The cohort is not just healthy kidney; it spans the diseases that destroy the organ: diabetic kidney disease, transplant rejection (acute and chronic), ANCA-associated vasculitis, and lupus nephritis, alongside the acute injury model of Chapter 7. This chapter asks what each disease looks like through the spatial lenses already established. The answers are readouts, observations of what the data show, not claims about disease mechanism, and they come with the platform caveats stated throughout this analysis.

Diabetic kidney disease: the damaged cortex’s program

The human DKD arm spans whole-transcriptome Visium (12 sections) and Visium HD (4 sections). The spatially variable program of these sections is the cortex’s own: renin (REN) and podocyte (NPHS2) lead the list.

Renin marks juxtaglomerular cells within the JGA, linking the spatial landscape to the renin–angiotensin–aldosterone system targeted by major antihypertensive therapies in diabetic kidney disease, while NPHS2 marks podocytes, a key component of the glomerular filtration barrier whose injury is central to diabetic glomerulopathy. The cortex’s spatial backbone, including renin and podocyte programs, is retained in whole-transcriptome space at both spot and bin resolution. These observations provide a spatial readout of DKD-associated cortical organization: disease-related changes are expressed within, and superimposed on, the kidney cortex’s intrinsic cellular programs.

Transplant rejection: compartment architecture at scale

The transplant arm is this analysis’s two-platform story in miniature. At region scale, 48 GeoMx ROIs from a chronic-rejection cohort reconstruct their own compartments (vessel ROIs pericyte-dominant, glomerular ROIs podocyte-enriched, capillary ROIs immune-enriched; Chapter 5). Chronic antibody-mediated rejection attacks the peritubular capillary bed; the compartment architecture that the modules reconstruct is the architecture under attack. At spot scale, the acute-rejection Visium sections surface tubular programs (PDZK1IP1, SLC3A1) as their spatial backbone. The two platforms see the same disease at different scales, and both readouts are the spatial program of the transplanted kidney under immune pressure.

Vasculitis and lupus: the immune program in space

The ANCA-vasculitis sections show a strong inflammatory/myeloid signal among their top spatially variable genes, including S100A8 and S100A9, which encode the calprotectin complex and are strongly associated with neutrophils and inflammatory monocytes. These cells are prominent contributors to ANCA-associated inflammation. They appear alongside a set of ribosomal genes; I will flag the potential technical contribution of that ribosomal signal explicitly (Chapter 9) rather than interpret it as disease biology. The S100A8/A9 signal therefore provides a spatial readout of myeloid inflammation. The lupus sections in the mouse are enriched for proximal-tubule metabolic genes, reflecting the cortical tubular program captured at spot resolution.

The shared caveats

Every disease readout in this chapter comes with the same caveats, stated together, and the same mechanical origin: each readout is the spatially variable gene list per section, computed by Moran’s I over the variable genes, restricted to the genes the platform actually measured (panel arms are flagged accordingly). The confounding described below is structural to the cohort, not a gap I could close with more computation, which is why I state it rather than hide it:

What the disease arm establishes, with those caveats in hand, is a coherent observation: disease reorganizes the spatial program of the kidney, including the segment, the compartment, and the immune niche, and that reorganization is readable across every platform that can see it. this analysis’s final chapters assemble what that means.