Abstracts

Breakfast will be available for attendees during this time. Seating is open.

Opening remarks, history of the Kambara Symposium, and overview of the day’s program.

Amitha Domalpally, MD, PhD

Moving AI from Code to Clinical Trials

Selection of ideal trial population remains a major bottleneck in ophthalmic clinical trials, contributing to delayed enrollment, increased costs, and high screen failure rates. Advances in artificial intelligence (AI), particularly imaging-based models, offers a solution through automated prescreening and cohort enrichment. This presentation on AI-based prescreening models will focus on implementation, tracing the path from algorithm development to real-world deployment within clinical trial settings. Using examples from diabetic retinopathy and geographic atrophy, the talk will discuss how AI can be used to identify imaging based inclusion criteria speeding trial recruitment. Key considerations including reference standards, AI performance, human oversight and workflow integration will be highlighted. The presentation will focus on practical lessons learned in translating AI from code into clinical trial operations.

Three short research presentations followed by a moderated Q&A.

Roomasa Channa, MD, FASRS

Real World Validation of AI Algorithm for GA area Measurement

Purpose: Optos ultrawide field imaging is widely used in routine care to monitor Geographic Atrophy (GA) progression due to ease of imaging and comfort to patients. Our AI algorithm for GA quantification was originally developed and validated using high quality clinical trial images from the AREDS2 OPERA study and showed high performance metrics. Real-world deployment, however, requires performance assessment on heterogeneous clinic images. This study evaluates the real-world performance of the Optos-based AI algorithm for automated GA measurement using clinical images from a large tertiary care hospital prior to implementation in practice.

Methods: Using ICD-10 codes for GA, UW Clinical Research Data Services identified eligible patients and transferred de-identified images from the Picture Archiving and Communication System (PACS) to Platform X, a secure NIST and HIPAA compliant cloud-based workstation for research. Because CPT codes do not distinguish Optos FAF from other fundus images, all images meeting ICD-10 criteria for GA were transferred. After identifying Optos images, the validated AI model was deployed within Platform X, and automated GA segmentations were generated. Each AI-derived mask was reviewed by a certified grader and classified as accurate, minor errors or major errors with reasons documented.

Results: A total of 8751 images met the ICD-10 based query. Of these 1066 images were identified as potential Optos FAF images of which 174 images had GA upon review. The remainder represented AMD without GA, atrophy from inherited retinal diseases, or non-Optos images and were excluded. Among the 174 GA images, the AI algorithm was accurate in 93 (53.4%), showed minor errors in 33 (18.9%), and major errors in 48 (27.6%). Minor errors reflected small areas of missed GA in 9 (5.2%) or minimal inclusion of optic nerve in 24 (13.8%). Major errors included substantial missed GA in 19 (10.9%), inclusion of peripheral atrophy in 7 (4.0%), and inclusion of non-GA areas in 22 (12.6%). While initial performance is promising, it provides a foundation for continued refinement to achieve higher accuracy.

Conclusion: The Optos AI algorithm produced acceptable GA segmentation (i.e. no or minor errors) in over 70% of clinical PACS images, demonstrating feasibility for real-world use. Automated segmentation may enable real-time GA monitoring in routine clinical practice, supporting efficient assessment of disease progression

Barb Blodi, MD, MALD

Comparison of Fundus Autofluorescence Systems for Geographic Atrophy Area Measurements

Purpose: Fundus Autofluorescence (FAF) imaging with the Heidelberg Spectralis is the predominant modality used in clinical trials to assess geographic (GA) lesion area and progression. Ultrawidefield FAF platforms such as Optos California and Zeiss Clarus 700 are now in use due to their multimodal capabilities and expanded field of view. We aimed to compare GA area measurements across all three platforms.

Methods: 107 eyes from 62 participants with GA were imaged on the same day using Optos California FAF (green excitation, 532nm), Zeiss Clarus 700 FAF (both blue and green imaging modes) and Heidelberg Spectralis FAF (blue excitation 488nm). GA area was segmented manually in all systems; Heidelberg images were also segmented semi-automatically using RegionFinder. Mean GA area and agreement with RegionFinder were compared across platforms.

