Transcriptomic Analysis in a Model of Aniridia-Associated Keratopathy for Target Discovery and Evaluation of Duloxetine Therapy

Abstract

Purpose

Aniridia-associated keratopathy (AAK) leads to loss of corneal transparency because of epithelial, inflammatory, and pathological vascular changes. Here, we sought to understand this process at the transcriptomic level while evaluating an experimental pharmacotherapy for potential modulatory effects.

Method

17 Pax6^+/− Small-eye (Sey) heterozygous mice with p.Gly208* Pax6 mutation and 10 wild-type 129S1/SvImJ mice at four months of age were examined to identify dysregulated genes and pathways in established AAK. We next evaluated the potential efficacy of 10 µM duloxetine administered as eye drops twice daily for 90 days, assessing outcomes at the transcriptomic level via microarray and protein level with Western blot and immunostaining.

Results

Transcriptomic analysis of the cornea revealed enrichment of Ccl21 gene family members associated with lymphangiogenesis, along with upregulation of genes involved in inflammation, cell adhesion, differentiation, motility, and keratinization, and downregulation of drug metabolism with significantly dysregulated genes emerging as potential therapeutic targets, including Gpha2, Chrnb3, Epgn, Cnfn, kallikreins and inflammation mediators Il18r1 and classical complement factors. Duloxetine therapy failed to regress AAK in adult corneas; however, transcriptomic profiling indicated duloxetine suppressed inflammatory genes and promoted anti-inflammatory and protective activity while modulating drug metabolism, suggesting potential beneficial effects in the cornea.

Conclusions

Transcriptomics reveals multiple unexplored pathways and genes altered in the AAK mouse model. Although clinical results with duloxetine are promising, our current regimen and delivery method did not improve established disease. Duloxetine's therapeutic potential requires further study.

Congenital aniridia is a rare eye disorder caused principally by PAX6 gene mutations, leading to incomplete eye development and multiple ocular abnormalities such as iris and foveal hypoplasia, early-onset glaucoma, and cataract.¹ The most vision-threatening complication, however, is aniridia-associated keratopathy (AAK), a progressive deterioration of the cornea marked by limbal stem cell deficiency, inflammation, neovascularization, and loss of corneal transparency.² AAK worsens with age, and despite surgical interventions, it often results in blindness with poor treatment outcomes owing to the inevitable immune rejection of transplanted allogeneic cells or tissues.³ This highlights the importance of identifying targets for AAK that can be modulated pharmacologically to counteract the pathology. Recently we performed deep phenotyping of a new mouse model of AAK, a Pax6^+/− heterozygous small eye (Sey) mouse model on the 129S1/SvImJ background⁴ first developed by Hickmott et al.,⁵ which carries the canonical c.622G>T nonsense mutation in Pax6, resulting in a premature stop codon and p.Gly208* protein truncation. This mouse model closely resembles the slow onset and progression of human AAK, crucially providing a defined window for evaluating potential pharmacotherapies to address the underlying PAX6 deficiency. Importantly, the model also indicates at least partial preserved corneal limbal epithelial stem cells, even in advanced stages,⁴ making it a promising model for evaluating pharmacotherapies aimed at restoring stem cell function or PAX6 protein production to restore corneal epithelial integrity and thereby its transparency.

Duloxetine is a selective serotonin-norepinephrine reuptake inhibitor (SSNRI) approved by the United States Food and Drug Administration and commonly prescribed as an oral medication for the treatment of major depressive disorder, generalized anxiety disorder, fibromyalgia, neuropathic pain, and central sensitization. Its mechanism of action involves the inhibition of serotonin reuptake in synapses, increasing the neurotransmitter levels in the central nervous system, thus managing to alleviate symptoms associated with these conditions.⁶ It was recently shown that duloxetine efficiently rescued endogenous PAX6 gene expression in nonsense-mutated haploinsufficient limbal epithelial cells in vitro.^7,8 Additionally, short-term duloxetine treatment was shown to alter PAX6 expression in primary human limbal stromal cells from aniridia patients.⁹ Duloxetine has further shown anti-inflammatory effects in both in vitro and in vivo models of lipopolysaccharide-induced inflammation, where short-term in vivo treatment upregulated PAX6 expression in the cornea.^10,11 Another study showed increased norepinephrine levels in the tear fluid of rats administered with SSNRIs, which may play a protective role in ocular surface homeostasis. In this context, SSNRIs may counteract elevated serotonin levels, thereby reducing the expression of inflammatory factors at the ocular surface.¹² Additionally, norepinephrine promotes cell activity and proliferation, further contributing to ocular surface protection.¹²

Recognizing the need to identify novel molecular targets and develop targeted therapeutic approaches for AAK, a condition in which inflammatory processes play an important role,¹³ this study primarily aimed to identify dysregulated transcripts and pathways in a mouse model mimicking the development of human AAK. Additionally, given the promising in vitro and in vivo evidence supporting the potential therapeutic effects of duloxetine and considering that most patients in need of effective therapy are adults with advanced stages of AAK, we sought to evaluate whether duloxetine administration could restore corneal transparency after establishment of progressive AAK in this AAK mouse model.

