Molecular Vision 2024; 30:348-367
<http://www.molvis.org/molvis/v30/348>
Received 28 April 2024 |
Accepted 05 October 2024 |
Published 08 October 2024
Samuel G. Novo,1 Adam P. Faranda,1 Justin C. D’Antin,2,3 Yan Wang,1 Mahbubul Shihan,1 Rafael I. Barraquer,2,3 Ralph Michael,2,4 Melinda K. Duncan1
1Department of Biological Sciences, University of Delaware, Newark, DE; 2Centro de Oftalmología Barraquer, Barcelona, Spain; 3Institut Universitari Barraquer, Universitat Autònoma de Barcelona, Barcelona, Spain; 4Institute for Medical Informatics, Statistics and Epidemiology, University Leipzig, Leipzig, Germany
Correspondence to: Melinda Duncan, Department of Biological Sciences, University of Delaware, Newark, DE 19716; Phone: (302) 831-2281; email: duncanm@udel.edu
Purpose: Cataracts are typically treated by phacoemulsification followed by intraocular lens implantation. Studies of mouse models of cataract surgery have revealed that lens epithelial cells rapidly remodel their transcriptome to express proinflammatory cytokines after lens fiber cell removal, but it is currently unknown whether this response is conserved in human lenses. This study seeks to fill this knowledge gap.
Methods: Human cadaver eyes from 70 to 89 year old individuals were prepared for the human capsular bag model of cataract surgery. The central epithelium was preserved in RNAlater during culture preparation, then the equatorial epithelium was either immediately preserved in RNAlater after the culture was created, or 24 h later. Gene expression profiles were generated by bulk sequencing of RNA isolated from these tissue samples. The transcriptomic response of human cadaver-derived lens epithelial cells to culture in this “capsular bag” model was characterized by bioinformatic analysis. The human response was directly compared to that of 24-month-old mouse lens epithelial cells subjected to fiber cell removal surgery.
Results: Human lens epithelial cells remodel approximately a third of their transcriptome by 24 h after surgery, and like mice, this response consists of induction of proinflammatory cytokine genes, upregulation of fibrotic markers and downregulation of genes controlling the lens epithelial phenotype.
Conclusions: These observations demonstrate that humans and mice have similar responses to cataract surgery and support the use of mice to study the response of lens epithelial cells to cataract surgery, suggesting that identified injury response mechanisms can be leveraged to elucidate new approaches to improve the outcomes of cataract surgery.
Cataract, defined as the clouding of the ocular lens, has historically been a major cause of human blindness. However, the incidence of cataract-induced visual disability has greatly decreased in recent decades due to the development and refinement of surgical cataract treatment combined with its increased global availability [1,2]. Phacoemulsification, the most common method used today, removes the central lens epithelium/capsule and lens fibers followed by placement of an intraocular lens (IOL) prostheses into the retained lens capsule to restore refraction [3,4].
For most patients, cataract surgery is highly successful, restoring vision while improving their overall quality of life [5]. However, like all surgeries, cataract surgery is not completely benign as it often induces acute ocular inflammation [6] which can precipitate vision compromising macular edema [7] and in rare cases, retinal detachment [8,9]. Further, the most common cataract surgery methods do not remove all of the lens epithelial cells (LECs) from the retained peripheral anterior lens capsule, and these cells undergo a wound healing response which causes them to proliferate, migrate across the denuded lens capsule, and differentiate into a mixture of dysmorphic lens fiber cells and myofibroblasts which disrupt vision if they reach the visual axis, a condition called posterior capsular opacification (PCO) [10,11].
It is recognized that transforming growth factor-beta (TGF-β) signaling in lens capsule associated cells (CACs) is critical for myofibroblast formation [12,13] and persistence [14] post cataract surgery (PCS), however much less is known about the mechanisms by which cataract surgery triggers this signaling. Notably, we have discovered that it takes 48 to 72 h for the canonical TGF-β pathway, as measured by elevations in pSMAD2/3 levels, to activate in CACs PCS [15], likely due to the need to upregulate the expression of αVβ8-integrin, an important regulator of latent TGF-β activation [16]. To understand how cataract surgery induces this pathway, we performed RNAseq profiling on CACs collected at 24 h following lens fiber cell removal in a mouse model which unexpectedly revealed that a large proportion of the LEC transcriptome was remodeled by 24 h PCS [15], a response which included the downregulation of many genes regulating the lens epithelial phenotype and induction of proinflammatory cytokine and fibrotic marker expression [15,17].
