RNA-Seq Study of Human Lens Epithelial Cells: Differentially Expressed Genes and Pathways in Steroid, Uveitic, Post-Vitrectomy, and Senile Cataracts

Abstract

Purpose: Secondary causes of cataract contribute to significant morbidity, but their pathogeneses are not well understood. This RNA sequencing study aimed to be the first to quantify and compare the transcriptome of the uveitic, steroid-induced, and post-vitrectomy cataract, using age-related cataracts (ARCs) as the study control.

Methods: Between March and July 2023 in Melbourne (VIC, Australia), human anterior lens capsules were prospectively collected during surgery from ARCs (n = 36), as well as steroid-induced (n = 23), uveitic (n = 25), and post-vitrectomy (n = 13) cataracts, and they were stabilized in RNAlater reagent. The Australian Genome Research Facility performed RNA isolation with RNeasy Mini and library preparation and sequencing using the Illumina workflow. Quality control was performed with the Agilent 2200 TapeStation. Bioinformatic analysis of RNA sequencing data identified differentially expressed genes (DEGs), defined as those with a log fold change ≥ 1 and false discovery rate (FDR) < 0.05.

Results: Differential gene expression analysis demonstrated significant differences between the transcriptome of age-related versus uveitic cataract (345 DEGs), steroid-induced versus uveitic cataract (117 DEGs), and age-related versus post-vitrectomy cataract (30 DEGs in the subgroup without removal of silicone oil [ROSO] and 1347 DEGs in the subgroup with ROSO). No DEGs were identified between age-related and steroid-induced cataracts.

Conclusions: To our knowledge, this is the first large-scale gene expression study focusing on these secondary cataracts. This dataset will assist in forming a broader knowledge base of secondary cataract pathogenesis and inform future research in this area, particularly in the selection of specific genes and investigating their impact on cataract development through animal model studies.

Cataract is a leading cause of blindness worldwide.¹ Age-related cataract (ARC), also known as senile cataract, is the most common form, occurring as a result of decreased antioxidants and water-soluble crystallins in the lens.^2–4 Cataract is also linked to various other conditions, of which there is limited understanding of the pathogenesis. Uveitic and steroid-induced cataracts have particularly significant morbidity, and the reason for acceleration or subsequent formation of cataracts in these groups remains ill understood.^5,6 Post-vitrectomy cataracts are another clinically observed phenomenon.⁷ It is postulated that increased exposure of the lens to oxygen leads to increased oxidative stress and nuclear sclerotic cataract, but existing literature has yet to identify the pathogenesis.^7–10 

This study used RNA-sequencing of ex vivo human lens epithelial cells (HLECs) to quantify and compare the transcriptome of three types of secondary cataracts (steroid-induced, uveitic, and post-vitrectomy cataracts) and ARCs. Comparisons were made between each of the secondary cataracts and ARCs and between the steroid and uveitic cataract groups. Analysis of these differential profiles may provide insight into the pathogenesis of these secondary cataracts. 

Methods

Between March and July 2023, a cohort of eligible patients undergoing routine phacoemulsification surgery with anterior capsulorhexis was prospectively and consecutively recruited to donate explanted anterior capsule tissue. This study complied with the tenets of the Declaration of Helsinki and received approval from the Alfred Human Research Ethics Committee (project 53905). The human tissue experiments complied with the guidelines of the ARVO Best Practices for Using Human Eye Tissue in Research. Informed consent was obtained from the participants prior to sample collection. 

Patients were recruited from the Royal Victorian Eye and Ear Hospital, Alfred Hospital, Sandringham Hospital, Victoria Parade Surgery Centre, and Cabrini Malvern Hospital, all located in Victoria, Australia. Collection of anterior capsule samples took place at the time of cataract surgery. Twenty surgeons and one collector participated in extracting and transferring the samples under sterile conditions. The anterior capsulorhexis was performed in the standard manner as part of routine phacoemulsification cataract surgery. The anterior capsule disc and associated lens epithelial cells were removed from the anterior chamber, blood and viscoelastic were washed off if possible, and the specimen was passed to the collector. Immediately upon collection, the anterior capsule samples were stabilized in RNAlater (Thermo Fisher Scientific, Waltham, MA, USA) in RNase and DNase-free Eppendorf tubes (Thermo Fisher Scientific). Stabilized samples were transferred within 7 days to the Australian Centre for Blood Diseases in Alfred Hospital for storage at −80°C. 

