A nanopathology pipeline for clinical research across scales using human tissue

Alana Burrell, Matthew Lawson, Paolo Ronchi, Xiaofan Xiong, Jonas Albers, Adam McLean, Michelle Willicombe, Linda Moran, Jennifer Pearson-Farr, Martin L. Jones, Graham Ross, Joost de Folter, Matthew Hartley, Virginie Uhlmann, Amy Strange, Yannick Schwab, Elizabeth Duke, Candice Roufosse, Lucy CollinsonmedRxiv 2025 · DOI

TL;DR

Researchers have developed a "nanopathology" pipeline that bridges the gap between clinical biopsies and high-resolution 3D imaging. By combining synchrotron X-ray tomography, confocal microscopy, and volume electron microscopy, they can now map the molecular identity and nanoscale structure of entire human tissue samples in 3D—a process that would take 50 years using traditional methods, now condensed into roughly 50 days.

What Problem Does This Paper Address?

In clinical pathology, there is a frustrating "either/or" trade-off. If you want to see the fine internal structures of a cell (ultrastructure), you typically use glutaraldehyde fixation for electron microscopy (EM), which destroys the molecular markers (antigens) needed to identify cell types. If you want to identify cells using immunofluorescence (IF), you use formalin, which often ruins the ultrastructure.

Furthermore, standard diagnostics are "needle-in-a-haystack" operations: pathologists typically examine less than 2% of a biopsy in 2D slices. This makes it nearly impossible to study rare cellular interactions or understand the complex 3D architecture of diseases like kidney transplant rejection in their full context.

Methods

The authors propose a multi-stage pipeline designed to preserve both structure and molecular information:

  1. Modified Fixation: They use 4% EM-grade formaldehyde in a phosphate buffer. This "Goldilocks" approach stabilizes the tissue for over a year while keeping it compatible with both EM and antibody labeling.
  2. HiTT (High Throughput Tomography): The whole, hydrated biopsy is imaged using synchrotron-based hard X-rays. This provides a 3D map of the entire tissue at sub-micron resolution in minutes, without needing heavy metal stains.
  3. Targeted Slicing & Labeling: Using the X-ray map, they identify specific Regions of Interest (ROIs). The tissue is cut into 60 µm slices. They use a "non-permeabilization" immunofluorescence technique—labeling only the cell surfaces to keep the internal ultrastructure pristine.
  4. Volume EM (SBF-SEM): The ROIs are then imaged using Serial Block Face Scanning Electron Microscopy, which physically shaves off thin layers of the sample to create a high-resolution 3D volume.
  5. Multimodal Registration: All datasets (X-ray, Light, and EM) are aligned into a single virtual 3D space using the MoBIE (Multimodal Big Image Data Sharing and Exploration) framework.

Datasets

The study utilized three human kidney transplant biopsies collected from patients at the Imperial College NHS Trust.

  • Patient 1: Polyoma virus nephropathy.
  • Patient 2: Acute tubular injury.
  • Patient 3: Borderline T cell-mediated rejection. While the sample size is small, the "dataset" produced for each biopsy is massive—terabyte-scale multimodal image stacks covering the entire biopsy volume.

Key Insights and Ideas

  • The "50 Years to 50 Days" Leap: Imaging an entire kidney biopsy at EM resolution would take half a century. By using X-rays to "scout" the tissue and only performing volume EM on specific ROIs, the authors reduced the timeline to 50 days, with a target of 5 days in future iterations.
  • Seeing "Patrolling" in Humans: For the first time, researchers observed immune cell behaviors in 3D human tissue that were previously only seen in live animal models. They documented CD16-positive monocytes "patrolling" the endothelial surface—literally lifting the lining of blood vessels to sample the space beneath.
  • The Hydrated Advantage: Because HiTT images tissue in its natural, hydrated state without stains, it preserves the sample's "molecular virginity" for downstream labeling, a major departure from traditional resin-embedded X-ray techniques.

Key Results

  • Whole-Biopsy Mapping: The pipeline successfully imaged 100% of the biopsy volume at micron resolution and targeted ROIs at nanometer resolution.
  • Glomerular Census: They identified significantly more glomeruli (filtering units) than expected—23, 14, and 32 respectively—noting that traditional 2D slicing often misses a large portion of these critical structures.
  • Cellular Interactions: The team successfully identified and 3D-modeled "synapses" between mononuclear cells and endothelial cells, as well as complex platelet "chains" within capillaries.
  • Quantification: They provided manual quantification of intraluminal cells (platelets, monocytes, neutrophils) per volume unit, showing a relative enrichment of mononuclear cells in glomerular loops compared to peripheral blood counts.

Limitations and Future Work

  • Throughput Bottlenecks: While the imaging is fast, shipping samples to synchrotrons and manual segmentation of cells are current bottlenecks. The authors suggest that laboratory-based X-ray sources and AI-driven (Deep Learning) segmentation will be essential for scaling.
  • Synchrotron Access: Relying on massive particle accelerators limits the pipeline's immediate clinical adoption.
  • Labeling Depth: The non-permeabilization method is great for structure but limits labeling to cell surfaces. Future work will need to find ways to label internal proteins without compromising the "nano" in nanopathology.
  • Biobanking: The authors propose a new form of "nanopathology biobanking," where tissues are fixed in a way that keeps them "future-proof" for any high-resolution imaging technology that emerges.

Summary updated

Abstract

Abstract Most human tissue collected in clinic for diagnosis of pathology is formalin fixed and paraffin embedded or snap frozen, both of which destroy ultrastructure, making them unsuitable for clinical research requiring high resolution imaging. For diagnosis of some pathologies, most commonly renal and ciliopathies, a portion of the tissue is preserved optimally for ultrastructural imaging using electron microscopy (EM), but molecular antigenicity is masked. To resolve this incompatibility, we propose a protocol for fixation of human tissue in the clinic in EM-grade formaldehyde in phosphate buffer, which stabilises tissue for at least a year and preserves both ultrastructure and molecular antigenicity. To leverage this tissue for clinical research, we developed a new ‘nanopathology’ pipeline and applied it to study intravascular immune cells in kidney biopsies from patients with a transplant. In contrast to routine diagnostic practice, which images only a few percent of the whole biopsy in 2D, the nanopathology pipeline images the microstructure of the entire biopsy with X-rays using high throughput tomography (HiTT), assigns identity to immune cells using non-permeabilisation immunolabelling and confocal fluorescence microscopy of 60 µm-thick slices in regions of interest (ROIs; glomeruli and peritubular capillaries), and then images those ROIs using volume EM, specifically serial block face scanning EM. Multimodal registration of all imaging datasets into the same virtual 3D space enables extraction of nanoscale features from patient tissue that had previously only been observed in in vitro or animal models. The established nanopathology pipeline could be optimised for application to clinical research investigating a range of pathologies in the future, and with sufficient speed-up could also find application in next generation diagnostic pathways.
BibTeX
@misc{burrell2025nanopathology,
  author = {Alana Burrell and Matthew Lawson and Paolo Ronchi and Xiaofan Xiong and Jonas Albers and Adam McLean and Michelle Willicombe and Linda Moran and Jennifer Pearson-Farr and Martin L. Jones and Graham Ross and Joost de Folter and Matthew Hartley and Virginie Uhlmann and Amy Strange and Yannick Schwab and Elizabeth Duke and Candice Roufosse and Lucy Collinson},
  title = {{A nanopathology pipeline for clinical research across scales using human tissue}},
  year = {2025},
  doi = {10.1101/2025.08.29.25334660},
  howpublished = {medRxiv},
}
Figure