Results: 61 eyes of 39 participants had complete and gradable images. Clarus blue imaging data were excluded due to a high ungradable rate of 21%. Agreement with RegionFinder was excellent for all imaging systems (intraclass correlation range 0.96-0.99). Clarus green and Heidelberg manual grading produced larger mean GA areas relative to RegionFinder ((0.49mm2 with 95% CI, 0.23 to 0.75 and 0.34mm2 with 95% CI 0.22 to 0.45)) whereas Optos California produced smaller measurements (-0.48mm2 with 95% CI -0.81 to -0.13).

Conclusions: GA area measurements are reproducible within each FAF platform but differ systematically between devices and wavelengths. Consistent use of a single device is recommended for longitudinal assessment and clinical trials.

Jeong Won Pak, PhD

OCT Biomarkers in Intermediate AMD in AREDS2

Purpose: The AMD Severity Scales, derived from color photographs, are widely used to identify eyes with intermediate AMD and progression risk to advanced AMD. Numerous OCT biomarkers have been proposed as risk markers, but the relationship between color-based severity and OCT-detected biomarkers is not well defined. This study compares color photograph severity levels with the prevalence of key OCT biomarkers to assess whether OCT identifies advanced disease earlier than color imaging.

Methods: Baseline OCTs from AREDS2 were graded for the presence of iRORA, cRORA, hyperreflective foci (HRF), drusen with hyporeflective cores, calcified drusen, drusenoid PED, and reticular pseudodrusen (RPD). Color photograph severity was categorized as Levels 1–6, 7 and 8 with increasing area of drusen and pigmentatry changes , Level 9 ( non central GA) , and level 10 (central GA). Neovascular AMD was excluded from biomarker correlations.

Results: A total of 579 eyes from 319 participants were included: 155 eyes (26.8%) were in Levels 1–6, 173 (29.9%) in Level 7, 51 (8.8%) in Level 8, 37 (6.4%) in Level 9, and 30 (5.2%) in Level 10. Across biomarkers, prevalence increased with higher severity levels of intermediate AMD. Hypocore drusen were observed in 13.5%, 25.4%, and 60.8% of Levels 1–6, 7, and 8, respectively, and calcified drusen in 8.4%, 21.4%, and 45.1%. RPD prevalence rose from 27.7% to 38.2% and 47.1% across Levels 1–6, 7, and 8. HRF increased from 18.7% (Levels 1–6) to 50.3% (Level 7) and 64.7% (Level 8). iRORA was present in 5.8% of eyes in Levels 1–6, 15.6% in Level 7, and 29.4% in Level 8, increasing to 54.1% and 20.0% in Levels 9 and 10. cRORA was infrequent in Levels 1–6 (3.9%) and Level 7 (11.6%), increasing in Level 8 (47.1%), and was seen in all eyes in Levels 9 and 10.

Conclusions: OCT biomarkers increase consistently across color-based AMD levels , with substantial structural abnormalities seen even in intermediate stages. Nearly half of Level 8 eyes already showed cRORA. These findings indicate that OCT detects advanced structural damage earlier than color photographs, highlighting the potential value of OCT-based criteria for refining intermediate AMD phenotypes and assessing progression risk.

Moderated question-and-answer period with Session I speakers.

Brief intermission. Please return promptly to remain on schedule.

Raunak Sinha, PhD

Species-specific tuning of photoreceptor function in the primate retina

The new world common marmoset is a valuable primate model for vision research but evolutionarily their ancestral lineage diverged from the old world macaques and human ancestors over 35-40 Mya. Despite many similarities, there are well-known differences in visual performance, retinal anatomy including photoreceptor density and neural circuitry between new and old world primates.

However, we have little or no knowledge about the functional and physiological differences in the retina at a cellular and circuit level between these two primate species. We therefore asked whether cone photoreceptors, that mediate daylight vision, have preserved their basic functional properties between the two primate species despite the evolutionary divergence long ago. We compared cone photoreceptor signaling in the retina and determined if regional differences between fovea – central specialized region responsible for highest acuity and color vision – and peripheral retina are prevalent in both macaque and marmoset retina using single cell electrophysiology.

