Location of corneal epithelial stem cellsArising from: Majo, F., Rochat, A., Nicolas, M., Jaoude, G. A. & Barrandon, Y. Nature 456, 250–254 (2008).

The longstanding concept that corneal epithelial stem cells residemainly in the limbus is supported by the absence of major cornealepithelial differentiation markers, that is, K3 and K12 keratins, inlimbal basal cells (these markers are expressed, however, in cornealbasal cells, thus distinguishing the mode of keratin expression incorneal epithelium from that of all other stratified epithelia), thecentripetal migration of corneal epithelial cells, the exclusive locationof slow-cycling cells in the limbal basal layer, the superior in vitroproliferative potential of limbal epithelial cells, and the transplantedlimbal cells’ ability to reconstitute corneal epithelium in vivo(reviewed in refs 1–4). Moreover, previous data indicate that cornealand conjunctival epithelia represent two separate cell lineages(reviewed in refs 1–4). Majo et al.5 suggested, however, that cornealand conjunctival epithelia are equipotent, and that identical oligo-potent stem cells are present throughout the corneal, limbal andconjunctival epithelia. We point out here that these suggestions areinconsistent with many known growth, differentiation and cellmigration properties of the anterior ocular epithelia.

Majo et al. suggested that corneal and conjunctival stem cells areequipotent because corneal epithelial cells could form goblet cells,and because cultured (thus somewhat ‘de-differentiated’) pig cornealand conjunctival cells shared a similar phenotype5. They may haveoverlooked, however, reports showing that cultured rabbit corneal/limbal epithelial cells, but not conjunctival cells, expressed K3/K12keratins6–8; conversely, conjunctival epithelial cells, but not cornealcells, formed goblet cells when transplanted into athymic mice8,9.Similar phenotypic specificity was preserved in cultured humanlimbal/corneal and conjunctival cells10. Moreover, human and rabbitstudies showed that limbal epithelial cells, but not conjunctival cells,could restore a true corneal epithelium (reviewed in ref. 4). Thesedata have established that limbal/corneal and conjunctival epitheliaare not equipotent and that they represent two distinct cell lineagesgoverned by their own stem cells (reviewed in refs 4, 9 and 10).

Majo et al. suggested that corneal epithelium contained stem cellsbecause corneal epithelium gave rise to large colonies, serially trans-planted mouse central corneal epithelium could regenerate, andtransplanted mouse limbal cells did not migrate centripetally5.Although their data showed that some pig corneal cells have signifi-cant proliferative potential, this property is not unique to stem cells:some transit amplifying cells such as hair matrix are known to be ableto divide numerous times. Hence, a more meaningful test is to com-pare the growth potential of corneal and limbal cells by seriallypassaging them under identical culture conditions. Such studies haveestablished that rabbit and human limbal cells have a much greaterproliferative capacity than corneal cells7,10. Moreover, Majo et al.’sdata (see figure 3b in ref. 5) showed that although corneal cells ofrabbit, pig and sheep grew well in primary culture, those of human10

and calf did not. Such a major species variation argues against theidea that corneal epithelium contains stem cells (which, if they exist,cannot be slow-cycling given that they are undetectable as label-retaining cells2). Regarding the ability of corneal epithelium to self-sustain, Huang and Tseng showed that, after limbal removal, rabbitcentral corneal epithelium can remain apparently intact for a longtime until it is wounded, indicating that central corneal cells have asignificant maintenance potential until it is perturbed11. Finally, Majoet al.’s negative finding that limbal cells do not migrate centripetallycontradicts many reports establishing that, in intact human12 andmouse eyes13,14 (that have not been surgically manipulated) corneal

epithelial cells undergo centripetal migration. Collectively, the exist-ing data strongly suggest that corneal epithelial stem cells residemainly, if not exclusively, in the limbus.

Finally, Majo et al.’s model hypothesized that both corneal andconjunctival epithelial cells migrated towards the limbus (the ‘tectonicplate confrontation model’). They may have overlooked, however,several reports showing that conjunctival cells do not migrate15, whilecorneal cells undergo centripetal, rather than centrifugal, migra-tion12–14. We conclude that this model, which suggests (1) that cornealand conjunctival epithelia are equipotent, (2) that identical oligopo-tent stem cells are distributed throughout the anterior ocular surfaceepithelium including the central corneal epithelium, and (3) thatcorneal and conjunctival epithelial cells migrate towards the limbus,is incompatible with existing data.Tung-Tien Sun1, Scheffer C. Tseng2 & Robert M. Lavker3

1Departments of Cell Biology, Dermatology, Pharmacology and Urology,

New York University School of Medicine, New York, New York 10016,

USA.

e-mail: sunt01@nyumc.org2Department of Tissue Technology, Ocular Surface Center, Ocular

Surface Research and Education Foundation, Miami, Florida 33173, USA.3Department of Dermatology, Northwestern University, Feinberg School

of Medicine, Chicago, Illinois 60611, USA.

Received 24 January 2009; accepted 19 November 2009.

1. Schermer, A., Galvin, S. & Sun, T. T. Differentiation-related expression of a major 64Kcorneal keratin in vivo and in culture suggests limbal location of corneal epithelialstem cells. J. Cell Biol. 103, 49–62 (1986).

2. Cotsarelis, G., Cheng, S. Z., Dong, G., Sun, T. T. & Lavker, R. M. Existence of slow-cycling limbal epithelial basal cells that can be preferentially stimulated to proliferate:implications on epithelial stem cells. Cell 57, 201–209 (1989).

3. Kenyon, K. R. & Tseng, S. C. Limbal autograft transplantation for ocular surfacedisorders. Ophthalmology 96 (5), 709–722; discussion 722–723 (1989).

4. Lavker, R. M., Tseng, S. C. & Sun, T. T. Corneal epithelial stem cells at the limbus:looking at some old problems from a new angle. Exp. Eye Res. 78, 433–446 (2004).

5. Majo, F., Rochat, A., Nicolas, M., Jaoude, G. A. & Barrandon, Y. Oligopotent stem cellsare distributed throughout the mammalian ocular surface. Nature 456, 250–254(2008).

6. Sun, T.-T. & Green, H. Cultured epithelial cells of cornea, conjunctiva and skin:absence of marked intrinsic divergence of their differentiated states. Nature 269,489–493 (1977).

7. Wei, Z. G., Wu, R. L., Lavker, R. M. & Sun, T. T. In vitro growth and differentiation ofrabbit bulbar, fornix, and palpebral conjunctival epithelia. Implications on conjunctivalepithelial transdifferentiation and stem cells. Invest. Ophthalmol. Vis. Sci. 34,1814–1828 (1993).

