Using scRNA-seq, we sequenced 17,914 cells from dissected limbal tissue (contains drainage structures) of 2-month-old mice. The dataset included 13,251 cells from strain C57BL/6J (B6) and 4,663 cells from strain 129/Sj (129). The data for both strains were integrated (Figure S1). Computational analysis revealed six distinct clusters of cells that were TM-containing, epithelia, pigmented epithelia (iris and ciliary body), endothelial, immune and neuron (Figure 1A). The identity of each of these cell clusters was based on various well-characterized marker genes (Figure S2A-B, Balasubramanian et al., 2024; Thomson et al., 2021; van Zyl et al., 2020). There were no major differences in the distributions of strain B6 and strain 129 cells within the relevant cell-type clusters (Figure S3A-C). Therefore, to improve statistical power, downstream analyses used the integrated dual-strain dataset.

Cells of cluster 1 expressed various TM genes, including Acta2 (α-SMA), Pitx2, Tfap2b, and Myoc (Figure S2A-B, Akula et al., 2020; Balasubramanian et al., 2024; de Kater et al., 1992; Ostojic et al., 2008; van Zyl et al., 2020). Unbiased sub-clustering of Cluster 1 revealed 10 distinct sub-clusters (Figure 1B), which included three TM cell clusters based on signature gene expression (Figure 1D-E). We named these clusters TM1, TM2, and TM3. Hierarchical clustering of these cell types indicated that TM1 and TM2 are more molecularly similar to each other than to TM3 (Figure 1C). The remaining seven clusters were identified as cells from other ocular tissue including the iris stroma, sclera, corneal keratocytes, corneal endothelium, ciliary muscle, pericytes, and Schwann cells (Figure S2C, Liu et al., 2003; Lopez et al., 2009; Monje et al., 2018; Nitzan et al., 2013; Patel and Parker, 2015; Stratton et al., 2017; Toyono et al., 2015; van Zyl et al., 2022).

Next, we integrated an additional limbal tissue scRNA-seq dataset (14,912 cells, Duke University) from 3 months old C57BL/6J mice with our initial 2 months old C57BL/6J dataset. Due to batch effects most likely based on differences in cell isolation techniques and environment (Columbia University vs. Duke University, see Methods), TM cells from these datasets occupied adjacent, partially overlapping, but not identical UMAP space (Figure S4A). Individual analysis of the Duke dataset also identified 3 TM cell subtypes with the same marker gene expression and enriched molecular pathways compared to the Columbia University data (Figure S4B-E, Table S1). In datasets from previous mouse limbal scRNA-seq studies, there was also a high degree of overlap of marker gene expression with our 3 TM cell clusters (Figure S5A-B, Table S2, Thomson et al., 2021; Ujiie et al., 2023; van Zyl et al., 2020). However, some TM subtype clusters reported in previous studies expressed genes that overlapped with both of our TM2 and scleral cell clusters. (Figure S5B-C, Thomson et al., 2021; van Zyl et al., 2020). To resolve this discrepancy, we performed IF for two of the discordantly annotated markers previously reported to be expressed in TM cells but present in our scleral cluster (CD34 and LY6C1). This immunolabeling showed that these markers are expressed in scleral but not TM cells (Figure S6A-B). In addition, in previous studies, cells with markers matching TM3 (e.g. Lypd1) were named uveal cells without IF confirmation (Thomson et al., 2021; Ujiie et al., 2023; van Zyl et al., 2020). IF revealed that our TM3 cells are neither located in the uvea nor abundant in the uveal adjacent TM but primarily reside in the anterior TM closer to the cornea (see below).

To further explore the sub-anatomical localization of the TM cell subtypes, we used a combination of IF and ISH to identify subtype markers. Each TM cell subtype had a higher percentage of cells expressing, and an increased average expression of, its respective subtype markers (all P < 1E-100) compared to other TM cells, Figure S7A). To analyze the distribution of subtypes, we first divided the TM into zones along its anterior/posterior and inner/outer axes (Figure S7B-E). The total area of TM occupied by each individual subtype marker was assessed in each zone on more than 150 tissue sections. This detected biases for the subtypes to be differentially localized in specific TM zones (Figure 2A-F). For instance, TM1 cells were significantly more frequent in the posterior and outer TM zones, while TM3 cells were overrepresented in the anterior and inner zones (all P < 0.01). The different marker molecules that were assessed for each TM cell subtype gave highly consistent results regarding zonal occupancy (Figure S8A-B, Figure S9). In addition, we confirmed the localization biases in 3D whole mounts of the mouse limbus (Kizhatil et al., 2014, for TM1: MYOC and TM3: α-SMA markers; Figure 2G-H). Collectively, these data indicate clear localization biases that differ for the TM1 and TM3 subtypes.