Material and Methods

Animal Model

In this study, we used a mouse model of slowly progressing AAK that we have previously thoroughly characterized⁴; the heterozygous 129S1.Cg-Pax6Sey/Mmmh strain on the 129S1/SvImJ background (MMRRC stock no. 050624-MU).^4,5 The Pax6 mutation of this model is a spontaneous point mutation in exon 8 of the Pax6 gene (G to T substitution), leading to a premature stop codon. Breeding and genotyping were conducted at Linköping University to obtain homozygous wild-type (Pax6^+/+) mice (Wt) and heterozygous mutant (Pax6^+/−) mice (hereafter referred to as Het mice). For colony maintenance, Pax6^+/+ females were bred with Pax6^+/− males. Mice were housed under controlled conditions (23°C, 40%–60% humidity, 12-hour light/dark cycle). To determine the mutational status of offspring in the mouse colony and allocate them to appropriate study groups, PCR was conducted on genomic DNA extracted from ear clip tissue collected at the time of weaning as previously described.⁴ All experiments strictly adhered to the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research, followed institutional and national ethical guidelines and were performed after obtaining ethical approvals from the Linköping regional animal ethics committee (Protocol no. 10940-2021 and 17409-2023).

Duloxetine Preparation and Administration

A 10 µM concentration of duloxetine was selected based on prior assessments of its toxicity and efficacy in vitro and in vivo.^8,11 A topical eye drop formulation was prepared using a viscous formulation of 2 mM duloxetine (chlorhydrate salt, SML0474; Sigma-Aldrich Corp., St. Louis, MO, USA), 0.75 % hydroxypropyl cellulose (Inresa Medical, Bartenheim, France), 1% Tween 80 (Cooper, Melun, France), 0.9 % NaCl pH = 7.0, and diluted to the lower concentration of 10 µM with the identical vehicle formulation. The viscous formulation was dispensed into eyedropper vials under sterile conditions. Twenty-seven mice, comprising 17 Het (seven female and 10 male) and 10 Wt (six male and four female) mice aged four months old, were used in the experiment. Previously we reported that there is no significant sex difference in phenotypic characteristics.⁴ The corneal phenotypes in the het mice ranged from early-stage AAK (Grades 1-2, 22 eyes) to late-stage AAK (Grades 3-4, 12 eyes) at the start of the experiment, according to our previously published AAK mouse model grading scale,⁴ thus representing the variable expression of AAK in humans. Nine Het mice with 11 eyes graded as early-stage AAK and seven eyes with late-stage AAK were administered duloxetine (Het-dul) and eight Het mice with 11 eyes having early-stage AAK and five eyes having late-stage AAK were given vehicle only (Het-veh). The administration regimen was twice daily to both eyes, and the experiment duration was 90 days. Wt mice were kept untouched as controls during the treatment period.

Characterization of Mice

Characterization of the corneal phenotype and AAK grading was performed on both eyes of each mouse before and after treatment, at the ages of four months (start of experiment) and seven months (end of experiment), using slit lamp photography (Micron III; Phoenix Technologies, Bowling Green, OH, USA) and in vivo confocal microscopy (IVCM) imaging of corneal layers (Heidelberg Retinal Tomograph 3 with Rostock Corneal Module, HRT3-RCM; Heidelberg Engineering, Heidelberg, Germany). At the end of the experiment, mice were euthanized for tissue extraction. Obtained tissues were dissected, processed, or preserved according to the specific assay for which they were individually allocated, as detailed below.

Immunofluorescence

Two mice per group were used for immunostaining. A standardized immunostaining protocol was used, including antigen retrieval, blocking, and antibody staining. Sections were blocked with BSA (Cat. no. A7906; Sigma-Aldrich Corp.) and goat serum (Cat. no. CST-5425S; Cell Signaling Technology, Danvers, MA, USA) to prevent nonspecific binding, incubated with primary antibodies overnight at 4°C in a humidified chamber and treated with secondary antibodies (Supplementary Table S3), for two hours at room temperature. Immunostained slides were mounted with SlowFade Diamond Antifade Mountant (cat. no. 536963; Thermo Fisher Scientific, Waltham, MA, USA) or quick-hardening media ProLong Gold Antifade Hardset Mountant (cat. no. P36930; Invitrogen, Carlsbad, CA, USA). Microscopy imaging was performed at the central cornea region. For whole mounts, corneas were sectioned, blocked, stained with primary antibodies overnight, and treated with fluorophore-conjugated secondary antibodies for two nights (Supplementary Table S4). Imaging was performed using an LSM 800 microscope with identical exposure and acquisition settings across samples (Zeiss, Oberkochen, Germany). Images were exported as eight-bit grayscale TIFF files for each channel. Mean gray value of defined regions of interest (ROIs) was quantified using ImageJ (version 1.54; FIJI Distribution). Three non-overlapping ROIs of equal size (100 × 100 µm) were selected per sample in representative cell-containing areas. For each ROI, the mean fluorescence intensity was measured, and values were used as one data point per ROI. The data were subsequently analyzed and graphed in Prism v.8.3.0 (GraphPad, San Diego, CA, USA).