8. Doran, T. I., Vidrich, A. & Sun, T.-T. Intrinsic and extrinsic regulation of thedifferentiation of skin, corneal and esophageal epithelial cells. Cell 22, 17–25 (1980).

9. Wei, Z. G., Sun, T. T. & Lavker, R. M. Rabbit conjunctival and corneal epithelial cellsbelong to two separate lineages. Invest. Ophthalmol. Vis. Sci. 37, 523–533 (1996).

10. Pellegrini, G. et al. Location and clonal analysis of stem cells and their differentiatedprogeny in the human ocular surface. J. Cell Biol. 145, 769–782 (1999).

11. Huang, A. J. & Tseng, S. C. Corneal epithelial wound healing in the absence of limbalepithelium. Invest. Ophthalmol. Vis. Sci. 32, 96–105 (1991).

12. Auran, J. D. et al. Scanning slit confocal microscopic observation of cell morphologyand movement within the normal human anterior cornea. Ophthalmology 102, 33–41(1995).

13. Collinson, J. M. et al. Clonal analysis of patterns of growth, stem cell activity, and cellmovement during the development and maintenance of the murine cornealepithelium. Dev. Dyn. 224, 432–440 (2002).

14. Nagasaki, T. & Zhao, J. Centripetal movement of corneal epithelial cells in the normaladult mouse. Invest. Ophthalmol. Vis. Sci. 44, 558–566 (2003).

15. Nagasaki, T. & Zhao, J. Uniform distribution of epithelial stem cells in the bulbarconjunctiva. Invest. Ophthalmol. Vis. Sci. 46, 126–132 (2005).

Competing financial interests: declared none.

doi:10.1038/nature08805

NATURE | Vol 463 | 25 February 2010 BRIEF COMMUNICATIONS ARISING

E10Macmillan Publishers Limited. All rights reserved©2010

Majo et al. replyReplying to: T.-T. Sun, S. C. Tseng & R. M. Lavker Nature 463, doi:10.1038/nature08805 (2010)

Our claim is not that there are no stem cells in the limbus, but thatthere is more to corneal renewal than the limbus and that the double-dome-shaped structure of the cornea and physical constraints have acrucial impact on cell dynamics1.

Sun and colleagues2 imply that in our paper3 we misused the term‘holoclones’ that we defined as stem cells4; the central cornea of the pigcontains numerous true holoclones, meaning that the cornea of thepig has extensive growth potential and the ability to be serial passagedin vitro. We agree that there are species differences among mammals;nonetheless, all corneas that we have investigated, including calf andhuman, contain colony-forming cells. Fifty cell doublings in pigcornea is not trivial and contradicts the model proposed by Sun andcolleagues5; we quote their abstract ‘‘we demonstrate the existence of ahierarchy of TA cells; those of peripheral cornea undergo at least tworounds of DNA synthesis before they become post-mitotic, whereasthose of central cornea are capable of only one round of division’’. Italso does not agree with Huang and Tseng’s experiment6 showing‘‘that, after limbal removal, rabbit central corneal epithelium canremain apparently intact for a long time until it is wounded, indicatingthat central cornea cells have a significant maintenance potential’’.

Our results show that corneal cells can form goblet cells when theymigrate onto a conjunctival environment (in mouse) or generate truegoblet cell colonies when cloned (in pig). Corneal differentiation isfound in human conjunctiva7, conjunctival cells may be successfullytransplanted in the human to replace cornea8, and there are reportsof cornea remaining transparent for years in limbal deficiency9.Furthermore, corneal cells10, like conjunctival cells (our unpublishedresults), can form hairy skin when exposed to an inductive skinmicroenvironment, indicating a greater plasticity than anticipatedand that stem cell fate strongly depends on stromal signals.

We are not aware of any paper that clearly demonstrates stem cellmigration from the limbus. Buck11 in his landmark paper has notdemonstrated basal cell migration; we quote his abstract: ‘‘the mediandistance migrated was about 17mm per day. This figure represents thedistance through which superficial and wing cells had migrated; thedistance migrated by basal cells was not determined’’. Nagazaki andZhao12 have presented evidence of movement in the cornea but notthat the migrating cells actually originated from the limbus (‘from’ isnot the same as ‘near’). An overcrowding of the corneal epithelium, asource of tension and sliding as previously emphasized by Sun andcolleagues13, or sequential activation of the b-actin promoter caneasily explain these observations. Similarly, the spiral stripe organiza-tion mixing clockwise and counterclockwise clones14 is highly remini-scent of centrifugal growth originating from a small number of stemcells originally located in central cornea. This biological model occurswidely in nature, for instance in the growth of a daisy, as the easiest and

most efficient way to fill space, a notion supported by mathematicalmodels15 and a clothoid growth model (Euler spiral).Francois Majo1, Ariane Rochat2,3, Michael Nicolas 1,

Georges Abou Jaoude4 & Yann Barrandon2,3

1Hopital Ophtalmique Jules Gonin, Avenue de France 15, 1004 Lausanne

CH, Switzerland.2Department of Experimental Surgery Lausanne University Hospital

(CHUV), 1011 Lausanne CH, Switzerland.3Laboratory of Stem Cell Dynamics, Ecole Polytechnique Federale de

Lausanne (EPFL), 1015 Lausanne CH, Switzerland.e-mail: yann.barrandon@epfl.ch4Laboratory of Informatics and Visualization, Ecole PolytechniqueFederale de Lausanne (EPFL), 1015 Lausanne CH, Switzerland.

1. Dupps, W. J. Jr & Wilson, S. E. Biomechanics and wound healing in the cornea. Exp. EyeRes. 83, 709–720 (2006).

2. Sun, T.-T., Tseng, S. C. & Lavker, R. M. Location of corneal epithelial stem cells. Nature463, doi:10.1038/nature08805 (2010).

3. Majo, F., Rochat, A., Nicolas, M., Jaoude, G. A. & Barrandon, Y. Oligopotent stem cellsare distributed throughout the mammalian ocular surface. Nature 456, 250–254(2008).

4. Barrandon, Y. & Green, H. Three clonal types of keratinocyte with different capacitiesfor multiplication. Proc. Natl Acad. Sci. USA 84, 2302–2306 (1987).

5. Lehrer, M. S., Sun, T. T. & Lavker, R. M. Strategies of epithelial repair: modulation ofstem cell and transit amplifying cell proliferation. J. Cell Sci. 111, 2867–2875 (1998).

6. Huang, A. J. & Tseng, S. C. Corneal epithelial wound healing in the absence of limbalepithelium. Invest. Ophthalmol. Vis. Sci. 32, 96–105 (1991).

7. Kawasaki, S. et al. Clusters of corneal epithelial cells reside ectopically in humanconjunctival epithelium. Invest. Ophthalmol. Vis. Sci. 47, 1359–1367 (2006).