Cataract phenotype and relevant clinical history were identified prior to surgery from electronic and paper health records, including Lens Opacification Classification System (LOCS) III, age, sex, smoking and alcohol status, medical history, surgical history, and current medication use.¹¹ For patients using glucocorticoids, route of administration was also recorded. Intraoperative findings and procedures were recorded post-surgery. 

Steroid-induced cataracts were diagnosed in patients with a documented history of substantial corticosteroid usage, without any prior history of uveitis affecting the operated eye, and with a predominant posterior subcapsular (PSC) lens phenotype, which is characteristic of steroid-induced cataract.^5,12 Uveitic cataracts were observed in patients with a confirmed history of non-infectious or infectious uveitis, as per the criteria outlined in the Standardization of Uveitis Nomenclature.¹³ These patients were permitted to have received local or systemic steroid therapy as part of their uveitis management. 

Post-vitrectomy cataracts were identified in patients who had undergone pars plana vitrectomy at least 6 months prior to their scheduled cataract surgery. These patients may have presented solely with cataracts in the vitrectomized eye, or they may have demonstrated distinctly more severe cataracts in the vitrectomized eye compared to the non-vitrectomized eye. Ocular pathologies unrelated to the lens, including glaucoma, asteroid hyalosis, Sjögren's disease, and corneal surgery, were not excluding factors. Diabetes, with or without retinopathy, was not an excluding factor. Staining of the anterior capsule with VisionBlue (DORC, Zuidland, the Netherlands) and concurrent surgical procedures, such as the removal of silicone oil in the post-vitrectomy group, were not disqualifying features. 

ARCs were categorized into two subgroups. The ARCs without diabetes subgroup included patients 65 years or older with no history of glucocorticoid use and no history of ocular disease or ocular surgery in the study eye. The ARCs with diabetes subgroup otherwise followed the same criteria as above. 

RNA Sequencing

This study used whole transcriptome (Ribo-Zero depletion; Illumina, San Diego, CA, USA) bulk RNA sequencing (RNA-seq) to identify coding and non-coding transcripts. RNA extraction was conducted using the RNeasy Mini Kit (QIAGEN, Hilden, Germany) on a QIAGEN QIAcube Connect system at the Australian Genome Research Facility (AGRF) in Adelaide (SA, Australia). Extractions were processed in multiple batches given the large number of samples, following standard operating procedures to ensure equal treatment to all. Subsequently, the isolated RNA samples were sent to the AGRF in Melbourne (VIC, Australia) for quality assessment using the Agilent 2200 TapeStation system (Santa Clara, CA, USA). Samples with a DV200 (percentage of RNA fragments > 200 nucleotides) below 60% were excluded from further analysis. DV200 is comparable or superior to the traditional RNA integrity number (RIN) as a quality assessment metric in next-generation sequencing analyses and is suitable for low-input RNA.¹⁴ The concentration of isolated RNA was measured utilizing the QuantiFluor RNA system (Promega, Madison, WI, USA). 

Library preparation utilized the Illumina Stranded Total RNA Prep with Ribo-Zero Plus, employing ribosomal RNA depletion through hybridization capture followed by magnetic bead separation. Samples were prepared on two plates (one plate capable of accommodating a maximum of 96 samples) following standard operating procedures to ensure equal treatment to both batches. Library quality assessment was performed on the Agilent 2200 TapeStation system using the D1000 assay. Quantification of the libraries was achieved using the NEBNext Library Quant Kit (New England Biolabs, Ipswich, MA, USA). Sequencing was performed on the Illumina NovaSeq X Plus instrument with the Illumina NovaSeq X Series 10B Reagent Kit using a 2 × 150 bp run configuration. Image analysis was performed in real time by NovaSeq Control Software, version 1.1.0.18335, and Real Time Analysis, version 4.6.2. Primary analysis data were generated using the Illumina DRAGEN BCL Convert 07.021.645.4.0.3 pipeline. FASTQ files were uploaded to Monash University's Vault storage. Further analyses were performed by our biostatistician (JS) and followed the nf-core/rnaseq pipeline v3.8.^15,16 Pre-processing quality control was performed using the FastQC tool. Raw reads were aligned to the GRCh38 reference genome using STAR 2.7.11b alignment and Salmon quantification methods.^17,18 

