Ocular surface epithelia contain ABCG2-dependent side population cells exhibiting features associated with stem cells

Abstract

When cell populations are incubated with the DNA-binding dye Hoechst 33342 and subjected to flow cytometry analysis for Hoechst 33342 emissions, active efflux of the dye by the ABCG2/BCRP1 transporter causes certain cells to appear as a segregated cohort, known as a side population (SP). Stem cells from several tissues have been shown to possess the SP phenotype. As the lack of specific surface markers has hindered the isolation and subsequent biochemical characterization of epithelial stem cells this study sought to determine the existence of SP cells and expression of ABCG2 in the epithelia of the ocular surface and evaluate whether such SP cells had features associated with epithelial stem cells. Human and rabbit limbal-corneal and conjunctival epithelial cells were incubated with Hoechst 33342, and analyzed and sorted by flow cytometry. Sorted cells were subjected to several tests to determine whether the isolated SP cells displayed features consistent with the stem cell phenotype. Side populations amounting to <1% of total cells, which were sensitive to the ABCG2-inhibitor fumitremorgin C, were found in the conjunctival and limbal epithelia, but were absent from the stem cell-free corneal epithelium. Immunohistochemistry was used to establish the spatial expression pattern of ABCG2. The antigen was detected in clusters of conjunctival and limbal epithelia basal cells but was not present in the corneal epithelium. SP cells were characterized by extremely low light side scattering and contained a high percentage of cells that: showed slow cycling prior to tissue collection; exhibited an initial delay in proliferation after culturing; and displayed clonogenic capacity and resistance to phorbol-induced differentiation; all features that are consistent with a stem cell phenotype.

Fig. 1

Hoechst plots of human and rabbit ocular surface epithelial cells. Human (A,B) and rabbit (C-F) limbal cell plots obtained using 2.0 μM dye (left-hand graphs) and conjunctival cell plots using 3.0 μM dye (right-hand graphs). (C,D) Controls. (E,F) Identical experiments to C and D, respectively, but with 10 μM fumitremorgin C added 5 minutes prior to the introduction of Hoechst. Inserts in D and F are a different experiment with rabbit conjunctiva in control condition or with 20 μM reserpine. The three main zones present in these plots, the SP, the main cohort of cells containing one copy of DNA (cells in G0-G1) and the cohort of most cells containing DNA in excess of one copy by virtue of being in the S, G2 or M phase of the cell cycle (S-G2/M) are indicated.

Fig. 2

ABCG2 immunolocalization in the ocular surface. A. Whole human limbus. To generate this micrograph, multiple frames were spliced together. Note that ABCG2 is intensively expressed in the recedes of the Palisades of Vogt (Pal. Vogt), in particular near the conjunctiva. The immunostaining pattern suggests a cluster organization. Individual clusters are marked by asterisks. (B) High magnification micrograph of a single Palisade of Vogt. (C) The section adjacent to that stained in B was processed with the omission of the anti-ABCG2 antibody as a control. (D) Human corneal central epithelium. BL, Bowman layer. (E-G) Selected images of the human conjunctiva. ABCG2 expression is extremely low in the very thick palpebral zone (E). It is moderately present in the thinner transition area (F) and at higher levels in the Goblet cell-rich palpebral-fornical zone (G). Asterisks indicate Goblet cells. Magnifications: A,G 120×; B,C,E,F 250×; D 400×.

Fig. 3

Reciprocal correlations between Hoechst emission and light scattering features in ocular surface cells. (A) Human limbal epithelium. The scatter plot for all the cells (All cells) shows a main cohort and a small group of cells (LSSC) displaying very low SSC values. Note that ~50% of the SP cells are LCCS cells and that some of the LSSC cells are not SP cells. (B) Representative experiment of the distribution of LSSC cells within the Hoechst plot in the rabbit conjunctiva. Note that the LSSC cells (zone highlighted with a gray mask) distribute equally between the SP and G0-G1 zone. Axis denominations are provided in capital italics within the frames.

Fig. 4

Phase-contrast micrographs of human limbal epithelial G0-G1 and SP cells. The non-SP cells were collected from the center of the G0-G1 spot. Bar, 30 μm.

