Mouse Conjunctival Forniceal Gene Expression during Postnatal Development and Its Regulation by Krüppel-like Factor 4

Abstract

Purpose.: To identify the changes in postnatal mouse conjunctival forniceal gene expression and their regulation by Klf4 during the eye-opening stage when the goblet cells first appear.

Methods.: Laser microdissection (LMD) was used to collect conjunctival forniceal cells from postnatal (PN) day 9, PN14 and PN20 wild-type (WT), and PN14 Klf4-conditional null (Klf4CN) mice, in which goblet cells are absent, developing, present, and missing, respectively. Microarrays were used to compare gene expression among these groups. Expression of selected genes was validated by quantitative RT-PCR, and spatiotemporal expression was assessed by in situ hybridization.

Results.: This study identified 668, 251, 1160, and 139 transcripts that were increased and 492, 377, 1419, and 57 transcripts that were decreased between PN9 and PN14, PN14 and PN20, PN9 and PN20, and PN14 WT and Klf4CN conjunctiva, respectively. Transcripts encoding transcription factors Spdef, FoxA1, and FoxA3 that regulate goblet cell development in other mucosal epithelia, and epithelium-specific Ets (ESE) transcription factor family members were increased during conjunctival development. Components of pathways related to the mesenchymal–epithelial transition, glycoprotein biosynthesis, mucosal immunity, signaling, and endocytic and neural regulation were increased during conjunctival development. Conjunctival Klf4 target genes differed significantly from the previously identified corneal Klf4 target genes, implying tissue-dependent regulatory targets for Klf4.

Conclusions.: The changes in gene expression accompanying mouse conjunctival development were identified, and the role of Klf4 in this process was determined. This study provides new probes for examining conjunctival development and function and reveals that the gene regulatory network necessary for goblet cell development is conserved across different mucosal epithelia.

The health of the transparent cornea, which accounts for more than 70% of the refractive power of the eye, is dependent on other components of the ocular surface, including the conjunctiva, lacrimal and accessory lacrimal glands, and meibomian glands. ^{1 –5} Different components of the ocular surface are connected by the tear film, a complex structure consisting of an outermost lipid layer secreted by the meibomian glands, central aqueous layer secreted by the lacrimal glands and the corneal and conjunctival epithelial cells, and an inner glycocalyx layer of the membrane-bound mucins on the superficial epithelial cells. ^5,6 Additional components of the aqueous include soluble mucins, antimicrobial peptides, antibodies, and different solutes that are secreted by conjunctival goblet cells and conjunctival and corneal epithelial cells, or are derived by diffusion from the conjunctival vasculature. ^{5 –8}  

The conjunctiva consists of the basal collagenous lamina propria covered with an epithelium comprising four different kinds of stratified squamous cells, including apical granule-rich cells, endoplasmic reticulum-rich cells, Golgi-rich cells and mitochondria-rich cells interspersed with mucin-secreting goblet cells. ⁹ In addition, the conjunctiva is highly innervated, with parasympathetic nerves regulating the goblet cell secretions and sympathetic nerves regulating the stratified squamous cell secretions. ^10,11 The conjunctival goblet cells play a critical role in ocular surface health by producing and secreting mucins, trefoil factors, and other components of the tear film by an apocrine mechanism, in which all the secretory granules in the cell are emptied on stimulation. Loss of the conjunctival goblet cells is associated with severe ocular surface disorders, such as dry eye, ocular cicatricial pemphigoid (OCP), and Sjögren's syndrome. ^{10,12 –15}  

Developmental studies of the ocular surface have mostly focused on the cornea, resulting in characterization of the involvement of the transcription factors Pax6, Klf4, Klf6, E2F, AP1, AP2α, Sp1, Sp3, Sp6, Shh, Cited2, and IκBζ in embryonic development, postnatal maturation, and maintenance of the cornea. ^{16 –30} Despite its critical contributions to the homeostasis of the ocular surface, development of the conjunctiva has been relatively understudied. In the mouse, initial conjunctival epithelial stratification and goblet cell development occur around eye opening, when the conjunctiva is first exposed to photo, oxidative, and environmental stresses. The conjunctival goblet cells share similarities in their structure and function with the intestinal, colonic, and airway mucosal epithelial goblet cells. ^31,32 In these tissues, goblet cell development is regulated by genetic programs involving the Notch pathway; transcription factors Hnf4α, Hnf1α, HNF1β, Klf4, Klf5, FoxA1, FoxA2, FoxA3, Spdef, and IκBζ; and increased cytokine levels. ^{31,33 –48} Similar studies of conjunctival goblet cells are limited, ^{10,27,29,49,50} resulting in sparse information on developmental changes in conjunctival gene expression. 

Microarrays have been used successfully in the study of developmental changes in gene expression ^{51 –53} and for comparative analysis of gene expression in diverse organs. ^49,54 In this study, we catalogued the changes in gene expression accompanying postnatal development of conjunctival forniceal cells by microarray-based gene expression profiling at postnatal day (PN) 9, PN14, and PN20, when goblet cells are absent, developing, and present, respectively, and identified the conjunctival Klf4 target genes by comparing the gene expression patterns between PN14 WT and Klf4-conditional null (Klf4CN) conjunctivas lacking goblet cells. ²⁷ These results provide the first detailed description of the dynamic changes in gene expression accompanying postnatal conjunctival maturation involving epithelial cell stratification and goblet cell development and define the role of Klf4 in this process. 

