Spatial Organellomics Maps Cell State Diversity and Metabolic Adaptation in Tissues

This study introduces spatial Organellomics (sOrganellomics), an imaging workflow that leverages multi-organelle features to map functional cell state diversity and metabolic adaptation in tissues, revealing novel hepatocyte organization patterns and linking architectural remodeling to bioenergetic heterogeneity in vivo.

The Gist

Imagine you are trying to understand a bustling city. For years, scientists have tried to figure out what makes different neighborhoods unique by reading the "residents' diaries" (their genes). They found out that people in the financial district talk about money, while people in the arts district talk about paint. This is like transcriptomics (reading the genetic instructions).

But there's a problem: Just because someone says they are an artist in their diary doesn't mean they are currently holding a paintbrush. Maybe they are sleeping, or maybe they are actually a chef right now. The "diary" doesn't always show you what the person is doing in the moment.

This new paper introduces a new way to look at the city called sOrganellomics (spatial Organellomics). Instead of reading the diary, the scientists decided to look at the architecture of the residents' homes.

The Big Idea: The "Home Interior" Analogy

Think of a cell like a house. Inside every house, there are different rooms (organelles) like the kitchen (mitochondria for energy), the pantry (lipid droplets for storage), and the recycling center (peroxisomes).

• The Old Way: Scientists used to ask, "What is written on the mailbox?" (Gene expression).
• The New Way (sOrganellomics): Scientists now ask, "What does the inside of the house look like?" Are the kitchens huge and full of pots? Are the pantries empty or overflowing? Are the recycling centers working overtime?

The authors discovered that the shape, size, and arrangement of these "rooms" tell you exactly what the cell is doing right now, often better than reading the genetic instructions.

What They Found in the Liver (The City's Power Plant)

The liver is like a massive power plant that processes food and detoxifies blood. For decades, scientists believed the liver was organized like a simple factory assembly line:

• Zone 1 (Near the entrance): Workers do Task A.
• Zone 2 (Middle): Workers do Task B.
• Zone 3 (Near the exit): Workers do Task C.

They thought everyone in Zone 1 looked exactly the same, and everyone in Zone 3 looked exactly the same.

The Surprise:
Using their new "home interior" camera, the scientists found out the liver is actually more like a vibrant, mixed neighborhood.

• Even within "Zone 1," there are different types of houses. Some have giant kitchens, some have tiny pantries, and some have high-tech recycling centers.
• These different "house styles" (cell states) are intermixed. They aren't in neat, separate rows. They are neighbors living side-by-side, each doing a slightly different job to help the whole community function.
• The Metaphor: It's not a factory line; it's a potluck dinner. Everyone brings a different dish (a different metabolic function), and they are all mixed together on the same table, not sorted into separate rows.

How the City Adapts to Stress

The scientists also watched how this neighborhood changed when the "city" faced different challenges:

1. Fasting (The City runs out of food):

   • The "houses" quickly remodel. The kitchens get bigger, the recycling centers get more active, and the pantries get emptied to release stored energy.
   • The Result: The neighborhood stays diverse. Different houses adapt in different ways, but they all work together to keep the city running. It's like a community coming together to share resources during a storm.

2. Western Diet (The City is overfed with junk food):

   • The "houses" get clogged. The pantries overflow with fat, and the recycling centers get overwhelmed.
   • The Result: The neighborhood loses its diversity. Everyone starts looking the same—cluttered and stressed. The unique "house styles" disappear, and the whole area becomes a uniform mess. This is a sign of early disease (like fatty liver disease).

Why This Matters

This new method is like giving scientists a high-resolution X-ray of the cell's interior.

• It's a "Structural Fingerprint": Just as you can tell if a house is a bakery or a library by looking at its layout, you can tell if a cell is healthy, fasting, or sick just by looking at its organelles.
• Early Warning System: They found that they could predict if a liver was getting sick (developing fatty liver disease) just by looking at these tiny architectural changes, often before the disease was visible to the naked eye or detectable by other methods.
• Real-Time Action: Unlike reading a diary (genes), which tells you what a cell might do, looking at the organelles tells you what the cell is actually doing right now.

In Summary

This paper teaches us that to truly understand a cell, we shouldn't just read its instructions; we should look at its furniture and layout. By mapping the "interior design" of cells in tissues, scientists can see how our bodies adapt to hunger, how they get sick from bad diets, and how different cells work together in a complex, mixed community to keep us alive. It turns the study of cells from reading a script into watching a live, dynamic movie.