As human LECs also produce growth factors and inflammatory cytokines when cultured in vitro in the capsular bag model of PCO [18], and the aqueous humor of cataract patients exhibits highly elevated levels of inflammatory cytokines by 20 h PCS [19], it is possible that the LECs remaining behind PCS could play a direct role in driving ocular inflammation, i.e., flare, PCS [20]. Further, as inflammation can drive fibrotic disease in other tissues [21,22], this result suggests that the early inflammatory response of LECs to cataract surgery could set the stage for their later conversion to myofibroblasts, thus leading to fibrotic PCO.
While the mouse has been long used as an animal model of cataract [23-25], mouse and human lenses do exhibit structural and molecular differences [26-28], although the similarities between the molecular responses to injury/surgery in these species have not been comprehensively evaluated using unbiased approaches. Here we perform bulk RNAseq profiling of the central and equatorial lens epithelium isolated from human cadaver lenses obtained from elderly donors and compared the resulting transcriptomes with our prior transcriptomic profiling of aged mouse lens epithelial cells [17]. The acute response of human equatorial LECs to the lens capsular bag model of cataract surgery [29] was assessed by transcriptome profiling of cells obtained either immediately after explant creation or following 24 h of culture. These data provide a comprehensive overview of the similarities and differences between the mouse and human LEC transcriptome and confirm that the mouse is a valid model to understand the acute response of LECs to cataract surgery.
Aged human donor eye globes (70–80 years of age) were obtained from the “Banc d´Ulls per a Tractaments de Ceguesa” eye bank (Barcelona, Spain) in concordance with the Tenets of the Declaration of Helsinki. The experimental protocol was approved by the Ethical Committee for Clinical Research of the Centro de Oftalmología Barraquer. Age, cause of death, sex, and post mortem time of each donor can be found in Appendix 1.
Human lens capsular bag cultures were created as previously described with all steps performed at room temperature [29,30]. Briefly, the corneoscleral disk was removed from the globe via a circular trephine and the iris-ciliary body-lens complex dissected, then transferred to a sterile Petri dish. Meanwhile, a silicone ring mount was placed in a Petri dish and immersed in Hanks balanced salt solution (HBSS) with added 2% antibiotic/antimycotic (10,000 U penicillin, 10 mg streptomycin, and 25 mg amphotericin B per milliliter). The iris-ciliary body lens complex was then placed on the silicone ring, anterior surface facing upward, attached by placing eight 30-gauge needles through the ciliary body and the iris removed. A capsulorhexis was then performed on the lens, with the anterior capsulotomy specimen (lens capsule with attached central epithelial cells) immediately collected in RNALater (Sigma-Aldrich, Madrid/Spain). The fiber cells were removed by hydrodissection and hydroexpression through the capsulorhexis and cortical fibers also placed in RNAlater. One capsular bag from each globe pair (lens capsule with attached equatorial epithelial cells) was then immediately separated from the zonules and placed in RNAlater, as a 0 h control. The fellow capsular bag was washed 3 times for three minutes each with HBSS and then cultured in a humidified CO2-incubator at 37 C and 5% CO2 with RPMI-1640 medium (supplemented with 5% fetal calf serum (FCS), 1% 10,000 units penicillin, 10 mg streptomycin and 25 µg amphotericin B per milliliter). After 24 h, the capsular bag (lens capsule with attached equatorial epithelial cells) was separated from the zonules and also submerged in RNAlater. The samples in RNAlater were then shipped to the University of Delaware for further analysis.
The capsular fragments with attached lens epithelial cells were removed from RNAlater, and RNA isolated using the RNeasy Micro Kit from Qiagen (Catalog Number: 74004). RNAseq libraries were prepared using the SMARTer® Stranded Total RNAseq Kit-Pico Input Mammalian (Takara Bio, Inc., Mountain View, CA) and sequenced by DNA Link (1000 S Hope St. unit 521 Los Angeles, CA 90015) on a Novaseq 6000 (San Diego, CA). This yielded approximately 35 million raw sequencing reads (paired end, 101 nucleotides long) per sample which were assessed for quality, and trimmed for adaptor content and low-quality base calls, using the Babraham Institute’s “FastQC” and “Trim Galore!” programs.