Materials and Methods

Conditional Deletion of the Klf4 Gene

Klf4CN mice were generated by mating Klf4 ^loxP/loxP, Le-Cre/⁻ mice with Klf4 ^loxP/loxP mice and genotyped as described earlier. ^27,33,55 All procedures described herein were approved by the Institutional Animal Care and Use committee of the University of Pittsburgh and Division of Laboratory Animal Resources (DLAR) and conformed to the ARVO Statement for the use of Animals in Ophthalmic and Vision Research. 

Staining for Goblet Cells, Laser Microdissection, and RNA Isolation

Goblet cells were visualized by staining 8-μm sections from paraformaldehyde-fixed, paraffin-embedded PN9, PN14, and PN20 WT and PN14 Klf4CN head tissue with periodic acid-Schiff's (PAS) reagent, as described earlier. ²⁷ A microscope (model BX60; Olympus America Inc., Lake Success, NY) equipped with a digital camera (Spot; Diagnostic Instruments Inc., Sterling Heights, MI) was used for light microscopy. For LMD, tissue blocks comprising the whole eye, eyelids, and conjunctivas were dissected and incubated in an RNA stabilizer (RNA-Later; Ambion, Austin, TX) for 15 minutes at room temperature, fixed in 2% paraformaldehyde in 0.1 M phosphate buffer (pH 7.3) for 15 minutes, and incubated in 30% sucrose solution for 1 hour. ⁵⁶ Fixed tissue blocks were embedded in OCT compound on dry ice and 8-μm-thick sections were cut on a cryotome (model CM3050Sl; Leica Microsystems, Nussloch, Germany). The sections were transferred to membrane slides (PEN; Leica Microsystems), and the conjunctival forniceal region was harvested by LMD (DMLA microscope and LMD software version 4.4; Leica). ^57,58 We directed the laser beam used for LMD through the conjunctival stroma, to avoid UV laser burns of our target tissue, the conjunctival epithelium. Even though this method yielded relatively undamaged conjunctival epithelial cells, it also resulted in minor contamination with the underlying stromal cells. Considering that a eukaryotic cell contains ∼20 pg total RNA, we collected approximately 10,000 cells from each stage, enough to isolate ∼100 ng total RNA for each sample. Microdissected cells were collected in sterile RNase-free microtubes containing lysis buffer, and total RNA was isolated (RNeasy-Micro kit; Qiagen, Germantown, MD). Each of these total RNA preparations was then analyzed (2100 bioanalyzer, with an RNA 6000 Pico chip kit; Agilent Technologies, Santa Clara, CA). Only high-quality RNA with an RNA integrity number (RIN) > 5 was used for microarray and quantitative RT-PCR analyses. 

Microarray Analysis

All analyses reported here were performed with three independent samples in each group. Total RNA isolated from the LMD samples was subjected to two rounds of amplification, labeled, and hybridized to a mouse genome microarray (430 2.0; Affymetrix, Santa Clara, CA). The microarrays were washed, developed, and scanned using standard protocols. The raw data were processed and analyzed using commercially available (Gene Chip Operating Software [GCOS] ver. 1.4 Affymetrix]), pathway analysis tool (Ingenuity Systems, Redwood, CA) and free (BRB ArrayTools [www.nci.nih.gov/brb-arraytools.html/]) software packages. Differentially expressed genes were categorized into various biological processes using the Panther analysis tool (www.pantherdb.org). Microarray data have been submitted to the National Center for Biotechnology Information Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE26076). Detailed protocols for sample preparation, microarray hybridization and analysis of the microarray data are provided as Supplementary Methods. 

Validation of Microarray Results

RNA samples (100 ng) were amplified as before and 1 μg amplified RNA was used for cDNA synthesis (Superscript VILO cDNA synthesis kit; Invitrogen, Carlsbad, CA). qRT-PCR assays were performed with SYBR green reagent (Super-Array Biosystems, Frederick, MD), using prestandardized, gene-specific primers in a commercial PCR system (StepOne Plus Real Time PCR system; Applied Biosystems, Inc. [ABI], Foster City, CA) with GAPDH as the endogenous control and were analyzed with allied software (StepOne software ver. 2.1; ABI). The sequence of primers used for qRT-PCR is provided in Supplementary Table S1. 

In Situ Hybridization

In situ hybridization was performed with 12-μm sections cut from fresh frozen tissue in optimal cutting temperature (OCT) compound (Sakura Finetec, Torrance, CA) that were fixed in 4% paraformaldehyde, treated with proteinase K (0.2 μg/mL PBS) for 5 minutes, and processed as described earlier. ³⁰ Riboprobes were synthesized using a digoxigenin (DIG) RNA labeling kit (Sp6/T7; Roche Molecular Biochemicals, Indianapolis, IN) with linearized plasmid cDNA templates for the respective genes. The color development reaction was allowed to proceed for 60 minutes, after which the reactions for both the sense and antisense riboprobes were terminated at the same time. 

Results

Development of Goblet Cells and Spatiotemporal Expression of Klf4 in the Mouse Conjunctiva

Developing mouse conjunctival sections stained by PAS reagent revealed that the mouse conjunctival goblet cells were absent at PN9, began to first appear at PN14 around the time of eyelid opening, and are well formed by PN20 (Fig. 1). The one- to two-cell layered PN9 conjunctival epithelium stratified and formed a three- to four- and five- to six-cell–layered squamous epithelium by PN14 and PN20, respectively (Fig. 1). In view of the significant role of Klf4 in conjunctival goblet cell development, ²⁷ we examined the spatiotemporal expression patterns of Klf4 in the developing mouse conjunctiva by in situ hybridization. Klf4 expression was low at PN9, moderate at PN14 and highest at PN20 in the WT conjunctiva (Fig. 2A–C). In contrast, it was faint in the PN14 Klf4CN conjunctiva, consistent with its successful disruption (Fig. 2D). Similarly processed PN20 WT sections hybridized with a Klf4 sense strand probe served as the negative control (Fig. 2E). 