8. Di Girolamo, N. et al. A contact lens-based technique for expansion andtransplantation of autologous epithelial progenitors for ocular surface reconstruction.Transplantation 87, 1571–1578 (2009).

9. Dua, H. S., Miri, A., Alomar, T., Yeung, A. M. & Said, D. G. The role of limbal stem cellsin corneal epithelial maintenance: testing the dogma. Ophthalmology 116, 856–863(2009).

10. Ferraris, C., Chevalier, G., Favier, B., Jahoda, C. A. & Dhouailly, D. Adult cornealepithelium basal cells possess the capacity to activate epidermal, pilosebaceous andsweat gland genetic programs in response to embryonic dermal stimuli. Development127, 5487–5495 (2000).

11. Buck, R. C. Measurement of centripetal migration of normal corneal epithelial cells inthe mouse. Invest. Ophthalmol. Vis. Sci. 26, 1296–1299 (1985).

12. Nagasaki, T. & Zhao, J. Centripetal movement of corneal epithelial cells in the normaladult mouse. Invest. Ophthalmol. Vis. Sci. 44, 558–566 (2003).

13. Lavker, R. M. et al. Relative proliferative rates of limbal and corneal epithelia.Implications of corneal epithelial migration, circadian rhythm, and suprabasallylocated DNA-synthesizing keratinocytes. Invest. Ophthalmol. Vis. Sci. 32, 1864–1875(1991).

14. Collinson, J. M. et al. Clonal analysis of patterns of growth, stem cell activity, and cellmovement during the development and maintenance of the murine cornealepithelium. Dev. Dyn. 224, 432–440 (2002).

15. Stewart, I. Mathematical recreations: Daisy, Daisy, give me your answer, do. Sci. Am.272, 96–99 (1995).

Competing financial interests: declared none.

doi:10.1038/nature08806

NATURE | Vol 463 | 25 February 2010 BRIEF COMMUNICATIONS ARISING

E11Macmillan Publishers Limited. All rights reserved©2010

LETTERS

Oligopotent stem cells are distributed throughout themammalian ocular surfaceFrancois Majo1,3{, Ariane Rochat1,3, Michael Nicolas1,3, Georges Abou Jaoude2 & Yann Barrandon1,3

The integrity of the cornea, the most anterior part of the eye, isindispensable for vision. Forty-five million individuals worldwideare bilaterally blind and another 135 million have severelyimpaired vision in both eyes because of loss of corneal transpar-ency1; treatments range from local medications to corneal trans-plants, and more recently to stem cell therapy2. The cornealepithelium is a squamous epithelium that is constantly renewing,with a vertical turnover of 7 to 14 days in many mammals3.Identification of slow cycling cells (label-retaining cells) in thelimbus of the mouse has led to the notion that the limbus is theniche for the stem cells responsible for the long-term renewal ofthe cornea4; hence, the corneal epithelium is supposedly renewedby cells generated at and migrating from the limbus, in markedopposition to other squamous epithelia in which each residentstem cell has in charge a limited area of epithelium5,6. Here weshow that the corneal epithelium of the mouse can be serially

transplanted, is self-maintained and contains oligopotent stemcells with the capacity to generate goblet cells if provided with aconjunctival environment. Furthermore, the entire ocular surfaceof the pig, including the cornea, contains oligopotent stem cells(holoclones)7,8 with the capacity to generate individual colonies ofcorneal and conjunctival cells. Therefore, the limbus is not theonly niche for corneal stem cells and corneal renewal is not dif-ferent from other squamous epithelia. We propose a model thatunifies our observations with the literature and explains why thelimbal region is enriched in stem cells.

To explore the function of limbal stem cells (Fig. 1a, b), portions(1.5 mm 3 0.3 mm) of the limbus of athymic mice were excised andreplaced with limbus from b-gal-ROSA26 mice (Fig. 1c). In mostcases, epithelial continuity was rapidly restored (Supplementary Fig.1). Transplanted eyes were usually collected 4 months later and onceafter 11 months. Notably, b-galactosidase (b-gal)-labelled cells never

1Laboratory of Stem Cell Dynamics, 2Laboratory of Informatics and Visualization, Ecole Polytechnique Federale de Lausanne (EPFL), 1015 Lausanne CH, Switzerland. 3Department ofExperimental Surgery Lausanne University Hospital (CHUV), 1011 Lausanne CH, Switzerland. {Present address: Hopital Ophtalmique Jules Gonin, Avenue de France 15, 1004Lausanne CH, Switzerland.

a b c

d

2.6 mm

Fornix Conjunctiva Limbus Cornea

Figure 1 | The limbus does not contribute to steady-state corneal renewal,but to corneal repair. a, Anatomy of a mouse eye. The limbus is the zone atthe junction of the cornea and the conjunctiva. Arrows denote mucin-producing cells (goblet cells). b, The limbal paradigm4 is shown. The corneadoes not contain stem cells and is renewed by transient amplifying cells(purple) generated by limbal stem cells (red) dividing asymmetrically.Transient amplifying cells then migrate centripetally to the central corneawhile renewing the corneal epithelium en route. The distance separating thelimbus from the corneal midline is 2.6 mm. c, The limbus does notcontribute to the steady-state renewal of the corneal epithelium. A limbalfragment of an athymic mouse was excised and replaced by an equivalentlimbal fragment obtained from a b-gal-ROSA26 mouse (left panel). Thetransplant was stitched with its corneal side facing the cornea of the recipientmouse (normal orientation). Cells did not migrate out of the transplant

3 months after the transplantation (b-gal staining; middle panel). Similarresults were obtained when the limbal transplant was stitched with itsconjunctival side facing the cornea of the recipient mouse (reverseorientation) or using SCID mice as recipients. Note the clear boundarybetween transplanted limbal (stained blue) and corneal (unstained) cells(right panel); white arrows indicate nylon stitches. See also SupplementaryFig. 1a. d, The limbus can contribute to corneal repair. An extensive cornealwound challenged the limbal transplant 4 months after transplantation (leftpanel). Labelled cells migrated out of the transplant and restored the cornealintegrity in 7 days (b-gal staining) (middle panel). The migrating cells didnot cross the corneal midline (right panel). These experiments indicate thatthe limbus contributes little to steady renewal of the cornea, but becomesinstrumental when the cornea is extensively wounded. Scale bars, all 50 mm.