We found that cone photoreceptor signaling differs in sensitivity, kinetics and adaptation between marmoset and macaque retina. Interestingly, the regional variations in cone function, between fovea and periphery, are much less pronounced in the marmoset retina compared to that in the macaque retina. Thus, our results reveal fundamental species-specific differences in photoreceptor function and its regional tuning which may stem from specific ecological and evolutionary demands on visual function for the two primate species.

Three short research presentations followed by a moderated Q&A.

Raymond Doudlah, PhD

Mechanisms of spatial constancy for 3D object vision in primates

To preform accurate and precise visually guided behaviors, the visual system must transform ambiguous retinal projections into three-dimensional (3D) scene representations. Furthermore, movements of the eyes to explore a scene result in dramatic changes in the retinal projections, even when objects in the environment do not move. In this study, we investigate how stable visual representations are formed by the visual system despite the volatility of the sensory input caused by eye movements. Our previous work implicated area V3A and the caudal intraparietal (CIP) area in the computation of 3D object pose (position and orientation), but did not consider the impact of gaze on these representations.

To investigate the mechanisms of visuospatial constancy, we trained monkeys to report the orientation of planar surfaces at a fixed distance while holding their gaze at different fixation distances (in front of, at, or behind the surface). The extracellular responses of V3A and CIP neurons were recorded while the animals performed the task. Across the two areas, we observed heterogeneous representations of lower- and higher-level visual selectivity, with similar proportions of neurons in V3A and CIP forming stable 3D pose representations across changes in vergence. The activity of a subset of neurons in both areas was modulated by vergence angle (fixation depth). Notably, neurons that did not carry vergence information show significantly greater visuospatial constancy.

To synthesize these findings with our previous work, we developed a hierarchical, computational model. The model first integrated primary visual cortex-like local absolute disparities into disparity gradient detectors. Then, units which multiplicatively mixed disparity gradient and vergence signals were created.

Finally, integrating mixed-selectivity units produced visuospatially stable object pose representations that were robust to the gaze distance, consistent with our experimental findings. Therefore, to account for the volatility of retinal signals resulting from eye movements, the visual system incorporates oculomotor and visual signals. These results suggest that vergence signals are integrated throughout V3A and CIP to resolve ambiguities in the interpretation of retinal signals to create stable visual representations for guiding interactions with a 3D world.

Whitney Stevens-Sostre, PhD

Deficiency of the voltage-gated potassium channel Kv1.1 dampens neuronal excitability of key output neurons in the mammalian retina
The mammalian retina is a specialized neural circuit of the central nervous system that enables vision: it detects light, processes images, and transmits visual information to higher brain centers. Retinal ganglion cells (RGCs) are the output neurons of the retina that form the optic nerve and relay visual information via spike patterns.

RGC dysfunction or death leads to irreversible vision loss, highlighting their essential role for vision. Ion channels are key modulators of the electrical properties of RGCs underlying their responses to visual stimuli. Little is known, however, about how voltage-gated potassium (Kv) channels that typically regulate neuronal excitability contribute to RGC function and synaptic connectivity within the mammalian retina. Single-cell RNA sequencing reveal that the murine ON-sustained (ON-sus) ‘alpha’ RGC (αRGC) subtype, that are the primary RGC type encoding for light increments, robustly express Kv1.1 channels.

Whether Kv1.1 channels have roles in circuit establishment of the retina or contribute to the passive and/or active membrane properties of αRGCs and synaptic function remains unknown. Using a murine model with constitutive Kv1.1 knock-out targeted to αRGCs, electrophysiology, pharmacology and high-resolution confocal imaging, we demonstrate a novel role for Kv1.1 channels in regulating the spiking pattern of ON-sus αRGCs.

