Blood vessel organoids trace an IGFBP7-galectin axis in diabetic microangiopathy

What the work claims

Pooni et al. argue that human blood vessel organoids (BVOs) can recapitulate the early structural lesions of diabetic microangiopathy and can be used to discover effectors that are genuinely tied to human disease. They report that a seven-day "diabetic medium" (20 mM D-glucose plus 1 ng/mL IL-6 and 1 ng/mL TNF-α) is enough to produce pericyte dropout and basement-membrane thickening in iPSC-derived BVOs. Proteomic profiling of the BVO secretome then singles out insulin-like growth factor binding protein-7 (IGFBP7) as a strongly upregulated protein. Follow-up CRISPR knockouts and patient cohort data are used to position LGALS1 and LGALS3 as downstream effectors of IGFBP7 and as candidate markers or targets in diabetic nephropathy and retinopathy.¹

How it works

The BVOs are built from the KOLF2 reference iPSC line and from peripheral-blood-mononuclear-cell iPSC lines of healthy non-diabetic donors and people with long-standing type 2 diabetes. The vascular network is allowed to self-assemble in a collagen I-Matrigel matrix, then extracted and cultured as free-floating BVOs. The diabetic medium is meant to mirror grade-4 hyperglycaemia and the low-grade inflammation seen in diabetic retinopathy, with 20 mM glucose as the metabolic stress and 1 ng/mL each of IL-6 and TNF-α as the inflammatory component. A 20 mM mannitol control accounts for osmotic pressure.

After seven days, immunofluorescence and flow cytometry show no change in overall endothelial cell number or BVO diameter, but a clear loss of pericyte coverage and a marked increase in collagen-IV deposition. Basement-membrane thickness is quantified from resliced confocal z-stacks. These are precisely the early microvascular lesions described in clinical diabetic specimens, so the model is being offered as a functional, human preclinical platform rather than a simple endothelial monolayer.

The secretome proteomics point to IGFBP7. In the conditioned medium of diabetic BVOs, IGFBP7 is sharply elevated, while soluble CD93 is reduced. The authors note that IGFBP7 has previously been reported as a ligand for CD93, so the diabetic microenvironment appears to dysregulate the CD93-IGFBP7 signalling axis. IGFBP7 is secreted by both endothelial and mural cells, with mural cells producing the highest levels. CRISPR-Cas9 knockout of IGFBP7 in the KOLF2 background produces a broad proteomic remodelling of adherens junctions, focal adhesions, and extracellular matrix proteins, and it specifically downregulates LGALS1 and LGALS3. Conversely, diabetic-medium treatment upregulates LGALS1 and LGALS3 in both healthy-donor and diabetic-donor BVOs. The proposed chain is therefore IGFBP7 elevation, downstream galectin increase, and microvascular regression.

The clinical translation effort uses two large cohorts. In UK Biobank, IGFBP7 is elevated in type 1 and type 2 diabetes, and further elevated in diabetic nephropathy. CD93 is elevated in both nephropathy and retinopathy. LGALS1 and LGALS3 track with nephropathy and retinopathy and with adverse cardiovascular and heart-failure outcomes. In the All of Us cohort, genetic variants in the IGFBP7-CD93-LGALS1/LGALS3 region are associated with prevalent or ever-diagnosed diabetes and future cardiometabolic outcomes. The variant with the largest effect size, rs570122247, is a missense variant in IGFBP7 that is more frequent in type 2 diabetes and associated with elevated five-year type 2 diabetes risk.¹

Where a skeptic should push

The most load-bearing assumption is that the seven-day diabetic medium is a faithful model of human diabetic microangiopathy rather than a generic high-glucose, pro-inflammatory stress. The authors show the expected structural changes, but the evidence that these changes arise through the same molecular drivers as clinical disease is still associative. The rescue experiments, if any, are not reported; there is no demonstration that blocking IGFBP7 or LGALS1/3 in the diabetic-medium BVOs restores pericyte coverage or vascular density. A knockout in the reference line shows what happens when IGFBP7 is absent from the start, not whether acute IGFBP7 inhibition reverses established microangiopathy.

The sample sizes for the patient-derived work are small: two healthy and two diabetic iPSC lines for endothelial-cell metabolic phenotyping, and the same four lines for BVO validation. A finding that is reproducible across four lines is encouraging, but it is not a population-wide claim. The UK Biobank and All of Us associations are valuable for triangulation, yet they cannot establish causality, and the proteomic signatures were measured in plasma or inferred from genetic variation rather than in BVO-secreted proteins.

Another caveat is the composition of the diabetic medium. Glucose is fixed at 20 mM and cytokines at 1 ng/mL. These concentrations are plausible, but they are single-point choices. A dose-response surface across glucose, IL-6, TNF-α, and other diabetic factors would strengthen the claim that the phenotype is specific to the diabetic milieu. Finally, the BVOs lack blood flow, immune cells, and systemic hormonal inputs, so the model captures structural microvascular responses but not the full organ context of kidney or retina.

Implication for vascular organoid drug-target models

For organoid-based drug discovery, the important message is that a relatively short, chemically defined diabetic stimulus can generate a measurable microvascular lesion in a human tissue model. That lowers the barrier to screening compounds for microvascular protection, provided the readouts are chosen carefully. Pericyte coverage, vascular length density, and basement-membrane thickness are concrete, image-based endpoints; secreted IGFBP7, LGALS1, and LGALS3 offer soluble biomarkers that could be scaled into higher-throughput assays. The study also illustrates how to connect an organoid phenotype to population data: the same proteins that change in BVO conditioned medium are associated with microvascular complications in UK Biobank, giving sponsors a translational argument that is stronger than organoid-only pharmacology.

The threat is over-interpretation. Because the model produces diabetic-like lesions quickly, it is tempting to treat every hit that normalises IGFBP7 or galectin secretion as a validated therapy. That would be premature. The cohort associations are not causal, and the BVOs do not yet show reversibility. A compound that lowers secreted IGFBP7 could be acting through non-specific cytotoxicity or by simply shrinking the organoid. The right next step is a rescue experiment in which a candidate drug is added after the diabetic injury has been established, with pericyte coverage and vascular function as primary endpoints.

A second threat is generalisation from one donor set. Diabetic microangiopathy varies with diabetes duration, glycaemic control, comorbidities, and genetic background. Four iPSC lines are enough to show feasibility, but a screening campaign will need dozens of donor lines to avoid a single-line artefact. The KOLF2 reference line is useful for mechanistic genetics, yet it is not diabetic; the model therefore tests the effect of the diabetic medium, not the full patient genotype.

The opportunity is real. If the IGFBP7-LGALS1/3 axis can be validated as causal in larger organoid panels and in animal models, BVOs could become a standard preclinical bridge between endothelial cell assays and in vivo nephropathy or retinopathy studies. The ability to measure both structural and secreted endpoints in the same human microvascular construct is a genuine advance over planar endothelial cultures.

The bottom line

This is a promising, well-designed model-development paper rather than a definitive target-validation study. The BVOs develop early diabetic microvascular lesions on a clinically plausible timescale, and the proteomic link to IGFBP7 and galectins is reinforced by large cohort data. What would confirm the model is a dose-response rescue experiment showing that inhibiting IGFBP7, LGALS1, or LGALS3 after injury restores pericyte coverage. What would weaken it is evidence that the same phenotype appears in response to any high-glucose or inflammatory stress without involvement of the proposed effectors. For now, the work is best read as a blueprint for building human microvascular disease models and generating testable targets, not as proof that IGFBP7 inhibition will protect diabetic patients.