Western Blot

Protein extraction was conducted by homogenizing extracted corneas with the TissueLyser LT bead mill system (Qiagen, Hilden, Germany) at 50 Hz for five minutes, followed by sonication in ice-cold water for 10 minutes and centrifugation at 12,000g for 20 minutes. Protein concentration was determined using the Pierce BCA Protein Assay kit (cat. no. 23225; Thermo Fisher Scientific) according to the manufacturer's protocol for a 10 µL sample volume. BIS-PAGE was performed using a 4%–15% gradient precast gel (4561084 Mini-PROTEAN TGX; Bio-Rad Laboratories, Ann Arbor, MI, USA), with 15 µg of total protein per sample mixed with 4× Laemmli buffer (1610747; Bio-Rad Laboratories), β-mercaptoethanol, and double-distilled water before loading. Proteins were transferred onto a PVDF membrane using Trans-Blot Turbo Mini Transfer Packs (1704156; Bio-Rad Laboratories), and standard immunostaining procedures were followed, with antibody dilutions specified in Supplemental Table S3. Bands were visualized using the ImageQuant LAS 500 Chemiluminescent Imaging System (GE Healthcare, now Cytiva, Marlborough, MA, USA) after incubation with Pierce ECL Western Blotting Substrate (32106; Thermo Fisher Scientific). Band densitometry analysis was performed using Image Lab v.6.1 software (Bio-Rad Laboratories).

Transcriptomic Analysis

RNA extraction was performed using the GeneJET RNA Purification Kit (K0731; Life Technologies, Carlsbad, CA, USA) according to the manufacturer's instructions, with a final elution volume of 20 µL. RNA concentration was measured using the NanoDrop One spectrophotometer (Thermo Fisher Scientific), whereas RNA quality was assessed with the Agilent 2100 Bioanalyzer and High Sensitivity RNA Analysis Kit (Agilent Technologies, Winooski, VT, USA), ensuring that all samples had a RIN >8.8. For high-throughput transcriptome analysis, the GeneChip WT Plus Reagent Kit (Thermo Fisher Scientific) was used in the MO 2.0 array system with 100 ng of total RNA per reaction, following the kit's protocol.

Bioinformatics Analysis

GeneChip scan analysis was performed using the Transcriptome Analysis Console (TAC v4.0.3.14; Applied Biosystems, Foster City, CA, USA) loaded with the MO 2.0 library. Besides the conversion of optical data into transcript quantification, TAC was used to perform Quality Control, principal component analysis, groupwise per-protein t-tests and WikiPathways pathway enrichment analysis. All raw data were exported into spreadsheets for further bioinformatics analyses with the STRING Protein- Protein Interaction Networks Functional Enrichment Analysis tool (STRING v.12.0), where the 150 most significantly differentially expressed genes (DEGs), with P < 0.05 as cutoff, were selected and used as inputs. For pathway enrichment analysis, the iPathwayGuide tool (Advaita Bioinformatics, Ann Arbor, MI, USA) and Database for Annotation, Visualization and Integrated Discovery (DAVID) web server^14,15 were also used, with a threshold of P < 0.05 and Log₂ fold change (FC) > 1.8 for Het versus Wt group and 1.4 for Het-dul versus Het-veh group.

Statistical Analysis

ImageJ software (version 1.54, FIJI distribution) was used for morphometric measurements (number of epithelial layers and number of basal cells per 100 µm linear distance) of histology images. Prism v.10.0.2 (GraphPad Software) was used for group comparisons. For quantitative data, the unpaired two-tailed t-test or two-way ANOVA was performed. Unless more stringent conditions are specified, a P value ≤0.05 was considered significant. In the transcriptomic analysis with TAC, a fold change >1.8 for Het-Veh versus Wt and a fold change >1.4 for Het-Dul versus Het-Veh and a t-test P < 0.05 was considered significant. In STRING pathway analysis, networks, functions, and pathways with a false discovery rate (FDR) < 0.05 were considered valid. The most significant biological processes were classified into five main categories, and their upregulation or downregulation was assessed based on the associated genes within each process. Manually selected representative and minimally overlapping instances of pathways of interest were chosen for illustration and discussion. For the iPathwayGuide analysis, significance was defined by a threshold of P < 0.05 combined with a Log₂FC > 1.4.

Results

Transcriptomics Reveals Dysregulation of Multiple Gene and Pathway Targets in AAK

To examine changes in the Het mouse cornea in detail at the gene expression level, whole-transcriptome microarrays were used to identify differentially expressed mRNAs in the cornea associated with AAK relative to healthy wild-type littermates. 365 genes were identified as DEGs in Het-Veh relative to Wt with P < 0.05 and >1.8-FC. Among these, 181 genes were downregulated, and 184 genes were upregulated in Het. Chrnb3 and Epgn were the most significantly downregulated genes with a fold change of −5.33 and −3.92, respectively, both being components of the cell surface receptor signaling pathway. Klk10 was the most significantly upregulated gene with a 6.75-fold increase and Cnfn, involved in epidermal cell differentiation, exhibited a 4.5-fold increase. The top 20 up- and down-regulated genes in this comparison are presented in Table 1, ordered by fold changes (Table 1; Supplementary Fig. S1A).

Table 1.