Fig. 5

Hoechst and scatter plots of human corneal cells. Note the complete absence of SP and the paucity of LSSC cells.

Fig. 6

Immunofluorescence micrographs of BrdU in rabbit ocular surface epithelia. (A,B) Palpebral conjunctiva. The location of the mucocutaneous junction (mcj) is indicated. Arrowheads point to some of the Goblet cells. Frame A shows a representative section from the rabbit subjected to 10 days of BrdU treatment. About 50% of the cells incorporate label (yellow-green fluorescence). Labeling extends to the suprabasal compartment indicating the rapid nature of cell turnover and stratification in this epithelium. Frame B shows an equivalent section from the animal subjected to a 38-day BrdU chase. Label is scant. Within the proliferative basal layer BrdU is present in cells in the Goblet cell-rich zone but it is essentially absent within the basal cells of the highly layered and Goblet-cell free zone proximal to the mucocutaneous junction. (C) Fornical conjunctiva of the same specimen shown in B. In this area the epithelium presents a lobular-like organization and displays substantial BrdU retention in basal cells. (D,E) Sections incorporating bulbar conjunctiva and the limbal (Li) and central corneal (Co) zones. There is substantial BrdU incorporation in all three domains (D). Following the 38-day chase, strong BrdU retention is seen only in the cells of the limbus (E). Bar, 100 μm.

Fig. 7

Distribution of BrdU in rabbit ocular surface epithelial cells sorted by flow cytometry. (A-D) Representative images of the experiments with limbal and conjunctival cells sorted following a 10-day label infusion. Frames A and B are limbal SP and LSSC cells. Note that in both populations the percentile of BrdU⁺ cells (yellow-green fluorescence) is low. Images were captured using dual red-green emission filtering. Unlabeled cells are shown as inserts for reference purposes. Frames C and D belong to the adherent cell fractions (Methods) from conjunctival G0-G1 and SP cells, respectively. For this experiment, green and red emissions were captured independently and superimposed, as graphically depicted on the left-hand side of C. Note that in the G0-G1 fraction the majority of cells incorporate BrdU whereas only a few SP cells are BrdU-positive. (E-H) Images from an experiment where the rabbit was infused for the full pump duration (14–16 days) followed by a 38-day label chase. Frames F and E correspond to conjunctival G0-G1 and SP respectively. Frames G and H are from the non-LSSC and LSSC limbal cells. Small inserts are images of BrdU-free nuclei.

Fig. 8

CFDA-SE retention of freshly plated limbal epithelial cells as a function of their SSC. The time elapsed since plating is indicated within each plot. The decrease in CFDA-SE fluorescence primarily reflects cell division. The general population of adherent cells exhibits large decreases in fluorescent intensity in the 24–48, 48–72 and 72–96 hour intervals whereas the great majority of the LSSC cells seem to retain most of the initial fluorescence for the first 72 hours. The experiment shown is representative of two experiments with limbal and two experiments with conjunctival epithelial cells.

Fig. 9

Clonogenic growth of rabbit limbal G0-G1 and SP cells on 3T3 feeder cells. G0-G1 cell collection was restricted to cells having an FSC range comparable to that of the SP cells. Colonies from G0-G1 (A-C) and SP cells (D-F) were generated from 2000 plated cells after 11 days (A,B) or after 14 days (C,D) of culture. PMA (100 μM) was included in these cultures for 4–72 hours post-plating interval. Note that essentially all colony formation by the G0-G1 cells is arrested whereas a substantial number of colonies grow from the SP cells, and the SP culture contains a substantial number of microcolonies (mc). Microscopic examination showed that most of these colonies consist of small, morphologically undifferentiated cells. (C,F) Fate of microcolonies. Dishes were complemented with 400 limbal epithelial cells and treated with PMA as described for B and E. After 14 days, colonies were present only in the SP dish. These large fast-growing colonies were removed by swiping (clear cell-free areas marked by asterisks on the grayish background generated by the 3T3 cell mat) and the culture was continued for two more weeks. The microcolonies continued to grow to form new large colonies.