To better understand the location of TM cell subtypes, we analyzed marker expression patterns in 50-60 exceptionally high-quality sections for each marker of each TM cell subtype (see Methods). For this analysis, the TM was divided into eight regions (Figure S7F). Not only did this refine the localization bias of TM1 and TM3 cells, but it discovered that TM2 cells are most concentrated in the mid-posterior two-thirds of the TM (mid regarding inner-outer axis, all P < 0.01, Figure 2I-K). Despite their biased distributions, some cells of each TM subtype were detected in most TM regions.

To determine the molecular functions of TM cells, we first conducted gene ontology (GO) analysis using genes that were differentially expressed genes (DEGs) in all TM cells as a group compared to other sequenced cells (Table S3). Additionally, we used gene set enrichment analysis (GSEA) to assess whether these GO pathways are relatively enriched or underrepresented in TM cells. Compared to other cells, TM cells were enriched for pathways associated with extracellular matrix (ECM) structure and function, including signaling associated with collagens, proteases, integrins, and glycosaminoglycans (Figure 3A, Figure S10A, Table S4). TM cells were also enriched for growth factor signaling. Conversely, underrepresented pathways in TM cells were associated with desmosomes, peroxisomes, and certain cytoskeletal and plasma membrane elements. These pathways help define TM cells relative to other cells in the region.

Next, we compared TM cell subtypes to each other (Figure 3B-D). TM1 cells were most enriched for ECM pathways, particularly structural components such as collagens, integrins, and other ECM elements (Figure 3B, Figure S10B). The expression of several ECM genes that are reported to impact IOP are heightened in TM1 cells, including Type VIII collagens, fibronectin, and Ltbp2 (Ali et al., 2009; Desronvil et al., 2010; Roberts et al., 2020). Both TM1 and TM2 cells have enriched expression of glycosaminoglycan (GAG) genes (Figure 3B-C, Figure S10B-C). GAGs in the JCT and intertrabecular spaces form a gel-like consistency and changes in GAG abundance correlate with altered AH outflow resistance (Knepper et al., 1981; Knepper et al., 1996; Knepper and McLone, 1985). TM2 cells are uniquely enriched for the laminin complex, a key component of basement membranes (Aumailley, 2013). Pathways related to clearing debris and monitoring the extracellular environment including phagocytic vesicles, antigen presentation, and lysosomal function are also enhanced in TM2 cells. In contrast, TM3 cells are underrepresented for ECM-related pathways compared to other TM cells but are enriched for pathways related to actin binding and mitochondrial metabolism (Figure 3D, Figure S10D). The actomyosin system plays a crucial role in maintaining TM cell contractility and in regulating outflow facility (Rao et al., 2005; Syriani et al., 2009; Thieme et al., 2000; Tian and Kaufman, 2005; Zhang and Rao, 2005). The enrichment of mitochondrial metabolism pathways in TM3 cells emphasizes their energetic needs and their expected heightened susceptibility to genetic and environmental contexts that compromise metabolic functions. Importantly, genetic variation in mitochondrial/ metabolic pathways contributes risk for IOP elevation and glaucoma (Aboobakar et al., 2023; Khawaja et al., 2016; Zhang et al., 2023).

We next compared growth factor signaling and ligand-target interactions between TM cells subtypes and other cells. TM1 cells were enriched for transforming growth factor beta (TGF-β) pathway signaling (Figure 3B). Receptor ligand activity was enriched in TM2 cells while the membrane raft pathway (a site of signal transduction) was enriched in TM3 cells (Figure 3C-D). Next, we used LRLoop (Xin et al., 2022), which was developed based on NicheNet (Browaeys et al., 2020), to predict significant ligand-target interactions between TM cells and other local endothelial cells that are relevant to AH drainage (Figure 4A-D). TGF-β ligands (Tgfb1, 2, & 3) were predominantly expressed in TM1 cells. By contrast, TM2 cells expressed molecules critical for vascular function, such as Angpt1, Vegfa, and Edn3 (Figure S11A-D, Genovesi et al., 2022; Saharinen et al., 2017; Shibuya, 2011). These TM-secreted molecules have been directly associated with SC maintenance and IOP (Comes and Borras, 2009; Fujimoto et al., 2016; Reina-Torres et al., 2017; Thomson et al., 2020).