The 20 Most Down- and Up-Regulated Transcripts in Het Mice Compared to Wt, Ordered by Fold Changes

Gene Symbol	P Value	Fold Change
Upregulated in Het vs Wt
Klk10	8.42E-06	6.75
Cnfn	0.002	4.5
Cyp2b19	0.0003	3.84
Ankrd34c	0.0428	3.53
Krt4	1.70E-06	3.33
Ndrg2	1.77E-08	3.26
Clca3b	2.16E-05	3.26
Snord57; Nop56	2.17E-05	3.02
Chit1	5.31E-07	2.97
Klk11; Klk12	1.37E-05	2.96
Vsig10l	0.0004	2.92
Il18r1	0.0005	2.87
Pglyrp1	2.13E-08	2.74
Duox1	6.68E-06	2.73
Mir669a-3	0.0008	2.58
Ccl21c; Ccl21e; Ccl21b	0.0012	2.47
Ccl21a	0.0009	2.4
Krt13	1.45E-06	2.38
Lgals9	0.0115	2.36
Ccl21f; Ccl21d	0.001	2.35
Downregulated in Het vs Wt
Chrnb3	2.79E-09	−5.33
Epgn	4.42E-05	−3.92
Snora44	0.0026	−3.73
Popdc3	6.50E-05	−3.47
Raver2	4.12E-06	−3.06
H2afz	0.0115	−2.93
Ifi205	0.0035	−2.9
Gpha2	0.0007	−2.87
Eif1ad14	0.0273	−2.73
Serpina1c	0.0002	−2.55
St6gal1	1.25E-07	−2.54
Cyp3a25	0.0004	−2.54
Snord91a	0.003	−2.45
Cstdc6	0.0345	−2.39
Gkn1	3.80E-05	−2.38
Snord72	0.0027	−2.38
Ighv1-63	0.012	−2.35
Mir3473a	0.0004	−2.33
Vmn1r172	0.0056	−2.3
Tcaf2	4.04E-06	−2.25

Pathway enrichment analysis based on the transcriptomic data revealed significant alterations in specific biological pathways in Het mice relative to Wt, that were influenced by a small number of genes (2 DEGs) each. Among these, the complement activation classical pathway was significantly enriched in Het-Veh relative to wild-type corneas (P = 0.01), characterized by the upregulation of factors C2 and C4b. Additionally, the glucuronidation pathway was enriched in Het-Veh mice (P = 0.01), with two members of the Ugt2a family (Ugt2a1 and Ugt2a2) being downregulated (Supplementary Table S2).

Of the DEGs in Het mice relative to Wt mice, the 150 most significant were retained for further analysis using the STRING platform. Biological process enrichment analysis revealed 110 gene ontology (GO)-terms being significantly enriched (FDR < 0.05). Among these, mesangial cell-matrix adhesion (GO: 0035759), positive regulation of myeloid dendritic cell chemotaxis (GO:2000529), negative regulation of leukocyte tethering or rolling (GO:1903237), and positive regulation of cell adhesion mediated by integrin (GO:0033630) were the most enriched. To gain an overview of the biological alterations associated with AAK in Het mice, all enriched GO processes were grouped into six major categories. Categories with representative biological processes are summarised and color-coded in Figure 1.

Figure 1.

STRING pathway enrichment analysis of DEGs in Het mice relative to Wt mice. Visual clustering by node relocation was performed manually, to aggregate proteins of selected significant gene ontology process and to minimize edge (connecting line) intersections. Groups of enriched biological processes and molecular functions (gene ontology-based) were constructed manually (colored boxes) according to the STRING functional enrichment results, with the color serving as a visual differentiator. Connecting lines represent protein-protein associations. Pink: experimentally determined interactions. Teal: Interactions from curated databases according to the STRING platform. Green: Predicted interactions based on gene neighborhood. Red: Predicted interactions based on gene fusions. Blue: Gene interactions based on gene co-occurrence. Yellow: Interactions based on text mining. Black: Co-expression. Light blue: Protein homology.

The regulation of each category was evaluated based on the expression patterns of DEGs involved in the corresponding biological processes. The analysis revealed that inflammation, cell adhesion, differentiation, cell motility, and keratinization were upregulated in the corneas of Het mice, whereas drug metabolism was downregulated relative to Wt mice. Interestingly, the Ccl21 family, comprised of chemokines associated with lymphangiogenesis, were central to, and extensively involved in, the identified GO terms. The Reactome pathway “formation of the cornified envelope” from DAVID analysis further supported the marked upregulation of keratinization processes in Het mice compared with wild-type controls, whereas “developmental biology” was also dysregulated in Het mice (Supplementary Table S3).

Administration of Topical Duloxetine Did Not Regress Established AAK at the Tissue Level

Slit lamp and IVCM observations confirmed that in both Het-dul and Het-veh groups, late-stage AAK eyes remained late-stage after treatment, whereas eyes with early-stage AAK remained in the early-stage after treatment (Fig. 2A; Supplementary Table S1). Late-stage AAK was characterized by blood vessel invasion into the central cornea whereas IVCM revealed that lymph vessels were present in both early- and late-stage AAK, a trait we have described previously.⁴ In the duloxetine treatment group, comparison of AAK grade before and after duloxetine treatment detected no significant difference in AAK grade (no regression or progression, P = 0.21; Fig. 2B; Supplementary Table S1). In the vehicle-treated group, there was similarly no change in AAK grade from pre- to post-treatment (P = 0.16, Fig. 2C), indicating that the AAK grade was generally established by four months of age and did not change thereafter. The change in AAK grade after 90 days of treatment (at seven months of age) was not significantly different between duloxetine or vehicle-treated groups (Fig. 2D; P > 0.9).