As SMARTer® Stranded Pico RNAseq libraries often have significant numbers of reads derived from rRNA, trimmed paired end reads were first aligned to the human 45S pre-spliced rRNA transcript (RN45SN5), and reads aligning to this transcript excluded from downstream analyses. All remaining paired end reads of at least 85 nucleotides were aligned to the Ensembl primary assembly of the human GRCh38 genome [31] using HISAT2 with its default parameters [32]. Read pairs aligning to genomic features in the Ensembl Human version 104 GTF file were quantified as gene level counts, using HTSeq-Count in union mode [33]. Length normalized abundance estimates (Fragments per Kilobase-Million (FPKM)) were calculated from gene level counts using the total length of all known exons for a given gene, after merging overlapping exons. Genes with at least 10 mapped reads in at least four samples were considered to have “detectable” levels of expression, those failing criteria were removed before normalization and differential expression testing using edgeR’s “filterByExpr” function [34]. Gene level counts were normalized, and pairwise differential gene expression statistics calculated using robust dispersion estimates with the “Trimmed Median of Means” (TMM) [35] and “exactTest” methods implemented in edgeR [36]. Biologically significant differentially expressed genes (DEGs) were defined as those exhibiting a statistically significant difference in expression using Storey’s Q value to adjust for false discovery rate (Q ≤0.05) [37], a difference in expression level greater than 2 FPKM between conditions, Fold Change (FC) greater than 2 in either the positive or negative direction and expressed at a level greater than 2 FPKM [38]. Tables of differentially expressed genes (DEG Tables) were generated for the following pairwise contrasts: freshly isolated equatorial versus central epithelium, and freshly isolated equatorial epithelium versus equatorial epithelium cultured 24 h in the lens capsular bag model. The resulting FastQ files and analyzed data were deposited in the Gene expression omnibus (GEO) under Accession number GSE186716.
iPathwayGuide (Advaita Bioinformatics, Plymouth, MI) was used for pathway analysis performed on all statistically significant DEGs defined as those exhibiting a fold change ≥ 2 and false discovery rate corrected p value (FDR) ≤0.05. The iPathwayGuide uses Impact Analysis, which considers the directed interactions of DEGs within a pathway (as defined by the Kyoto Encyclopedia of Genes and Genomes [39] (KEGG, Release 96.0+11–21, Nov 20) and if more pathway participants are observed in the DEG list than would be estimated by chance [40]. Gene ontology comparisons were made against the October 14, 2020 release of the Gene Ontology Consortium database [41].
FastQ files generated from RNAseq data previously obtained from mouse lens epithelial cells isolated from 24-month-old mice subjected to the in vivo mouse model of cataract surgery [42] both the time of lens fiber cell removal and 24 h later [17] (Geo accession number GSE166619) were reanalyzed using the same pipeline as the human LEC samples above except that it was mapped to the mouse genome (Ensembl GRCm39, GTF v104) and the mouse Rn45s pre-spliced transcript used for ribosomal filtering. This resulted in expressed gene tables representing the naïve 24 month old mouse LEC transcriptome and the changes to this transcriptome that manifest 24 h after fiber cell removal surgery.
Ensembl’s v104 homology mappings were used to identify mouse homologs of human genes. For genes where multiple homologs were detected, homolog pairs used for further analysis were chosen based on Ensembl orthology confidence score and percentage of sequence identity. Genes with recognized homologs in mouse and human were then ranked based on their group mean abundance (FPKM). For each cell type (naïve human central and equatorial epithelium, equatorial epithelium cultured for 24 h; mouse, naïve lens epithelium, lens epithelium 24 h after fiber cell removal), the genes with known homologs in the other species were rank ordered based on mean abundance, and then partitioned into three categories: only within the top 10% of abundance in human LECs, in the top 10% of abundance in both mouse and human LECs, and only in the top 10% of abundance in mouse LECs. Human central LEC and human equatorial LEC profiles were each compared to profiles generated from the entire mouse lens epithelium in separate analyses. The genes meeting differential expression criteria in the human and mouse comparisons were compared with the genes annotated to participate in either KEGG Pathway “Cytokine / Cytokine-Receptor Interaction” (KEGG ID: map04060) or a set of well characterized genes that are known to be associated with fibrosis and/or epithelial to mesenchymal transition (EMT; MSigDB Hallmark EMT) [43].