Colleen McDowell, PhD

Function and role of CGRP nerve innervation in aqueous humor outflow

The aqueous humor (AH) outflow pathway through the trabecular meshwork (TM) and Schlemm’s canal (SC) is known to be innervated by sensory neurites. However, the function of these neurites in regulating aqueous humor outflow is not known. CGRP positive C-fibers are the major type of sensory neurons innervating the TM and SC. Upon activation these neurons secrete CGRP locally, and CGRP receptors are known to be expressed in TM and SC cells. The goal of this project was to determine the role of CGRP in regulating aqueous outflow homeostasis in the TM and SC.

High and low-flow regions in the TM and SC were determined in 3 and 9-month-old wild-type C57BL/6J mice using fluorescent tracer latex beads (0.2 µm). Anterior segment flat mounts were co-labeled with antibodies against CGRP (afferent neurons), CALCRL (CGRP receptor), and PECAM1 (SC endothelium), and imaged by confocal microscopy to acquire tiled Z-stacks of entire outflow area. Density of CGRP neurites in each flow region was determined by volumetric analysis in ImageJ. Primary human TM cells in culture (n=3 cells strains) were treated with or without CGRP (0.1 – 3M) for 24-48 hours and processed for immunocytochemistry or western blot analysis.

Significantly more CGRP positive neurites were identified in high-flow regions compared to low-flow regions (p<0.05, n=10 eyes). CALCRL and RAMP1 were shown to be expressed in both primary TM and SC cells in culture (n=3 primary cell strains each) and CALCRL expression was significantly associated with high-flow regions compared to low-flow regions in mice (p<0.01, n=5 eyes). In situ hybridization was used to detect expression of CGRP co-receptor RAMP1 in the TM and expression was shown in cells closely located to CGRP positive neurites in this region. CGRP treatment in primary TM cells resulted in increased phagocytosis (p<0.01), decreased collagen1, fibronectin, and smooth muscle actin expression compared to control (p<0.05), and attenuated TGFβ2 induced expression of these proteins (p<0.05). These data suggest CGRP modulates important homeostatic mechanisms of TM and SC function and may influence the development and regulation of high and low-flow regions of aqueous humor outflow.

Moderated question-and-answer period with Session II speakers.

Brief intermission. Please return promptly to remain on schedule.

Maureen Neitz, PhD

Long-read Sequencing and Droplet Digital PCR as Complementary, Powerful Diagnostic Tools for Xq28 Opsin-based Vision Disorders.

The human opsin gene cluster at Xq28 contains highly similar OPN1LW and OPN1MW genes (≥98% identity), which are essential for red-green color vision. Mutations in these genes are also responsible for X-Linked Cone Dysfunction Disorders, including high myopia, Bornholm eye disease, and blue cone monochromacy. The number of opsin genes at Xq28 varies, but X chromosomes with 3 or more opsin genes are common. Molecular diagnosis of affected males and female carriers of Xq28 cone opsin-based vision disorders requires knowledge of the first two genes in the array, as these are the only opsin genes that are expressed and contribute to vision. Current molecular methods cannot accurately analyze this complex locus, limiting the diagnosis of X-linked cone dysfunction disorders and carrier detection in XX individuals.

To comprehensively resolve this locus, we performed Oxford Nanopore long-read sequencing (ONT) of 206 individuals. We used assembly-based methods to resolve the opsin gene order. We also developed a droplet digital PCR assay to determine the number of OPN1LW, OPN1MW genes and locus control regions (LCRs) in individuals. We validated our assays by determining the frequency of X chromosomes carrying the genetic signature of an inherited red-green color vision defect in the sample of 206 individuals. We applied our method to a family with Bornholm eye disease, and to a complex pedigree segregating two types of inherited red-green color vision deficiency. We also used the droplet digital PCR assay to reliably identify female carriers and affected males with blue cone monochromacy caused by deletion of the Xq28 locus control region.

Our ONT-based targeted assembly approach provides the first comprehensive reference-free method for analysis of the complex Xq28 opsin gene region, addressing a critical diagnostic gap and enabling reliable carrier detection of opsin-based vision disorders. Together, the ONT-based targeted assembly and ddPCR are complementary, powerful new diagnostic tools for identifying the genetic causes of Xq28 opsin gene-based vision disorders in XY individuals and determining carrier status in XX individuals.