Figure 2.

Representative IVCM and slit lamp images of Wt and Het mouse eyes. (A) The top row illustrates the corneal stability in Wt mice over a 90-day period. In Het mice, a 90-day duloxetine treatment (second to fourth row) did not alter the progression of AAK compared to vehicle-treated Het mice with a similar AAK grade (Fifth to seventh row), where the disease was stable or progressed naturally. In later stages of AAK, blood vessels (black arrows) were visible in the central cornea. In both early- and late-stage AAK in both groups, lymph vessels were visible in the central cornea (white arrows). (B) Comparison of AAK grade before and after duloxetine treatment (P = 0.21, n = 18). (C) Comparison of AAK grade before and after vehicle treatment (P = 0.16, n = 16). (D) Change in AAK grade after 90 days of treatment between duloxetine or vehicle-treated groups (P > 0.9, n > 16).

Duloxetine Eye Drop Therapy Did Not Affect Corneal Epithelial Structure, PAX6 or Stem Cell Marker Expression

Hematoxylin and eosin (H&E) images and quantitative analysis of corneas from Wt and Het mice revealed that treatment with duloxetine did not lead to any observable improvement in epithelial cellular organization, number of epithelial layers in Het mice compared to those treated with vehicle (Fig. 3). No epithelial basement membrane could be detected in corneas from Het mice, regardless of treatment. Similarly, immunostaining for PAX6, a critical transcription factor for corneal epithelial maintenance, revealed no change in expression in Het mice treated with 10 µM of duloxetine relative to the vehicle-treated group, with PAX6 confined to the basal epithelial layer in Het mice but present throughout the epithelium in Wt mice. Furthermore, immunohistochemical analyses of P63 and GPHA2, key markers of limbal stem cell populations, showed that duloxetine failed to enhance the expression of these stem cell markers in Het mice. P63 expression remained in a single layer of basal epithelial cells in Het mice regardless of treatment, whereas GPHA2 expression in Het mice shifted from the basal to superficial epithelium similarly in both het groups. Ki-67 expression indicated an active and proliferative epithelium in Het mice (but not in Wt) that was not rescued by duloxetine. Additionally, expression of MUC5AC, a marker of conjunctivalization, was present in the epithelium of both duloxetine and vehicle-treated groups, however in the duloxetine-treated corneas the expression shifted from the basal to more superficial epithelial layers (Fig. 3). Quantitative analysis of fluorescence intensity in immunostained sections confirmed the lack of difference in duloxetine-treated from vehicle-treated corneas.

Figure 3.

(A) H&E staining and immunolocalization of PAX6, P63, GPHA2, Ki-67 and MUC5AC in the central corneal zone from Wt and Het mice treated with duloxetine and vehicle showed no difference in marker expression or histology after 90-day duloxetine treatment. (First row) H&E images from Wt and Het mice. White arrows indicate basement membrane and black arrows blood vessels in the corneal stroma of Het mice. (Second row) Reduced PAX6 expression in Het mice remained after treatment. (Third row) P63 expression was reduced in Het mice. (Fourth row) GPHA2 expression shifted from basal in Wt to superficial in Het mice. (Fifth row) Ki-67 indicated basal proliferation in Het mice but not in Wt. (Sixth row) MUC5AC expression demonstrated lack of rescue of conjunctivalization in Het mice, but duloxetine shifted expression from the basal to superficial epithelial layer. (B) The relative fluorescent intensity was measured using ImageJ. ****P < 0.0001. n = 6 from two biological replicates per group. (C) Number of basal cells per 100 µm linear distance, n = 2. (D) Number of epithelial layers, n = 6 from two biological replicates per group. Labeled regions denote tissue compartments: E, epithelium; S, stroma.

Duloxetine Treatment Did Not Ameliorate Corneal Neovascularization in Established AAK

Immunofluorescence of corneal flat mounts stained with CD31, a marker for blood vessels, and LYVE-1, a marker for lymphatic vessels, confirmed in vivo imaging results, showing no evidence of vascular regression, blood or lymphatic, after 90 days of duloxetine treatment in Het mice across all AAK grades. (Fig. 4). These results indicate that the applied 90-day course of duloxetine treatment applied was ineffective in regressing the pathological vascularization associated with AAK.

Figure 4.

Whole-mounted corneas with immunofluorescence staining for blood vessels (CD31, top row), and lymphatic vessels (LYVE-1, bottom row). Neither Het-Dul nor Het-Veh exhibited regression of blood or lymph vessels.

Duloxetine Modulates Homeostasis, Wound Healing, and Immune Pathways at the Transcriptomic Level

Comparing Het-Dul with Het-veh, 253 genes were differentially expressed with p < 0.05 and >1.4-fold change. Among these, 20% were downregulated and 80% were upregulated by duloxetine treatment. Crisp1 and Ltf were the most significantly upregulated genes, with 4.45- and 3.86-fold increases, respectively. Conversely, Ankrd34c and Igkv5-43 were the most significantly downregulated genes, with −4.28- and −2.33-fold change, respectively. The top 20 differentially expressed genes between Het-Dul and Het-veh are summarized in Table 2, ordered by fold changes (Table 2; Supplementary Fig. S1B).

