Elucidating the Corneal Endothelial Cell Proliferation Capacity through an Interspecies Transcriptome Comparison

The regenerative capacity of corneal endothelial cells (CECs) differs between species; in bigger mammals, CECs are arrested in a non‐proliferative state. Damage to these cells can compromise their function causing corneal opacity. Corneal transplantation is the current treatment for the recovery of clear eyesight, but the donor tissue demand is higher than the availability and there is a need to develop novel treatments. Interestingly, rabbit CECs retain a high proliferative profile and can repopulate the endothelium. There is a lack of fundamental knowledge to explain these differences. Gaining information on their transcriptomic variances could allow the identification of CEC proliferation drivers. In this study, human, sheep, and rabbit CECs are analyzed at the transcriptomic level. To understand the differences across each species, a pipeline for the analysis of pathways with different activities is generated. The results reveal that 52 pathways have different activity when comparing species with non‐proliferative CECs (human and sheep) to species with proliferative CECs (rabbit). The results show that Notch and TGF‐β pathways have increased activity in species with non‐proliferative CECs, which might be associated with their low proliferation. Overall, this study illustrates transcriptomic pathway‐level differences that can provide leads to develop novel therapies to regenerate the corneal endothelium.


Introduction
The cornea is an avascular and transparent tissue located in the anterior segment of the eye that allows light to enter it.It is composed of five layers: epithelium, Bowman's layer, stroma, Descemet's membrane, and endothelium.The corneal endothelium is the innermost layer of the cornea and is composed of a monolayer of corneal endothelial cells (CECs) that reside in contact with the stroma on the Descemet's membrane. [1]The corneal endothelium function is to actively pump ions and metabolites from the stroma into the aqueous humor of the eye, thereby maintaining the cornea slightly dehydrated and transparent. [2]he regenerative capacity of CECs differs between species.In humans [3] and sheep, [4] CECs are arrested in a quiescent state, lacking regenerative capacity through cell division.Iatrogenic damage after surgery, genetic diseases such as Fuchs' endothelial cell dystrophy, or infections in species with non-proliferative CECs can cause a decrease in their number, compromising the tissue function and leading to corneal opacity and impaired vision, also referred as corneal endothelial dysfunction.In contrast, rabbit CECs possess a high proliferative capacity and can repopulate the endothelium in response to damage. [5,6]he standard treatment for corneal endothelial dysfunction is endothelial keratoplasty, [3,7] but the increasing number of transplantations are causing a global donor cornea shortage leaving 12.7 million patients waiting for treatment. [8]There is a need to find alternative treatments to address corneal endothelial dysfunction.Promoting the proliferation of CECs could be an elegant approach to stimulate tissue regeneration and recovery of corneal transparency, but the mechanisms governing the differences between proliferative and non-proliferative CECs remain unknown.Therefore, to study the possible regulators of CEC proliferation, we performed an RNA sequencing (RNAseq) comparison between human, sheep, and rabbit CECs.
This study presents an interspecies transcriptome comparison of CECs originating from species with proliferative and quiescent endothelium.We identified pathways with different activity, which could be the driving force for CEC proliferation and regeneration.These findings can be used to improve current CEC expansion protocols and identify novel drug targets to promote corneal endothelial proliferation and regeneration.

Homology Mapping of Endothelial Samples Detected 9757 Protein Coding Transcripts Commonly Expressed Across Species
Corneal endothelium was isolated from human, sheep, and rabbit corneas, following RNA isolation.RNA integrity numbers ranged from 6.2 to 9.3.All samples were sequenced following an adapted CEL-Seq2 protocol on Illumina Nextseq500, high output 2 × 75 bp run mode, at a sequencing depth of 20 million reads per sample.After mapping to the species reference genome, a comparable number of protein coding genes, 16852, 17624, and 15258, were detected in human, sheep, and rabbit, respectively.Homology mapping detected 9757 full homologue protein coding genes across species which were then used for further pathway-level analyses (Figure 1), homology gene reads were normalized across samples using quantile normalization.