We next worked to relate our findings to human data. We compared marker genes for each of our TM cell subtypes to sequenced human TM cells from two previous studies (Patel et al., 2020; van Zyl et al., 2020). These previous studies differed in the number of human TM cell subtypes that they reported and in the identities of their clusters. Van Zyl et al. identified three TM cell subtypes that they called JCT, beam A, and beam B (van Zyl et al., 2020), whereas Patel et al. identified two subtypes that they called fibroblast-like and myofibroblast-like (Patel et al., 2020). TM1 marker genes were most enriched in Van Zyl et al’s human JCT cells (Figure S12A). Similarly, Dcn was expressed in mouse TM1 cells and was localized to fibroblast-like TM cells adjacent to SC by Patel et al. TM1 cell nuclei were typically shorter with greater sphericity than the elongated endothelial-like nuclei of TM3 cells (Figure S13A-D). This nuclear morphology was more consistent with a JCT than beam cell identity. Together, these data support a JCT identity for TM1 cells. TM2 cells were more closely related to human beam A and beam B cells than to the JCT cells reported by Van Zyl et al (Figure S12B). TM2 cells also shared expression of the RSPO2 marker gene with the human myofibroblast-like cells reported by Patel et al. Together, this suggests that TM2 cells are myofibroblast-like beam cells. TM3 cells also shared marker genes with Patel et al’s myofibroblast-like cells (TAGLN and ACTA2), suggesting TM3 cells are also beam cells. There was relatively equal TM3 marker gene enrichment across all human TM cell subtypes of Van Zyl et al (Figure S12C). Nuclei immediately adjacent to TM2 and TM3 cell marker staining had an elongated, endothelial-like shape (Figure S13A-C). This nuclear morphology was also consistent with the location of TM2 and TM3 cells on beams and analogous to the endothelial-like morphology of human beam cells (Stamer and Clark, 2017).

Next, we evaluated the expression of genes implicated in elevated IOP and glaucoma by genome-wide association studies (GWAS). TM1 and TM2 cells had similar expression levels of GWAS genes associated with IOP elevation and primary open-angle glaucoma (POAG) respectively (Figure S14A). TM3 cells had slightly less expression of these GWAS genes compared to TM1 and TM2. To better understand the genes driving GWAS gene enrichment in different TM subtypes, we studied pathways that are significantly enriched for GWAS genes (Figure S14B-C, see Table S6 for complete list). TM1 cells were the most enriched for genes involved in ECM organization, cell-substrate adhesion, and growth factor stimulus (Figure S14D). All 3 TM subtypes had similar expression of GWAS genes associated with hormone stimulus, Vegf signaling, and Rho kinase signaling. TM3 cells expressed lipid and carbohydrate mitochondrial metabolism pathway genes at higher expression levels and in more cells than did TM1 and TM2 cells (Figure S14E). The same was true for actin binding-associated GWAS genes in TM3 cells versus TM1 and TM2. Genetic variation in lipid and carbohydrate mitochondrial metabolism pathways is associated with IOP elevation and glaucoma (Khawaja et al., 2016).

To better understand the overlap and differences in transcriptional control of TM cell gene expression, we performed single-cell resolution multiome sequencing. Using limbal strip tissue isolated from a separate cohort of mice than those used for scRNA-seq, we simultaneously performed both single nucleus RNA sequencing (snRNA-seq) and snATAC-seq. We thus captured both gene expression and open chromatin data throughout the genome in the same cell. Using the snRNA-seq data, we clustered cells into various cell types. The generated clusters and their marker genes closely agreed with our scRNA-seq analyses (Figure 5A-B, Figure S15A). Notably, there was a significant correlation between gene expression (RNA-based clustering) and chromatin accessibility (ATAC-based clustering) with adjusted Rand index of 0.20 (p < 0.001, permutation test, Figure S15B-D). This included an increase in chromatin accessibility at the promoters of marker genes in their respective TM cell subtypes (Figure 5C-E). To discover the key transcriptional regulators within TM cells, we examined the correlation between the expression of each transcription factor (TF) and their chromatin accessibility at predicted binding sites within all TM cells. Various TFs that either activated or repressed transcriptional activity exhibited good correlations with gene expression across TM cells (Figure 5F). Such activating TFs include Tcf21, Arid3b, and Myc, while repressing TFs include Meox2, Junb, and Sox6. Notably, LMX1B, an important gene in glaucoma GWAS (Choquet et al., 2018; Gao et al., 2018; Gharahkhani et al., 2021; Khawaja et al., 2018; MacGregor et al., 2018; Shiga et al., 2018), is among the most strongly activating TFs in TM cells.

We evaluated the utility of our new TM cell atlas to better understand the functions of Lmx1b in TM cells and how Lmx1b mutations lead to glaucoma. To do this, we analyzed scRNA-seq data generated for limbal cells from mice with a dominant mutation in Lmx1b (Lmx1b^V265D/+). These mutant C57BL/6J mice develop early onset IOP elevation (Cross et al., 2014; Tolman et al., 2021; Zhang et al., 2024). We compared gene expression in V265D mutant TM cells after integration with our wild-type data (all data B6 background and postnatal day 60, Figure 6A, Figure S16A). Results show that Lmx1b is minimally expressed in TM1 and TM2 but is highly expressed in TM3 cells (Figure 6B). Next, we analyzed DEGs and pathway differences between WT and V265D/+ cells across each TM cell subtype. Across all TM subtypes, various genes related to ECM function and growth factor signaling were downregulated in mutants, including glycosaminoglycans and insulin growth factor binding proteins (Figure 6C, Table S6 for complete list of pathways). Pathways enriched in mutant TM cells were associated with ribosome and calcium signaling. We then examined the differential responses to the Lmx1b mutation by analyzing pathways dysregulated in each TM subtype that were not common across all TM cells (Figure 6D-F, Figure S16B). V265D mutant TM1 cells had a downregulation of collagen synthesis/metabolism gene expression relative to WT cells, particularly in the production and turnover of fibrillar collagens. TM2 cells with Lmx1b mutations had upregulation of immune signaling pathways. The highly Lmx1b expressing TM3 cells exhibited perturbed mitochondrial metabolism, with a downregulation of genes in complex I and ATP synthase, and upregulation of genes involved in ubiquitin pathways. These results document altered mitochondrial function and proteostasis in TM3 cells.