Table 2.

The 20 Most Down- and Upregulated Transcripts in Het Mice Treated With Duloxetine Compared to Het Mice Treated With Vehicle, Ordered by Fold Changes

Gene Symbol	P Value	Fold Change
Upregulated in Het-Dul vs Het-Veh
Crisp1	0.0012	4.45
Ltf	0.0009	3.86
Rnu1b6	0.0014	3.5
Rnu1b1; Rnu1b2	0.0014	3.5
Muc5ac	0.0112	2.8
Igkv6-17	0.0031	2.64
Snora74a	0.0028	2.5
Retnla	0.0177	2.43
Gp2	0.011	2.41
Cxcl17	0.0026	2.34
Alox12e	0.0035	2.27
Clmn	4.26E-06	2.11
Tfcp2l1	0.0022	2.04
Igkv6-13	0.0082	2.04
Snora5c; Tbrg4	0.0354	1.94
Krt18	7,90E-06	1.92
Alox15	0.0473	1.91
Ptx4	0.012	1.88
Slc15a2	0.0011	1.87
mt-Ts1	0.0003	1.85
Down-regulated in Het-Dul vs Het-Veh
Ankrd34c	0.0087	−4.28
Igkv5-43	0.0143	−2.33
Phf11	0.0023	−1.91
Napepld	0.0351	−1.88
Hhip	0.0117	−1.83
Cyp24a1	0.0168	−1.69
Olfr657	0.0162	−1.66
Dsg1c	0.0122	−1.65
Speer1j	0.0059	−1.62
Hist1h2bh	0.0063	−1.58
Mir669a-3	0.0106	−1.56
Prr23a3	0.0296	−1.51
Ttc30a2	0.0149	−1.49
Senp8	0.0088	−1.48
Mir501	0.0012	−1.46
Fam25c	0.0229	−1.46
Igkv1-122	0.035	−1.46
Cyp4f16	0.0226	−1.45
Prss2	0.0278	−1.45
Klk8	0.0184	−1.43

At the pathway level, treatment with duloxetine induced significant enrichment in the blood clotting cascade (P = 0.0098), a critical pathway for hemostasis and wound healing, as well as “metapathway biotransformation” (P = 0.024) and oxidation by cytochrome P450 enzymes (P = 0.036). The latter two pathways are pivotal in the metabolism of xenobiotics, indicating duloxetine's potential impact on metabolic and detoxification processes (Supplementary Table S2).

To further investigate the potential effects of duloxetine treatment on biological processes in the AAK mouse model, a comprehensive pathway and functional enrichment analysis was conducted with STRING; however, no significant biological processes were identified. Given this limitation, we used iPathway analysis, in which four significant (FDR < 0.05) and biologically meaningful processes were identified (Fig. 5A). The analysis revealed that negative regulation of multicellular organismal process (GO: 0051241) was negatively regulated whereas immune system response (GO:0002376), cellular development process (GO:0048869) and cell differentiation (GO:0030154) were positively regulated. The dysregulated genes involved in each biological process are summarized quantitatively in Figure 5B. From KEGG pathway analysis, “Spliceosome” was enriched in the duloxetine-treated corneas (Supplementary Table S3).

Figure 5.

Pathway enrichment analysis and enrichment of significantly differentially regulated genes in Het mice treated with duloxetine relative to Het mice treated with vehicle. (A) Correlation between the core network and the top four significantly abundant gene biological processes differentiated by color. (B) Bar graphs indicating individual DEGs for the most significant biological processes. The perturbation is represented with negative values in blue and positive values in red. The box plot on the left summarizes the overall gene perturbations in each biological process. The box represents the first quartile, the median and the third quartile, whereas circles represent the outliers.

PAX6 Expression Levels Remain Unaltered After Duloxetine Administration In Vivo

Given the central role of PAX6 in aniridia-associated keratopathy (AAK), we performed both Western blot analysis and transcriptomic evaluation to assess the impact of duloxetine treatment on PAX6 expression. However, neither the protein levels nor the transcript levels of PAX6 in the corneal tissue where significantly different between treatment groups after the 90-day treatment (Fig. 6). These findings indicate that the duloxetine regimen given was unable to rescue or enhance PAX6 expression in the treated mice.

Figure 6.

Quantification of in vivo PAX6 expression levels in the cornea. (A) PAX6 protein levels from Western blot in mouse corneas after 90 days of topical duloxetine administration (n > 5) with representative bands and corresponding β-actin normalization bands. (B) Transcriptomic expression levels of Pax6 gene assayed by microarray (n > 3).

Discussion

We used a Pax6^+/– Sey mouse model on the 129S1/SvImJ background, a model that mimics the slow onset and progression of AAK,^4,5 to examine biological processes regulated by PAX6 haploinsufficiency in the cornea at the transcriptomic level, and to investigate whether 90 days of twice daily topical duloxetine treatment could regress the already established AAK.