Vol 456 | 13 November 2008 | doi:10.1038/nature07406

250

©2008 Macmillan Publishers Limited. All rights reserved

migrated out of the grafts onto the cornea (Fig. 1c), whether the graftswere transplanted with their corneal (normal orientation) or con-junctival (reverse orientation) side facing the cornea of the recipientmouse (n 5 39 and 28, respectively). Similar results were alsoobtained using SCID mice as recipients, thus excluding the remotepossibility that the migration of the transplanted limbal cells wasadversely affected in athymic mice. Next we chemically or physicallywounded the cornea of recipient mice4 in which limbal grafts hadbeen in place for up to 6 months. As expected, labelled cells migratedout of the grafts along with unlabelled limbal cells of the recipientmice resulting in a mosaicism of the healed corneal epithelium(Fig. 1d), whether the graft was in a normal (n 5 39) or a reverse(n 5 32) orientation (data not shown). The healed cornea was usuallytransparent but corneal opacity was observed when the stroma wasaltered (n 5 13). Cells migrating from the limbus usually covered thedistance to the central part of cornea (2.6 mm) in a week and had atypical corneal phenotype (Fig. 1d). Collectively, these experimentsindicated that the absence of a contribution of limbal stem cells tocorneal renewal did not result from an incapacity of the transplantedcells to migrate, and confirmed that limbal stem cells could efficientlyheal an extreme corneal wound2,4.

In humans, limbal deficiency is a condition thought to result from acomplete loss of limbal stem cells9. As a consequence, the cornealepithelium is replaced by a conjunctival-like epithelium containingmucus-producing goblet cells, and vision is impaired. To mimic lim-bal deficiency in the mouse, we cauterized the entire circumference ofthe limbus of wild-type mice (12 eyes). The burn entirely destroyed thelimbal epithelium and the underlying matrix, reaching the ciliarybodies in many places, making it unlikely that any limbal stem cellshad escaped destruction (Supplementary Fig. 2a). The cornea ofburned eyes remained transparent for the duration of the experiment(a minimum of 4 months), and neither corneal ulcers nor stromalblood vessels were observed (Supplementary Fig. 2b), strongly sug-gesting self-maintenance of the corneal epithelium.

To evaluate further the ability of the cornea to self-renew, full-thickness grafts including the corneal epithelium and its stroma wereobtained from the central part of the cornea of b-gal-ROSA26 donormice older than 6 months (up to 1 year), using a 1.5 mm diameterophthalmic trephine. The grafts were then transplanted at the limbal

region of athymic mice as described earlier. Most of the grafts(n 5 11) were rapidly incorporated into the limbal region of therecipient mice, and remained viable for months with no signs ofnecrosis (Fig. 2a). We next challenged the grafts at the time of trans-plantation or several weeks later by extensive corneal wounds(Fig. 2b). In most cases (47 out of 51) cells quickly migrated out ofthe grafts onto the denuded corneal stroma and corneal integrity wasrestored within 7 days. Notably, healed corneas were transparent andthe regenerated epithelium remained b-gal-labelled for the entireduration of the experiments, up to 9 months (Fig. 2b andSupplementary Fig. 3). To exclude the remote possibility that thecorneal cells were reprogrammed to a limbal phenotype because theywere transplanted in a limbal/conjunctival microenvironment, cor-neal transplants were implanted into the central cornea of recipientathymic mice; transplants (n 5 9) were again able to restore theintegrity of a wounded cornea (Fig. 2c). Transplanted cells oftencovered the entire distance (5.2 mm) to the opposite corneal sidebefore limbal cells even had time to move out (Fig. 2b and

a

b

c

d

e

Figure 2 | The cornea of the mouse has extensive self-renewal capacity andcontains oligopotent stem cells. Full-thickness biopsies were taken from thecentral cornea of 6-month-old ROSA26 (a–c) and ROSA26-GFPU (d, e) miceand transplanted at the limbus of a recipient athymic mouse as described inFig. 1. a, The corneal transplant (asterisk, right panel) remained healthy for4 months (b-gal staining); black arrow denotes that the transplantedepithelium is in continuity with the adjacent corneal epithelium. See alsoSupplementary Fig. 1b. b, The entire corneal epithelium of the recipientmouse was removed to challenge the corneal transplant 4 months aftertransplantation. After 7 days, cells that had migrated out of the transplanthad almost entirely restored the corneal epithelium, whereas the limbal cellsof the recipient contributed little. The b-gal-positive cells crossed the cornealmidline almost reaching the opposite side (compare with Fig. 1d). See alsoSupplementary Fig. 3. c, The corneal transplant was challenged by a cornealwound. Same results as in b. Arrowheads (middle panel) denote the sharpboundary with the limbus of the recipient mouse. The white arrows point toa ligature knot (b, c, right panels), the black arrow indicates the direction ofmigration and the asterisk indicates the transplant. These experimentsindicate that reprogramming of corneal cells by the limbal environment isunlikely and that the cornea of the mouse contains epithelial cells withextensive growth capacity. d, e, The corneal epithelium can be seriallytransplanted. The corneal transplant was challenged by wounding of thecornea and of the conjunctiva. Two months later, a biopsy was obtainedfrom the healed cornea and transplanted onto a second athymic mouse.Corneal and conjunctival wounds challenged the transplant for a secondtime (d). The eye was collected 42 days later (b-gal staining). Labelled cellsrepaired the wounded cornea and conjunctiva (black arrows in the middlepanel of e indicate goblet cells) (e). These experiments indicate that thecentral cornea of the mouse contains bona fide oligopotent epithelial stemcells. Scale bars, all 100mm, except left panel e, 200mm.

NATURE | Vol 456 | 13 November 2008 LETTERS

251

©2008 Macmillan Publishers Limited. All rights reserved

Supplementary Fig. 3), a further indication of the robust migratorycapability of corneal cells10. Markedly, the boundary between thelabelled cornea and the adjacent unlabelled limbus was sharplydemarcated for the duration of the experiments (up to 9 months;Fig. 2c and Supplementary Fig. 3), another sign that limbal stem cellscontribute little to corneal renewal.

