Abstract
Advancements in single-cell RNA sequencing (scRNA-seq) have enabled high-resolution tissue-specific transcriptomic profiling. Unlike studies on more readily obtainable tissues, research specifically requiring primary human retinal tissue has faced hurdles due to sample accessibility constraints. While single nucleus RNA sequencing (snRNA-seq) expands the utility of frozen post-mortem tissues, interpreting dynamic cellular states from this data requires consideration, as it primarily reflects the nuclear transcriptome and may not fully capture rapid post-transcriptional regulation or cytoplasmic events. Furthermore, inconsistencies in sampling location and a predominance of white donors in existing retinal atlases introduce spatial and racial biases, limiting the generalizability of current findings. These knowledge gaps limit the completeness of our understanding of retinal biology, particularly concerning complex processes like aging and the progression of diabetic retinopathy (DR).To address these limitations, this work performed integrative research of the human retina using scRNA-seq across 18 unfrozen retinal samples from Chinese donors, encompassing a wide age range and representing three clinical states: non-diabetic, diabetic without retinopathy, and DR. To minimize potential heterogeneity introduced by physical partitioning of the retina, our approach analyzed retinal tissue largely as a whole, without prior dissection into specific anatomical regions. Combined with the inclusion of both living and post-mortem donors, this method establishes an ethnically representative transcriptomic resource capturing a broad cellular profile.
The thesis progressed through three interrelated studies. First, we established a high-resolution single-cell atlas consisting of both living and post-mortem human retina tissues, identifying specialized cellular states including the ELF1-mlCone state, characterized by enhanced synaptic machinery and metabolic reprogramming. Then, we discovered distinct transcriptional dynamics between living and post-mortem samples. It underscored the limitations of post-mortem tissues for faithfully representing active cellular transitions.
The second study focused on microglial heterogeneity across DR progression, revealing three distinct microglial states, homeostatic, stress-response, and inflammatory, existing along a functional continuum rather than as discrete activation states. Dynamic bifurcating trajectories and three major functional modules that showed disease-specific activation patterns were identified. Then, we also uncover sophisticated neural-immune interactions, particularly between photoreceptors and microglia.
In the third study, we characterized age-related retinal changes, identifying cellular shifts (e.g., decreased rod-to-cone ratios) and pronounced transcriptional alterations, especially in Müller glial cells, which upregulated inflammatory and matrix remodeling genes. Pathway and transcription factor analyses revealed shared aging mechanisms and distinct cellular vulnerabilities. To elucidate aging signatures beyond conventional analysis, we employed a machine learning (ML) model, an approach not extensively utilized in prior human retinal aging transcriptomic research, which uncovered subtle yet consistent patterns that standard statistical methods might overlook. Our ML models achieved high classification accuracies (80-96%, average 89%) across retinal cell types, identifying both cell type-specific aging vulnerabilities and shared aging hallmarks.
To summarize, this work provides the profiling of human retinal cellular dynamics, disease progression, and aging changes in a Chinese population. The atlas serves as a resource for future studies and aims to refine the current understanding of human retinal biology. It can also serve to inform potential therapeutic strategies for retinal degeneration in underrepresented populations.
| Date of Award | 15 Nov 2025 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Weihua Meng (Supervisor) & Richard Rankin (Supervisor) |