Immune profile of squamous metaplasia development in autoimmune regulator-deficient dry eye

Abstract

Purpose: Squamous metaplasia of the ocular surface epithelium in severe Sjögren syndrome (SS) dry eye has been implicated to be associated with chronic engagement of immune-mediated inflammation. While the detailed immunopathological mechanism underlying keratinization of the ocular muco-epithelium in this setting remains unclear, mice deficient in the autoimmune regulator gene (Aire) demonstrate SS-like pathological changes in the exocrine organs and ocular surface including squamous metaplasia. Using this murine model, we sought to determine the specific immune events that predict squamous metaplasia of the cornea in Aire deficiency.

Methods: Lissamine green staining, goblet cell density, and corneal small proline-rich protein 1B (SPRR1B) were compared in Aire-sufficient and -deficient mice at 4, 8, and 16 weeks of age. Corneal, limbal and conjunctival infiltration of CD4(+) and CD8(+) T cells as well as CD11c(+) and MHC class II (I-A(d+)) dendritic cells (DCs) were examined at the same time points. Ordinary least squares regression was used to model SPRR1B's relationship with lissamine green staining, goblet cell density, and immune cell infiltration.

Results: Lissamine green staining was present in Aire-deficient mice by four weeks of age and increased over time. Compared to Aire-sufficient controls, conjunctival goblet cell density (GCD) decreased and corneal SPRR1B increased in Aire-deficient mice with significant differences noted at both 8 and 16 weeks. Immune-mediated CD4(+) T cell infiltration of the conjunctiva and limbus peaked at eight weeks and then decreased. In contrast, corneal T cell infiltration continued to increase over time, reaching a maximum cell number at 16 weeks. CD11c(+) myeloid-derived DCs were found in the conjunctiva and limbus at all time points. As the mice aged, there was a notable increase in corneal CD11c(+) cell counts. Interestingly, the dynamic of activated MHC class II(+) DCs was nearly identical to that of CD4(+) T cells, peaking first in the limbus at eight weeks with maximum infiltration of the cornea by 16 weeks. Regression analysis showed that squamous metaplasia biomarker, SPRR1B, is strongly related to the lissamine green staining of the ocular surface. Corneal infiltration of activated DCs was most prognostic of corneal SPRR1B expression while the presence of precursor DCs, activated DCs, and CD4(+) T cells in the limbus were also significant predictors of SPRR1B.

Conclusions: Aire-deficient mice represent a useful model to study Sjögren-like autoimmune-mediated ocular surface disease. Results of the current study suggest that squamous cell precursor protein, SPRR1B, provides an important readout to evaluate ocular surface damage and specific events related to immune-mediated inflammation. Results also define an appropriate time frame for interventional studies to develop more effective therapies for keratinizing ocular surface disease.

Figure 1

Infiltration and destruction of the lacrimal gland and ocular epithelium in Aire-deficient mice. A: Immunohistochemical analysis of CD4⁺ and CD8⁺ T cells was negative in Aire^+/− while Aire^−/− mice showed intensive multifocal aggregates of both cell types. Of note, characteristic interlobular connective tissue that exists in the normal lacrimal gland (asterisks) disappears in the Aire^−/− lacrimal gland following lymphocytic infiltration. Bar, 100 μm. B: Time course of lissamine green staining in the corneas of Aire^+/− and Aire^−/− mice demonstrates a compromised epithelial integrity caused by Aire deficiency. Mild to severe keratoconjunctivitis sicca (KCS) shown as punctate to confluent green staining were observed in Aire^−/− mice whereas corneas of Aire^+/− largely remained unstained. C: Progression of lissamine scores is shown on a chart. Data are shown as mean±SD. The asterisk in this panel indicates that p<0.05, Aire^+/− versus Aire^−/− at each time point.

Figure 2

Time course of goblet cell density in Aire^+/− versus Aire^−/− mice. A: Periodic acid–Schiff staining revealed abundant goblet cells (arrow) in Aire^+/− at all time points (representative example shown) and in Aire^−/− at four weeks of age. A gradual decrease in goblet cell density was observed in the conjunctiva of Aire^−/− mice over time. Bar, 100 μm. B: Time course of goblet cell density is shown as mean±SD. An asterisk indicates that p<0.05, Aire^+/− versus Aire^−/− at each time point.

Figure 3

Expression of squamous cell biomarker, SPRR1B. A: Immunofluorescence showed increased SPRR1B staining (green) in Aire^−/− at all time points with no staining in age-matched Aire^+/− controls. Bar, 100 μm. B: Quantification of SPRR1B immuno-intensity over time. Data are shown as mean±SD. An asterisk in indicates that p<0.05, Aire^+/− versus Aire^−/− at each time point.