Corneal Endothelial Cell Transcriptomes Differed Highly across Species
All endothelial samples showed a normal and comparable distribution of gene transcript counts after normalization (Figure 2A).Pearson's correlation analysis revealed that endothelial samples differed across species, with human and sheep having a higher correlation than rabbit (Figure 2B).Principal component analysis further confirmed these findings revealing that samples from the same species cluster together and rabbit samples cluster more distantly than sheep and human endothelial samples (Figure 2C).These findings are in line with the biological difference assessed in this study, being human and sheep species with non-proliferative CECs, versus rabbit, a species with proliferative CECs.The samples of all species showed expression of typical endothelial markers such as ALCAM (CD166), TJP1 (ZO-1), NCAM1, SLC4A4, and ATP1A1, [3,9,10] confirming the isolation of endothelium.Moreover, the absence of stromal markers KERA, CD34, and LUM, [9,11] absence of epithelial markers WNT7A, and PAX6, [9,12] and absence of immune markers ITGAM, [13] CCR7, [14] and CD19 [15] confirmed the absence of contaminating cell types (Figure 2D).
Interestingly, the transcriptomic analysis showed highly similar expression patterns between human peripheral endothelial biopsies (8.5-10 mm of corneal diameter) and the full human analysis shows that human and sheep samples have a higher correlation than rabbit corneal endothelium, confirming that rabbit corneal endothelial cells highly differed compared to human and sheep.C) Principal component analysis shows that samples originating from the same species cluster together, confirming the reproducibility of the sample isolation, and further confirms that rabbit corneal endothelial cells highly differed compared to human and sheep corneal endothelium.D) All samples had high expression of typical corneal endothelial cell markers, namely ALCAM, TJP1, NCAM1, SLC4A11, and ATP1A1; and absence/low expression of stromal markers (KERA, CD34, and LUM), epithelial markers (WNT7A and PAX6), and immune cell markers (ITGAM, CCR7, and CD19), confirming a successful isolation of the corneal endothelial cell transcriptome.
corneal endothelium, with a minimum Pearson's correlation coefficient of p = 0.935 (Figure S1A, Supporting Information).We also observed that important progenitor-associated markers such as LGR5, PITX2, and SOX2 had a comparable expression level (Figure S1B) in all human corneal endothelial samples.This is in contrast to the findings by Yam et al., [16] where LGR5 was detected in the corneal endothelial periphery at the protein level.

Pathway Level Analysis Revealed 52 Pathways with Different Activity between Proliferative and Non-Proliferative CECs
To elucidate possible drivers of CEC proliferation across the species studied, we performed a pathway level analysis benchmarking our data against WikiPathways.After filtering out disease-related pathways and pathways comprising less than 10 genes, we performed a non-parametric Wilcoxon rank-sum test to identify up-or down-regulated pathways between: 1) human and rabbit, 2) sheep and rabbit, and 3) sheep and human CECs.Only pathways with a p-value lower than 0.05 and an absolute effect size greater than |1| were retained for analysis in each of the comparisons.
Our analysis revealed that sheep and rabbit CECs were most different, with different activity in 121 pathways (Figure 3B), followed by human and rabbit with 94 pathways with different activity (Figure 3A).Human and sheep corneal endothelium showed different activity in only 54 pathways.These results show that more pathways have a different activity when comparing species with non-proliferative and proliferative CEC (121 in sheep vs rabbit, and 94 in human vs rabbit), whereas species with non-proliferative CEC had fewer pathways with different activity, namely 54 in human versus sheep (Figure S2, Supporting Information).
To further understand the differences between species with proliferative (rabbit) and non-proliferative CECs (human and sheep), we selected the pathways that showed different activity when comparing both human versus rabbit as well as sheep versus rabbit, as these pathways might be related to their differing proliferative potential (Figure 3C).Our data analysis revealed a total of 52 pathways (Table S1A, Supporting Information) with different activity in both human versus rabbit and sheep versus rabbit.The detected pathways were diverse and included signaling, immune, and metabolic pathways.In all detected pathways, the effect score correlated with the proliferation potential of the CECs, and pathways were more or less active in both human and sheep CECs when compared to rabbit CECs.Only 9 pathways were more active in rabbit CECs when compared to human and sheep, and 43 were more active in human and sheep when compared to rabbit (Figure 3D).These results might indicate that the possible drivers of CEC quiescence are pathways with increased activity in sheep and human.