We previously reported that remnant lens epithelial cells (LECs) massively remodel their transcriptome by 24 h after lens fiber cell removal in a mouse model of cataract surgery [15-17,44]. The most impacted pathways were related to the inflammatory response, most notably, cytokine-cytokine receptor interactions, although the mRNA levels of key fibrotic markers are also upregulated by this time [15,17]. However, the relevance of these observations to the response of human LECs to cataract surgery, and the progression of post-surgical ocular inflammation in humans, was unclear as there are known species differences in lens gene expression between mouse and human [45-47] while injury-induced inflammation in other tissues seems to be more acute in mice than humans [48,49]. Therefore, we evaluated the aged human LEC transcriptome and its acute response to an ex vivo culture model previously developed to investigate the molecular mechanisms of posterior capsular opacification (PCO) [29,50,51].
Lenses from human eye bank eyes obtained from a Spanish population (n=3, 70–89 years of age, 2 male, 1 female; see Appendix 1) were subjected to continuous curvilinear capsulorhexis to isolate the central lens epithelium while the equatorial lens epithelium was prepared for culture in the capsular bag model of PCO [29]. RNA was prepared from capsulorhexis specimens and capsular bags containing equatorial epithelial cells isolated either immediately after dissection or after 24 h of culture, then RNAseq performed. Principal component analysis (Figure 1) revealed that all three biologic replicates for each cell population cluster together well despite the fact that these samples vary in sex, postmortem time and cause of death, with the greatest variance detected between equatorial epithelial cells analyzed before and after culture. These data have been deposited in the Gene Expression Omnibus under accession number GSE186716. Comparisons between the freshly isolated central and equatorial lens epithelium transcriptomes are found in Appendix 2. Comparisons between the freshly isolated equatorial lens epithelium transcriptome and the equatorial epithelium following 24 h of culture in the capsular bag model of PCO transcriptomes are found in Appendix 3.
Consistent with other studies [52,53], regional differences were found in the bulk lens epithelial transcriptomes, with central and equatorial human LECs differentially expressing 720 genes by at least twofold with false discovery rate (FDR) corrected p values of ≤0.05. Further filtering for biologic significance revealed 464 biologically significant DEGs (Appendix 2). Functional predictions via Advaita iPathway analysis revealed that the “calcium signaling pathway” (p value 1.165E-6) was the most impacted pathway, while consistent with known regional differences in the adult lens epithelium, the equatorial epithelium expressed numerous genes regulating epithelial cell proliferation (p value 3.600E-4), cell adhesion (p value 4.400E-8), cell junction assembly (p value 8.700E-8), and cation transport (p value 2.500 E-8) at higher levels than the central lens epithelium (Figure 2). As we also noticed that many DEGs expressed at higher levels in the equatorial epithelium than central epithelium such as Fibrillin 1 (FBN1) which are known components of the ciliary zonule, we compared these DEGs with a list of proteins found in the zonule in a previous study [54] and found that equatorial LECs express genes encoding 12 known zonule components at higher levels than the central LECs, while genes encoding 4 zonule components are expressed at lower levels in the equatorial epithelium (p value of 2.8E-3; Table 1).
Despite these differences, the central and equatorial epithelium share 76% of their most abundantly expressed (top 10% highest expression as measured by FPKM) genes in common (1465 out of 1921, Appendix 4). Many of these shared, highly expressed genes exhibit enriched expression in the lens as assessed by iSyTE (Appendix 2), and are known to be important for lens biology, including crystallins, genes encoding components of the lens capsule, and developmentally important transcription factors (Table 2).
The mouse is a common animal model to study lens biology and cataract development. Mouse genetics methods are robust and the mouse lens expresses most genes relevant to human cataract etiology. While differences in the molecular composition of mouse and human lenses have been investigated for specific classes of proteins [44-46], a global comparative transcriptomic analysis was lacking until now. Here we compare the transcriptome of 70–89 year old human lens epithelial cells to that of 24 month old (aged) mice, as mice of this age have been proposed to be biologically equivalent to 70 year old humans [55].