Bioinformatics Analysis

Downstream analyses were carried out by two analysts (JS and SB), encompassing differential gene expression (DGE) analysis and Gene Ontology (GO) enrichment analysis. DGE analysis was conducted using Monash Bioinformatics Platform Degust software (coded with R; R Foundation for Statistical Computing, Vienna, Austria), using the edgeR quasi-likelihood test.¹⁹ Clinically relevant demographic parameters were recorded as part of data collection and examined to identify significant covariates. These included sex, smoking, alcohol intake, intraoperative use of VisionBlue, past ophthalmic history (including ocular surgery), stroke or transient ischemic attack, and diabetes. We adjusted for the following covariates in all DGE analyses. Gender and VisionBlue were included, as they exhibited spatial aggregation by sample group or significantly influenced gene expression when comparing models with and without them (Supplementary Table S6); diabetes was also included to account for potential differences in control samples with and without diabetes. There were no significant batch effects (RNA extraction batch, sequencing batch) on gene expression confirmed visually in principal component analysis plots (Supplementary Fig. S1) and statistically by including these as covariates in all analyses; the number of DEGs did not change with the inclusion of these batch effects. The main comparison groups included steroid cataracts versus ARCs, uveitic cataracts versus ARCs, steroid cataracts versus uveitic cataracts, and post-vitrectomy cataracts versus ARCs. Differentially expressed genes (DEGs) were defined with a false discovery rate (FDR) q-value less than 0.05 and absolute log fold change (|log₂FC| ≥ 1. Although this is a commonly used threshold in the literature, it is important to note that the log fold change cut-off merely identifies DEGs with higher magnitude differences of expression between groups and may miss genes with low expression but high biological impact.^20–22 GO enrichment analysis was performed using significant genes (FDR q < 0.05) against the full set of genes in the GO knowledgebase (http://geneontology.org).^23,24 No log₂FC cut-off was applied to maximize the number of genes included in the analysis, ensuring robust enrichment results. This approach also accounts for genes with smaller expression changes that may still contribute to biologically meaningful pathways. 

Results

Sample Characteristics

The study initially included capsulorhexis samples from 97 eyes for RNA isolation. DEGs and bioinformatic analyses were gathered from 89 samples (Table 1). One sample was excluded before library preparation and sequencing due to RNA degradation (DV200 = 27%) (Supplementary Table S2). Seven samples were excluded from the final bioinformatic analyses after a secondary review of health records revealed that they did not meet the inclusion criteria. This included four samples from the steroid cataract group that lacked the PSC lens phenotype associated with steroid cataracts (Supplementary Table S3); two samples from the non-diabetic ARC group, which showed male sex in the health records but demonstrated the female-specific gene (all other samples disclosed XIST expression profiles consistent with documented sex; see Supplementary Table S1); and one sample from the diabetic ARC group that had pseudoexfoliation syndrome, which could influence lens gene expression (Supplementary Table S2). Table 1 summarizes the patient characteristics for all samples included for the DGE and GO enrichment analyses. In the steroid cataract group (n = 23), the most common reason for glucocorticoid therapy was lung transplant (26.3%), followed by asthma (21.1%) (Table 2). In the uveitic cataract group, the majority of samples were collected from study eyes with anterior uveitis (68%) of non-infectious aetiology (76%) (Table 3). Intraoperative VisionBlue was used in 33.7% of patients (n = 30) (Table 1). Contemporaneous removal of silicone oil was performed in 38.5% of post-vitrectomy cataracts (n = 5) (Table 1). 

Table 1.

 

Baseline Demographic and Clinical Characteristics of Participants by Cataract Group

 

Table 2.

