The cornea's matrix proteins are part of its immune system

What the work claims

This is a primary in vivo and in vitro study from Maiti, Koduri, Shin and colleagues at NYU Grossman School of Medicine, with collaborators at the University of Pittsburgh and Regeneron. Mice with corneas scarified and infected with 10,000 plaque-forming units of HSV-1 (McKrae strain) were compared across wild type, lumican-null and biglycan-null genotypes. By day 3 post-infection, knockout animals had significantly higher clinical keratitis scores than wild type (lumican-null 2.42 plus or minus 0.38, biglycan-null 1.43 plus or minus 0.17, wild type 1.27 plus or minus 0.18, n of 6 to 7 per genotype across three independent experiments), and by day 5 the lumican-null mice remained severely scored (2.44 plus or minus 0.3 versus 0.57 plus or minus 0.3 for wild type) while the biglycan-null difference had drifted to non-significance (1.86 plus or minus 0.51, p = 0.054). Viral titers told the same story: at day 3, wild type corneas carried 8 plus or minus 2.76 plaque-forming units against 27 plus or minus 3.27 in lumican-null and 26.17 plus or minus 3.23 in biglycan-null; by day 5 the gap widened to 15 plus or minus 8.66 versus 102.5 plus or minus 14.93 and 98 plus or minus 16.55.¹

The mechanistic claim has three layers. First, early innate recruitment collapses without these matrix proteins: at day 3, neutrophils averaged 285 plus or minus 40.13 cells per wild type cornea but 74.2 plus or minus 7.39 in lumican-null animals, and monocytes averaged 479.2 plus or minus 26.52 in wild type against 90.2 plus or minus 6.97 in lumican-null and 293 plus or minus 40 in biglycan-null corneas, with plasmacytoid dendritic cells similarly reduced. Second, the proteins are themselves induced by infection: lumican and biglycan rise in infected corneas and appear in the basal epithelial layer where uninfected corneas show none, and infected human corneal epithelial cells secrete both. Third, this induction is functional: adding recombinant lumican or biglycan to human corneal epithelial cells exposed to UV-inactivated HSV-1 significantly amplifies CCL2 secretion, while neither protein alone induces CCL2 - they are amplifiers of virus-triggered signaling, not primary triggers. A human induced pluripotent stem cell corneal organoid, which has a stratified KRT3-positive epithelium over a COL5A1-positive, lumican and biglycan-rich stroma but no resident or infiltrating immune cells, was infected with 1,000 plaque-forming units of mCherry-tagged HSV-1 (KOS strain); at 24 hours the virus formed discrete epithelial foci with epithelial tropism and induced CCL2 in epithelial cells adjacent to and away from infection sites, confirmed by immunostaining and by CCL2 rising in the culture medium.¹

How it works

Lumican and biglycan are small leucine-rich repeat proteoglycans, best known as structural organizers of corneal stromal collagen that keep the tissue transparent. The same laboratory and others have spent years building a second, less expected picture of them as immune modulators: both proteins promote TLR2 and TLR4 signaling (lumican also binds CD14 and caveolin 1, increasing TLR4 residence at the macrophage membrane), while the authors' earlier work showed both bind CpG DNA and restrict TLR9 responses. In this study, the working model is that lumican and biglycan normally tune several Toll-like receptor inputs at once, and that their infection-induced appearance in the basal epithelium amplifies NF-kappaB-driven CCL2 production, which recruits monocytes and, indirectly, neutrophils and plasmacytoid dendritic cells during the first three days - the window in which viral clearance is decided.¹

The timing evidence is the interesting part. Multiplex cytokine analysis at day 3 showed broad deficits in the knockout corneas - pro-inflammatory cytokines, chemokines and type I interferons all lower than wild type - yet bulk RNA-sequencing at day 5 found no major transcriptional differences in innate immune programs between genotypes. The knockout corneas eventually mount a partial response, but it arrives after the recruitment window has closed. The authors read this as lumican and biglycan being required for timely amplification of early signals rather than for the response itself, and they flag the underlying receptor-level mechanism as unresolved: global knockouts cannot say whether the epithelial CCL2 effect is driven by epithelial-derived or stroma-released protein.¹