We then performed serial transplantations of b-gal and GFP(green fluorescent protein) double-labelled central corneas (Fig. 2d)

that were monitored by ultraviolet illumination on living mice andby b-gal staining on enucleated eyes. Transplanted corneal cells wereagain able to restore the corneal integrity for a minimum of 4 months(sum of serial transfers). Furthermore, when the transplants werechallenged by either a conjunctival or a fornical wound, labelledcorneal cells efficiently migrated towards the fornix generatingb-gal-positive goblets cells, a landmark of the conjunctiva (Fig. 2d).Hence, corneal epithelial cells could adopt either a conjunctival or a

Hum

anR

atR

abb

itS

heep

Cal

fP

ig

Cj L Cp Ci Cc

Conjunctivaholoclone

Corneaholoclone

RT–PCR Pig

Cornea Conjunctiva

Bmi1Zfp145Trp63

K12Muc5ac

K4b-actin

K19

Nested PCR

Cl.1

5

Cl.1

6

Cl.2

1

Ctr

l

Cl.0

3

Cl.0

4

Cl.0

5

Ctr

l

Forn

ix

Rat

eye

3T3

H2O

Normalized expression level

r = 0.978

Con

junc

tiva

Cornea

14

12

10

8

6

4

2

00 2 4 6 8 10 12 14

ab

c d

p63 p63

K3 K3

PAS PAS

Figure 3 | The cornea of mammals contains clonogenic keratinocytes.a, The ocular surface was dissected and cut into several fragments: theconjunctiva (Cj), the limbus (L), the peripheral cornea (Cp), theintermediate cornea (Ci) and the central cornea (Cc). Cornea shown is frompig and each fragment is on average 4.5 mm2. Each fragment was cultivatedas described8,28; see also Supplementary Information. Several clonogenickeratinocytes were distributed through the entire ocular surface includingthe central cornea, although with species differences (for example, betweenpig and human). b, The cornea and conjunctiva of the pig containholoclones7 that formed colonies containing p631 and K31 corneal cells andlarge PAS1 cells (arrows, bottom panels), suggestive of goblet cells. Scale

bars, 50 mm c, The corneal and conjunctival holoclones (cl.) expressed p63, aputative marker of corneal stem cells29, Bmi1 and Zfp145 (also known asZbtb16), implicated in stem cell renewal12, K3 and K12 (also known as Krt3and Krt12), markers of corneal differentiation30, and K19, K4 (also known asKrt19, Krt4) and Muc5ac, all markers of conjunctival differentiation21. Ctrl,control. d, DNA microarray analysis comparing three corneal clones(AP6cl15, AP6cl16 and AP1cl21) to three conjunctival clones (AP6cl3,AP1cl4 and AP1cl5). Only two genes out of 20,201 were differentiallyexpressed. These experiments demonstrate that stem cells of the conjunctivaand cornea of the pig are oligopotent and share the same ocular signature.

LETTERS NATURE | Vol 456 | 13 November 2008

252

©2008 Macmillan Publishers Limited. All rights reserved

corneal phenotype depending on which environment (conjunctivalor a corneal) they were exposed to (Fig. 2e).

Together, our results challenge the prevailing opinion that thelimbus is the sole niche for corneal stem cells, because long-term selfrenewal and the capacity to sustain serial transplantation are pro-perties of stem cells and not of transit amplifying cells—the growthcapability of which is limited to a few divisions by definition11.Furthermore, the capacity of corneal stem cells to generate gobletcells, the hallmark of conjunctiva, strongly suggests oligopotence.Our results further highlight the importance of function to dem-onstrate ‘stemness’ in agreement with recent observations in theepidermis and hair follicles6,12, the bone marrow13 and the gut14.Nonetheless, the limbal region undoubtedly contains stem cells thatare efficient in restoring the corneal surface in the extreme situationof a massive corneal injury or for cell therapy as demonstrated by theimpressive clinical results obtained with the transplantation of limbalholoclones2,15.

The inefficacy of the cultivation of mouse keratinocytes and theirinclination to undergo spontaneous immortalization seriously lim-ited thorough single cell analysis. We therefore isolated keratinocytesfrom the ocular surface of several other mammals and cultivatedthem in conditions strictly identical to human cell therapy15,16.Despite the species differences (for example, between pigs andhumans) the entire ocular surface of these mammals contains manykeratinocyte colony-forming cells that can be serially passaged(Fig. 3a). The pig was selected for clonal analyses because of therobust growth of its ocular surface cells. Eyes, the integrity of whichwas controlled under a dissecting microscope or with a slit lamp,were obtained from two pigs (AP1 and AP6) soon after being killed.Several colonies (487 per mm2 in the conjunctiva to 1,025 per mm2 inthe cornea) were obtained from 3–6 mm2 fragments of bulbar con-junctiva and of the central cornea, located at least 1.5 and 6 mm fromthe limbus, respectively. A total of 277 single cells were then isolatedfrom the central cornea7,17 and 53 clones were obtained, of which 26were passaged; 13 clones were holoclones, 8 were meroclones and 5were paraclones. One holoclone and three meroclones were thenpassaged once a week for more than 16 weeks (over 150 doublings).In a separate experiment, 125 single conjunctival cells were isolatedand 29 clones were obtained, of which 9 were holoclones. Both cor-neal and conjunctival holoclones maintained a diploid karyotype(2n 5 38 chromosomes) in early passages, whereas various degreesof aneuploidy was observed in late passages. Notably, corneal andconjunctival holoclones had a similar, if not identical, phenotype inculture (Fig. 3b). Both initiated colonies containing p631 cells, ker-atin 31 cells and PAS1 (periodic acid schiff) cells, reminiscent ofgoblet cells (Fig. 3b), and both had a similar ocular signature bypolymerase chain reaction analysis with reverse transcription (RT–PCR; Fig. 3c) and DNA microarray analysis (Fig. 3d). The cornealepithelium of the pig thus contained bona fide oligopotent stem cellsthat expressed lineage markers as other tissue stem cells, for example,neural stem cells expressing markers of astrocyte differentiation 18.Our finding that corneal and conjunctival stem cells have equalpotency explains the successful use of both to treat corneal defects19–21

and further supports the notion that stromal signals are determinantfor epithelial stem cell fate22. Stromal dysfunction may then lead toaberrant epithelial differentiation, as shown by the formation of der-moid cysts on the ocular surface of epidermis in the wounded corneaof conditional Notch1 null mice23, and of a corneal epithelium in theconjunctiva of humans24.

To unify our observations with previous results4, we propose amodel that integrates cell dynamics and various constraints to whichthe corneal epithelium is submitted. In this model (Fig. 4), the limbus isa zone of equilibrium in which the expanding conjunctival and cornealepithelia are confronted in a mechanism reminiscent of tectonic plates.Rupture of the equilibrium—for example, because of an increasedrigidity of the corneal stromal or because of an extensive cornealwound—results in migration of limbal stem cells onto the cornea4,25,26.

Most importantly, our model explains why resting stem cells accu-mulate at the limbus and why the limbus is a principal source of stemcells for cell therapy in humans2,27. It also explains why the limbus is aprivileged site of cancer, as are other transitional zones—for example,the junction of the vaginal epithelium with the endocervical epithe-lium, the epidermis with the oral epithelium at the lips, and the oeso-phagus with the gastric epithelium.

METHODS SUMMARY

Detailed methods and surgical procedures are described in the Methods.