 

Indications for Glucocorticoid Therapy in the Steroid Cataract Group

 

Table 3.

 

Anatomical and Etiological Categories of Samples in the Uveitic Cataract Group

 

RNA Isolation and Sequencing

Total RNA was extracted from 96 samples. The mean total RNA extracted was 182.8 ng in the non-diabetic control group, 147.7 ng in the diabetic control group, 187.0 ng in the steroid-induced cataract group, 177.5 ng in the uveitic cataract group, and 192.5 ng in the post-vitrectomy cataract group (Table 4). Supplementary Tables S1 to S5 show the extraction results for each study group. Library preparation was performed with 80 ng of input and 13 PCR cycles per sample. 

Table 4.

 

Summary of Total RNA Extraction Results

 

Bioinformatic Analyses

A total of 89 samples underwent DGE analysis with statistical correction for significant covariates. In the comparisons of interest, no DEGs were identified between steroid cataracts and ARCs (Fig. 1); 369 DEGs were identified between uveitic cataracts and ARCs (Fig. 1B); 219 DEGs were identified between uveitic cataracts and steroid cataracts (Fig. 1C); 30 DEGs were identified between post-vitrectomy cataracts (without removal of silicone oil at time of cataract surgery) and age-related cataracts (Fig. 1D); and 1359 DEGs were identified between post-vitrectomy cataracts (with removal of silicone oil at time of cataract surgery) and ARCs (Fig. 1E). See Supplementary Tables S7 to S10 for full lists of the significantly up- and downregulated DEGs. For each comparison group, 10 potentially noteworthy positively or negatively enriched gene ontology pathways have been indicated in Figures 2A to 2D. In the uveitic cataract versus ARC group, 369 DEGs were identified: 305 upregulated DEGs and 64 downregulated DEGs. Upregulated DEGs include genes involved in immune response (e.g., HLA-DRA, CD44, GBP1). Downregulated DEGs include genes involved in regulation of apoptosis (e.g., TRIB2, SOSTDC1) and cell architecture (e.g., TMEM221, FRAS1, LAMB1). 

Figure 1.

 

(A–E) Comparison of gene expression profiles of all samples (A) with a principal component analysis plot and paired subgroups (B–E) with volcano plots displaying DEGs. DEGs were defined by FDR q < 0.05 and |log₂FC| ≥ 1. No DEGs were identified in the steroid-induced cataract versus ARCs groups. S, steroid cataract; U, uveitic cataract. For a full list of the DEGs, see Supplementary Tables S7 to S10.

Figure 2.

 

(A–D) GO enrichment results showed significant (FDR q < 0.05) positively enriched pathways (underlying DEGs were upregulated) or negatively enriched pathways (underlying DEGs were downregulated). Negatively enriched pathways are indicated by an asterisk. The size of the circles and numbers within the circles represent the number of DEGs that intersected genes in each GO term. The colors of the circles indicate FDR significance level.

Discussion

To our knowledge, this is the first study to identify DEGs among uveitic cataracts, steroid-induced cataracts, post-vitrectomy cataracts, and ARCs. Although the central aspect of the anterior lens capsule is readily accessible during routine cataract surgery and serves as the standard for RNA-seq studies of the lens, we acknowledge this may not represent the site of most prominent cataract pathology for our experimental groups. We were unable to obtain HLECs from the equatorial region of the lens capsule, where lens epithelial cell proliferation predominantly occurs in the germinative zone, nor did we collect posterior capsule samples, where the PSC lens phenotype manifests in steroid cataracts.^25–27 Genes with epithelial features including alpha crystallins (CRYAA and CRYAB), beta crystallins (e.g., CRYBB1, CRYBB2, CRYBA1), PAX6 (regulator of lens epithelial cell identity), FOXE3 (associated with lens epithelial proliferation), and PCNA (proliferative cell nuclear antigen) were detected consistently across all samples. It is likely that these genes would have greater expression in equatorial HLECs; however, without a control sample in this analysis it is difficult to comment. Our study demonstrates that central anterior HLECs appear to have differential gene expression in different types of cataract; however, there may be additional patterns that could be detected in equatorial cells. Our group attempted technical studies to collect HLECs from the posterior capsule, and the quality and quantity of the cells removed was low. Similarly, our methodology did not allow for the collection of lens fiber cell samples, where we would expect notable aberrations in gene expression for age-related and post-vitrectomy cataracts, known for their nuclear sclerotic lens phenotype. Additionally, assembling a cohort of capsulorhexis samples devoid of cataract pathology as controls proved unfeasible. 