Lunch will be available for attendees. Afternoon sessions will begin promptly at the scheduled time.

Shaoqin “Sarah” Gong, MS, PhD

Nanoparticle-mediated systemic delivery for CRISPR editing in CNS disorders

The clinical translation of gene therapy, particularly CRISPR/Cas9 genome editing for central nervous system (CNS) disorders, remains hindered by the blood-brain barrier (BBB), which restricts the systemic delivery of macromolecular therapeutics and prevents widespread access to target cells throughout the brain. To address this challenge, we developed a biomimetic nanoplatform (NP) for systemic administration to achieve BBB penetration for efficient brain-targeting RNA delivery and gene editing. Mechanistically, the NP platform leverages a rational biomimetic design that engages receptor-mediated transcytosis to enable transport across the BBB.

Notably, the platform features a modular, plug-and-play adaptability, enabling rapid adaptation to deliver diverse genetic cargos, including mRNA, CRISPR/Cas9 systems, base editors, prime editors, and so on, positioning it as a universal platform for precision neurogenetic medicine. Firstly, to evaluate functional mRNA delivery in vivo, NPs were administered systemically in Ai14 Cre-reporter mice, resulting in robust Cre mRNA delivery and widespread, stable tdTomato expression throughout the brain. Cell-type analysis demonstrated broad transfection across major brain cell populations, supporting the capacity of NPs for brain-wide gene delivery. Furthermore, the NP platform was applied to co-deliver Cas9 mRNA and App659-targeting sgRNA in the Alzheimer’s disease (AD) mouse model.

Following five doses of systemic administration, NP-mediated Cas9 mRNA/sgRNA delivery achieved efficient App659 gene editing and markedly reduced APP protein levels in the cortex and hippocampus of APP-SAA knock-in mice. This molecular correction translated into significant therapeutic benefits, alleviating AD-associated pathologies. Specifically, treated mice exhibited significant reductions in amyloid-β plaque burden and tau phosphorylation. In addition, NP treatment suppressed neuroinflammation and mitigated neuronal loss, consistent with attenuation of key AD pathological hallmarks. More importantly, in established behavioral paradigms, NP-treated mice showed improved memory performance and locomotor activity.

Collectively, these findings support NPs as a clinically translatable nanomedicine platform that enables brain-targeting RNA delivery and in vivo CRISPR genome editing following systemic administration, providing a versatile and scalable strategy with potential applicability across a broad spectrum of CNS disorders.

Three short research presentations followed by a moderated Q&A.

Praveen Joseph Susaimanickam, PhD

Identification and enrichment of human retinal organoid-derived red/green cone precursors with optimal properties for foveal reconstruction

Background: Cell replacement therapies aimed at restoring foveal vision require a robust source of red/green (long/medium wavelength, or L/M) cone photoreceptors with intrinsic properties conducive to functional integration into host retina. Recent evidence has shown that cones present within mature human retinal organoids (ROs) can generate light responses comparable to macaque foveal cones. However, only cone precursors from early developing ROs possess a capacity for cell-autonomous axonogenesis. Therefore, we sought to identify and enrich for a population of early L/M cone precursors with intrinsically superior axon dynamics that would provide an ideal donor cell source for future foveal reconstruction efforts.

Methods: We developed a dual L/M cone/rod reporter (L/M-CRR) line to unequivocally identify early L/M cone precursors from human ROs. To do so, we used CRISPR/Cas9 to link a tdTomato transgene to the endogenous THRB2 promoter in the WA09 NRL+/eGFP rod reporter line. Differentiated ROs were characterized at early (day 50) and intermediate (day 100) developmental stages using a combination of live fluorescence imaging, immunocytochemistry, and flow cytometry, followed by fluorescence-activated cell sorting and bulk RNAseq analysis to delineate the unique molecular signature of early L/M cone precursors.