Figure 1.

 

Timeline of conjunctival goblet cell development. PAS-stained sections from (A) WT PN9, (B) WT PN14, (C) WT PN20, and (D) Klf4CN PN14 conjunctival fornices show that the goblet cells first appeared in the conjunctival fornix at eyelid opening at PN14 (arrows), increased in number at PN20, and were absent from the Klf4CN conjunctiva at PN14 (n = 3).

Figure 2.

 

Developmental expression of Klf4 in WT and Klf4CN conjunctiva. The expression of Klf4 mRNA was examined by in situ hybridization with Klf4-specific antisense probe in (A) PN9 WT, (B) PN14 WT, (C) PN20, and (D) PN14 Klf4CN conjunctiva. (E) Similar hybridization with a Klf4 sense strand probe at PN20 served as the negative control. A–E, arrow; Conjunctival forniceal epithelium, n = 3.

Microarray Analysis

We collected the PN9, PN14, and PN20 conjunctival forniceal cells by LMD, isolated good-quality total RNA (Fig. 3), and analyzed their gene expression patterns by microarray, as described in Supplementary Methods. Systemic variation in microarray-analyzed gene expression was examined by using redundant panels targeting the housekeeping genes GAPDH and pyruvate carboxylase. The mean coefficient of variation across the 12 samples (four groups with three independent samples in each) was 28%, so that a twofold increase from the mean is equivalent to 3.5 SD. Mean correlation among replicates was high within each group and lower across different groups, indicating the repeatability and reliability of the microarray data (Supplementary Fig. S1). Microarray data are available at http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE26076. 

Figure 3.

 

Collection of conjunctival forniceal cells and quality of the total RNA isolated. (A) LMD of conjunctival forniceal cells. PAS-stained WT PN14 sections before and after LMD. Arrows, right: the denuded areas from where the cells have been collected. Roughly 40% of the distance from the fornix to the eyelid junction was collected. Bottom: conjunctival fornix before and after LMD at higher magnification, illustrating that, along with most of the forniceal epithelial cells, some underlying stromal cells were also collected. (B) Quality of the total RNA isolated from the LMD samples. Representative gel image of the total RNA isolated from LMD-harvested PN14 WT and Klf4CN conjunctival forniceal epithelial cells. The RNA integrity number (RIN) is shown, along with their usage status for each of the samples.

Changes in Gene Expression Profiles in the Developing WT and Klf4CN Conjunctiva

The numbers of increased and decreased transcripts during conjunctival development and in the Klf4CN compared with WT conjunctiva, detected by pair-wise comparison of microarray results, are shown by Venn diagram (Fig. 4A). The complete list of genes falling in each Venn area is provided in the Supplementary Data). There were 668, 251, and 1160 transcripts that were increased and 492, 377, and 1419 transcripts that were decreased between PN9 and PN14, PN14 and PN20, and PN9 and PN20 WT conjunctivas, respectively. Comparison of the PN14 WT and Klf4CN conjunctival gene expression patterns revealed 139 increased and 57 decreased transcripts, among which 80 transcripts were increased and 19 decreased during different stages of conjunctival development (Fig. 4A; Supplementary Table S2). Venn area A represents the rapid early changes (139 increased and 145 decreased transcripts; Supplementary Table S3) identified only in the PN9 versus PN14 comparison, and Venn area C represents the rapid late changes (75 increased and 149 decreased transcripts; Supplementary Table S4) detected only in the PN14 versus PN20 comparison (Fig. 4A). Venn area B contains transient changes (i.e., rapid early changes in the PN9 versus PN14 comparison that were definitively reversed in the PN14 versus PN20 comparison and are listed in Supplementary Table S5). Venn area G represents the gradual changes spanning the PN9 to PN20 period (581 increased and 898 decreased transcripts; top 10% of transcripts are listed in Supplementary Table S6) identified only in the PN9 versus PN20 comparison (Fig. 4A). Venn area D (450 increased and 309 decreased; top 50 transcripts are listed in Supplementary Table S7), representing the early changes, contained 2.6 times as many genes as Venn area F (97 increased and 190 decreased; top 50 transcripts are listed in Supplementary Table S8) representing the late changes, suggesting that the eyelid-opening stage (around PN12) is an actively morphing period when gene expression patterns are rapidly changing in the conjunctiva (Fig. 4A). Finally, Venn area E contains transcripts that are progressive (i.e., show a twofold change between PN9 and PN14 and a subsequent twofold change in PN14 versus PN20 comparison; Supplementary Table S9). 

Figure 4.

 

Microarray comparison of developmental changes in gene expression. (A) Representation of the microarray data in a Venn diagram. Within each Venn area, the number of increased transcripts is shown in red and the number of decreased transcripts in green. Values in parentheses give the number of transcripts modulated in the Klf4CN versus WT comparison. Please see the Supplementary Tables for the full list of transcripts in each Venn area. (B) Categorization of the affected transcripts according to their biological functions. Significantly affected transcripts are categorized into various biological processes using the Panther analysis tool (www.pantherdb.org). Corresponding gene ontology (GO) terms are shown in parentheses in the legend.