Notch and TGF-𝜷 Signaling Pathways Are More Active in Species with Non-Proliferative CECs
For further analysis, we did not consider pathways describing the effect of experimental procedures or external activities, such as "photodynamic activation of NRF2".Based on our research question, we selected pathways that could have implications on the CEC proliferative capacity.Our results revealed that Notch and TGF- signaling pathways were more active in human and sheep CECs compared to rabbit CECs.These pathways have been demonstrated to induce senescence and endothelial to mesenchymal transition in CECs, [17][18][19] and could play a key role in the differences in CEC proliferative capacity between species.To better understand the implications of these pathways we performed a pathway visualization with Cytoscape (Figure 4).
Our analysis on the Notch signaling pathway showed that positive regulators of the pathway such as DTX3, APH1A, PCAF, and MAML1 were more expressed in human and sheep, while Notch co-repressor genes, namely HDAC1, HDAC2, and CTBP2 had a similar expression across species, indicating indeed an increased activity of this pathway in human and sheep CECs.We hypothesize that the downregulation of the Notch pathway could play a role in promoting the proliferative capacity of CECs in rabbits while inhibiting proliferation in human and sheep CECs.
Next, TGF- pathway visualization showed that human and sheep CECs had increased expression of TGF- receptors, TGFBR1 and TGFBR2, and TGF- ligand TGFB1 when compared to rabbit CECs.In addition to this, rabbit CECs had increased expression of LTBP1, a TGFB1 inhibitor, and a reduced expression of THBS1, an LTBP1 inhibitor.These results together with an increased expression of downstream genes FOS and SERPINE1 in human and sheep CECs compared to rabbit show an increased TGF- signaling in species with non-proliferative CECs.We hypothesize that the increased activity in the TGF- signaling pathway in sheep and humans may contribute to the growth arrest of CECs.
Overall, we found that the decreased activity of Notch and TGF- signaling in rabbit CECs could play a key role in their proliferation capacity.Based on previous findings, increased TGF- has been shown to suppress proliferation in CECs in vitro [19] and to induce senescence in CECs. [18]Moreover, previous studies have shown that TGF-2 present in the human aqueous humor suppresses CEC proliferation. [20,21]Furthermore, the use of the TGF- inhibitor SB431542 has been shown to aid human CEC proliferation in vitro by reducing the CEC fibroblastic transition. [22,23]Based on our results and previous findings, we suggest that TGF- inhibition can play a crucial role in regulating CEC proliferation and could be a target to develop novel regenerative medicine therapies.Furthermore, our data show that Notch signaling is more active in human and sheep CECs and could have biological implications in their non-proliferative profile.Notch pathway-associated genes such as HES1, NOTCH2, NOTCH1, and JAG1 have been previously detected at the transcriptomic level in human corneal endothelial samples, both at bulk RNAseq level [24] (DDBJ accession number E-GEAD-399), and single cell level [9,25] (GEO accession numbers GSE186433 and GSE155683) in line with our results.Studies performed in murine corneal endothelium have shown that the repression Notch signaling improves proliferation and reduces endothelial to mesenchymal transition of CECs, [17] suggesting Notch signaling might play a role in CEC proliferation.Nevertheless, further research is required to understand if Notch signaling can have a similar effect on CECs and could be a suitable pathway target to promote corneal endothelium regeneration.Interestingly, this finding is similar to research performed in the corneal epithelial  layer.Previous studies on cornea epithelial cells have shown that a downregulation of Notch and inhibition of the signaling pathway can cause an increase in their proliferative capacity and can contribute to epithelial regeneration. [26,27]

