CFH risk variant linked to complement changes in non-AMD eyes

A high-risk version of the complement factor H gene is associated with a local cascade of protease activity, depletion of a complement inhibitor and membrane attack complex deposition in eyes without clinical age-related macular degeneration, according to a study combining human postmortem tissue, cultured retinal pigment epithelial cells and genetically modified mice. The findings center on granzyme B, an enzyme traditionally associated with cytotoxic immune cells but able here to enter the extracellular environment, cleave vitronectin and undermine protection against the membrane attack complex. In donor eyes carrying the high-risk CFH Y402H CC genotype, researchers found more granzyme B-positive mast cells, less vitronectin in the inner choroid and greater deposition of the terminal complement product C5b-9. The human evidence comes from postmortem snapshots and therefore does not establish that these changes cause AMD or occur in a defined sequence.

The investigation focused on the Y402H polymorphism in CFH, a gene whose protein regulates the alternative complement pathway. Individuals homozygous for the risk-associated C allele have previously been reported to have a five- to sevenfold higher risk of AMD than those with the protective TT genotype. The variant lies in a ligand-binding region of factor H and is thought to reduce the protein’s ability to attach to polyanionic surfaces such as Bruch’s membrane. Reduced regulation can permit persistent complement activation and the generation of inflammatory fragments, including C5a. Those fragments can recruit immune cells and stimulate inflammatory responses, but the work reported here asks how a locally active protease might connect complement dysregulation to remodeling of extracellular matrix in the retina and choroid.

To investigate systemic granzyme B concentrations, the researchers examined people with dry AMD and compared them with a younger, non-AMD reference group. Both age-matched AMD genotype groups had higher mean plasma concentrations than the younger controls: 89.49 pg/mL with SEM 23.65 for the CC group and 64.48 pg/mL with SEM 24.35 for the TT group, compared with 22.3 pg/mL with SEM 7.5 in controls. The difference between the two AMD genotypes was not statistically significant in a Mann–Whitney U test. This result did not identify a systemic marker that distinguished the two genotypes, and the comparison does not establish when or where the genotype-associated tissue changes begin. Small sample numbers, including eight participants in each AMD group, limit the strength and generalizability of the plasma comparison, while the age difference between the AMD and control cohorts prevents treating the systemic increase as a genotype-specific effect.

The clearest human findings came from immunofluorescence analysis of non-AMD donor eyes. Five eyes with the high-risk CC genotype were compared with five age-matched protective TT eyes. Choroidal mast-cell density was significantly higher in the CC group, as was the density of mast cells expressing granzyme B. Granzyme B-positive mast cells averaged 4.2 cells per 0.1 square millimeter in CC eyes, compared with 1.9 cells per 0.1 square millimeter in TT eyes. The difference in degranulating mast-cell density only approached the threshold for statistical significance, at p = 0.0635, so the result should not be described as a demonstrated difference in degranulation across the groups. Nevertheless, high-magnification images showed extracellular granzyme B and tryptase granules around a degranulating cell in one CC eye, while dual labeling indicated that 87% of resident choroidal mast cells coexpressed the two proteins. A subset of granzyme B-positive cells, averaging 24.2% when the groups were combined, was tryptase-negative, consistent with other immune sources being present. The tissue evidence therefore supports mast cells as a major local source of granzyme B in this experimental context.

Vitronectin supplied a possible functional link between extracellular granzyme B and complement regulation. An in vitro cleavage assay demonstrated that active granzyme B could generate specific vitronectin fragments after incubation with the protein. In human donor tissue, protective TT eyes displayed strong vitronectin immunoreactivity along Bruch’s membrane and throughout the choriocapillaris and Sattler’s layer, whereas the inner choroid of high-risk CC eyes showed significantly less staining, with a reported p value of 0.0144. CC eyes also more often contained vitronectin concentrated in sub-retinal pigment epithelial deposits, while TT eyes had fewer and smaller deposits consistent with age-related change. The observations indicate redistribution as well as depletion. They are consistent with the proposed model that proteolysis removes soluble or matrix-associated vitronectin while leaving aggregation-prone material near the retinal pigment epithelium, although the tissue study alone cannot prove that granzyme B directly caused the pattern in donors.

The contribution of granzyme B was tested more directly in aged mice. The researchers compared 13- to 17-month-old granzyme B knockout animals with age-matched wild-type C57BL/6 mice, with four animals in each group. Choroidal vitronectin staining was generally faint in wild-type eyes but robust in knockout eyes, particularly around the choriocapillaris and larger choroidal blood vessels. Quantified signal was significantly higher in knockout tissue, at p = 0.0007. This animal comparison supports a role for endogenous granzyme B in regulating or turning over perivascular vitronectin. It does not reproduce the complete human genetic-risk environment, and the knockout establishes the influence of granzyme B on vitronectin abundance more directly than it establishes an AMD phenotype or a treatment effect. The study also does not determine whether elevated granzyme B alone is sufficient to produce the human pattern.