Categorization of the affected genes according to their biological functions showed consistent changes in PN14 versus PN9, PN20 versus PN14, PN20 versus PN9, and Klf4CN versus WT comparisons (Fig. 4B). Metabolic processes, cellular processes, cell communication, developmental processes, and transport were the most affected biological processes in both the developing and the Klf4CN conjunctiva. Other important processes affected included immune system process, response to stimulus, cell adhesion, and cell cycle (Fig. 4B). 

Validation of Microarray Results

Selected transcripts that were unaffected, increased, or decreased in each of these comparisons were measured by real time qRT-PCR, using cDNA generated with three different independent samples of total RNA than the ones used for microarray analyses. These qRT-PCR results were largely similar to those obtained from microarray analyses, thus validating the results obtained through microarray analyses (Fig. 5). 

Figure 5.

 

Validation of microarray results. Real-time qRT-PCR was used to compare the expression levels of selected transcripts in (A) PN14 WT and Klf4CN, (B) PN9 and PN14, (C) PN14 and PN20, and (D) PN9 and PN20 conjunctival forniceal cells harvested by LMD. Relative expression levels were calculated from three independent samples, each with two replicates. Mean relative expression levels detected by qRT-PCR (red bars) are compared with the values derived from the microarray data (blue bars). Logarithmic scale is used on the y-axis axis. Error bars, SE (n = 3).

Significant Changes in Gene Expression during Conjunctival Development

A list of the top 20 most increased and decreased transcripts between PN9 and PN20 is provided in Table 1. The zymogen granule membrane protein GP2 that binds enterobacterial type 1 fimbria FimH ^59,60 was the most increased transcript between PN9 and PN20 suggesting that the conjunctival mucosal immune system develops around this time. Consistent with this, two members of the tetraspanin family with putative roles in mucosal immunity, ⁶¹ CD274 (PD-L1) ⁶² and polymeric immunoglobulin receptor (Pigr), ⁶³ also were significantly increased during this time period. Other significantly increased transcripts relevant to conjunctival physiology include the goblet cell marker Muc5Ac and glucosaminyl transferase Gcnt3, required for glycosylation of mucins (Table 1). Expression of several mucins and glycosylation-related genes, although increased in the course of conjunctival development, was decreased in the Klf4CN conjunctiva, consistent with the appearance of goblet cells during this period and the absence of goblet cells in the Klf4CN conjunctiva (Supplementary Table S10). Expression of a large number of solute carrier family members that play important roles in different mucosal epithelial tissues was increased during conjunctival development (Supplementary Table S11). Several members of the Wnt signaling pathway were significantly decreased during conjunctival development (Supplementary Table S12), consistent with the lack of involvement of Wnt pathway in enteroendocrine cell differentiation. ⁶⁴ The cytokines Il15 and Il33, associated with goblet cell hyperproliferation and development, ⁴⁸ and Jak2 and Stat1, members of the Jak-STAT pathway through which these cytokines exert their influence, were altered in the developing WT and the Klf4CN conjunctiva. 

Table 1.

 

Top 20 Most Increased and Decreased Transcripts in the PN20 Compared with PN9 Conjunctiva along with Changes in Other Developmental Stages and in the Klf4CN Conjunctiva

Gene Symbol	Gene Name	PN 20/PN9	PN14/PN9	PN20/PN14	Klf4CN/WT
Most Increased Transcripts
Gp2	Glycoprotein 2 (zymogen granule membrane)	87.04	25.55	3.25	0.91
Slc5a1	Solute carrier family 5, member 1	56.37	30.10	1.87	0.26
Apbh	Androgen-binding protein eta	49.75	5.30	24.16	133.20
Muc5ac	Mucin 5, subtypes A and C, tracheobronchial/gastric	38.25	7.16	3.59	0.00
Gcnt3	Glucosaminyl (N-acetyl) transferase 3, mucin type	29.21	6.66	4.47	0.15
Reg3g	Regenerating islet-derived 3 gamma	26.50	21.22	6.27	3.46
Gnat1	Guanine nucleotide binding protein, α transducing 1	26.33	33.64	1.42	1.30
Aqp4	Aquaporin 4	20.32	12.23	1.68	0.63
Elovl4	Elongation of very long chain fatty acids-like 4	19.78	25.24	0.87	2.48
Slc44a4	Solute carrier family 44, member 4	19.62	9.99	1.93	0.60
Asb11	Ankyrin repeat and SOCS box-containing 11	17.55	13.07	1.24	0.61
Olfm4	Olfactomedin 4	17.31	9.64	1.79	1.79
Cd274	CD274 antigen	17.10	11.36	1.50	0.47
Pigr	Polymeric immunoglobulin receptor	16.69	14.42	1.22	0.93
Tspan1	Tetraspanin 1	15.74	8.43	1.69	0.57
Slc24a1	Solute carrier family 24, member 1	15.19	16.24	1.80	0.98
Cldn2	Claudin 2	15.18	5.63	2.80	1.11
Tspan8	Tetraspanin 8	14.96	2.30	14.84	0.92
Clic5	Chloride intracellular channel 5	14.69	1.51	6.14	0.40
Slc6a14	Solute carrier family 6, member 14	14.25	7.09	1.99	1.54
Most Decreased Transcripts
Igf2	Insulin-like growth factor 2	0.063	0.22	0.28	0.76
Fbn2	Fibrillin 2	0.069	0.24	0.30	0.47
Acan	Aggrecan	0.088	0.21	0.43	1.39
Lrrc15	Leucine rich repeat containing 15	0.092	0.44	0.23	1.01
Gjc1	Gap junction protein, gamma 1	0.103	0.34	0.18	1.51
Agtr2	Angiotensin II receptor, type 2	0.107	0.49	0.21	1.15
Col11a1	Collagen, type XI, alpha 1	0.110	0.38	0.26	0.97
Mdk	Midkine	0.115	0.27	0.40	1.22
Smoc2	SPARC related modular calcium binding 2	0.115	0.21	0.55	0.67
Gpx7	Glutathione peroxidase 7	0.121	0.50	0.22	0.65
Tchh	Trichohyalin	0.124	0.28	0.21	0.18
Upk3b	Uroplakin 3B	0.126	0.20	0.59	0.41
Podxl2	Podocalyxin-like 2	0.130	0.43	0.31	0.62
Actg2	Actin, gamma 2, smooth muscle, enteric	0.134	0.49	0.27	0.79
Lgals1	Lectin, galactose binding, soluble 1	0.134	0.39	0.34	0.75
Meg3	Maternally expressed 3	0.134	0.33	0.39	0.53
Aebp1	AE binding protein 1	0.135	0.37	0.35	0.57
Col9a3	Collagen, type IX, alpha 3	0.139	0.30	0.15	0.12
Loxl1	Lysyl oxidase-like 1	0.140	0.39	0.34	0.77
Selm	Selenoprotein M	0.144	0.38	0.37	0.73

