This study presents the first RNA-seq study of human Tenon’s fibroblast-to-myofibroblast activation, a critical TGFβ-dependent process involved in postoperative wound healing following glaucoma filtration surgery. Myofibroblast activity drives wound healing both physiologically and pathologically. The progression of fibrotic diseases and cancers is dependent on myofibroblasts that remain pathologically activated rather than undergo apoptosis and vascular pathologies 9. TGFβ signalling notably plays a critical role in maintenance of tissue homeostasis and is hence highly regulated at several steps in both intracellular and extracellular microenvironments 10. As this analysis was conducted in myofibroblasts at a 5-day timepoint post-TGFβ treatment, where HTFs expressed a myofibroblast phenotype, we expected our findings to reflect the direct and indirect effects of TGFβ on activation and inhibitory signalling in myofibroblasts.
Growth factors and reactive oxidants are released from damaged cells and trigger the activation and proliferation of fibroblasts 11. CTGF is one such growth factor that has been found to play an important role in TGFβ-dependent myofibroblast differentiation and ECM production, indicated by its upregulated expression being highly enriched in the TGFβ-pathway (Fig. 8d). As an important downstream mediator of TGFβ, CTGF has been found to influence myofibroblast differentiation in orbital fibroblasts through increasing of fibronectin and a-SMA protein expression12. Hence CTGF may provide a safer target for suppression therapy compared to TGFβ. ACTA2 appears to be important in myofibroblast contraction and migration. ACTA2 is involved in the regulation of multiple genes relevant to fibrosis, including COL1, GFAP, TIMP1, TGFβ, and ET1 13. We found NOX4, a gene encoding for reactive oxygen species NADPH oxidase 4 enzyme, was upregulated by TGFβ. NOX4-dependent generation of hydrogen peroxide is a requirement for TGFβ induced myofibroblast differentiation, as well as ECM production14. TXNDC5 facilitates proper protein folding in the endoplasmic reticulum and mediates redox reactions via interacting with NADPH oxidase 11. These genes promote conjunctival fibrosis by activation of SMAD3-dependent TGFβ-signaling and lead to differentiation of HSCs into myofibroblasts, this results in considerable myofibroblast proliferation and ECM production 11. Notably, we observed the downregulated expression of SOD3 in our TGFβ treated samples. As known ROS scavengers, SODs act to reduce oxidative stress in both extra- and intra- cellular environments, with SOD3 specifically acting in the ECM where it has been found to reduce intracellular ROS levels 15. Hence our finding is consistent with findings of SOD3 deficiency contributing to liver fibrogenesis and TGFβ1 mediated EMT 16.
TGFβ utilizes components of the ECM such as fibronectin and integrins to communicate and control cell behaviour leading to myofibroblast differentiation. EDN1, FN1, TNC and ITGA11 were upregulated by TGFβ and have been reported in myofibroblast differentiation. EDN1 has been shown to induce resistance to apoptosis in fibroblasts and contribute to fibrogenesis by the abnormal persistence of the myofibroblast phenotype 17. TGFβ can induce expression of fibronectin 1 extra domain A (FN1 EDA) in fibroblasts. FN1 is not expressed in healthy tissue but is expressed during wound healing, fibroblasts detect FN1 EDA and its presence is required for TGFβ-mediated myofibroblast formation 18. TNC encodes for tenascin-C, an ECM glycoprotein that has been shown to engage integrins to elicit cell specific responses such as fibrotic responses including collagen synthesis and differentiation of myofibroblasts 19. TNC was upregulated in the TGFβ sample and has been shown to induce myofibroblast differentiation and migration 20. Integrins are the main cell-adhesion transmembrane receptors and bind proteins in the ECM such as fibronectin and transduce biochemical and mechanical signals21. ITGA11 was the most significantly upregulated integrin in our data and has been shown to co-localize with a-smooth muscle actin-positive myofibroblasts and was correlatively induced with increasing fibrogenesis in human fibrotic organs. ITGA11 knockdown has been shown to markedly reduce TGFβ-induced differentiation and fibrotic parameters 22. Our findings are consistent with other studies that reported TGFβ upregulation of ITGA11 and CTGF predominantly binding ITGA1123. We suggest that ITGA11 is a key regulator of myofibroblast differentiation in HTF and this association may hold potential as a therapeutic target.
Elastic fibre proteins play a pivotal role in guiding and facilitating elastogenesis involved in wound healing. The elastic fibre proteins implicated in this process include fibulins, fibronectins, fibrillins and LTBPs24. In our TGFβ sample, we recorded upregulated gene expression of FBLN5, FBN1, FN1, and LTBP1, alongside downregulated gene expression of LTBP4 and FLBN7. Of particular interest was FBLN5 which was in the top 20 upregulated DEGs. FBLN5 encodes the matricellular protein fibulin-5, which performs a vital role in elastogenesis and dysregulated expression of FBLN5 has been reported to occur in pseudoexfoliation glaucoma25. These findings correlate with the observed upregulated expression of FBLN5 being highly enriched in the EMICA pathway.