Although we were able to demonstrate DGE in HLECs, this does not mean that the DEGs were necessarily responsible for the cataract. Many of the effects we demonstrated, such as increased expression of inflammation-associated genes in the uveitic cataracts, may be bystander effects unrelated to cataractogenesis. Trypan blue was analyzed as a covariate in the DGE analysis to account for its use in cases with reduced lens capsule visibility, which may correlate with cataract severity or surgical complexity. Significant gene expression differences were found between samples stained with Trypan blue and those that were not (Supplementary Table S6). It is possible that Trypan blue has a direct effect on lens gene expression, or the intraoperative use of Trypan blue may serve as a surrogate marker for specific cataract phenotype or morphology. These differences are worth investigating in the future and have been adjusted for in this study. Every effort was made to collect uncontaminated specimens during surgery. Although some contamination with blood or aqueous may have occurred, it was likely minimal, ensuring that the majority of RNA analyzed originated from lens epithelial cells. Although unlikely, it remains possible that some DEGs identified were derived from aqueous or blood. 

Finally, no power analysis was performed to determine the optimal sample size. However, this study exceeds the general recommendation of 10 to 12 biological replicates per experimental group, which helps to ensure valid biological interpretation and mitigate the impact of both biological and technical variation on DGE results.^28,29 Despite this, we acknowledge the possibility that the study may still be underpowered. 

With these limitations, our findings support several established hypotheses and observations of these cataract types. Notably, to our knowledge, this is the first ex vivo study of HLECs that has demonstrated evidence in support of the growth factor theory for steroid cataract pathogenesis. The growth factor gradient hypothesis, initially proposed by Jobling and Augusteyn⁵ and subsequently supported by various animal model studies,^26,30–32 suggests that the balance between growth factor concentrations in the vitreous and aqueous humor plays a crucial role in normal LEC proliferation and lens polarity. Interference with this growth factor gradient in the anterior and posterior ocular media is a potential explanation for the observed aberrant fiber cell differentiation in steroid cataracts. In the comparison of uveitic versus steroid cataracts, our findings reveal relative upregulation of several growth factor genes by steroid cataracts, including platelet-derived growth factor subunit B (PDGFB; log₂FC = −1.33, FDR = 0.01), fibroblast growth factor 10 (FGF10; log₂FC = −1.14, FDR = 0.02), and inhibin subunit alpha (INHA; log₂FC = −1.03, FDR = 0.02), which encodes a member of the transforming growth factor-beta (TGF-β) family. The FGF family, especially FGF2, and the TGF-β family have been linked extensively to posterior capsule opacification (PCO) after lens extraction surgery and other processes of epithelial–mesenchymal transition.^33–35 In rat lens epithelial explants and cultured whole lenses, TGF-β has been shown to induce punctate opacities, apoptosis, and localized capsule wrinkling that can be found in PSC cataract and PCO. Injection of TGF-β into the vitreous of adult rat eyes induced changes similar to those in human cortical and PSC cataracts, and transgenic studies in mice with over-expression of TGF-β1 have reported anterior subcapsular plaques histologically indistinguishable from human anterior subcapsular cataract.^32,36 Similarly, several members of the FGF family (e.g., FGF4 and FGF7 in transgenic mice) have been shown to cause cataractous changes by inducing abnormal lens fiber cell differentiation. However, FGF10 has not specifically been described to have a role in cataractogenesis but has been implicated in diseases of the cornea and lacrimal and salivary glands.^31,32 We were unable to find existing links between PDGFB and cataract. 