If granzyme B reduces functional vitronectin, complement inhibition should decline and the terminal membrane attack complex should accumulate. The human tissue analysis produced a corresponding pattern. Protective TT eyes had minimal, patchy staining for C5b-9, the membrane attack complex also known as MAC, whereas CC eyes showed prominent deposits around the choriocapillaris and at the interface between Bruch’s membrane and the retinal pigment epithelium. Quantitative analysis confirmed higher MAC immunoreactivity in the high-risk group, with a reported p value of 0.0079. These findings extend previous observations of increased choroidal MAC in high-risk CFH genotypes. Together, the lower vitronectin signal and higher MAC deposition support impaired local complement control, although the cross-sectional donor data cannot establish the temporal order of the two events in an individual eye. The findings also do not show that MAC deposition caused the observed mast-cell infiltration or vitronectin redistribution.

Cell experiments offered a potential explanation for how complement signaling initiates the pathway. Exposure of ARPE-19 retinal pigment epithelial cells to 5 micrograms per milliliter of C5a for 24 hours significantly increased expression of GZMB, TNFα and IL8 messenger RNA, with reported p values of 0.024, 0.0007 and 0.007, respectively. ELISA measurements showed that granzyme B protein remained at similar levels within cell lysates but increased significantly in culture supernatants after stimulation, with a reported p value of 0.0006. The result distinguished increased secretion from increased intracellular storage. C5a-treated cells also showed disruption of the continuous hexagonal pattern of the tight-junction protein ZO-1 after 48 hours, with intracellular accumulation of the protein and fewer cells retaining an intact barrier; the difference in cells displaying intact staining had a reported p value of 0.0037. These findings provide direct experimental evidence that C5a can stimulate granzyme B release and stress RPE cultures. As a simplified culture system, however, ARPE-19 cells do not reproduce the full immune, vascular and extracellular environment of a human eye.

Bulk RNA sequencing of ARPE-19 cells exposed to 100 nanomolar granzyme B for 18 hours identified 1,464 differentially expressed genes. A volcano plot displayed a twofold-change threshold, while a separate greater-than-1.5-fold criterion was applied to the selected epithelial, vascular, barrier, immune, stress and trophic genes shown in the detailed analysis. RPE identity markers RBP1, TJP1 and BEST1 remained stable across control and stimulated groups, but functional enrichment pointed to tissue morphogenesis, epithelial differentiation and vasculature development. Some structural and cell-fate genes increased, including GJA1, PPL, BMP4 and ID3, while CCN2, VEGFA and EDN1 declined. Damage-sensing and stress-associated genes, including TLR3, IL-33 and CYP1B1, increased, while several broad inflammatory or survival cytokines were reduced. The pattern is compatible with attempted barrier compensation alongside diminished vascular and trophic support, but transcriptional signatures alone do not demonstrate retinal atrophy or predict its development in patients.

Collectively, the experiments support a model in which the CFH Y402H risk genotype is linked to chronic complement activation, increased C5a signaling, granzyme B release by RPE cells and infiltrating mast cells, and extracellular remodeling. Cleaved vitronectin may then provide less inhibition of membrane attack complex assembly while contributing to sub-RPE deposits, creating conditions associated with early tissue changes. The study also notes that prior work found granzyme B can cleave C5 to generate active C5a, raising the possibility of a self-reinforcing loop, although that specific feedback mechanism was not directly tested here. Postmortem tissue, limited donor numbers, unequal control ages, the inability to assess biological-sex effects and the reduced fidelity of cell culture and mouse models constrain interpretation. The work identifies extracellular granzyme B as a mechanistic link and a potential therapeutic target, but it does not demonstrate that inhibiting it would prevent AMD or protect vision in patients.

Subject of Research: Granzyme B, vitronectin cleavage and complement dysregulation in early age-related macular degeneration

Article Title: Factor H polymorphism elevates granzyme B and impairs complement inhibition in human and mouse models

Article References: Chakravarthy, H., Xi, J., Yoo, H.-S., Buss, A., Li, B., Bell, R., Haegert, A., Le Bihan, S., Granville, D. J., & Matsubara, J. A. (2026). Factor H polymorphism elevates granzyme B and impairs complement inhibition in human and mouse models. iScience, 29(11), Article 117957. https://doi.org/10.1016/j.isci.2026.117957

Image Credits: AI Generated

DOI: 10.1016/j.isci.2026.117957

Keywords: age-related macular degeneration, CFH Y402H, complement factor H, granzyme B, vitronectin, membrane attack complex, retinal pigment epithelium, choroidal mast cells, extracellular matrix, C5a, Factor, polymorphism