Klf4 Target Genes in the Mouse Conjunctival Fornix

Comparison of the PN14 WT and Klf4CN conjunctival forniceal gene expression patterns identified 139 and 57 transcripts increased and decreased, respectively, in the Klf4CN conjunctiva (Fig. 4A). A list of the top 20 most increased and decreased transcripts between PN14 WT and Klf4CN conjunctival cells is provided in Table 2 (complete list in Supplementary Table S2). Among these, Muc5Ac, Gcnt3, and Slc5a1 were also present in the list of most highly increased transcripts in the course of conjunctival development between PN9 and PN20 (Tables 1, 2). Transcripts encoding four other mucins (Muc5b, Muc4, Muc16, and Muc20) and Lypd2, a member of the Ly6 family of proteins, were significantly decreased in the Klf4CN conjunctiva (Table 2). Transcripts encoding the goblet cell markers Muc5Ac, Lman1l, and Gcnt3; transcription factor Spdef, which regulates goblet cell development in airway and colonic epithelial tissues; the mucosal epithelium enriched serine protease Prss22; and the solute carrier family member Slc5a1 were significantly decreased in the Klf4CN conjunctiva and increased during conjunctival forniceal development between PN9 and PN20. ^{14,36,37,41,65,66}  

Table 2.

 

Top 20 Most Significantly Increased and Decreased Transcripts in the Klf4CN Conjunctiva

Gene Symbol	Description	Fold Change
		Klf4CN/WT	PN14/PN9	PN20/PN14	PN20/PN9
Most Increased Transcripts
Apbh	Androgen-binding protein eta	13.50	5.30	24.16	128.1
Sprr2a	Small proline-rich protein 2A	7.41	0.58	1.12	0.65
Ppbp	Pro-platelet basic protein	7.18	0.71	0.85	0.60
Fam19a2	Family with sequence similarity 19, member A2	6.41	0.69	1.03	0.71
Elovl4	Elongation of very long chain fatty acids-like 4	6.25	25.24	0.87	22.06
Ms4a4c	Membrane-spanning 4-domains, subfamily A, member 4C	4.98	0.41	1.02	0.42
Alox15	Arachidonate 15-lipoxygenase	4.41	1.28	1.12	1.44
Siah1a	Seven in absentia 1A	4.27	0.68	3.46	2.37
Umod	Uromodulin	4.11	1.39	0.67	0.93
Expi	Extracellular proteinase inhibitor	4.09	0.68	1.80	1.22
Cyp2d9	Cytochrome P450 family-2, subfamily-d, polypeptide-9	4.03	0.55	0.89	0.49
Far2	Fatty acyl coa reductase 2	3.98	35.18	0.43	15.19
Soat1	Sterol O-acyltransferase 1	3.82	3.91	0.93	3.65
P2ry10	Purinergic receptor P2Y, G-protein coupled 10	3.81	1.02	1.05	1.07
Crisp1	Cysteine-rich secretory protein 1	3.71	2.15	1.92	4.13
Depdc1b	DEP domain containing 1B	3.60	0.24	3.54	0.85
Krt79	Keratin 79	3.49	15.20	0.42	6.32
Retnla	Resistin like alpha	3.48	2.95	0.72	2.12
Cap1	CAP, adenylate cyclase-associated protein 1	3.44	1.15	0.90	1.03
Cytip	Cytohesin 1 interacting protein	3.30	0.99	0.82	0.81
Most Decreased Transcripts
Muc5ac	Mucin 5, subtypes A and C, tracheobronchial/gastric	0.01	7.16	3.59	25.73
Lypd2	Ly6/Plaur domain containing 2	0.02	6.34	0.47	2.99
Muc5b	Mucin 5, subtype B, tracheobronchial	0.10	2.27	1.89	4.31
Krt78	Keratin 78	0.11	4.36	1.88	8.17
Muc4	Mucin 4	0.14	5.31	0.71	3.79
Gcnt3	Glucosaminyl (N-acetyl) transferase 3, mucin type	0.15	6.66	4.47	29.74
AW112010	Expressed sequence AW112010	0.17	8.67	1.73	14.95
Fam83e	Family with sequence similarity 83, member E	0.17	1.77	2.01	3.56
Avil	Advillin	0.19	1.21	2.05	2.49
Efcab6	EF-hand calcium binding domain 6	0.19	1.44	0.84	1.21
Slc28a3	Solute carrier family 28, member 3	0.20	6.07	1.56	9.45
Krt12	Keratin 12	0.20	3.03	0.36	1.09
5430404G13Rik	RIKEN cdna 5430404G13 gene	0.21	1.32	0.32	0.43
Crygs	Crystallin, gamma S	0.22	2.19	0.93	2.04
Prss32	Protease, serine, 32	0.22	5.72	1.20	6.87
Mip	Major intrinsic protein of eye lens fiber	0.23	1.01	1.35	1.36
Upk1b	Uroplakin 1B	0.23	1.69	0.70	1.18
Muc20	Mucin 20	0.23	3.60	0.87	3.13
Slc5a1	Solute carrier family 5, member 1	0.25	30.10	1.87	56.36