Our findings implicated perturbed metabolism within TM3 cells as responsible for IOP elevation in an Lmx1b glaucoma model. Nicotinamide (NAM) is known to boost nicotinamide adenine dinucleotide (NAD), supporting healthy metabolism/ mitochondrial function (Schondorf et al., 2018; Williams et al., 2017b). Thus, we hypothesized that the metabolic abnormalities in TM3 cells underlie the IOP elevation that causes glaucoma, and that NAM treatment may protect the Lmx1b mutant TM. Beyond LMX1B, this hypothesis is relevant more commonly to POAG treatment because metabolism-relevant genes are implicated by GWAS (and their expression is enriched in TM3 cells, Figure S14E). To test our hypothesis, we treated mice by adding NAM to their drinking water starting at postnatal day 2 and continuing into adulthood. Results showed that NAM supplementation significantly protected against anterior chamber deepening (ACD, a symptom of IOP elevation, Figure 7A-B) and IOP elevation (Figure 7C-D). These results support NAM as a treatment to prevent TM damage and IOP elevation in LMX1B-mediated glaucoma with potential general relevance to POAG.

Here we provide evidence for three TM cell subtypes in mice that are robust across datasets from different institutions, using different tissue processing and sequencing protocols. Although others (Thomson et al., 2021; Ujiie et al., 2023) subdivided TM cell subtypes beyond the three defined in the present study, at this point we decided against further subdivision (See Figure S5) for the following reasons. We lacked confidence in the higher resolution clustering, as it was not consistent across datasets, did not produce unique marker gene sets, and could have resulted in over-clustering due to stochastic or stress-induced differences between subpopulations of cells. Moreover, the number of reported mouse TM cell clusters in previous studies were inconsistent. Although three TM subtypes were reported by Van Zyl et al. and Ujiie et al. (Ujiie et al., 2023; van Zyl et al., 2020), the molecular features of cells in these clusters were not the same across these studies. Differences between studies may be due to batch effects, different computational methods and possibly over-clustering. Most importantly, there was very limited validation by IF, IHC, or other methods in previous studies. Only the JCT type of TM cell in these previous studies was assessed by means other than scRNA-seq (IF and ISH, van Zyl et al., 2020). Additionally, Thompson et al. treated cells with Y27632, a ROCK inhibitor, which could alter TM cell transcriptomes. It is also important to point out that both Thompson et al. and Van Zyl et al. used albino mouse strains, which can impact TM cell development and function (Libby et al., 2003). Finally, Ujiie et al. sequenced TM cells at 3-4 weeks, which is prior to full maturation of the TM structure (Smith et al., 2001), while our ongoing developmental studies show substantial molecular maturation of TM cells between P21 and P60.

As Van Zyl et al’s data is deposited in the Broad Institute’s single-cell portal, we further compared their data with ours. Although both Van Zyl et al. and our current paper identified 3 TM subtypes, our work does not fully agree. Some of the cells named beam A and beam Y by Van Zyl et al. express markers present in our TM2 cell subtype, while others express markers that are absent in TM2 cells but present in our scleral cells. Our IF studies confirm that these discordant markers are not expressed in TM cells, but instead show that they are in fact scleral cells. These scleral cells highly express fibroblast markers including Mfap5, Clec3b, and Tnxb, suggesting they are scleral fibroblasts. Cells initially named uveal cells by Van Zyl et al. (and subsequently by Ujiie et al. and Thompson et al.) express markers of our TM3 cells. Consistent with our gene expression data, our IF shows that these Van Zyl markers are expressed by cells that are primarily located in the TM region that is closer to the cornea and not in the uvea or the TM region adjacent to the uvea. Thus, our current study resolves some previous inconsistencies in classifying mouse TM cell subtypes. It lays a firmer foundation for continued understanding of TM cell types and their biology. Moving ahead, sequencing cells at greater depths will enhance TM cell characterization and improve comparisons between datasets.

