenic), and ECM-remodeling enzymes. The most intriguing finding is the expression of genes characteristic of macrophages, including genes encoding mannose receptor, lipocalins, chemokines, IL-1 receptor 1, MHC class I antigens, and others in the cultured keratocytes. Therefore, within the cornea, soon after mobilization of stromal cells by injury, the keratocytes may be induced to become macrophage-like, to enable efficient removal of infectious organisms and cell debris and to aid the inflammatory process. Although, fibroblasts in serum-rich medium continue to express a subset of keratocyte genes, the macrophage-related genes are, for the most part, quiescent. The fibroblast expression pattern may reflect healing events that occur after the initial immune-inflammation response. The myofibroblasts clearly downregulate most of the immune-related, cell-adhesive, and ECM-remodeling genes and continue expressing interstitial ECM collagens and proteoglycans typifying late-healing responses. 

Our study has provided an unprecedented view of the functional plasticity of resident stromal cells, commonly known as keratocytes (not to be confused with dermal epithelial keratinocytes). The keratocytes are responsible for the synthesis of corneal crystallins, stromal proteoglycans, and interstitial fibrillar collagens. ^{7 14} However, in culture, the cells downregulate these corneal housekeeping genes, as we have shown, and switch on a repertoire of macrophage genes, implying that as a first response to injury or infection of the cornea, keratocytes become macrophage-like. An earlier demonstration of phagocytosis in cultured corneal keratocytes and antigen-presentation functions further supports a macrophage identity for stromal keratocytes. ^{36 37} Thus, stromal keratocytes may have a dual function in the cornea: one of synthesizing stromal proteins for a transparent resting cornea and a second of immunoprotection on injury and infection. This discovery raises new questions about the identity of corneal stromal keratocytes. Although, these cells have macrophage characteristics, they do not appear to be of bone marrow–derived monocyte lineage. First, these keratocytes do not express the bone marrow monocyte derivative marker F4/80 antigen, as indicated by our gene expression and immunofluorescence results (Fig. 6) and studies by Seo et al. ³⁷ Second, we detected MHC class I, but not MHC class II transcripts. Seo et al. detected MHC class II antigens only after inducing keratocytes with interferon-γ. Our keratocyte primary cultures expressed the known markers prostaglandin D2 synthase, keratocan, and lumican, another stromal proteoglycan (Fig. 6) , allaying our concern that these may be cells other than those commonly known as keratocytes. In addition, they expressed genes of neural crest origin (ephrin B2, paired box gene 6), suggesting that these are indeed derived from the neural crest during embryonic development ³⁸ and are distinct from the recently discovered bone marrow–derived cells in the cornea. ^{39 40} To maintain a transparent, refractive surface for vision, the cornea must restrict inflammation and create an immunosuppressive microenvironment without compromising tissue integrity against injury and infection. ³ Consistent with this need, MHC class II, but not MHC class I, antigens are generally low in the cornea, to minimize effecter T-cell–mediated inflammation. A further adaptation to this requirement is the macrophage-like nature of the stromal keratocytes, constitutively MHC class II negative, yet capable of MHC class I antigen presentation and innate defense. 

 Supported by Grants NIH EY11654 and EY6111340-LO-A (SC) from the National Eye Institute.

 Submitted for publication March 28, 2004; revised May 3, 2004; accepted May 6, 2004.

 Disclosure: S. Chakravarti, None; F. Wu, None; N. Vij, None; L. Roberts, None; S. Joyce, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Hierarchical clustering of adult cornea, P10, keratocyte, fibroblast, myofibroblast samples based on the variation in expression of 5002 genes/ESTs. Red: expression levels above the mean (in black) expression of a gene across all samples; green: below-mean expression levels.

Table 1.

 

Percent Similarity in Gene Expression between Cultured Stromal Cells and Corneal Samples P10 or Adult

Functional Groups (Number of Genes)	Keratocyte			Fibroblast			Myofibroblast
		P10	Adult	P10	Adult	P10	Adult
Apoptosis (99)	88.9	79.8	88.9	84.8	87.9	84.8
Cell proliferation (236)	87.3	80.5	86.9	80.1	83.9	80.1
Protein modification (252)	84.1	81.0	84.1	81.0	83.3	77.4
Nucleotide metabolism (46)	82.6	65.2	84.8	67.4	87.0	67.4
Cytoskeleton (254)	77.6	72.8	75.2	69.7	73.2	71.3
Lipid metabolism (122)	76.2	66.4	74.6	63.9	73.0	66.4
Carbohydrate metabolism (102)	74.5	68.6	76.5	71.6	73.5	75.5
Cell differentiation (47)	74.5	66.0	70.2	72.3	74.5	74.5
Organogenesis (174)	68.4	60.3	68.4	64.4	64.9	62.6
Immune response (72)	68.1	61.1	76.4	62.5	81.9*	66.7
Inflammation (18)	66.7	55.6	72.2	55.6	88.9*	72.2
Cell adhesion (125)	60.8	45.6	66.4	51.2	62.4	53.6
ECM (87)	58.6	44.8	58.6	47.1	59.8	43.7
Growth factor (37)	54.1	51.4	62.2	64.9	56.8	54.1
Angiogenesis (26)	53.8	38.5	61.5	50.0	50.0	50.0

 Percent gene expression similarity calculated as [(total genes in a group − number of differentially expressed genes)/total genes in a group] × 100.

*  Percent similarity between stromal myofibroblasts and P10 higher than other cell types and corneal samples.