Comparison of the conjunctival Klf4 target genes with the previously identified corneal Klf4 target genes ⁶⁷ yielded a surprisingly small list of common genes. Most of these changes were concordant between the cornea and the conjunctiva, with only seven genes showing discordant changes (Table 3). Thus, Klf4 appears to perform distinct, nonoverlapping functions in these two physically connected components of the ocular surface. These results also show that Klf4 has fewer target genes in the conjunctiva than in the cornea. ⁶⁷ A reason for this could be that in the case of the cornea, RNA was isolated from the whole cornea consisting of several cell types, ⁶⁷ whereas here we used LMD to collect conjunctival forniceal cells consisting mostly of epithelial cells mixed with a few underlying stromal cells. 

Table 3.

 

Klf4 Target Genes with Concordant and Discordant Changes in the Klf4CN Cornea and Conjunctiva

Concordant Increases	Concordant Decreases	Discordant Changes
A130040M12Rik	2200001115Rik	Gadd45a *
Asrgl1	Akr1c12/13	4930523C07Rik
Car2	Arhgef4	Agr2
Ccl6	Bnc1	AW112010
Ccnd2	Clic3	Pcdhb9
Cebpb	Clic5	Runx2
Cp	Cryab	Sox4
Cytip	Efcab6
Elavl2	Fam184b
Expi	Jakmip1
Fabp4	Paqr5
Flt1	Ppp1r3c
H2-Ab1	Prss32
H2-Eb1	Ptgr1
H2-Q7	Rasgrf1
Il33	Serpinb8
Il7r	Slc28a3
Krt17	Spaca1
Laptm5	Upk1b
Mmp13
Mpeg1
Nlrp10
Ppbp
Ptn
Pvrl3
Serpina3n
Sprr2a1 /// Sprr2a2
Tfrc
Tnc
Trps1

 The filtering criteria used in the present study were applied to the data from our previous report on the cornea, to select 454 and 322 unique characterized corneal Klf4 target genes with increased and decreased expression, respectively. A comparison of those corneal Klf4 target genes with the present conjunctival Klf4 target genes identified 56 genes that appeared in both the data sets. A χ² test (P = 1.46 × 10⁻²³) revealed that there were more than four times the expected number (based on random chance) of concordant changes. In contrast, the number of discordant changes was at or lower than that predicted by random chance alone.

*   Gadd45a was increased in Klf4CN conjunctiva, but decreased in Klf4CN cornea. All other discordant changes were decreased in Klf4CN conjunctiva, but increased in Klf4CN cornea.

Transcription Factors Affected in the Developing WT and the Klf4CN Conjunctiva

Families of transcription factors significantly affected during the postnatal development of the mouse conjunctiva between PN9 and PN20 are listed in Table 4, with the complete list of increased and decreased transcription factors provided in Supplementary Table S13. Prominent among them are the members of the forkhead box (Fox) and SRY-related HMG box (Sox), Ets, and Krüppel-like family members. Foxa3, Foxa1, Foxk1, Foxo3, and Foxp1 were increased, and Foxp2 was decreased during conjunctival development (Supplementary Table S13). Sox4, Sox11, and Sox12, members of the group C Sry-related HMG box proteins co-expressed in embryonic neuronal progenitors and in mesenchymal cells in many developing organs, ^68,69 were all decreased between PN9 and PN20 (Table 4; Supplementary Table S13). In contrast, Sox21, which counteracts the Sox1, Sox2, and Sox3 group required for stem-cell maintenance, ⁷⁰ was increased between PN14 and PN20, favoring differentiation. These changes in Sox family transcripts indicate a gradual tapering off of support for progenitor cell maintenance, promoting their differentiation into the stratified squamous epithelial cells. 

Table 4.

 

Most Significantly Affected Transcription Factor Families

Transcription Factor Family	Increased during Conjunctival Development	Decreased during Conjunctival Development
Krüppel-like factors	Klf4, Klf6, Klf9	Klf13, Klf7, Klf10, Gli2
Epithelial-specific Ets (ESE) factors	Spdef, Ehf, Elf3, Elf5	Ets1
Forkhead box family proteins	Foxa3, Foxa1, Foxk1, Foxo3, Foxp1	Foxp2
High-mobility group proteins	Hmgb1, Hmgb2	Hmg20b
Sox family members	Sox21	Sox4, Sox11, Sox12
Interferon regulatory factors	Irf1	Irf3

The epithelial-specific Ets (ESE) transcription factors Spdef, Elf3, Ehf, and Elf5, with critical roles in airway and intestinal epithelial development, ^{36 –38,71,72} were significantly increased during conjunctival development (Table 4; Supplementary Table S13). In contrast, Ets1 was the most significantly decreased transcription factor during conjunctival development. Transcripts encoding several transcription factors that favor epithelial–mesenchymal transition (EMT), ⁷³ including two Twist family members (Twist1 and Twist2), two ZEB family members (Zeb1 and Zeb2), and Snail family member Snai2/Slug were decreased during conjunctival development (Table 4; Supplementary Table S13). The concomitant increase of Klf4, an inhibitor of EMT, ⁷⁴ and suppression of the Wnt pathway (Supplementary Table S12), a promoter of EMT, ⁷⁵ is consistent with changes favoring mesenchymal-epithelial transition (MET) during conjunctival development. 