As a group, all TM cells share various molecular properties that distinguish them from other ocular cells. Our analyses determine the expression of IOP, and glaucoma genes identified by GWAS in TM cell subtypes as well as in other limbal cells. Pathways enriched across all TM cell subtypes based on RNA-seq are largely related to ECM (including collagens, integrins, and glycosaminoglycan pathways), cell-cell signaling, and growth factor signaling. This fits with an essential role of the TM in both ECM synthesis/turnover and in signaling to the SC to maintain IOP (Balasubramanian et al., 2024; Fuchshofer and Tamm, 2009; Keller et al., 2009; Stamer and Acott, 2012; Thomson et al., 2021). Additionally, genes associated with intermediate filaments, which function in determining cell shape and structure and can also act as mechanical stress absorbers (Herrmann et al., 2007), are underrepresented in TM cells. Lower levels of intermediate filaments may enable TM cells to alter their shape to respond to changes in IOP.

Mouse TM1 cells closely resemble sequenced human JCT cells. Like human JCT cells, the majority of TM1 cells are located nearest to the SC inner wall, particularly adjacent to the posterior of the mouse SC. Human JCT cells are generally accepted to be embedded in a more diffuse ECM and to not cover the collagenous trabecular beams (Keller and Acott, 2013; Stamer and Clark, 2017; Vranka et al., 2015). In vivo, a proportion of TM1 cells have a shorter, more spherical nuclear morphology, consistent with JCT cell identity. Importantly, the JCT region is critical in determining resistance to AH outflow and in regulating IOP (Acott and Kelley, 2008; Ethier, 2002; Johnson, 2006; Stamer and Acott, 2012). JCT cells have fibroblast-like properties, including the secretion of ECM proteins and degradation enzymes to support continuous ECM remodeling (Acott et al., 1988; Keller et al., 2009; Stamer and Clark, 2017). The subset of TM1 cells that lie distant from SC have more beam cell-like, elongated nuclear morphology. This suggests that either our TM1 cluster contains more than 1 TM cell subtype or that the morphology of a single TM1 cell subtype is influenced by local environment. This needs to be resolved by future studies. Compared to the other TM cell subtypes, TM1 cell pathways are enriched in both ECM structural molecules, such as collagens and integrins, and TGF-β signaling, which increases ECM production (Munger and Sheppard, 2011). Notably, excessive TGF-β signaling and ECM production can lead to fibrosis and elevate IOP (Biernacka et al., 2011; Frangogiannis, 2020; Meng et al., 2016). These data suggest that TM1 cells play an important role in the constant remodeling of the ECM. However, dysregulation of these pathways can be pathogenic.

TM2 cells molecularly resemble previously sequenced human TM beam cells (Patel et al., 2020; van Zyl et al., 2020). This includes overlap of expression of top TM2 signature gene expression with sequenced human beam cells. Consistent with a beam cell identity, TM2 cells are located more towards the mid and inner TM than most TM1 cells. TM2 cells have an elongated nuclear morphology, consistent with the previously established beam cell morphology of cells in the regions they occupy (Overby et al., 2014; Smith et al., 2001; Stamer and Clark, 2017). Beam cells have endothelial-like properties, such as maintaining the patency of the AH outflow tract through the production of antithrombotic molecules. Proteases and molecules classically known as antithrombotic and thrombolytic maintain fluid flow through the TM by preventing build-up of protein aggregates to maintain fluid drainage (Acott and Kelley, 2008; Stamer and Acott, 2012). Such molecules include glycosaminoglycans, like heparin, the serine protease tPA and matrix metalloprotease that are all highly expressed in TM2 cells. Beam cells are also phagocytic and this activity along with turnover of ECM further cleans the TM (called filtering) to enhance fluid drainage (Stamer and Clark, 2017). Human Beam cells also participate in antigen presentation and modulation of inflammation through cytokines and major histocompatibility proteins (Shifera et al., 2010; Stamer and Clark, 2017), consistent with TM2 cells being enriched for these immune pathways/ molecules. Additionally, our data suggest that signaling from TM2 cells to SC endothelial cells is mediated by soluble molecules including Vegfa, Edn3, and Angpt1, which are critical for SC maintenance and function. This links TM2 cells to ocular development and glaucoma. VEGF receptor and ANGPT signaling genes modulate SC development/ function as well as contribute to developmental and primary open-angle glaucoma (Aspelund et al., 2014; Kabra et al., 2017; Khawaja et al., 2018; Kizhatil et al., 2014; MacGregor et al., 2018; Souma et al., 2016; Thomson et al., 2020).

TM3 cells also have a location and elongated nuclear morphology resembling beam cells with enriched localization in the inner, anterior TM. TM3 cell-enriched processes include actin binding, and modulation of the actomyosin system, which is known to modulate outflow resistance and IOP (Rao et al., 2001; Yu et al., 2008). As such, TM3 cells express a large percentage of glaucoma-associated genes involved in actin binding at a high level. Importantly, TM3 cells are enriched for various mitochondrial metabolic and antioxidant pathways, including those associated with IOP elevation and glaucoma (Aboobakar et al., 2023; Khawaja et al., 2016). Inhibiting metabolic pathways alters AH outflow resistance (Reina-Torres et al., 2020). Since maintaining contractile tone is an energy-intensive process (DeWane et al., 2021), this could explain the enrichment of energy metabolism pathways in TM3 cells compared to other TM cells.