Aged human and mouse LECs express similar numbers of genes at sufficient levels to likely affect lens biology. The 10% most abundant mRNAs found in the aged human central epithelium were mapped to their direct mouse gene homologs. The resulting set of 1787 mapped homologs was then compared with the genes expressed in the top 10% of the aged human equatorial lens epithelium as well as the 24 month old mouse lens epithelium. Of the genes exhibiting the highest expression in the human central epithelium, 36% were also ranked in the top 10% of the human equatorial epithelium and aged mouse lens epithelium (637 total genes ranked in the top 10% of expression in all three cell types). An additional 32% of top 10% expressed genes (575 out of 1,787) whose mRNAs were highly abundant in both the central human and aged mouse LECs but not the human equatorial epithelium (See Table 3 for representative examples, entire list is found in Appendix 5). A similar comparison was made between the aged human equatorial LEC transcriptome and aged mouse lens epithelium which revealed that 40% of the highly expressed genes with known human/mouse homologs were among the most expressed in both cell populations (See Table 4 for representative examples, entire list is found in Appendix 6).
These data show that the bulk transcriptome of the aged mouse lens epithelium is generally more similar to equatorial human LECs than central as would be expected since a smaller lens diameter would result in a larger proportion of the lens epithelial cells in the lens epithelium being “equatorial” due to tissue geometry. The aged mouse lens epithelium thus appears to be a good representation of the aged human equatorial lens epithelium which is the cell population that gives rise to Soemmering’s ring and visual axis opacification (VAO)/PCO following cataract surgery.
We have previously shown that mouse lens epithelial cells acutely trigger inflammatory responses following lens fiber cell removal [15,17,56]. Here we evaluate whether human equatorial LECs respond similarly when placed in the human capsular bag model of posterior capsular opacification. As shown in Figure 1, principal component analysis revealed that 24 h of culture caused a major shift in the human equatorial LEC transcriptome compared to that found in freshly isolated equatorial epithelial cells. Further, all three cultured equatorial LEC samples had very similar gene expression profiles despite differences in sex, age, cause of death and postmortem time which indicates that the injury response of human LECs is reproducible. Comparison of the human equatorial LEC transcriptome following 24 h of culture with that from freshly isolated equatorial LECs revealed that 6269 genes were differentially expressed while filtering this list under criteria previously proposed to predict biologically significant DEGs [38] revealed a total of 4042 biologically significant DEGs (Appendix 3).
iPathway analysis of the genes whose expression changes in human equatorial epithelial cells after 24 h of culture revealed the “Cytokine-cytokine receptor interaction” is the pathway most impacted by 24 h of culture (Figure 3A,B; p value=1.120e-8). Consistent with this, many of the genes with the greatest fold change increase in expression play known roles in tissue inflammation or innate immunity (Table 5). Notably, many of the inflammatory cytokine genes upregulated in human LECs after fiber cell removal encode proteins whose concentrations elevate in human aqueous humor within 20 h of cataract surgery (19; Table 6), however other such cytokines either do not change in expression in LECs in the first 24 h of culture, or are not expressed by human LECs at all (Appendix 7).
Other significantly impacted pathways included those participating in “neuroactive ligand-receptor interactions” (Figure 3C,D; p value=6.959e-7) while the “epithelial to mesenchymal transition” pathway was also predicted to be impacted (Figure 3E; p value 0.042). Notably, as would be expected in cells transitioning to a mesenchymal phenotype, several genes known to be important for lens function also downregulate in human equatorial LECs after 24 h of culture (Table 7).
We previously reported that both the aged and young mouse LECs transcriptome is drastically remodeled by 24 h after fiber cell extraction in an in vivo model of cataract surgery [15,17]. Comparison of these DEGs with the list of genes differentially expressed at least twofold in human equatorial epithelial cells following 24 h of culture revealed that human and mouse LECs commonly upregulate 555 homologous genes and downregulate 290 homologous genes (Appendix 8). As the KEGG pathway “cytokine-cytokine receptor interaction” was the most impacted pathway in both human (Figure 3A,B) and mouse LECs [15,17] at 24 h after fiber cell removal, we next assessed the similarities between this response in human and mouse LECs and found that 20 genes in this pathway were upregulated in LECs from both species following injury including various tumor necrosis factor receptors and interleukin receptors (Table 8). However, while the KEGG pathway “neuroactive receptor-ligand interaction was predicted to be an impacted pathway in injured human LECs (Figure 3), this was not seen in injured aged mouse LECs [17] although some genes mapping to this pathway such as F2RL1, GAL, GABBR2, and C3 were upregulated in LECs of both species at 24 hours post lens fiber cell removal.