Transplantations were authorized by the veterinarian authorities of the Canton

de Vaud (authorization numbers 1524 and 1686) and performed by an ophthal-

mic surgeon (F.M.). ROSA26 or GFPU (modified GFP) donor mice were at least

4 months old, as were recipient mice (athymic or SCID). Keratinocytes were

cultivated onto a feeder layer of lethally irradiated 3T3-J2 cells28 and clonal ana-

lyses were performed as described7,17. Histological and immunological analyses

were performed using standard procedures. Antibodies were mouse monoclonal

anti-rabbit cytokeratin 3 (clone AE5, Chemicon), mouse monoclonal anti-

human p63 (clone 4A4, Dako), mouse monoclonal anti-human cytokeratin 19

(clone RCK 108, Dako) and anti mouse monoclonal anti-mucin 5AC (clone

45M1, Neomarkers). Pig primers (listed in Supplementary Information) were

selected using the primer 3-input software. PCRs were performed according to

standard procedures and PCR products were sequenced to confirm the specificity

of the reaction. DNA microarrays analyses were performed by the Lausanne DNA

Array Facility (DAFL) using standard protocols and the Affymetrix GeneChip

Porcine Genome Array. Raw and normalized microarray data are accessible

through the NCBI Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/

geo) public database (series record GSE12604).

Full Methods and any associated references are available in the online version ofthe paper at www.nature.com/nature.

Received 9 May; accepted 12 September 2008.Published online 1 October 2008.

1. Whitcher, J. P., Srinivasan, M. & Upadhyay, M. P. Corneal blindness: a globalperspective. Bull. World Health Organ. 79, 214–221 (2001).

2. Pellegrini, G. et al. Long-term restoration of damaged corneal surfaces withautologous cultivated corneal epithelium. Lancet 349, 990–993 (1997).

a b

c

Conjunctiva Limbal region Cornea

Model of efforts in the limbal region

Figure 4 | A unifying model of ocular surface renewal. a, Epithelial stemcells of equal potency are distributed throughout the entire ocular surface.b, Suprabasal cells at the limbal region of the pig express K31 corneal-typedifferentiation whereas basal cells are negative. Differentiated corneal cellsappear to slide over limbal basal cells, as suggested by their orientation. Scalebar, 100mm. c, Schematic representation of the same region in b. Thecombination of blinking, variable stroma elasticity and pressure in the eyeballand within the corneal dome-shape (dark arrows) results in opposite forces(white arrows) in the conjunctival and corneal epithelia. Hence, the cornealand conjunctival epithelia, which are the site of intense cell multiplication, arecontinuously expanding but in opposite directions and confront at thelimbus. Stem cells, sorted by an elutriation-like phenomenon, accumulateboth on the conjunctival and corneal side of the limbus.

NATURE | Vol 456 | 13 November 2008 LETTERS

253

©2008 Macmillan Publishers Limited. All rights reserved

3. Haddad, A. Renewal of the rabbit corneal epithelium as investigated byautoradiography after intravitreal injection of 3H-thymidine. Cornea 19, 378–383(2000).

4. Cotsarelis, G., Cheng, S. Z., Dong, G., Sun, T. T. & Lavker, R. M. Existence of slow-cycling limbal epithelial basal cells that can be preferentially stimulated toproliferate: implications on epithelial stem cells. Cell 57, 201–209 (1989).

5. Jones, P. H., Simons, B. D. & Watt, F. M. Sic transit gloria: farewell to the epidermaltransit amplifying cell? Cell Stem Cell 1, 371–381 (2007).

6. Clayton, E. et al. A single type of progenitor cell maintains normal epidermis.Nature 446, 185–189 (2007).

7. Barrandon, Y. & Green, H. Three clonal types of keratinocyte with differentcapacities for multiplication. Proc. Natl Acad. Sci. USA 84, 2302–2306 (1987).

8. Pellegrini, G. et al. Location and clonal analysis of stem cells and theirdifferentiated progeny in the human ocular surface. J. Cell Biol. 145, 769–782(1999).

9. Puangsricharern, V. & Tseng, S. C. Cytologic evidence of corneal diseases withlimbal stem cell deficiency. Ophthalmology 102, 1476–1485 (1995).

10. Buck, R. C. Cell migration in repair of mouse corneal epithelium. Invest.Ophthalmol. Vis. Sci. 18, 767–784 (1979).

11. Potten, C. S., Schofield, R. & Lajtha, L. G. A comparison of cell replacement in bonemarrow, testis and three regions of surface epithelium. Biochim. Biophys. Acta 560,281–299 (1979).

12. Claudinot, S., Nicolas, M., Oshima, H., Rochat, A. & Barrandon, Y. Long-termrenewal of hair follicles from clonogenic multipotent stem cells. Proc. Natl Acad.Sci. USA 102, 14677–14682 (2005).

13. Kiel, M. J. et al. Haematopoietic stem cells do not asymmetrically segregatechromosomes or retain BrdU. Nature 449, 238–242 (2007).

14. Barker, N. et al. Identification of stem cells in small intestine and colon by markergene Lgr5. Nature 449, 1003–1007 (2007).

15. Rama, P. et al. Autologous fibrin-cultured limbal stem cells permanently restorethe corneal surface of patients with total limbal stem cell deficiency.Transplantation 72, 1478–1485 (2001).

16. Ronfard, V., Rives, J. M., Neveux, Y., Carsin, H. & Barrandon, Y. Long-termregeneration of human epidermis on third degree burns transplanted withautologous cultured epithelium grown on a fibrin matrix. Transplantation 70,1588–1598 (2000).

17. Rochat, A., Kobayashi, K. & Barrandon, Y. Location of stem cells of human hairfollicles by clonal analysis. Cell 76, 1063–1073 (1994).

18. Merkle, F. T., Tramontin, A. D., Garcia-Verdugo, J. M. & Alvarez-Buylla, A. Radialglia give rise to adult neural stem cells in the subventricular zone. Proc. Natl Acad.Sci. USA 101, 17528–17532 (2004).

19. Shapiro, M. S., Friend, J. & Thoft, R. A. Corneal re-epithelialization from theconjunctiva. Invest. Ophthalmol. Vis. Sci. 21, 135–142 (1981).

20. Thoft, R. A. Keratoepithelioplasty. Am. J. Ophthalmol. 97, 1–6 (1984).21. Tanioka, H. et al. Establishment of a cultivated human conjunctival epithelium as

an alternative tissue source for autologous corneal epithelial transplantation.Invest. Ophthalmol. Vis. Sci. 47, 3820–3827 (2006).

22. Ferraris, C., Chevalier, G., Favier, B., Jahoda, C. A. & Dhouailly, D. Adult cornealepithelium basal cells possess the capacity to activate epidermal, pilosebaceousand sweat gland genetic programs in response to embryonic dermal stimuli.Development 127, 5487–5495 (2000).

23. Vauclair, S. et al. Corneal epithelial cell fate is maintained during repair by Notch1signaling via the regulation of vitamin A metabolism. Dev. Cell 13, 242–253(2007).

