Spatial organization of myofibroblastic and complement-secreting CAFs in neuroendocrine tumors

This study utilizes near single-cell spatial transcriptomics to reveal that anatomically distinct neuroendocrine tumors share a conserved stromal architecture characterized by two dominant, spatially segregated cancer-associated fibroblast states: myofibroblastic cells enriched in collagen-dense regions and complement-secreting cells localized at tumor-stroma interfaces.

The Gist

Imagine a neuroendocrine tumor (NET) not just as a lump of bad cells, but as a chaotic, bustling city under construction. For a long time, doctors have been obsessed with studying the "mayors" of this city—the tumor cells themselves—to understand the disease. They look at how fast the mayors are multiplying or what kind of uniforms they wear.

But this new study says: "Wait a minute! We need to look at the construction crew, the security guards, and the neighborhood around the city too."

That "neighborhood" is called the stroma. It's the support system made of connective tissue, blood vessels, and immune cells. The researchers wanted to map out exactly who lives where in this neighborhood and what they are doing.

Here is the breakdown of their findings, using some everyday analogies:

1. The High-Definition Map (The Technology)

Usually, looking at a tumor is like looking at a blurry photo where you can't tell one person from another. This study used a new technology called Visium HD.

• The Analogy: Imagine taking a photo of a crowded stadium. A normal photo just shows a sea of colors. This new technology is like having a high-resolution drone camera that can zoom in and assign a specific ID badge to every single person in the crowd, while also telling you exactly where they are standing.
• The Result: They looked at 8 different tumors from different parts of the body (pancreas, colon, appendix, bile duct) and created a detailed "neighborhood map" for each.

2. The Construction Crew: The Fibroblasts

The main characters in this story are Cancer-Associated Fibroblasts (CAFs). Think of these as the construction workers of the tumor city. They build the scaffolding (extracellular matrix) that holds the tumor together.

The researchers found that even though these tumors come from different body parts, the construction crews all speak the same "language" and have the same four main job titles. However, they don't all do the same job in the same place.

The Four Types of Construction Workers:

A. The "Bricklayers" (myCAFs)

• What they do: They are the muscle-bound workers. They pack down the ground, lay down thick layers of collagen (like packing concrete), and make the tumor hard and dense.
• Where they live: They hang out in the deep, tough, "desmoplastic" (scar-like) areas of the tumor.
• The Metaphor: They are the ones building the fortress walls. They make the tumor so dense that it's hard for medicine to get inside.

B. The "Security Guards with Megaphones" (csCAFs)

• What they do: This is the big discovery of the paper. These workers secrete complement proteins. Think of these proteins as "flares" or "megaphones" that shout to the immune system, "Hey! Something is wrong here!"
• Where they live: They don't hang out in the deep concrete. They stand right on the borderline, right next to the tumor cells.
• The Metaphor: They are the perimeter guards who are constantly ringing the alarm bell, trying to alert the immune system to attack the tumor.

C. The "Firefighters" (iCAFs)

• What they do: They release inflammatory signals (chemicals that cause swelling and heat) to fight infection or stress.
• The Reality: The researchers found these, but they were rare and quiet. It's like having a few firefighters in the city, but they aren't the main event.

D. The "ID Checkers" (apCAFs)

• What they do: They try to show the immune system what the bad guys look like (antigen presentation).
• The Reality: Like the firefighters, they were present but not the dominant force.

3. The Big Discovery: A Universal Neighborhood Plan

The most exciting part of the paper is that this layout is the same everywhere.

Whether the tumor is in the pancreas, the colon, or the appendix, the "city plan" is identical:

• The Bricklayers (myCAFs) are always in the tough, inner core.
• The Security Guards (csCAFs) are always on the edge, right next to the tumor cells.

It's like walking into a house in Latvia, a house in the US, and a house in Japan, and realizing they all have the exact same kitchen layout and bedroom arrangement, even though the families inside are different.

4. Why Does This Matter?

For years, scientists thought the "neighborhood" around a tumor was just random noise. This study proves it's a highly organized, predictable system.

• The "Bricklayers" might be making the tumor hard to treat because they pack it so tight that drugs can't get in.
• The "Security Guards" are interesting because they are trying to call for help. Maybe we can figure out how to turn up their volume so the immune system attacks the tumor more effectively.

The Bottom Line

This paper is like a city planner's report for neuroendocrine tumors. It tells us that no matter where the tumor starts in the body, it builds a very specific, organized neighborhood. By understanding the roles of the "Bricklayers" and the "Security Guards," doctors might be able to design better treatments that either break down the fortress walls or amplify the alarm bells to help the body fight the cancer.

Technical Summary

1. Problem Statement

Neuroendocrine tumors (NETs) are a heterogeneous group of neoplasms currently classified and graded primarily based on tumor-intrinsic features (e.g., mitotic count, Ki-67 index, and lineage markers like Chromogranin A and Synaptophysin). However, the stromal microenvironment (TME) of NETs remains poorly defined across different anatomical sites (pancreas, colon, appendix, bile duct).

• Knowledge Gap: While Cancer-Associated Fibroblasts (CAFs) are known to be heterogeneous (including myofibroblastic, inflammatory, antigen-presenting, and complement-secreting states) in carcinomas like Pancreatic Ductal Adenocarcinoma (PDAC), their transcriptional states, spatial distribution, and conservation across different NET subtypes are unknown.
• Objective: To characterize stromal heterogeneity across anatomically distinct NETs and map the spatial organization of specific CAF programs using high-resolution spatial transcriptomics.