Developmental Expression Patterns of Selected Transcripts in the Conjunctiva

We examined the developmental expression of the protein disulfide isomerase Agr2 and transcription factors FoxA3, Spdef, and Mitf by in situ hybridization, using cryosections from embryonic day 16.5, PN1, PN9, and PN14. We included Agr2, Spdef, and FoxA3, in view of their known involvement in goblet cell development in other mucosal epithelia, and Mitf because it was the only transcription factor with a known function in eye development ⁷⁶ that was decreased from PN9 to PN14 during conjunctival development. Expression of these transcripts was largely confined to epithelial cells, and followed a similar trend, as detected by microarray analysis (Fig. 6). At E16.5 and PN1, these transcripts were expressed at low levels, more or less uniformly across the conjunctival epithelium. Postnatal expression of FoxA3 and Agr2 gradually increased from PN1 to PN9 and PN14, with a noticeable enrichment of the corresponding transcripts in the conjunctival fornix, consistent with the tendency of the mouse conjunctival goblet cells to cluster in this area (Fig. 6). Expression of Spdef also increased gradually from PN1 to PN9 and PN14. However, unlike FoxA3 and Agr2, there was no enrichment of Spdef transcripts in the conjunctival fornix at PN9 or PN14. Expression of Mitf reached a peak on PN9 and displayed relatively decreased expression at PN14, consistent with the microarray results. Sections probed with similarly labeled sense strand riboprobes served as negative controls in these experiments (Fig. 6). 

Figure 6.

 

Spatiotemporal expression patterns of selected developmentally regulated transcripts in the mouse conjunctiva. Expression of Foxa3, Agr2, Spdef, and Mitf was examined by in situ hybridization at E16.5, PN1, PN9, and PN14 (n = 3). Sections probed with the similarly labeled corresponding sense strand served as negative control for each gene. Arrow: the conjunctival fornix in each panel.

Canonical Pathways Affected in the Course of Conjunctival Goblet Cell Development

Pathway analysis of the increased and decreased transcripts between PN9, PN14, and PN20 (1588 panels, targeting 1152 unique characterized genes) identified 84 canonical pathways as significantly (P < 0.05) enriched in at least one group. Among these, 33 canonical pathways showed higher enrichment in at least one of the three analyses (P < 0.001) (Table 5). Functionally related pathways are indicated by similar color code in Table 5. Several pathways related to various forms of signaling, endocytosis (light green), mucosal immunity (light pink), neural development and function (light blue), and retinoid receptor activation (light gray) were significantly affected during conjunctival development. 

Table 5.

 

Canonical Pathways Affected in the Course of Development

 

 For clarity, P values are omitted if P > 0.05, although the number of pathway genes is included. Green: endocytosis-related pathways; blue: neuronal development–and/or function–related pathways; gray: retinoid receptor–related pathways; pink: mucosal immunity–related pathways; bold: pathways validated by BRB-ArrayTools.

Discussion

By comparing the WT conjunctival forniceal gene expression patterns at PN 9, PN14, and PN20, when goblet cells are absent, developing, and present, respectively, we have catalogued the gene expression changes accompanying conjunctival goblet cell development. Similar comparison of the gene expression patterns of PN14 WT and Klf4CN conjunctival fornices (with and without goblet cells, respectively) identified 139 and 57 conjunctival Klf4 target transcripts increased and decreased by more than twofold, respectively, in the Klf4CN compared to WT conjunctiva. By this approach, we have identified a large number of novel transcripts whose levels are modulated in the developing conjunctiva, suggesting that diverse pathways related to signaling, endocytosis, MET, and neural development and function are modulated at eyelid opening. 

Several pathways are significantly affected during postnatal conjunctival forniceal development (Table 5). Consistent with the appearance of mucin producing goblet cells during eye opening, O-glycan biosynthesis pathway components were significantly increased during conjunctival development. Endocytosis of various solutes and soluble macromolecules from the ocular surface seem to play a crucial role in the conjunctival epithelial homeostasis, judging by the number of endocytosis-related pathways whose component genes are increased during conjunctival development (Table 5). The relatively large macropinosomes provide an efficient route for nonselective endocytosis of solute macromolecules, facilitating antigen presentation by dendritic cells. Consistent with this, transcripts involved in dendritic cell maturation pathway, the Fcγ receptor–mediated phagocytosis pathway, and tight junction signaling were increased, lending support for development of the conjunctival mucosal immune system at eyelid opening (Table 5). Genes representing three different pathways related to retinoid signaling were increased (Table 5), suggesting a role for retinoids in conjunctival development and/or function. Increased expression of genes related to neural development and function (Table 5), although consistent with neural regulation of conjunctival epithelial function, ³² was puzzling, considering that no neuron cell bodies are present in the conjunctiva. These transcripts may represent those localized to the axons surrounding developing goblet cells, as described previously, ^32,77,78 or those expressed in conjunctival cells surrounding the axons. 