Results: THRB2-driven tdTomato fluorescence faithfully demarcated L/M cone precursors throughout RO development, although fluorescence declined in later ROs as L/M cones matured. Transcriptomic profiling revealed that day 50 sorted L/M cone precursors were specifically enriched for genes associated with neural development, axon extension and guidance, and cell migration, which included a gene encoding the cell surface protein CD166/ALCAM. Magnetic-activated cell sorting using an anti-CD166 antibody resulted in specific enrichment of early, highly axonogenic L/M cone precursors assessed by time-lapse imaging.

Conclusions: The L/M-CRR line enables definitive identification and transcriptomic characterization of L/M cones throughout early to mid-stage RO development. Our investigation also revealed that CD166/ALCAM can be used to independently identify and enrich for a subset of early L/M cone precursors that selectively display axon dynamicity conducive for retinal circuit integration. Our studies provide the first insights into early human L/M cone development and establishes a method to isolate transplantable L/M cone precursors with enhanced integrative properties for treating central vision loss.

Ismail Zaitoun, PhD

Vascular Degeneration in a Mouse Model of Retinitis Pigmentosa

Purpose: Retinitis pigmentosa (RP) is a degenerative retinal disease leading to progressive vision loss. It is characterized by progressive photoreceptor degeneration and frequent retinal vascular degeneration, a hallmark of the disease observed in both patients and animal models. However, while research has focused heavily on photoreceptors, the vasculature in RP remains less thoroughly examined. Therefore, the purpose of this study was to assess the retinal and choroidal vasculature in the eyes of rd1 mice, a model of RP.

Methods: Eyes from rd1 and C57BL/6J mice were used. Eyes were collected at sequential postnatal time points, processed for immunohistochemistry analysis on whole-mounts and sections with specific cell markers, and visualized on a confocal microscope.

Results: The rd1 mice displayed photoreceptor loss beginning in the central retina and gradually progressing outwards toward the periphery, with complete loss by around 6 weeks of age. Vascular remodeling was observed in both the retina and choroid. Retinal vessel regression was most evident in the intermediate and deep plexuses across the retina, occurring in parallel with photoreceptor degeneration. Interestingly, the choriocapillaris exhibited degenerative changes in a center-to-periphery pattern starting at 6 weeks of age, with decreased density in the center while the middle and periphery regions remain relatively preserved, but extensive advanced degeneration occurred with aging. However, the choroidal veins and arteries were maintained over time.

Conclusions: Progressive degenerative remodeling of the retinal and choroidal vasculature occurs alongside photoreceptor degeneration in the rd1 mouse model.

Nader Sheibani, MS, PhD

Pericyte VEGF in Ocular Vascular Development and Neovascularization

Pericytes produce vascular endothelial growth factor A (Vegfa), which is important for promoting endothelial cell survival. VEGF also has inhibitory effects on proliferation and migration of pericytes through enhanced VEGFR2 and PDGFRβ heterodimerization. Heterodimerization of these receptors on perivascular supporting cells in culture, mediated by VEGF, is shown to mitigate signaling through these receptors and promote a quiescent phenotype. However, the detailed cellular mechanisms, and the significance of these interactions in vivo, needs investigation. The cell autonomous activities of pericyte Vegfa expression during vascular development and neovascularization remains unknown.

Here we utilized mice conditionally lacking Vegfa in pericytes (VegfaPC) to examine its impact on retinal vascular development and pathological ocular neovascularization. Vascular integrity was also assessed in older mice using fundus imaging and fluorescein angiography. The lack of Vegfa pericyte expression delayed the initial spreading of the retinal superficial layer. Mice lacking Vegfa pericyte expression had similar numbers of retinal endothelial cells and arteries as their wild-type littermates.

However, the number of pericytes was significantly reduced in younger VegfaPC mice but increased in more mature mice. In addition, pericyte Vegfa deficiency did not impact responses during oxygen-induced ischemic retinopathy and laser induced choroidal neovascularization. Thus, pericyte VEGF expression plays a role during early stages of retinal vascular development with limited influence on mature retinal vascularization, its integrity, and neovascularization.

Moderated question-and-answer period with Session III speakers.

Closing remarks and acknowledgements