Conclusion
The proliferative capacity of CECs differs between species.While in larger mammals such as humans and sheep, CECs are arrested in a quiescent state, rabbit CECs retain a high proliferative capacity and can repopulate the endothelium upon injury.Nevertheless, there is a lack of fundamental knowledge to explain why CECs can proliferate in some species and remain quiescent in others.Gaining information on the transcriptomic differences between species with proliferative and non-proliferative CECs could allow the identification of crucial drivers of CEC proliferation to develop novel approaches to regenerate the human corneal endothelium.In this study, we performed a cross-species RNA sequencing analysis between animals with non-proliferative CECs, namely human and sheep, and animals with proliferative CECs, namely rabbit.Our data present a reference transcriptomic dataset of CECs for human, sheep, and rabbit samples.Our analysis proposes a pipeline for transcriptomic homology mapping across species and subsequent pathway analysis, which could be applicable for other species comparisons.Furthermore, our analysis revealed that 52 pathways are more active when comparing two species with proliferative CECs (sheep and human) with a species with quiescent CECs (rabbit), which could play a potential role in the regulation of CEC proliferation.
While our analysis reveals interesting findings at the transcriptomic level there are inherent limitations, which require further attention and discussion.First, it is crucial to secure public access to updated reference genomes.While human and sheep reference genomes have been recently updated with new release versions in 2022 and 2021 respectively, improving their coverage compared to previous versions, the rabbit reference genome (OryCun2.0)has not been updated since 2009, which might limit the coverage compared to other reference genomes.It is crucial to promote the development of updated reference genomes to ensure a similar coverage of the compared species.Secondly, the gene homology mapping across species relies on publically available datasets.Our analysis could only identify 9757 homologue genes across species, 63% of the total identified protein coding genes for rabbit and ≈55% of the protein coding genes in sheep and human.It is possible that genes without a homologue play a role on the proliferation capacity of CEC, which will not be portrayed in our results.Further research in identifying gene homologues across species can directly impact this research by providing platforms for more extensive data analysis.
It is important to highlight that the age-range of each species could have an effect on the reported results.While sheep and rabbit samples originate from young adults, human samples originate from older adults.To minimize the age-related effect, in our analysis we only selected pathways that have differential expression when comparing both sheep (young adult and nonproliferative) versus rabbit (young adult and proliferative) and human (old adult and non-proliferative) versus rabbit (young adult and proliferative).This approach was taken to identify differences that are likely to be due to the proliferation capacity of the CEC intrinsic to each species.The addition of another species with proliferative CEC into this dataset, such as mouse or rat CECs could strengthen the reported findings.
One of the main caveats in the field, which we also encountered in our study, is how to perform data normalization across each species' transcriptomic data.Each species' gene transcript length may vary, and should be taken into account when performing data normalization.Nevertheless, there is currently no existing method to perform such correction on transcriptomic datasets of three different species. [28,29]Taking into consideration such limitations, we performed a quantile normalization, which took into account library size and distribution, and performed a non-parametric Wilcoxon rank-sum test to identify more or less active pathways across species.Analyzing the data with typical packages such as DESeq2, which performs a parametric test on gene expression, would lead to the appearance of many false positive and negative results, which might impair the conclusions taken from such a study.Several studies are focusing on correcting such differences by mapping to a single reference genome, in the case of close species, with the risk of enriching for genes in the sample from the same species reference genome. [28,30]Future studies will shed light on this problem.
Besides analyzing the intrinsic interspecies transcriptome differences, a highly interesting follow-up study to our findings would be to compare the transcriptomic pathway changes in each species' CECs upon injury.After a scratch test of the cornea endothelium, the human, rabbit, and sheep CECs can be analyzed and compared to the transcriptome profiles presented in this study followed by an interspecies comparison of the differentially expressed genes upon injury within each species.This follow-up study would be a more targeted approach to identify differentially expressed genes upon injury and could provide additional information on the CEC post-injury CEC proliferation and healing capacity across species.
Overall, our study provides an automated pipeline for crossspecies transcriptome comparisons and pathway activity analysis, which can be applied to any interspecies study.Our data suggests that species with non-proliferative corneal endothelium are more similar than those species with a proliferative corneal endothelium.Furthermore, our data analysis reveals that 52 pathways have different activity when comparing species with proliferative CECs (rabbit) to species with quiescent CECs (human and sheep).