24. Kawasaki, S. et al. Clusters of corneal epithelial cells reside ectopically in humanconjunctival epithelium. Invest. Ophthalmol. Vis. Sci. 47, 1359–1367 (2006).

25. Di Iorio, E. et al. Isoforms of DNp63 and the migration of ocular limbal cells inhuman corneal regeneration. Proc. Natl Acad. Sci. USA 102, 9523–9528 (2005).

26. Barbaro, V. et al. C/EBPD regulates cell cycle and self-renewal of human limbalstem cells. J. Cell Biol. 177, 1037–1049 (2007).

27. Nakamura, T., Inatomi, T., Sotozono, C., Koizumi, N. & Kinoshita, S. Successfulprimary culture and autologous transplantation of corneal limbal epithelial cellsfrom minimal biopsy for unilateral severe ocular surface disease. Acta Ophthalmol.Scand. 82, 468–471 (2004).

28. Oshima, H., Rochat, A., Kedzia, C., Kobayashi, K. & Barrandon, Y. Morphogenesisand renewal of hair follicles from adult multipotent stem cells. Cell 104, 233–245(2001).

29. Pellegrini, G. et al. p63 identifies keratinocyte stem cells. Proc. Natl Acad. Sci. USA98, 3156–3161 (2001).

30. Schermer, A., Galvin, S. & Sun, T. T. Differentiation-related expression of a major64K corneal keratin in vivo and in culture suggests limbal location of cornealepithelial stem cells. J. Cell Biol. 103, 49–62 (1986).

Supplementary Information is linked to the online version of the paper atwww.nature.com/nature.

Acknowledgements We are grateful to F. Radtke and S. Vauclair for Notch1lox/lox

K14Cre ROSA26R mice, to A. Smith for critical reading of the manuscript, toT. Hoang-Xuan, L. Zografos, F. Munier and S. Kinoshita for continuous support andto L. Schnell, J. Vannod and S. Vermot for excellent technical help. The work wassupported by grants to Y.B. from the Swiss National Science Foundation (grant3100A0-104160), the EPFL, the CHUV and EuroStemCell. The early part of thework was supported by grants to Y.B. from the Institut National de la Sante et de laRecherche Medicale (INSERM), the Association pour la Recherche contre leCancer and the Association Francaise contre les Myopathies. F.M. was supportedby fellowships from the Federation des Aveugles et Handicapes Visuels de France,the INSERM and then the CHUV.

Author Contributions F.M., A.R. and M.N. performed and assisted in the design ofthe experiments and the interpretation of results, A.R. and G.A.J. made the figuresand G.A.J. and Y.B. contributed the new concept. Y.B. designed the experimentsand wrote the paper.

Author Information Reprints and permissions information is available atwww.nature.com/reprints. Correspondence and requests for materials should beaddressed to Y.B. (yann.barrandon@epfl.ch).

LETTERS NATURE | Vol 456 | 13 November 2008

254

©2008 Macmillan Publishers Limited. All rights reserved

METHODSAnimals. Experiments were performed according to the Swiss guidelines and

authorized by the veterinarian authorities of the Canton de Vaud (authorization

numbers 1524 and 1686). Wild-type, OF1 (inbred), athymic (Swiss nu/nu,

inbred) and SCID mice were from Iffa Credo-Charles River. ROSA26 mice31

and GFPU mice32 were from the Jackson laboratory. The donor male or female

mice (ROSA26 or GFPU) were at least 4 months old, as were the recipient mice

(athymic or SCID) when operated. Mice were anesthetized with a mixture of

ketamine (0.104 mg per g of body weight) and xylazine (0.033 mg per g of body

weight) in PBS. Operated mice systematically received 0.07 mg per g body weightof buprenorphine as post-operative analgesia.

Surgical procedures. All surgical procedures were performed under a sterile

environment using a MZ 690 surgical microscope (Leica) or a dissecting bin-

ocular microscope (Zeiss) equipped with fluorescence and enhanced GFP filters.

Instruments used were an ophthalmic cautery, Vannas scissors, a Troutman-

O’Brien needle holder, Bonn and Barraquer forceps, a hockey spatula and a

1.5 mm diameter corneal trephine (all from Moria Surgical). Sutures were

Biosord 10.0 (Alcon Surgical) and 11/0 nylon (Mani Inc) for transplants and

Vicryl 8.0 (Ethicon) for lids. Special care was taken to keep the operated eyes

moisturized during surgery. After they were collected, transplants were kept in

DMEM supplemented with 10% calf serum (Hyclone) before they were stitched

in place. Clinical monitoring of the transplants (transparency, shape of the eye,

presence of ulcers) was performed under a dissecting microscope on anesthe-

tized animals. Adverse reactions included loss of transplants, infection, forma-

tion of an opaque pannus with corneal vascularization and phtisis bulbi. Most of

these reactions were the consequence of the surgical procedure.

Wounding of the ocular surface. Corneas were wounded according to standard

protocols4. In brief, the entire surface of the cornea to the exclusion of the limbalregion was exposed for 30 s to 20% ethanol applied with a cotton-tip applicator

before it was washed with sterile PBS. Complete removal of corneal epithelial

tissue was achieved using a Demarres spatula (Moria Surgical) with a 45u cutting

angle. Efficacy of wounding was monitored using 0.05% ophthalmic fluorescein

solution. Fornix and conjunctiva were wounded using 20% ethanol and an

ophthalmic knife as described above.

Cell culture. Biopsies of the ocular surface were dissociated using 0.05% trypsin

in 0.01% EDTA. Cells were cultivated onto a feeder layer of lethally irradiated

3T3-J2 cells in 3:1 DMEM/F12 medium, supplemented as described17. Cultures

were fed every 3 to 4 days in medium supplemented with 10 ng ml21 of human

recombinant EGF (Upstate Biotechnology) starting at the first feeding, and were

usually passaged once a week. Analyses of karyotypes were by Chrombios

(Raubling).

Distribution of clonogenic epithelial cells in mammalian ocular surface. The

ocular surface was dissected from eyes of different mammals soon after culling

and cut into several fragments, the size of which was 2–6 mm2. Each fragment

was enzymatically dissociated and the cells were cultivated onto a feeder layer of

irradiated 3T3-J2 cells8,28. Seeding density per dish was half of the cell suspensionfor the rat, 1/40 for the rabbit and 1/20 for other species. Cultures were fixed 9 to

12 days later. The shape, growth and differentiation of the colonies (as shown by

the differing intensity of the rhodamine B staining) varied from species to spe-

cies. Pig cells had the most robust growth whereas calf corneal cells stratified had

the least. Human corneas were dissected from eyes obtained from children

(1 year old and 4 years old) undergoing removal of an eye for retinoblastoma.