In addition to being enriched in mitochondrial/ energy metabolism processes, TM3 cells express Lmx1b at significantly higher levels than both the other TM cell subtypes and other limbal cells. Importantly, in heterozygous mutant V265D/+ mice, TM3 cells had pronounced gene expression changes that implicate mitochondrial dysfunction and oxidative stress, but that were absent or much lower in other cells including TM1 and TM2. Together, these data suggest that mutation of Lmx1b has a primary effect on metabolism and oxidative stress in TM3 cells that subsequently leads to IOP elevation and then glaucoma. Relevantly, homozygous conditional knockout of both Lmx1b and Lmx1a as well as siRNA knockdown of Lmx1b alone impact mitochondrial functions in neurons and may contribute to Parkinson’s disease (Bergman et al., 2009; Doucet-Beaupre et al., 2016; Jimenez-Moreno et al., 2023). Our data extend these published findings by showing that inheritance of a single dominant mutation in Lmx1b similarly affects metabolism in TM3 cells. In addition to modulating mitochondria, LMX1B was recently identified as a transcription factor that regulates responses to cell stress. This includes promotion of autophagy under stress conditions to enable recycling of cellular resources to maintain cellular functions and enable survival (Jimenez-Moreno et al., 2023). Thus, mutation of Lmx1b may also result in a deficiency of the beneficial autophagic response to stress (Jimenez-Moreno et al., 2023).

Although further experiments are needed, dysfunctional autophagy may further prolong and exacerbate TM cell stress, resulting in more extensive depletion of cellular resources, exacerbated cellular dysfunction, and IOP elevation. Other consequences of the Lmx1b mutation that were evident in the V265D/+ cells may also contribute to IOP elevation, including the overexpression of genes associated with ribosomes and calcium signaling, and depletion of genes involved with ECM synthesis and metabolism. However, these processes were not limited to TM3 cells or even to cell types that express detectable Lmx1b, suggesting that they are secondary damaging processes that are subsequent to the initiating, Lmx1b-induced, metabolic disturbances and cell stress in TM3 cells. Thus, we hypothesize that mitochondrial abnormalities in TM3 cells are of primary importance to IOP elevation in V265D/+ glaucomatous mice.

Based on our findings in TM3 cells, we tested if mitochondrial and metabolic disturbances are important in IOP elevation by administering nicotinamide (NAM). NAM is a form of vitamin B3 that is well-established to boost NAD⁺ levels, promote mitochondrial health and energy metabolism, relieve oxidative stress and thus promote cellular resilience (Gomes et al., 2013; Imai and Guarente, 2014; Jang et al., 2012; Mouchiroud et al., 2013; Verdin, 2015; Williams et al., 2017b, Jang et al., 2012; Kang and Hwang, 2009; Song et al., 2021). Supporting our hypothesis of primary metabolic dysfunction in TM3 cells, NAM treatment strongly protected V265D/+ mice from IOP elevation. Thus, treatments that support normal metabolism may protect from IOP elevation in individuals with LMX1B variants including both developmental glaucoma and POAG patients (Chen et al., 1998; Choquet et al., 2018; Choquet et al., 2017; Gao et al., 2018; Gharahkhani et al., 2021; Khawaja et al., 2018; MacGregor et al., 2018; Sweeney et al., 2003; Vollrath et al., 1998). As NAM and other NAD⁺ boosting treatments are also known to be directly neuroprotective in glaucoma (De Moraes et al., 2022; Hui et al., 2020; Williams et al., 2018; Williams et al., 2017a; Williams et al., 2017b), our current data suggest that such treatments will have a double benefit for patients. As mitochondrial defects are becoming broadly associated with glaucoma risk and progression (Abu-Amero et al., 2006; Lo Faro et al., 2021; Petriti et al., 2024), these treatments could have wide-ranging potential to complement and augment existing IOP-lowering therapeutics through a completely distinct mechanism.

Our results provide a thorough molecular characterization of mouse TM cells, providing much needed molecular information on subtype specializations in regulating IOP and the roles of specific subtypes in glaucoma. This comprehensive TM atlas provides a new foundation to guide development of new glaucoma treatments. We validate its utility by implicating metabolic changes in a single TM cell subtype in both mouse and human glaucoma, with success of a metabolic treatment targeting this subtype in a glaucoma model.

Experimental animals were either C57BL/6J (Stock# 664) or 129/Sj (unique substrain of 129, Tolman et al., 2021) inbred mice. The Lmx1b^V265Dmutation was discovered in an N-ethyl-N-nitrosourea (ENU) mutagenesis screen (Cross et al., 2014; Thaung et al., 2002; Tolman et al., 2021). These mice were then backcrossed to the C57BL/6J (Stock #000664) for at least 30 generations. All mice were treated in accordance with the Association for Research in Vision and Ophthalmology’s statement on the use of animals in ophthalmic research. The Institutional Animal Care and Use Committee of either Columbia University or Duke University approved all experimental protocols performed at their respective institutions.