Figure 4

T cell distribution in the ocular surface of Aire-deficient mice. Assemblage of low magnification images showing anatomic locations analyzed in Aire^−/− (A) and Aire^+/− (B) mice. Cj=conjunctiva; Lm=limbus; Co=cornea (paracentral and central). Immunohistochemistry of CD4⁺ (C,D) and CD8⁺ T (E,F) cells demonstrates both epithelial (white arrow) and stromal (black arrow) infiltration of the ocular surface in Aire^−/−. Aire^+/− controls exhibited no cell-mediated inflammation at the locations analyzed. Bar, 100 μm. G: Time course of CD4⁺ and CD8⁺ cell density in Aire^−/− and Aire^+/−. Data are shown as mean±SD. An asterisk indicates that p<0.05, CD4⁺ Aire^+/− versus CD4⁺ Aire^−/−. The symbol, †, indicates that p<0.05, CD8⁺ Aire^+/− versus CD8⁺ Aire^−/−.

Figure 5

Dendritic cell distribution across the ocular surface of Aire-deficient mice. A: Immunofluorescence study of CD11c⁺ cells (red) demonstrates dendritic antigen presenting cells (APCs) in the basal epithelium (open arrowhead) and sub-epithelial stroma (arrow) of the conjunctiva (Cj), limbus (Lm), and central cornea (Co). Although CD11c⁺ cells are found in both Aire^+/− and Aire^−/− mice, significantly more are seen in Aire-deficient mice. CD11c⁺ cells are present throughout the whole layer of the cornea including the posterior stroma (solid arrow head) in Aire^−/− while they are largely absent in the posterior cornea in Aire^+/−. The dotted line represents the epithelial basement membrane in the limbus, and the blue nuclear counterstaining (DAPI) is used for orientation. Bar, 100 μm. B: Immunohistochemical study of MHC class II surface antigen (I-A^d) reveals an intense infiltration of activated dendritic APCs in the epithelium (open arrow head) and stroma (arrow) of the conjunctiva (Cj), limbus (Lm), and central cornea (Co) of Aire^−/− mice. By comparison, I-A^d+ cells are rarely apparent in the conjunctiva and limbus and are completely absent in the central cornea of Aire^+/− mice. The asterisk denotes goblet cells. Bar, 100 μm. C: Quantification of CD11c+ and I-A^d+ cells is shown over time. Data are shown as mean±SD. An asterisk indicates that p<0.05, CD11c⁺ Aire^+/− versus CD11c⁺ Aire^−/−. The symbol, †, indicates that p<0.05, I-A^d+ Aire^+/− versus I-A^d+ Aire^−/−.

Figure 6

Corneal SPRR1B expression is related to lissamine green staining but not goblet cell density. A: Fifty-eight percent of the variability in corneal SPRR1B (expressed as mean intensity score) is explained by the amount of lissamine green staining (LG staining; R²=0.58, p=0.0030, n=11) versus only 1% by goblet cell density (GCD) in the right panel (B; R²=0.01, p=0.67, n=19). Open circles represent raw data, and the x-axis plots SPRR1B immuno-intensity in arbitrary units while the y-axis plots the lissamine green score (A) or GCD (B; cells/section). The straight line is the best-fit linear regression (ordinary least squares)

Figure 7

Immune cell infiltration of the ocular surface predicts corneal SPRR1B expression. For each of the 12 subfigures, we plotted SPRR1B as fluorescent intensity on the y-axis in arbitrary units while immune cell counts (CD4, CD8, CD11c, and I-A^d) are plotted on the x-axis in three different locations (Co, cornea; Lm, limbus; Cj, conjunctiva). Each graph shows raw observations (open circles) as well as a regression line. Each regression line was fit using ordinary least squares to ensure the validity of statistical tests with respect to normality. In order to minimize highly influential points, we transformed both X and Y using the cube root and logarithm, respectively. Curvilinear graphs appear when the regression model is transformed back to the original scale. Centered quadratic terms were included only when statistically significant at the 0.05 level, which only occurred for CD8 cells in the conjunctiva, row 2, column 3. Significant p values (shown in blue) were computed to test the null hypothesis that SPRR1B did not depend on each of the 12 immune cell counts. Note that these p values were adjusted for multiple comparisons using the Holm method. Squared multiple correlation coefficients indicate the fraction of the variance explained by the regression model. The most compelling predictors of corneal SPRR1B were corneal I-A^d+ cells (two stars) and limbal CD11c⁺ cells (one star). CD4⁺ and I-A^d+ immune cells in the limbus as well as generalized presence of immune cells in the conjunctiva also showed strong and statistically significant relationships to SPRR1B expression.