2. Methodology

The study employed a multi-step computational and experimental workflow using 10x Genomics Visium HD spatial transcriptomics on eight treatment-naive primary NET tissues (5 PanNETs, 1 appendiceal, 1 colorectal, 1 biliary).

• Sample Preparation: Formalin-fixed paraffin-embedded (FFPE) tissues were processed using the Visium HD protocol. RNA quality was verified (DV200 > 30%).
• Cell Segmentation: To achieve near-single-cell resolution, custom cell segmentation masks were generated using QuPath and StarDist on H&E images. These masks were used to aggregate 2 µm spatial bins into individual cell polygons, creating a cell-boundary-aware dataset.
• Data Processing & QC:
  • Data was processed via Space Ranger v4.0 and analyzed in Seurat v5.
  • A unique gene program score-based QC was implemented to filter out polygons with lateral RNA diffusion (e.g., tumor markers in stromal regions) or mixed lineage signals.
  • Stromal-enriched subsets were integrated across all eight samples using Reciprocal PCA (RPCA) to identify shared transcriptional states while minimizing batch effects.
• CAF Subtyping:
  • The shared fibroblast cluster was subsetted and scored against four literature-derived CAF programs: myCAF (myofibroblastic), iCAF (inflammatory), apCAF (antigen-presenting), and csCAF (complement-secreting).
  • A competitive scoring framework (z-score standardized) was used to assign subtypes, requiring a clear margin between the top-scoring program and the runner-up.
  • Pseudobulk differential expression (DESeq2) and Gene Set Enrichment Analysis (GSEA) were performed to validate subtype-specific pathways.

3. Key Results

A. Conserved Stromal Architecture

• Integration of stromal cells from eight distinct NETs revealed ten transcriptionally distinct clusters.
• Most clusters were shared across all anatomical sites, indicating a conserved stromal framework in NETs.
• Three specific clusters (3, 9, 10) were sample-predominant but mapped to specific spatial boundaries (tumor-stroma interface or stroma-exocrine pancreas boundary), representing niche-restricted states rather than integration artifacts.
• Non-fibroblast stromal compartments (myeloid, endothelial, smooth muscle, B/plasma cells) were successfully resolved.

B. Identification of Four CAF States

Within the shared fibroblast compartment, four CAF states were identified:

1. myCAFs (Myofibroblastic): The dominant population in most samples. Characterized by high expression of ECM genes (COL1A1, COL1A2, TAGLN, PDGFRB).
2. csCAFs (Complement-secreting): A substantial subset characterized by complement cascade genes (C3, C7, CFD, C1S, C1R).
3. iCAFs (Inflammatory): Detected but less abundant, characterized by chemokine/inflammatory genes.
4. apCAFs (Antigen-presenting): Detected but weak, characterized by CD74 and MHC-II machinery (note: limited by probe set availability for classical MHC-II genes).

C. Spatial Segregation of CAFs

Spatial mapping in desmoplastic tumors (NET078 and NET109) revealed distinct localization patterns:

• myCAFs: Enriched in collagen-dense, ECM-rich regions deep within the stroma, consistent with a matrix-remodeling role.
• csCAFs: Preferentially localized to the tumor-stroma interface (adjacent to tumor nests), suggesting a role in immune modulation near the tumor cells.
• iCAFs and apCAFs: Consistently detected but generally weaker and less spatially distinct in this cohort.

D. Validation

• Pseudobulk DE analysis confirmed that myCAFs showed broad upregulation of ECM remodeling genes (e.g., SPARC, COL5A1) beyond the scoring set.
• GSEA confirmed myCAFs were enriched for "Extracellular Matrix Organization" and "Epithelial-Mesenchymal Transition," while csCAFs were enriched for "Complement Cascade" pathways.
• The findings align with previous GeoMx DSP studies by the same group, validating that signals previously attributed to immune contamination (MHC-II and complement genes) likely originated from genuine fibroblast subtypes.

4. Significance and Contributions

• First High-Resolution Stromal Map: This study provides the first near-single-cell spatial reference for the stromal microenvironment across multiple anatomical sites of NETs.
• Conservation of CAF Programs: It demonstrates that despite anatomical diversity, NETs share a conserved stromal architecture dominated by matrix-producing (myCAF) and complement-enriched (csCAF) fibroblast states.
• Novel Insight into csCAFs: The study highlights complement-secreting CAFs as a major, spatially distinct population in NETs, localized specifically at the tumor interface. This suggests a potential mechanism for local immune regulation distinct from the inflammatory (iCAF) programs seen in carcinomas.
• Methodological Advancement: The study successfully applied Visium HD with custom cell segmentation to FFPE NET tissues, overcoming challenges of RNA diffusion and low transcript capture to resolve stromal heterogeneity.
• Clinical Implications: Understanding the spatial segregation of CAFs (e.g., matrix barriers vs. immune interfaces) could inform therapeutic strategies targeting the stroma, such as drug delivery enhancement or immune modulation, which are currently limited in NET treatment.

5. Limitations

• Transcript Depth: FFPE Visium HD has lower transcript capture per cell compared to fresh scRNA-seq, potentially limiting sensitivity for low-abundance cytokines (iCAF/apCAF markers).
• RNA Diffusion: Despite segmentation and filtering, lateral RNA diffusion from tumor cells to adjacent stromal polygons remains a potential confounder.
• Cohort Size: The small sample size (n=8) limits statistical power for correlating stromal subtypes with clinical outcomes or treatment response.
• Probe Limitations: The absence of classical MHC-II genes in the probe set required computational supplementation for apCAF identification.