Experimental Section
Research-Grade Tissue and Ethical Statement: This study was performed in compliance with the tenets of the Declaration of Helsinki.All human research-grade corneal tissue (n = 5) was obtained from the ETB-BISLIFE Multi-Tissue Center (Beverwijk, the Netherlands), with informed consent from the next of kin.All human corneas were stored up to 22 days in organ culture media at 31 °C and had a minimum endothelial cell density of 2300 cells mm −2 .Organ culture media comprised the following: minimum essential medium supplemented with 20 mm HEPES, 26 mm sodium bicarbonate, 2% (v/v) newborn calf serum (Thermo Fisher Scientific), 10 IU mL −1 penicillin, 0.1 mg mL −1 streptomycin and 0.25 μg mL −1 amphotericin.All eyes from rabbits (n = 5) and sheep (n = 5) were enucleated from the animals used for other experimental procedures approved by the ethics committees of Maastricht University (Maastricht, the Netherlands), Utrecht University (Utrecht, the Netherlands), and Merck Sharp &  1.
Isolation of Human Corneal Endothelium: Of the five human corneal endothelial samples, four were discarded peripheral endothelial rims (1.5 mm wide) resulting from tissue preparation for Descemet's membrane endothelial keratoplasty (DMEK) according to clinical practice [31] and one (Human 5) was a whole endothelium stripped from a cornea deemed unsuitable for transplant (Table 1).
The four endothelial rims were lysed in 300 μL TRIzol within 48 h of DMEK preparation.From the whole cornea, the complete endothelium was manually stripped following lysis in 300 μL TRIzol.Briefly, the cornea was vacuum-fixed in a punch base (e.janach) endothelial-cell side up and trephined with a 10 mm Ø corneal punch at a fixed depth of 100 μm (e.janach).To delimit the endothelial trephined line, the cornea was stained with a trypan blue solution (0.4%) for 30 s, and washed with balanced salt solution ophthalmic irrigation solution.The corneal endothelium was then gently lifted using a DMEK cleavage hook (e.janach), fully stripped using angled McPherson tying forceps, and immediately transferred to TRIzol.All sample lysates were stored at −80 °C until further use.
Isolation of Sheep Corneal Endothelium: All samples (n = 5) were obtained from Texel sheep used for a dorsal titanium implant study from the animal facility at Maastricht University (Table 1).Five eyes were enucleated from 1 to 2-year-old female sheep within 4 h of sacrifice, and corneoscleral disks were excised from whole ocular globes.Corneal endothelial cells were enzymatically treated and scraped from the corneoscleral disks because stripping of the endothelium proved unsuccessful due to the tissue characteristics.For this purpose, corneoscleral disks were rinsed with Dulbecco's phosphate buffered saline, vacuum-fixed in a punch base (e.janach) endothelial-side up and treated with ≈150 μL of StemPro Accutase (Thermo Fisher Scientific) for 20 min at ambient temperature.Corneal endothelial cells were then gently scraped using a DMEK cleavage hook (e.janach) and rinsed off the corneoscleral disk with ≈2 mL of minimum essential medium (Thermo Fisher Scientific).The cell suspension was then transferred to a 15 mL tube, centrifuged at 800 × g for 5 min, and lysed in 300 μL TRIzol.All sample lysates were stored at −80 °C until further use.
Isolation of Rabbit Corneal Endothelium: Eyes were obtained from New Zealand White rabbits used for other experimental purposes from the Utrecht University animal facility (n = 3) and MSD animal testing facility (n = 2) (Table 1).The eyes were enucleated 30 min to 4 h after sacrifice, corneoscleral disks were excised from whole ocular globes, and the corneal endothelium stripped.In brief, corneoscleral disks were rinsed with PBS, vacuum-fixed in a punch base endothelial-cell side up and trephined with a 10 mm Ø corneal punch at a fixed depth of 100 μm.To delimit the endothelial trephined line, the cornea was stained with a trypan blue solution (0.4%) for 30 s, and washed with balanced salt solution ophthalmic irrigation solution.The corneal endothelium was then gently lifted using a DMEK cleavage hook, fully stripped using angled McPherson tying forceps, and immediately transferred to 300 μL TRIzol.All sample lysates were stored at −80 °C until further use.
RNA Isolation and Bulk RNA Sequencing: RNA isolation was performed following TRIzol isolation procedure, and RNA integrity was assessed on a 2100 Bioanalyzer (Agilent).RNA sequencing was performed at Single Cell Discoveries (Utrecht, the Netherlands).Libraries were prepared following the CEL-seq2 protocol [32] to enable sample multiplexing.Paired-end sequencing was performed on Illumina Nextseq500, high output 2 × 75 bp run mode, at a sequencing depth of 20 million reads per sample.
Data Analysis and Pre-Processing: BCL files resulting from sequencing were transformed to FASTQ and read 1 was used to identify the Illumina library index and CEL-seq sample barcode.Read 2 was aligned to the specific RefSeq reference genome of each species, namely: hg38 for human, OryCun2.0 for rabbit, and ARS-UI_Ramb_v2.0for sheep using the STAR genome aligner. [33]he data from all samples were loaded in R (version 4.1.2) [34]for analysis.Gene homology mapping from rabbit and sheep to human was performed using the orthogene R-package (version 1.1.0)with the g:profiler method. [35]The three datasets were combined and only genes with homologs in all three species were included in the downstream analyses.Once the data were combined, quantile normalization was performed using the limma package (version 3.52.2). [36] Pathway Analysis: The human pathway collection from WikiPathways (release 10-07-2022) was used for assessing pathway activities in the three species. [37]A non-parametric Wilcoxon rank-sum test in R was used for identifying pathways with different pathway activities between the species (p < 0.05, |effect size| > 1).A Benjamini Hochberg test was used to assess false discovery rate (FDR).This test predicted low and acceptable FDRs at a p-value of 0.05: human versus sheep = 0.0359, sheep versus rabbit = 0.0468 and human versus rabbit = 0.0259.Pathway effect size was calculated as the differences of the averaged gene counts for each pathway in the species compared.Disease-related pathways and pathways comprising less than 10 genes were not considered for data analysis.Pathway visualization was performed using Cytoscape (version 3.9.1). [38]