Where a skeptic should push

The load-bearing assumption is that the immune phenotype reflects signaling roles of these proteins during infection rather than developmental consequences of their absence. Lumican-null and biglycan-null corneas develop with altered collagen architecture, and altered stromal mechanics could change viral spread or immune access for purely structural reasons. The authors say exactly this in their limitations section: tissue-specific and conditional epithelial knockouts are needed to separate the two stories, and the current data cannot assign the epithelial CCL2 effect to epithelial versus stromal sources of the proteins.¹

Second, the receptor mechanism is asserted more than demonstrated here. The TLR2/TLR4-promoting and TLR9-restricting activities come largely from the group's prior cell-culture and sepsis work; in this study the lumican-null paradox - dampened overall inflammation but evidence of exaggerated per-cell TLR9 signaling in plasmacytoid dendritic cells - is acknowledged as a tension, not resolved. Third, the human organoid experiments are confirmatory, not causal: the organoid recapitulates epithelial tropism and CCL2 induction, but no one has yet perturbed lumican or biglycan in the human organoid to show the chemokine response depends on them. The biglycan-null day-5 keratitis score missing significance (p = 0.054) is a reminder that the two knockouts are not equivalent, and sample sizes (n of 4 to 7 per group) are modest, though the three-experiment replication helps.¹

The matrix is an immune organ organoids lack

The non-obvious implication for organoid models of human organs is that extracellular matrix composition is a hidden experimental variable in every innate-immune readout taken on an organoid. Most epithelial and mesenchymal organoids are grown in Matrigel, a undefined, xenogeneic, batch-variable basement-membrane extract whose proteoglycan content nobody characterizes per lot, or in minimal synthetic matrices chosen for stiffness and integrin ligands. This paper demonstrates that two matrix proteins, at concentrations the tissue itself regulates during infection, can be the difference between a CCL2 response that recruits an army and one that recruits almost nobody. A compound screen that measures chemokine induction, antiviral state, or immune-organoid co-culture recruitment in an ECM-naive or Matrigel-grown system is therefore measuring its own matrix as much as its compound - and matrix lot changes can masquerade as compound effects or batch drift.¹

The opportunity is to make that variable a dial. Defined matrices supplemented with physiological lumican and biglycan - or engineered to sequester and release them - would let organoid infection models reproduce the amplification step this paper identifies, turning innate-immune readouts from qualitative to quantitative. It also sharpens the corneal organoid's legitimate use case: as the authors show, it faithfully reproduces epithelial viral tropism and local chemokine induction, which makes it a reasonable first-line model for antiviral entry inhibitors or epithelial-response modulators, screened at 24 to 72 hours, at far higher throughput than mouse eyes.¹

The threat is overclaiming what the model certifies. The clinical endpoint in HSV-1 keratitis is opacity driven by stromal inflammation and immune infiltration - precisely the axes the organoid lacks, because it has no resident or infiltrating immune cells and cannot model the recruitment failure that is the core of this paper's phenotype. An anti-keratitis compound that looks effective in the organoid has cleared only the epithelial gate. And the generalization caveat runs in both directions: the causal claims rest on global mouse knockouts and one human organoid line showing descriptive agreement, so the field should resist presenting the lumican-biglycan-CCL2 axis as established human biology until conditional models and perturbed human organoids exist.

The bottom line

Established: in mice, lumican and biglycan are required for normal early innate immune recruitment and viral clearance in HSV-1 keratitis, with a pronounced early defect in neutrophils, monocytes, plasmacytoid dendritic cells and cytokine induction; both proteins are induced in the infected corneal epithelium; and recombinant lumican or biglycan amplify virus-triggered CCL2 from human corneal epithelial cells. Established at a descriptive level: a human induced pluripotent stem cell corneal organoid reproduces epithelial viral tropism and CCL2 induction. Unestablished: which cell type's lumican and biglycan drive the effect, which receptors mediate it in the cornea, whether the human organoid's CCL2 response actually depends on its matrix proteoglycans, and anything about adaptive immunity or stromal opacity, which the model cannot address. What would confirm the axis as a therapeutic target is conditional epithelial knockout reproducing the CCL2 deficit, plus lumican or biglycan perturbation in the human organoid changing its chemokine output.