As the inflammatory response of LECs may contribute to the formation of capsular bag associated myofibroblasts [21] indicative of fibrotic PCO [11], it was expected that the 24H equatorial human LECs differentially express genes implicated in epithelial to mesenchymal transition (EMT; Figure 3E) similar to our prior observation that mouse LECs upregulate genes associated with EMT by 24 h after lens injury [15,17]. Further, direct comparison of the transcriptomic changes triggered in mouse and human LECs by 24 h PCS found that 23 genes of the MSigDB “HALLMARK EMT” gene set were commonly regulated in both species including the classical lens EMT markers α-smooth muscle actin and tenascin C (Table 9).
Mice have long been the most common animal model used to study mammalian lens development and cataract pathogenesis due to the availability of genetic tools that allow functional testing of genes in vivo. Such studies are generally believed to yield conclusions that further our understanding of human lens biology as mouse lenses express most (if not all) genes implicated in human cataractogenesis [57,58]. Further, mouse cataract surgery models have been developed for the study of human posterior capsular opacification (PCO) [42,59–61] which appear to be appropriate animal models as, like humans [50,62], the remnant capsular bags become populated with a mixture of α-smooth muscle actin expressing myofibroblasts and crystallin expressing dysgenic lens fiber cells at extended times following fiber cell removal [16,61,63,64]. However, while the literature reports several instances where mouse and human lenses express different genes [45-47], the global similarities between the mouse and human lens epithelial cell (LEC) transcriptome had not been investigated, and even less was known about how/whether human LECs, like mouse LECs, acutely alter their transcriptome in response to cataract surgery. This study sought to fill these knowledge gaps.
The most common surgical therapy for cataracts is phacoemulsification, a procedure that removes the central anterior lens capsule with attached lens epithelial cells and lens fiber cells, leaving the remaining lens capsule and attached equatorial LECs behind to anchor an intraocular lens implant [3,4]. Here, bulk RNAseq profiling of the central versus equatorial epithelial cells revealed that these cell populations have very similar transcriptomes which share most of their most abundant transcripts in common including those encoding transcription factors critical for LEC identity such as PITX3 [65], FOXE3 [66] and PAX6 [67], as well as numerous proteins important for LEC/lens function such as CRYAB (αB-crystallin) [68,69], VIM (vimentin) [70], SFRP1 (secreted frizzled related protein 1) [71], FTL (ferritin light chain) [72] and AQP1 (aquaporin 1) [73]. There are, however, key transcriptomic differences between the central and equatorial cell populations in the human lens epithelium. For instance, equatorial epithelial cells express higher levels of genes important for cell cycle progression such as CDK6 [74], while central epithelial cells express higher amounts of mRNAs encoding cell cycle inhibitors such as CDKN1C [75]. These observations are consistent with experimental findings that revealed that adult lens central epithelial cells seldom proliferate while equatorial epithelial cells continue proliferating throughout life leading to a slow increase in lens size across the lifespan [76]. Further, consistent with the known differentiation of equatorial epithelial cells into lens fibers, equatorial LECs express higher levels of many genes important for lens fiber cell biology than the central epithelium including those participating the Notch and FGF signaling pathways which regulate lens fiber generation [77,78]. Additionally, numerous genes associated with cell adhesion and cell junction assembly are expressed at higher levels in the human equatorial LECs compared to the central LECs which is consistent with prior reports that found that equatorial LECs upregulate genes associated with cell adhesion [79,80] to set the stage for the proper organization of differentiating lens fiber cells.
Notably, equatorial LECs from aged individuals also express appreciable levels of mRNAs that encode extracellular matrix (ECM) proteins that are components of the zonules, acellular filaments that tether the lens to the ciliary muscle, suggesting that the equatorial lens epithelium plays a role in zonule maintenance. Such a role is consistent with the clinical observation that “dead bag syndrome,” the failure to form Soemmering’s ring following cataract surgery, is associated with zonular fragility and dislocation of the IOL/capsular bag complex from its connections to the ciliary body [81,82]. However, it is currently unclear how LECs which are enclosed by the lens capsule can contribute to the maintenance of zonules which are largely intercalated into the outer surface of the lens capsule matrix [83,84].