Figure 1 .
Figure 1.Schematic representation of the experimental pipeline and data analysis.

Figure 2 .
Figure2.A) Human, sheep, and rabbit transcriptomic datasets show comparable gene count distributions after data normalization.B) Correlation analysis shows that human and sheep samples have a higher correlation than rabbit corneal endothelium, confirming that rabbit corneal endothelial cells highly differed compared to human and sheep.C) Principal component analysis shows that samples originating from the same species cluster together, confirming the reproducibility of the sample isolation, and further confirms that rabbit corneal endothelial cells highly differed compared to human and sheep corneal endothelium.D) All samples had high expression of typical corneal endothelial cell markers, namely ALCAM, TJP1, NCAM1, SLC4A11, and ATP1A1; and absence/low expression of stromal markers (KERA, CD34, and LUM), epithelial markers (WNT7A and PAX6), and immune cell markers (ITGAM, CCR7, and CD19), confirming a successful isolation of the corneal endothelial cell transcriptome.

Figure 3 .
Figure 3. Volcano plots of the pathway analysis comparing species with non-proliferative corneal endothelial cells (human and sheep) with species with proliferative corneal endothelial cells (rabbit).Human versus rabbit A) have 94 pathways with different activity and sheep versus rabbit B) have 121 pathways with different activity.C) Venn diagram depicts common pathways with different activity level across species, the pathways that have a different activity in species with non-proliferative corneal endothelial cells versus species with proliferative corneal endothelial cells, namely human versus rabbit and sheep versus rabbit are highlighted with the red circle.D) Heatmap represents the effect size of pathways with different activity across human versus rabbit and sheep versus rabbit.Pathway effect size was calculated as the differences of the averaged gene counts for each species compared, e.g., human average-rabbit average (human vs rabbit).A positive effect size means higher pathway activity in sheep or human whereas a negative effect size means higher activity pathway in rabbit.

Figure 4 .
Figure 4. Visual representation of averaged read counts of human, sheep, and rabbit (left to right) of the A) Notch and B) TGF- signaling pathways shows an increased activity in species with non-proliferative CECs, namely human and sheep, and a lower activity in the samples with proliferative CECs, namely rabbit.

Table 1 .
Donor and animal tissue information.COD = cause of death.N/A = not applicable.Boxmeer, the Netherlands) and were performed in accordance with Dutch national and European guidelines.All animal ocular tissue was used within 4 h after sacrifice.All donor and animal tissue information is specified in Table