Transport across the conjunctiva plays a critical role in ocular surface homeostasis. ³² Our analysis has identified a large list of solute carrier family members whose expression is modulated during conjunctival development (Supplementary Table S11). The solute carrier gene superfamily comprises 55 families, out of which members of 18 different families were increased during conjunctival development, 5 of which were affected in the Klf4CN conjunctiva. Chief among them are members of the neurotransmitter transporter Slc6 family, consistent with the neural development– and function-related pathways modulated during conjunctival development, as described above. In addition, mitochondrial transporter Slc25 family members were significantly affected. The sodium/glucose co-transporter Slc5a1 increased 30-fold between PN9 and PN14, and decreased by approximately 4-fold in the Klf4CN conjunctiva, raising the possibility that it is directly regulated by Klf4 (Supplementary Table S11). 

Even though Notch, Math1, Foxa1, Foxa3, and neurogenin-3 are known to regulate colonic goblet cell development, ³¹ molecular mechanisms regulating conjunctival goblet cell development remain poorly understood. ^{79 –81} This report on developmental changes in gene expression in the mouse conjunctival fornix where goblet cells tend to congregate serves as a useful reference point for identifying novel therapeutic targets to modulate goblet cell densities in human diseases such as ocular cicatricial pemphigoid (OCP), dry eye, and Sjögren's syndrome, where goblet cell densities are reduced, or allergic conjunctivitis, where goblet cell densities are increased. Considering that the goblet cell densities are affected in gastrointestinal and respiratory epithelial disorders as well, this knowledge is expected to be useful, not only in the ocular surface but also in other mucosal epithelial tissues, where goblet cells play critical roles. 

By comparing the WT and Klf4CN PN14 conjunctival forniceal gene expression, we identified a list of transcription factors whose expression was affected in the Klf4CN conjunctiva lacking goblet cells (Table 4; Supplementary Table S13). Among them, factors such as FoxA1, FoxA3, and Spdef have been shown to be essential for goblet cell development in other mucosal epithelia, validating our approach. In addition, our analysis has identified other transcription factors such as Grhl3, Creb3l4, and Bnc1, hitherto not thought to be involved in conjunctival development. It is important to analyze their contributions in conjunctival development and function in the near future. 

Comparison of the conjunctival Klf4 target genes with those in the cornea yielded a small number of common genes, consistent with significant differences in the function of Klf4 in these two related tissues (Table 3). Cyclin-mediated cell cycle regulation, one of the most affected pathways in the cornea ⁶⁷ was not affected in the conjunctiva, suggesting that Klf4 plays a major role in the rapid renewal of epithelial cells in the cornea but not in the conjunctiva. On the other hand, glycosylation-related transcripts that were decreased in the Klf4CN conjunctiva were relatively unaffected in the Klf4CN cornea. ⁶⁷ Protein disulfide isomerase Agr2, significantly decreased in the Klf4CN conjunctiva, was increased in the cornea. ⁶⁷ These changes in expression may be an indirect outcome of the absence of mucin-producing goblet cells in the Klf4CN conjunctiva or a reflection of the direct involvement of Klf4 in glycoprotein biosynthesis in the conjunctiva, but not the cornea. 

Although the data provided here are valuable, some limitations remain. First, we focused on the mouse conjunctival fornix, where goblet cells tend to congregate. Thus, the data may not be representative of the events occurring in the palpebral or bulbar conjunctivae, away from the fornix. Second, LMD yielded a mixed population of cells consisting mostly of the conjunctival forniceal epithelial cells and a small but variable fraction of the underlying stromal cells, potentially contributing to some variation in the data. Finally, we have focused on the conjunctival gene expression changes at the level of transcripts. Although the changes in transcript levels detected by the current microarray analyses are important, they may not completely and accurately reflect the changes in the proteome, as has been demonstrated in the yeast. ⁸² Therefore, it is important to examine changes in the proteins of interest in further analyses. 

In summary, through microarray analysis, we have generated the first catalog of gene expression changes accompanying postnatal conjunctival forniceal development in the mouse. We have also identified several transcriptional regulators differentially expressed during conjunctival forniceal development. In addition, by identifying the Klf4 target genes in the mouse conjunctival fornix around the time of initial goblet cell differentiation, we have shown that Klf4 regulates goblet cell development directly, by controlling the expression of genes with critical roles in goblet cell physiology, and indirectly, by controlling the expression of other transcription factors known to influence goblet cell development, such as Spdef, FoxA1, and FoxA3. Thus, we now have a better understanding of the genetic network of transcription factors regulating conjunctival goblet cell development. We anticipate that the information in this report will provide new probes for studying conjunctival development, opening new avenues for understanding the complex genetic programs underlying conjunctival development and the ocular surface disorders associated with defective conjunctiva. 

Supplementary Materials

Text s01, DOC - Text s01, DOC 

Footnotes

 Supported by the National Eye Institute (NEI) K22 Career Development Award EY016875 (SKS), startup funds from the Department of Ophthalmology, Core Grant 5P30 EY08098-19 for vision research, Research to Prevent Blindness, the Eye and Ear Foundation Pittsburgh, and the NEI Intramural Research Program.

Footnotes

 Disclosure: D. Gupta, None; S.A.K. Harvey, None; N. Kaminski, None; S.K. Swamynathan, None

The authors thank Gloria Limetti and Cindy Stone (Balaban Laboratory, Department of Otolaryngology, University of Pittsburgh) for help with histology and Kira Lathrop (Imaging Core Module, Department of Ophthalmology, University of Pittsburgh) for help with microscopy.