Nicotinamide (NAM or niacinamide; Sigma-Aldrich, St. Louis, Missouri) was dissolved in the standard institutional drinking water to a dose of 550 mg/kg/day (low dose) based on the average volume mice consume. Untreated groups received the same drinking water without NAM. Water was changed once per week. Treatment was started at postnatal day 2. Births were checked daily between 9am-12pm to determine the pup’s age.

Columbia animals: Eyes were pooled from 4 animals to generate 1 sample (8 eyes total per pool). Two independent samples were processed. Mice were 2 months of age. Sex was balanced in each group.

Duke animals: One sample was generated from a pool of 4 animals (8 eyes total). All mice were male. Mice were 3 months of age.

Dissections: After euthanizing the animals, eyes were enucleated into fresh Dulbecco’s Modified Eagle Medium (DMEM; Gibco) and subsequently dissected in DMEM. All dissection tools and surfaces were treated with RNAse zap (Ambion). For dissections, we removed the posterior eye by cutting between the limbus and sclera. The lens was then removed. Finally, we made a small circular cut in the anterior cornea, removing the middle 1/3^rd of the area. Only limbal tissues were collected and all other tissues were removed. The remaining limbal tissue was minced and prepared for digestion.

Single-cell RNA sequencing (Columbia University): Tissues were digested enzymatically using Papain and Deoxyribonuclease I (Worthington Biochemical Corporation, Cat. no. LK003153) for 20 min at 37°C and stopped using Earl’s balanced salt solution (EBSS, Thermo Fisher Scientific, Cat no. 24010-043). Cells were triturated using an 18-gauge needle, centrifuged at 300g at 4 °C, washed with cold DMEM, and filtered using a 100 μm Flowmi Cell Strainer (H13680-0040, Bel-Art. Cells were resuspended in cold DMEM and placed on ice immediately. Cells were counted using Countess II automated cell counter and processed for 10x library preparation immediately (see below). All data will be deposited in the Gene Expression Omnibus (GEO) database and single cell portal.

Single-cell RNA sequencing (Duke University): Single cell dissociation was performed using Collagenase IV (Worthington Biochemical Cat. no. LS004188), Dispase II (Sigma Cat. No. D4693) and DNAse I (Roche Cat no. 50225500) for 60 min and then trypsin/EDTA for 10 min at 37°C. Cells were triturated using a 1 ml pipette tip, centrifuged at 300g at 4°C, washed with cold DMEM containing 10% FBS, and filtered using a 70 μm cell strainer (Fisher). Cells were resuspended in cold DMEM containing 5% FBS and placed on ice immediately and sent to Duke Genome Core Facility. Cells were counted using Countess II automated cell counter (the cell viability was 88%) and then immediately processed immediately (see below). All data will be deposited in the Gene Expression Omnibus (GEO) database and single cell portal.

Multiome (Columbia University): Limbal strip samples for multiome analysis were snap-frozen on dry ice immediately following dissection and stored at −80C until processing. To isolate nuclei, tissue was homogenized in 100 μL of chilled lysis buffer (10 mM Tris-HCl, pH 7.4, 10 mM NaCl, 3 mM MgCl2, 0.1% NP40, 0.1% Tween20, and 0.01% digitonin – from (Corces et al., 2017) supplemented with 1% BSA). Afterwards, samples were incubated for 5 min on ice, gently agitated by pipetting, and incubated for an additional 10 minutes on ice. We then added 1 mL of chilled wash buffer (10 mM Tris-HCl, pH 7.4, 10 mM NaCl, 3 mM MgCl2, 0.1% Tween 20, 1% BSA) to the cells and centrifuged at 500 rcf for 5 min at 4°C. Cells were resuspended in chilled Diluted Nuclei Buffer (PN-2000153, 10x Genomics) and filtered through a 40 μm Flowmi Cell Strainer (H13680-0040, Bel-Art). Cells were resuspended in cold DMEM and placed on ice immediately. An aliquot of nuclei suspension was stained with SYTOX green to count nuclei using Countess II automated cell counter and processed for 10x library preparation immediately (see below). All data will be deposited in the Gene Expression Omnibus (GEO) database and single cell portal.

Single-cell and single-nucleus RNA-seq was performed at the single-cell sequencing core in Columbia Genome Center (Columbia University) or Duke Molecular Genomics Core (MGC, Duke University). Single cells or nuclei were loaded into chromium microfluidic chips with v3 chemistry and barcoded with a 10x chromium controller (10x Genomics). RNA from the barcoded cells was reverse-transcribed, and sequencing libraries were constructed with a Chromium Single Cell v3 reagent kit (10x Genomics). Sequencing at both institutions was performed on NovaSeq 6000 (Illumina).