Differential whole-transcriptome analysis identified key genes and processes in AAK corneas relative to wild-type littermate corneas that may improve our understanding of the pathogenesis of AAK and provide potential therapeutic targets of interest. The previously reported selective and persistent lymphangiogenesis in this mouse model⁴ was in this study found to be consistent with the high prevalence of Ccl21 family members in enriched biological processes in AAK, including Ccl21a, Ccl21a-2, Ccl21b, Ccl21d, Ccl21e, and Ccl21f. Chemokine C-C motif ligand 21 (CCL21) is a chemokine primarily involved in guiding the migration of immune cells, such as T cells and dendritic cells, to secondary lymphoid organs.¹⁶ Beyond its role in immune cell trafficking, CCL21 has been implicated in lymphangiogenesis through enhancing the expression and secretion of vascular endothelial growth factor-C (VEGF-C), a key lymphangiogenic factor.¹⁷ CCL21 or VEGF-C may therefore represent potential targets for antigen presentation and lymphangiogenesis, features also observed in clinical observations of patients with congenital aniridia, who exhibit an invasion of corneal lymph vessels and heightened dendritic and other immune cells.^18,19

Other candidate genes implicated in AAK were also identified, such as Gpha2, a putative limbal stem cell marker,^4,20 which was one of the most downregulated genes in AAK (−2.87-fold, P = 0.0007). Previously, we have showed a pathological expression of GPHA2 at the protein level in Het mice,⁴ consistent with the present findings. Here, we further confirm GPHA2 as a target of interest and putative marker for limbal stem cell deficiency, warranting closer investigation for its potential role in AAK pathogenesis.

Cholinergic receptor nicotinic beta 3 subunit (Chrnb3) was the most downregulated gene in AAK (−5.33-fold, P = 2.79 × 10⁻⁹). This gene encodes, a member of the nicotinic acetylcholine receptor family, which plays a key role in synaptic transmission within the nervous system²¹ and is also implicated in regulating inflammation and tissue repair.²² Although its function within the cornea remains largely unexplored, its known functions in neuroprotection and anti-inflammatory processes suggest that Chrnb3 may influence critical pathways involved in cell survival in AAK. This hypothesis aligns with our observations of nerve disorganization in the cornea of heterozygous (Het) mice, where we identified a loss of stromal/epithelial compartmentalization of nerves.⁴

The second most suppressed gene in AAK (−3.92-fold, P = 4.42 × 10⁻⁵) was epithelial mitogen (Epgn), also known as epigen, a member of the epidermal growth factor (EGF) family. The role of epigen in the cornea is not well studied, but given abundant epithelial expression of EGFR-triggering signaling pathways for cell migration, proliferation, and differentiation essential for rapid epithelial repair following injury,²³ downregulation of Epgn in AAK could lead to impaired wound healing.

Raver2, also downregulated in AAK, stands out as noteworthy. Our finding align with those of Das et al.,²⁴ who reported that RAVER2 expression parallels that of PAX6, suggesting its evolutionarily conserved role in corneal development, which suggests that RAVER2 deficiency may contribute to, or mirror, AAK pathogenesis. Interestingly, Xiao and colleagues²⁵ recently demonstrated that PAX6 in corneal epithelial cells directly binds to the promoter region of Raver2, regulating the Raver2/sVEGFR-1 (sFlt-1) axis, a key mechanism by which corneal epithelial cells maintain corneal avascularity.²⁶ Our finding thus highlights the Pax6/Raver2/sFlt-1 axis as a potential therapeutic target for mitigating the corneal neovascularization (CoNV) in AAK.

Among the most upregulated genes in AAK, three members of the kallikrein (KLK) family, including Klk10, Klk11, and Klk12, were identified. Although the role of the KLK family in the cornea remains largely unexplored, their involvement in chronic inflammation has been documented.²⁷ Interestingly, it has been shown that in atopic dermatitis, overexpression of KLKs disrupts the normal skin barrier.²⁸ Therefore the upregulation of the KLK family in Het mice may be linked to the keratopathy, inflammation and impaired epithelial barrier function observed in our mouse model.⁴

Cornifelin (Cnfn) was also among the most upregulated genes in corneas with AAK. Cnfn expression is typically low in uninjured tissue but increases following epithelial damage,²⁹ playing crucial roles in epithelial differentiation, structural integrity, barrier function, and cell adhesion.³⁰ Interestingly, CNFN was upregulated in an in vitro model of the human corneal epithelium when exposed to irritants such as Triton X-100.³¹ Given this evidence, the upregulation of Cnfn in Het mice may reflect a compensatory response to keratopathy-associated epithelial damage, aimed at maintaining cell adhesion and barrier function.

Notably, Il18r1 also exhibited significant upregulation in the corneas of Het mice. IL-18R1 plays a vital role in immune activation by triggering the MAPK pathway.³² Beyond its proinflammatory functions, Il-18 has been implicated in choroidal angiogenesis regulation, exerting antiangiogenic and anti-permeability effects.³³ Given these roles, the elevated expression of IL-18R1 may contribute to inflammatory and vascular activity associated with AAK, as well as reduced PAX6 levels through stimulation of MAPK.

Pathway enrichment analysis revealed activation of the classical complement pathway, with C2, C4b, C1qb, and C3ar1 being significantly upregulated in AAK. Although complement activation enhances the early innate immune response and aids in maintaining homeostasis, insufficient regulation of this process can result in unintended host tissue damage.^34,35 The involvement of complement factors in VEGF upregulation and the increased risk of choroidal neovascularization inAMDpatients has been previously documented.³⁶ Moreover, complement-targeted therapies have been explored as potential treatment strategies for AMD.³⁷ Given the chronic inflammation in AAK, the classical complement pathway represents an interesting target.