A sample including the cornea, the limbus region and its adjacent conjunctiva

was immediately dissected, placed in cold culture medium and transported to

the laboratory where it was quickly processed as described. The corneal sample

was then cut into five fragments (conjunctiva, limbal, peripheral, intermediate

and central cornea); fragment sizes ranged from 4.5 to 6 mm2. Each fragment was

enzymatically dissociated and cells were then plated onto an irradiated feeder

layer of 3T3-J2 cells and cultivated as described previously. Cultures were fixed at

12 days and stained with rhodamine B. Each Petri dish contained 30% of the cells

present in each fragment.

Single-cell isolation. Single cells were isolated with an elongated Pasteur pipette

under an inverted microscope and individually cultivated on a feeder layer of

irradiated 3T3 cells as described33. Cultures were fed every 3 to 4 days asdescribed previously. Clonal analyses were performed as described7,17.

Histology. Tissue samples were fixed in Bouin’s liquid, embedded in resin

(Historesin, Leica) and 4-mm sections were stained with haematoxylin. b-gal

staining was performed on 4% paraformaldehyde-fixed samples according to

standard protocols. Samples were then post-fixed in Bouin’s liquid and embed-

ded in resin (Historesin, Leica). Four-millimetre sections were counterstained in

nuclear fast red (CI 60760) and mounted in Eukitt. Sections were examined

under an Axioskop miscrocope (Zeiss) equipped with a digital camera.

Captured images were processed using Zeiss Axiovision 3.1 software. PAS stain-

ing was performed on cultured cells according to standard procedures.

Immunostaining. Tissue samples were snap-frozen in liquid nitrogen and 8-mm

frozen sections were obtained using a cryostat (Leica) with the cutting chamber

temperature set at 230uC. Sections or cultured cells were fixed for 10 min in

50:50 acetone/methanol (v/v) at 220uC. Antibodies were highly cross-adsorbed

mouse monoclonal anti-rabbit cytokeratin 3 (clone AE5, Chemicon), mouse

monoclonal anti-human p63 (clone 4A4, Dako), and mouse monoclonal anti-

human cytokeratin 19 (clone RCK 108, Dako), used at dilutions of 1:500, 1:500

and 1:200, respectively. Samples were fixed for 10 min in 4% paraformaldehyde

at 4 uC for staining with anti-mouse monoclonal anti-mucin 5AC (clone 45M1,

Neomarkers). Secondary antibodies were goat anti-mouse Alexa Fluor 488 or

Alexa Fluor 568 (Molecular Probes Inc), diluted to 1:200.

RT–PCR. Primers were selected to anneal with messenger RNA of selected rat or

human genes using the primer 3-input software. Total RNAs were extracted

using Trizol reagent (Invitrogen). Fifty to one-hundred nanograms of total

RNA were amplified using the OneStep RT–PCR kit (Qiagen) according to

the manufacturer’s instructions. For nested PCR of keratin 19, 1ml of the first

amplification was added to a new reaction mix containing the inner primer for

keratin 19 and amplified as described before for a total of 39 cycles. PCR products

were resolved by agarose gel electrophoresis, stained with ethidium bromide and

visualized under ultraviolet light. b-actin was used as a control for equal loading

throughout the experiments. PCR products were sequenced to confirm the

specificity of the reaction (Fasteris SA). Sequences were compared to human

sequences from the GenBank database and were highly similar (90%).

List of primers. The following primers were used: b-actin forward, 59-

TCATGTTTGAGACCTTCAACACCC-39; b-actin reverse, 59-GTACTTGCGC-

TCAGGAGGAG-39; Bmi1 forward, 59-TGTGCGTTACTTGGAGACCA-39;

Bmi1 reverse, 59-TCATTCACCTCCTCCTTTGG-39; K4 forward, 59-

CTCCAGCAAAAACCTTGAGC-39; K4 reverse, 59-ACCTCGGCAATAATGC-

TGTC-39; K12 forward, 59-GGTCCAGGTGAGGTCAATGT-39; K12 reverse,

59-GACAGTTGGCAGCAGTACCC-39; K19 outer, forward 59-CGCGACTACA-

GCCACTACT-39; K19 outer, reverse, 59-GCTCACTATCAGCTCGCACA-39;

K19 inner, forward, 59-TTTGAGACGGAACAGGCTCT-39; K19 inner, reverse,

59-TCAGTAACCTCGGACCTGCT-39; Muc5ac forward, 59-CTGCCAGG-

ACTGCATCTGTA-39; Muc5ac reverse, 59-CCTCACACAGGCATCTCTCA-39;

Trp63 forward, 59-ACTGCCAGATTGCAAAGACC-39; Trp63 reverse, 59-

GCTGCTGTTGCACATGAAAT-39; Zfp145 forward, 59-AAAGCAGAGGA-

CCTGGATGA-39; Zfp145 reverse, 59-TCATGGCTGAGAGACCAAAA-39.

DNA microarray analysis. Quantitative analysis of RNA expression was per-

formed using Affymetrix gene chip cDNA microarrays. Three central corneal

clones and three conjunctival clones were used to prepare total RNA using the

RNeasy Mini Kit (QIAGEN Inc). RNA quality was measured using the Agilent

Bioanalysation system to ensure the integrity of the RNA. Complementary DNA

synthesis, hybridization to the Affymetrix GeneChip Porcine Genome Array

(23,937 probe sets representing 20,201 genes) and analysis was performed by

the Lausanne DNA Array Facility using standard protocols. To identify differ-

entially expressed genes from each group, P values were calculated using

Bioconductor limma package34 and probe sets with a false discovery rate35 ,0.05

were considered significant.

31. Friedrich, G. & Soriano, P. Promoter traps in embryonic stem cells: a geneticscreen to identify and mutate developmental genes in mice. Genes Dev. 5,1513–1523 (1991).

32. Hadjantonakis, A. K., Gertsenstein, M., Ikawa, M., Okabe, M. & Nagy, A.Generating green fluorescent mice by germline transmission of green fluorescentES cells. Mech. Dev. 76, 79–90 (1998).

33. Barrandon, Y. & Green, H. Cell size as a determinant of the clone-forming ability ofhuman keratinocytes. Proc. Natl Acad. Sci. USA 82, 5390–5394 (1985).

34. Smyth, G. K. Linear models and empirical bayes methods for assessing differentialexpression in microarray experiments. Stat. Appl. Genet. Mol. Biol. 3, article–3(2004).

35. Hochberg, Y. & Benjamini, Y. More powerful procedures for multiple significancetesting. Stat. Med. 9, 811–818 (1990).

doi:10.1038/nature07406

©2008 Macmillan Publishers Limited. All rights reserved