Sections: Microscopy of sectioned postnatal eye slides was performed using an LSM SP8 confocal microscope (Leica) and 40×1.1 NA water immersion objective.

In situ: Images were taken using a Nikon Eclipse 90i confocal laser-scanning microscope (Melville). Z-stack images of 1 µm thickness, using the 40X objective lens, were taken of each section and converted into maximum projection images that were used for analysis.

Whole mounts: Microscopy was performed using an LSM SP8 confocal microscope (Leica) using a 63×1.4 NA glycerol immersion objective. The Mark and Find mode was used to automate collection of images, generating a folder full of Z stacks (Z = 0.3 µm) at various individual overlapping positions around the limbus. We examined n = 6 eyes total. We imaged a certain distance around the ocular circumference, with an equal representation from each ocular quadrant (1.5-2.5 mm total).

3D rendering: We followed the guidelines detailed in Kizhatil et al for visualizing SC and TM in 3D. Individual confocal Z stacks (.lsm files) were processed directly using Imaris 9.5 (Oxford Instruments, Carteret, NJ). The maximum intensity projection setting in Surpass mode of Imaris was used for 3D rendering. Controls and experimental image sets were treated identically. Images were oriented so that structures of interest were visible. For whole-mounted eyes, multiposition Z stacks generated using the Mark and Find setting on the confocal microscope were first imported into Imaris and converted into Imaris files (.ims files). The resulting Imaris files were then stitched to generate a comprehensive Z stack encompassing the limbus using ImarisStitcher 9.5 (BitPlane AG, Zurich, Switzerland). The stitch was exported as an Imaris file and processed appropriately. For sections, images were taken in the Z-plane in focus. When making figures, the snapshot feature of Imaris was used to collect images at high resolution (1024×1024 pixels, 300 dpi).

We defined the TM as the region between the inner wall of Schlemm’s canal (SC, outer TM) and the anterior chamber (inner TM), extending from the anterior edge of the pars plana (posterior TM) to the posterior edge of Descemet’s membrane of the corneal endothelium (anterior TM). The SC was defined using a panel of validated markers that varied to allow for different antibody species and different secondary antibodies to be used in conjunction with the SC markers (See Table S7 for list of antibodies, Kizhatil et al., 2014). Sections were only included in the scoring analysis if they met the following criteria:

Absence of obvious sectioning artifacts or abnormalities in the SC or TM region.

Additional analysis was performed by selecting a subset of examined sections with staining for select markers (TM1: MYOC antibody, TM2: CRYM antibody, Inmt in situ, TM3: α-SMA antibody). These sections were of exceptionally high quality and had: 1) Anterior-posterior TM length between 125-200 µm. 2) Inner-outer TM distance is between 15-30 µm.

With these selected sections, we subdivided the TM into eight different zones along both inner/outer and anterior/posterior axes (Figure S7F). For each individual section, we used Imaris to measure the anterior-posterior distance and divide the TM into three equal sectors (anterior, central, and posterior. Within each sector, the average depth of the TM was measured in Imaris. This calculation was used to divide the central and posterior sectors into three zones each (inner, intermediate, and outer). Because the anterior TM depth is thinner, we divided the anterior sector into two zones (inner and outer). In total, there were eight distinct zones.

The percent of total marker area was calculated for each of the eight zones for each individual section as described above. Percent of marker area in each zone was compared as shown in Figure 2 by one-way ANOVA followed by Tukey’s honestly significant difference test.

For CD34 and LY6C1 expression, we examined greater than 60 sections from 4 eyes (CD34) and 3 eyes (LY6C1) respectively. The observed locations of these markers within the iridocorneal angle are described.

To analyze the expression of markers within the TM, we segmented the TM based on anatomy. Because various anatomical features including the pars plana were removed during the whole mount process, we used the SC (marked by PECAM) as the major anterior-posterior landmark. The TM was defined as the group of cells inner to the SC and not outside the anterior or posterior boundary of the SC.

Anterior segment whole-mounts were used to examine nuclear morphology in 3D (TM1: MYOC, TM3: α-SMA, all antibodies). The TM was segmented as described above and the ‘Surface’ feature in Imaris used to demarcate TM cell nuclei. Nuclei that were analyzed were immediately adjacent to the immunolabeled TM marker and had to be well spaced from other nuclei. To ensure accuracy we used manual inspection to exclude: 1) Nuclei that were very close to other nuclei or groups of nuclei and 2) nuclei of cells that were labeled with a Schlemm’s canal endothelial cell marker (PECAM) and/ or that projected into the lumen of SC. The ‘Sphericity’ tool was used to determine the degree of sphericity in Imaris. For each cell subtype, we assessed a total 60 nuclei from 6 eyes, assessing the first 10 randomly selected nuclei that met our inclusion criteria in each eye. Groups were compared by Student’s t-test.