Unlike human cataract surgery, a capsulorhexis is typically not performed in the mouse in vivo cataract surgery model due to the small size of the eye. As both the central and equatorial epithelium are retained in the mouse eye following surgery, potential differences between these cell populations could complicate translating mechanistic conclusions from the mouse model to human. To gauge the likely extent of such differences, we compared our prior transcriptomic profiling of the 24 month old mouse lens epithelium with that of the aged human central and equatorial epithelium by identifying genes that have homologs in both mouse and humans. As it was not possible to directly compare transcript abundance between species, we then rank ordered expressed genes based on transcript abundance in each species/cell type then compared which genes contributed to transcripts in the top 10% of abundance. This analysis revealed that approximately 30% of the most abundant genes are shared between all three cell types (human central and equatorial LECs, aged mouse LECs) while another 30%–40% of the most abundant genes are shared between either human central LECs and mouse or human equatorial LECs and mouse, although bulk transcriptome of the aged mouse lens epithelium is more similar to the human equatorial epithelium as would be expected since the human equatorial epithelium makes up 68% of the total lens epithelium [85]. It should be noted that these abundance comparisons likely underestimate the similarities between the mouse and human LEC transcriptome as our decision to compare the 10% most abundantly expressed genes is an arbitrary cutoff. Overall, the similarities in gene expression between the human central and equatorial epithelium and the total mouse lens epithelium suggest that analysis of the response of the entire lens epithelium in the mouse cataract surgery model likely does not greatly bias the results. However, the differences observed do emphasize the need to validate results obtained in the mouse model with human tissue whenever possible if seeking to relate findings from mouse models to human clinical care.
Mouse LECs rapidly remodel their transcriptome when separated from their underlying lens fiber cells in both in vivo mouse cataract surgery models [15,17,56] and during explant culture [86], leading these cells to express elevated levels of cytokines/modulators of the innate immune response and genes associated with fibrotic tissue, while downregulating the expression of classical markers of the lens epithelium. While this response appears to be conserved in embryonic chick LECs as well since they upregulate proinflammatory cytokines after just 1 h in explant culture [87], it was unknown whether human LECs left behind following cataract surgery respond similarly.
Here we found that human equatorial lens epithelial cells derived from human cadaver lenses also massively reprogram their transcriptome by 24 h of culture in the lens capsular bag organ culture model of PCO and the global nature of this reprogramming mirrors that seen in the mouse cataract surgery model. Notably, human equatorial LECs were found to downregulate the expression of many LEC preferred genes, while upregulating the expression of fibrotic markers, genes mapping to the KEGG pathway “cytokine-cytokine receptor interactions” as well as other genes known to participate in the innate immune response to tissue injury. Notably, two of the top three most upregulated genes in human equatorial epithelial cells after 24 h in culture, PTGS2/COX2 [88], and CSF3 [89], both participate in tissue inflammation and are also highly upregulated in mouse LECs at both 6 h [56] and 24 h [15,17] after lens fiber cell removal surgery, suggesting that the acute transcriptional response of LECs to lens fiber cell removal is evolutionarily conserved. Further, the cytokines whose levels elevate in human aqueous humor at 20 h following cataract surgery [19] partially overlap with the inflammatory cytokine genes that upregulate in human (and mouse) equatorial LECs after 24 h of culture suggesting that remnant LECs may influence ocular inflammation following cataract surgery by changing the balance of aqueous humor cytokines. However, this overlap is not complete as some cytokines which upregulate in aqueous humor are not expressed in cultured human equatorial epithelial cells. These additional proteins may come from the immune cells which begin to invade the in vivo eye by 18 h PCS [15,90], or other ocular tissues which are influenced by cataract surgery.
The present study revealed that human and mouse lens epithelial cells have similar transcriptomes which change similarly in response to lens fiber cell removal such as what is performed during cataract surgery. However, the study design does limit some of the conclusions that can be made.