Figure 2.

 

Differential expression by twofold or more of 865 genes/ESTs compared with the adult cornea using the following criteria. An expression index difference of more than 100 arbitrary units, a “present call” in ≥50% of the samples, a more than twofold change at the 90% confidence level and P < 0.05. Specific biological processes affected by the genes are indicated in solid colored bars on the left. The names of genes involved in the synthesis of immune and inflammation related (black), ECM (brown), homeodomain containing (red), cell adhesion and cytoskeletal (green), growth-related and corneal transparency/housekeeping (purple) proteins are in the indicated colored fonts.

Figure 3.

 

Immunostaining for specific markers in primary cultures and corneal sections. Actin is stained with Phalloidin (red). Positive immunofluorescence staining of prostaglandin D2 synthase (Ptgds), ephrin B2 (Efnb2), and cadherin 5 (Cdh5) in keratocytes is shown in red; lymphocyte antigen 6 complex locus A (Ly6a) in fibroblasts (green); and annexin A8 (Anxa8) in myofibroblasts (red) plated on coverslips and in paraffin-embedded sections of adult mouse cornea. The epithelium (ep), stroma (st), and endothelium (en) of the cornea are labeled. Nuclear (Hoechst) staining is blue. Bar, 50 μm.

Figure 4.

 

Genes differentially expressed in primary cultures of corneal keratocytes, fibroblasts, and myofibroblasts. ( ) Genes previously known for their expression in macrophages.

Table 2.

 

Genes Differentially Expressed in Response to TGF-β

Symbol	Gene	Accession No.	Expression Index of Genes
						Adult	P10	Keratocyte	Fibroblast	Myofibroblast
Anxa8	Annexin A8	AJ002390	2317	2365	55	94	443
Ape	Apolipoprotein E	D00466	835	1455	3238	329	106
Ass1	Argininosuccinate synthetase 1	M31690	172	147	318	451	1107
Crabp2	Cellular retinoic acid binding protein II	M35523	28	53	118	39	380
Ccl2	Chemokine (C-C motif) ligand 2	M19681	1	1	843	613	97
Ccl7	Chemokine (C-C motif) ligand 7	X70058	1	2	290	196	31
Ccl9	Chemokine (C-C motif) ligand 9	U49513	11	7	645	201	32
C4	Complement component 4 (within H-2S)	X06454	119	162	944	263	123
Ctgf	Connective tissue growth factor	M70642	56	216	3730	2292	4304
Csrp2	Cysteine-rich protein 2	D88792	54	68	198	173	583
Cyp2f2	Cytochrome P450, family 2, subfamily f, polypeptide 2	M77497	267	179	440	744	59
Fgfr2	Fibroblast growth factor receptor 2	M23362	483	814	317	269	134
Fbln2	Fibulin 2	X75285	114	336	1157	1184	2671
Gas1	Growth arrest specific 1	X65128	612	867	380	466	126
H2-T23	Histocompatibility 2, T region locus 23	Y00629	398	406	245	208	117
Il6st	Interleukin 6 signal transducer	X62646	22	91	457	707	253
Ltbp1	Latent transforming growth factor beta binding protein 1	AF022889	26	116	302	250	112
Mmp12	Matrix metalloproteinase 12	M82831	12	27	1079	437	41
Mmp3	Matrix metalloproteinase 3	X66402	103	41	2611	1427	229
Mgst1	Microsomal glutathione S-transferase 1	AW124337	86	261	568	521	232
Pdgfra	Platelet derived growth factor receptor, alpha	M57683	89	135	1181	1399	362
Col11a1	Procollagen, type XI, alpha 1	D38162	26	243	262	279	671
Ank	Progressive ankylosis	AW049351	456	402	319	430	1140
Ptgs2	Prostaglandin-endoperoxide synthase 2	M88242	2	17	419	628	1328
	RIKEN cDNA 0610037N01 gene	AW045637	40	306	201	270	455
	RIKEN cDNA 1110003O22 gene	AI847054	306	609	1337	1603	260
	RIKEN cDNA 1190017B18 gene	AI848851	92	69	83	64	387
	RIKEN cDNA 5830484A20 gene	AA717815	340	829	704	891	1699
Sat1	Spermidine/spermine N1-acetyl transferase	L10244	1097	691	692	819	1777
Taf1a	Box binding protein (Tbp)-associated factor, RNA polymerase	Y09972	45	63	119	225	336
Tfpi2	Tissue factor pathway inhibitor 2	D50586	5	11	53	29	203
Wisp1	WNT1 inducible signaling pathway protein 1	AF100777	44	65	312	289	1279
Zfp36l1	Zinc finger protein 36, C3H type-like 1	M58566	669	699	417	334	195

Figure 5.

 

Model of multistep corneal wound repair process. Red: upregulated genes; green: downregulated genes. Specific genes expressed in culture shown on the right of the yellow arrow and their functional implications in the cornea with respect to injury and healing hypothesized on the right.

Figure 6.

 

Immunofluorescence showing positive staining (red) of lumican (a corneal ECM proteoglycan) and negative staining for F4/80 (a marker for bone-marrow–derived macrophages) in primary mouse keratocyte culture. Nuclear (Hoechst) staining is shown in blue. Bar, 50 μm.

The authors thank Aravinda Chakravarti, Stephen Desiderio, Giovanni Parmigiani (The Johns Hopkins University, Baltimore, MD) and John Hassell (University of South Florida) for an insightful critique of the manuscript.