Another limitation to this study is that no robust findings were present in the direct comparison of steroid cataracts versus ARCs. The absence of DEGs may be indicative of a true absence of differences between the steroid and age-related cataract transcriptomes but more likely stems from the misclassification of ARCs as steroid-induced ones. Table 1 illustrates an overlap in mean age between the two groups, suggesting that some patients in the steroid group might exhibit an age-related component to their cataracts. Additionally, PSC cataracts in patients with a history of glucocorticoid therapy may not necessarily relate to the medication. This lens phenotype, well-documented in the literature as typical for steroid-induced cataracts, can also be associated with other cataract causes, such as advanced age and diabetes.¹² Future studies may benefit from employing a stricter age criterion to enhance the categorization of the steroid cataract group. 

Expectedly, DEGs and enriched GO biological pathways in the uveitic cataract versus ARC comparison showed upregulated inflammatory responses and regulated cell death processes, such as apoptosis and ferroptosis (Fig. 2A, Supplementary Table S7). In this group, approximately 60 DEGs have connections to both cataract and uveitis in existing literature. These include complement C3 (log₂FC = 4.63, FDR = 9.39 × 10⁻⁷), FBJ murine osteosarcoma viral oncogene homolog (FOS; log₂FC = 2.30, FDR = 1.13 × 10⁻⁴), and Tribbles pseudokinase 2 (TRIB2; log₂FC = −1.14, FDR = 1.38 × 10⁻⁷). C3 activation is the common endpoint of the three complement pathways and is a crucial component of innate immunity. In experimental autoimmune anterior uveitis (EAAU), an experimental model that resembles human anterior uveitis, massive amounts of increased levels of iC3b, a cleavage product of C3, have been observed within the Lewis rat eye during the peak of EAAU.^37,38 Copy number variations and gene polymorphisms of C3 have also been significantly associated with patients with Bechet's disease and Vogt–Koyanagi–Harada syndrome.³⁹ The FOS gene family includes four members (FOS, FOSB, FOSL1, and FOSL2) that encode for the proto-oncogene transcription factor complex AP-1. Studies of transgenic mice with over-expression of ΔFosB (a truncated form of FOSB) in the lens have demonstrated formation of a PSC cataract secondary to misalignment of the fibers in the suture region.^40,41 TRIB2 is a lesser described gene in ocular disease. It is involved in inducing apoptosis of mainly hematopoietic cells, but its upregulation has also been demonstrated in inflammatory diseases, including autoimmune uveitis.⁴² Polymorphisms of TRIB2 have been shown to significantly increase the risk of high-myopia-induced cataract in the Han Chinese population.⁴³ 

Interestingly, substantial differences were present when comparing ARCs to post-vitrectomy cataracts with removal of silicone oil (ROSO; 1359 DEGs) and without ROSO (30 DEGs). The profound effect of ROSO on HLEC gene expression may suggest silicone oil–specific effects on cataractogenesis or may be the result of damage or contamination of the HLECs during collection that is unrelated to cataract. At the time of writing, no study has compared cataract formation after vitrectomy with silicone-oil tamponade versus other tamponade agents. 

This dataset will assist in forming a broader knowledge base of secondary cataract pathogenesis and inform future research in these areas, particularly in the investigation of specific gene expression and its impact on cataract development through animal model studies. 

Acknowledgments

The authors thank everyone who has supported and contributed to this study. Special thanks to Sean Mullany, MD, for sharing his expertise in designing an RNA-seq study of human lens epithelial cells. We are also immensely grateful to Tiffany Khong at the Myeloma Research Group for her assistance with the initial sample collection and preparation, and we thank Julie Humphrey from Alfred Health and Kelly Mikunda from Centre for Eye Research Australia for their ongoing administrative support in the last 2 years. Our sincere thanks go to all of the surgeons and nursing staff at Alfred Health, the Royal Victorian Eye and Ear Hospital, and Eye Surgery Associates who assisted in sample collection. We also thank the Australian Genome Research Facility and the Monash Bioinformatics Platform for their guidance in RNA sequencing and data analysis. 

Finally, we extend our deepest appreciation to all the patients who graciously agreed to participate in this research. 

Disclosure: C. Fei, None; M.R. Dong, None; S. Byars, None; J.S. Shah, None; A.J. Hall, None; L.L. Lim, None