1) The mouse transcriptomes used for study were derived from inbred mice which were housed in controlled environments, had no apparent pathologies, and lens tissue was isolated immediately after death. In contrast, the human lenses studied were genetically diverse, came from individuals who died natural deaths at a range of ages from diverse causes, with tissue collected at a range of post-mortem times. While the principal component analysis (PCA) revealed that samples obtained from the same cell population isolated from different individuals have quite similar global transcriptomes (Figure 1), it is likely that some genes with bona fide expression differences across cell types were not revealed due to the intrinsic variability of the samples studied and the analysis of samples from only three individuals.
2) The mice and humans used for comparison were “age matched” based on a “frailty” index [55] which assesses a variety of physiologic parameters to set “equivalent ages” between these two species. However, it is recognized that different tissues age at different rates even within an individual [91], and the comparative rates of lens aging between mice and humans have not been directly assessed outside of propensity to develop cataract [92].
3) The human central epithelial samples were preserved in RNAlater immediately after capsulorhexis, but the “uncultured” equatorial epithelial samples were collected approximately 5 - 10 min after hydroexpression of the lens fiber cells. This may have led to an overestimate of the gene expression differences between the central and equatorial epithelium in this study as LECs can trigger gene expression changes as soon as 30 min following a stress [93], which leads to differential expression of approximately 10% of the LEC transcriptome by 6 h PCS [56]. Notably, some genes whose expression rapidly changes in mouse LECs upon lens fiber cell removal, such as the immediate early genes FOS and FOSB which can be upregulated as soon as 30 min after a stress [94], and the inflammatory mediators CXCL14 and CCL2 were detected at lower levels in the central lens epithelium which was immediately preserved versus the equatorial epithelium whose collection was delayed due to its placement within the capsular bag model.
4) The human ex vivo capsular bag model of PCO requires pinning the ciliary body/capsular bag complex into place followed by incubation in cell culture media supplemented by 5% serum which exposes the equatorial LECs to both the factors present in serum and those that may be produced by the injured ciliary complex, while the injury response of mouse lens epithelial cells was assessed in the eye of a living mouse. These differences likely result in the remnant lens epithelial cells being exposed to a different growth factors/cytokines milieu which could lead to different transcriptomic responses.
This work revealed the baseline similarities and differences between the mouse and human transcriptome and compared their responses to lens fiber cell removal models that mimic modern cataract surgery. Notably, this work confirmed that human LECs, like those in mice, induce the expression of pro-inflammatory cytokines and fibrotic marker genes by 24 h after fiber cell removal, further validating the mouse as a model to study the acute lens injury responses that likely set the stage for the development of posterior capsular opacification, a common negative consequence of modern cataract surgery.
Appendix 1. Supplemental Table 1.
Appendix 2. Supplemental Table 2.
Appendix 3. Supplemental Table 3.
Appendix 4. Supplemental Table 4.
Appendix 5. Supplemental Table 5.
Appendix 6. Supplemental Table 6.
This study was supported by National Institutes of Health Grant (EY028597) to MKD; INBRE program grant P20 GM103446 and the State of Delaware supported the University of Delaware Center for Bioinformatics and Data Science and the BioMix Compute Cluster. Conflict of interest statement: None *SGN & APF contributed equally to this work GEO accession number: GSE186716;
No table matches the in-text citation "Supplemental Table 1". Please supply the missing table or delete the citation.
No table matches the in-text citation "Supplemental Table 2". Please supply the missing table or delete the citation.
No table matches the in-text citation "Supplemental Table 3". Please supply the missing table or delete the citation.
No table matches the in-text citation "Supplemental Table 4". Please supply the missing table or delete the citation.
No table matches the in-text citation "Supplemental Table 5". Please supply the missing table or delete the citation.
No table matches the in-text citation "Supplemental Table 6". Please supply the missing table or delete the citation.
No table matches the in-text citation "Supplemental Table 7". Please supply the missing table or delete the citation.
No table matches the in-text citation "Supplemental Table 8". Please supply the missing table or delete the citation.
No box matches the in-text citation "box E3". Please supply the missing box or delete the citation.
No box matches the in-text citation "Box 6". Please supply the missing box or delete the citation.
No box matches the in-text citation "Box E3". Please supply the missing box or delete the citation.
No box matches the in-text citation "Box E3". Please supply the missing box or delete the citation.
No box matches the in-text citation "Box E3". Please supply the missing box or delete the citation.
No box matches the in-text citation "Box 6". Please supply the missing box or delete the citation.