Typically, TGFβ is secreted in a form that is covalently bound to members of the latent TGFβ-binding protein (LTBP) family, observed as a large latent complex. It is via interactions between LTBPs, fibronectin and fibrillin that this complex can be deposited into the ECM 26. Hence LTBPs function to regulate the bioavailability of TGFβ, by either localising latent TGFβ in the ECM, or secreting latent TGFβ from cells 27. In Lu et al.’s 2017 study of human sclerodermal skin fibroblasts, the knockdown of LTBP4 resulted in the inhibition of collagen expression through canonical TGFβ/SMAD signalling, while also reducing extracellular levels of TGFβ 28. However contradictorily, findings from Su et al.’s recent 2023 LTBP4 knockdown study in a murine model of renal fibrosis revealed that an LTBP4 deficiency promoted mitochondrial dysfunction, increased inflammation, oxidative stress via increased ROS production and fibrosis 29. Hence, we suspect that the regulatory function of LTBP4 is complex and contextual, as it has been found capable of performing both anti- and pro-fibrotic functions. This makes LTBP4 a prime candidate for further research. In dermal fibroblasts, TGFβ exposure resulted in increased deposition of fibrillin-1 and fibronectin into the ECM because of myofibroblast activation 30. This associates our observed upregulation of FBN1 and FN1 in our TGFβ treated HTFs with myofibroblast activation.
TGFβ mediated epithelial to mesenchymal transition is important for myofibroblast transition. Epithelial to mesenchymal transition causes several changes to cellular phenotype and cytoskeleton remodelling and increased cell migration. We identified several important genes interacting with TGFβ for this process. NREP was upregulated in the TGFβ sample and regulates the expression of TGFβ1 and has been shown to regulate myofibroblast differentiation. NREP is thought to stimulate the expression of TGFβ1 by the methylation of NREP promoter and activating TGFβ1 5’/3’ UTR 31. Knockdown studies of SCUBE3, which was upregulated in the TGFβ sample, have implicated its involvement with lung cancer tumorigenesis and cancer metastasis. SCUBE3 binds to TGFβ type 2 receptor and activates TGFβ signalling triggering epithelial-mesenchymal transition 32. DACT1 was upregulated in the TGFβ sample; it has been shown to inhibit the Wnt signalling pathway. Therefore, DACT1 may release inflammatory factors that were previously suppressed by Wnt and promote local inflammation 33. Of the Wnt signalling pathways that were highly enriched in our gene set, the Wnt/Ca2+ pathway (Fig. 7b) is known to play a critical role in profibrotic and proinflammatory processes including actin polymerisation, cell adhesion, cell migration and NFAT signalling 34. A key component of this pathway involves the activation of the serine-threonine phosphatase enzyme calcineurin (CaN). RCAN2, encoded by RCAN2, has been identified as an inhibitor of the Wnt/Ca2+ pathway by binding CaN and subsequently inhibiting CaN’s protein phosphatase activity35. Findings from our GO analysis is consistent with these previous findings as the molecular function of RCAN2 was found to be involved in protein phosphatase regulator activity (Fig. 6e, GO:0019888). Hence, we suspect that our observation that RCAN2 is downregulated in the TGFβ sample is associated with myofibroblast activation through promotion of the Wnt/Ca2+ signalling pathway.
The regulation of cell cycle process and cell senescence pathways were significantly enriched and largely driven by genes CDKN2, CCNA1, CDKN2A, CDKN2D, and CDK4. TGFβ causes G1 phase cell cycle arrest and CDKN2B complexes with CDK4 to prevent CDK4 activation and producing cell cycle arrest 36. These genes are known cell cycle regulators and CDKN2b/p15 has been reported to take part in multiple pathologies such as primary open angle glaucoma and cardiac fibrosis36. Interestingly, TGFβ treatment resulted in downregulation of the majority of cell cycle and proliferation marker genes (Fig. 5). The only exception to this observation was the upregulated expression of both the tumour suppressor gene CDKN2B (Figs. 4b and 4c) and the cell cycle proliferator gene TMPO (Fig. 5). Lung carcinogenesis has been linked to upregulated expression of TMPO 37, while knockdown of TMPO in glioblastomas has been found to inhibit apoptosis-dependent cell proliferation and arrest cell cycle progression at the G2/M Phase 38. Collectively these findings are consistent with other studies reporting the role of TGFβ as a tumour suppressor and inhibitor of cell proliferation, such that cell cycle arrest and apoptosis evasion occur 39.
CDK6 is a known cell cycle regulator for G1 to S phase transition. Dysregulated expression of CDK6 has been reported in a variety of malignancies. Palbociclib is a commercially available selective inhibitor of CDK6 which has shown significant inhibition of myeloproliferative neoplastic progenitors and cells, and amelioration of bone marrow fibrosis in myelofibrosis 40,41. These findings are consistent with other studies and highlight the biological effects of TGFβ at the cellular level.