When evaluating the therapeutic potential of duloxetine in our AAK model, the absence of overt signs of irritation, ocular redness, or behavioral changes suggest that the drug and its viscous eye drop formulation was well tolerated without inducing toxicity for the dosage and duration used in this study. Phenotypic assessment revealed no observable improvement in keratopathy or regression of vascularization relative to vehicle treatment. Progression or stability of AAK grade is attributed to the inherent characteristics of the mouse model, where not all Het mice develop advanced AAK.⁴

Duloxetine did not show a discernible effect on corneal tissue structure or expression of epithelial proliferation (Ki-67), vascular (CD-31 and LYVE-1), or limbal stem cell markers (P63 and GPHA2). MUC5AC, however, was overexpressed with duloxetine treatment, which was corroborated by transcriptomic analysis indicating upregulation of Muc5ac (2.8-fold, P = 0.0112) and Gp2 (2.41-fold, P = 0.011). Although there is little known concerning GP2 in the cornea, goblet cell populations in the conjunctiva expressing GP2 are reported to possibly contribute in the defence against microorganisms and antigens.³⁸

Transcriptomic analysis revealed that duloxetine treatment induced a distinct gene expression profile in the cornea, marked by downregulation of immune-related genes including Igkv5-43 having a role in adaptive immune response³⁹ and Napepld involved in inflammation and pain regulation,⁴⁰ and upregulation of anti-inflammatory and protective genes like Ltf and Retnla. Lactotransferrin for instance, maintains corneal health through its antimicrobial and anti-inflammatory properties.⁴¹ Dysregulated DEGs by duloxetine mainly suppressed immune-related responses, potentially leading to anti-inflammatory effects in the cornea, consistent with previously reported anti-inflammatory effects of duloxetine.¹¹

From a molecular therapeutic perspective, the anatomical and physiological barriers of the eye pose significant challenges in maintaining effective and sustained drug concentrations. Our analysis revealed widespread dysregulation across all three phases of drug metabolism in the corneas of Het mice. Specifically, Cyp family genes involved in phase I (oxidation via cytochrome P450 enzymes) were significantly dysregulated, whereas phase II (conjugation via UDP-glucuronosyltransferases or UGTs, and glutathione-s-transferases alpha or Gsta genes)⁴² and phase III (excretion via ABC transporters and mitochondrial ribosomal proteins) processes^43–46 were downregulated (factors such as ABCC, MRP, and ABCG2).

This suggests impaired detoxification, metabolic processing of xenobiotics, and cellular membrane integrity, as well as impaired cellular defense to oxidative stress in Het mice; however, treatment with duloxetine appeared to partially restore this metabolic imbalance. Notably, duloxetine modulated key enzymes in the Metapathway biotransformation, cytochrome P450 oxidation pathways, and Mgst1 (Microsomal Gst1)⁴⁷ central to detoxification, metabolism of bioactive molecules, and inflammation control in the cornea.

Importantly, Fmo2, a marker of limbal stem cells, was found to be dysregulated in untreated Het mice, but its expression was further upregulated by duloxetine. This is significant, because Fmo2 levels typically decrease in corneal injury,⁴⁸ and their restoration may suggest a role for duloxetine in preserving the limbal stem cell population. In contrast, however, Gpha2, another limbal epithelial marker that is downregulated in Het mice, was not restored but remained unchanged after duloxetine treatment.

Conclusions

In summary, duloxetine, administered using the current regimen and delivery method, was unable to enhance PAX6 expression at the transcriptomic or protein level in the cornea, or rescue the established AAK. The age of mice in this study corresponds to human ages older than 40–50,⁴ whereas it has been noted that in human studies a potential window for treatment exists preferentially during childhood.³⁸ However, a recent case series by Avila and colleagues⁴⁹ reported on six aniridia patients who received a high 6 mM concentration of duloxetine as eye drops over a 12-month period. Although clinically evident reduction in corneal keratinization was observed in some cases, the treatment's efficacy was variable.⁴⁹ Thus, although duloxetine has shown promising results in vitro, in vivo, and in first clinical use,^8,11 its efficacy in treating established AAK in adult corneas requires further research. Higher drug concentrations may be needed, as well as treatment for a prolonged period. Our results also motivate the need for further studies in younger mice, ideally initiating treatment during early postnatal stages following eyelid opening up to a starting age of 1 month, when disease-associated processes are not yet fully established. This suggests adjustment of treatment duration, higher duloxetine concentration, or alternative administration route may be necessary to achieve observable changes. Despite this, transcriptomic analysis nonetheless indicated multiple anti-inflammatory and immune-mediating effects of duloxetine as a topical pharmacotherapy in the cornea and the results here represent valuable information for future studies and other conditions. Indeed, duloxetine has been given to patients with chronic neuropathic pain at the ocular surface, without a detailed knowledge of its effects in the cornea.⁵⁰ Moreover, the transcriptomic results provided here shed further light on the multiple pathways, processes and genes that are perturbed in the mouse model of AAK and to date remain unexplored in the cornea.