TGFβ1 and TGFβ3 were upregulated in the TGFβ treatment group. The 3 TGFβ isoforms have highly homologous receptor-binding domains and similar effects on target cells but with divergent primary amino acid sequences in the latency-associated peptide (LAP) domains. Isoform-specific function may be related to different expression in cell types and different extracellular environments 42. TGFβ1 was enriched in pathways such as TGFβ-pathway, myofibroblast differentiation, and collagen fibril organisation whereas TGFβ2 and TGFβ3 were not. This suggests that HTF induced fibrosis is largely a TGFβ1 driven process. TGFβ3 appears to play a role in regulation of epithelial to mesenchymal transition and TGFβ-pathway rather than myofibroblast differentiation which is consistent with previous reports of TGFβ3 expression being restricted to mesenchymal cells42.
Our findings highlight the important molecular mechanisms of TGFβ-signaling, particularly positive and negative regulation of pathway-restricted Smad protein phosphorylation (Fig. 6c; GO:0060393 & GO:0060394). The main genes upregulated in this biological process were GDF10, PMEPA1, LDLRAD4, INHBA, LEFTY2, BMP6 and SMAD7. Canonically TGFβ signal transduction occurs via activation of downstream mediators Smad2 and Smad3, which are negatively regulated by Smad7 as part of a negative feedback loop 43. Smad7 inhibits phosphorylation of R-Smads(2/3) by competitively binding TβR-I activated by TβR-II 44. We speculate that the upregulated expression of SMAD7 is associated with the upregulated expression and enrichment of the SMURF1 gene as their encoded proteins regulate each other’s functions. For Smad7 to perform its inhibitory function it is required to interact with Smurf1 to be exported from the nucleus to the cytoplasm 45. INHBA encodes inhibin beta-A, which functions as a subunit of activin A and is classified as a ligand of the TGFβ superfamily 46. INHBA attenuation in renal fibroblasts has been found to significantly reduce their expression of profibrotic markers, inhibiting cell migration and proliferation 47. To promote the TGFβ signalling pathway, INHBA is able to homodimerise and bind the activin type I/II receptor complex, which then can phosphorylate R-Smad proteins 48,49. Hence, we suspect that upregulated INHBA expression correlates with upregulated expression of ACVR1 (encodes type I receptor, ALK2) and ACVR2B (encodes type II receptor) to promote Smad-dependent TGFβ signalling.
PMEPA1, LEFTY2, and LDLRAD4 are all negative regulators of TGFβ-signaling. PMEPA1 and LDLRAD4 are part of the PMEPAI-family genes and encode transmembrane proteins that have been found to facilitate TGFβ signalling inhibition by competitively targeting SARA through sequestration of the Smad2/3 protein complex and attenuating R-Smad complex recruitment to the TβR-1 for phosphorylation 50,51. Once activated by TGFβ, LEFTY2 inhibits the phosphorylation of Smad2 and downstream heterodimerisation Smad4. Lefty also regulates the extracellular matrix components and inhibits the profibrotic effects of CTGF 52–54.
While the results achieved in this study advanced our understanding of the molecular mechanisms of ocular fibrosis following GFS, this study has its limitations. Although current literature supports RNA-seq analysis of HTF as a homogenous population, single cell RNA-seq would enable high resolution transcriptome analysis and potentially identify heterogeneity or subtypes within the HTF and myofibroblast population. This high-resolution transcriptome analysis could then lend to the identification of tissue biomarkers for ocular fibrosis, and together these tools could be instrumental in performing detailed clinical phenotyping. This would enable clinicians to not only identify groups of patients that are likely to scar more severely than others, but also work toward developing a tiered approach to antifibrotic therapy that is more personalised to the patient. Another limitation of the study was that specimen collection was performed at the time of GFS, from patients who were receiving medical therapy for glaucoma (Supplementary table 2). In using these glaucoma medications, most, if not all the patients were exposed to the preservative benzalkonium chloride, which is frequently associated with adverse reactions 55. Recorded adverse side effects of benzalkonium chloride include but are not limited to trabecular meshwork cell apoptosis, conjunctival and anterior chamber inflammation, corneal cytotoxicity, tear film instability, cataract development and macular oedema 56–58. Yet the effects of these medications on gene expression profile remain largely unknown. The small sample size of this study was also a limitation. Although a sample size of 3 patients is within current practice for genomic studies a larger sample size would increase the statistical power of the findings. Future knock down studies would be useful to validate the DEGs identified in this study to identify their suitability as targets for drug or gene therapy.
In investigating a model of HTF transition into myofibroblasts at the transcriptomic level, we gained greater insight into the novel molecular interactions and causal pathways of ocular fibrosis. This study establishes a gene signature of myofibroblast activation from HTFs that is uniquely characterised by genes involved in regulation of myofibroblast differentiation, collagen fibril organization, cell cycle arrest, TGFβ-signaling pathways, and wound healing organization. The results of this study establish an essential milestone for the future development of effective antifibrotic therapies targeting scar tissue formation following GFS.