Technical Summary

1. Problem Statement

Current methods for mapping cellular diversity in tissues rely heavily on transcriptional profiling (e.g., spatial transcriptomics). While powerful, these approaches have limitations:

• Functional Disconnect: Gene expression levels do not always correlate with protein abundance or immediate functional states.
• Loss of Structural Context: Transcriptomics often sacrifices subcellular spatial resolution, missing critical functional insights embedded in organelle architecture.
• Scalability: Existing image-based profiling methods (like Cell Painting) are largely restricted to cultured cells and struggle to scale to intact, complex 3D tissues due to segmentation challenges and cellular diversity.

The authors posit that organelle architecture (abundance, morphology, and spatial distribution of intracellular compartments) is a direct, functional readout of cell state that complements molecular profiling but remains underutilized in native tissues.

2. Methodology: Spatial Organellomics (sOrganellomics)

The authors developed a comprehensive imaging and analysis workflow called sOrganellomics to quantify multi-organelle architecture at single-cell resolution in intact tissues.

• Imaging Protocol:
  • Sample: Fixed tissue sections (liver and pancreas) and intravital live imaging.
  • Multiplexing: 3D confocal microscopy labeling for cell boundaries (Phalloidin), mitochondria (Mito-Dendra2 transgene or antibodies), peroxisomes (PMP70/PEX14), and lipid droplets (LipidTOX).
  • Zonation Markers: In the liver, Aldolase B (periportal) and CYP2E1 (pericentral) were used to map the Portal Vein (PV) to Central Vein (CV) axis.
• Computational Pipeline:
  • Segmentation: Utilized Cellpose v2 (custom-trained models) for automated instance segmentation of cells and individual organelles.
  • Feature Extraction: Extracted ~100 interpretable features per cell, including organelle density, area, perimeter, aspect ratio, circularity, solidity, and distance to the nucleus.
  • Dimensionality Reduction & Clustering: Applied Principal Component Analysis (PCA) followed by Gaussian Mixture Modeling (GMM) to identify distinct cell states based on organelle signatures.
  • Benchmarking: Compared sOrganellomics against published spatial transcriptomic and proteomic datasets using matched cell budgets and feature dimensions to evaluate clustering robustness (using metrics like Adjusted Rand Index, silhouette score, and Davies-Bouldin index).
  • Predictive Modeling: Trained Multilayer Perceptron (MLP), Random Forest, and Logistic Regression models to predict cell states, nutritional conditions, and disease stages from organelle features.
  • Cross-Modal Alignment: Used a probabilistic framework (Wasserstein distance and Hungarian algorithm) to align organelle-defined states with independent spatial proteomic data to infer underlying metabolic programs.

3. Key Contributions

1. Development of sOrganellomics: A scalable workflow that bridges high-resolution imaging and machine learning to map functional cell states in intact tissues without relying on gene expression.
2. Redefining Liver Zonation: Challenged the classical model of liver zonation (homogeneous layers along the PV-CV axis). Demonstrated that hepatocytes form intermixed communities of distinct organelle-defined states within zones, creating a "sub-zonal diversity."
3. Dynamic Remodeling of Cell States: Showed that nutritional stress (fasting vs. Western Diet) does not just shift existing states but reshapes the landscape of organelle configurations, generating new cell states and altering tissue heterogeneity.
4. Structure-Function Coupling: Linked organelle architectural changes directly to metabolic function in vivo using intravital imaging, demonstrating that structural remodeling correlates with mitochondrial membrane potential and bioenergetic activity.
5. Predictive Power: Established organelle architecture as a robust "structural code" capable of predicting nutritional status and early stages of Metabolic Dysfunction-Associated Steatotic Liver Disease (MASLD) with high accuracy.

4. Key Results

A. Distinguishing Cellular Identities

• sOrganellomics successfully separated hepatocytes from pancreatic cells and further resolved pancreatic endocrine vs. exocrine populations based solely on organelle architecture, proving that subcellular organization encodes tissue identity.

B. Sub-Zonal Hepatocyte Diversity

• Five Distinct States: In control livers, unsupervised clustering revealed five stable hepatocyte categories (H1–H5) with unique mitochondrial, peroxisomal, and lipid droplet signatures.
• Intermixed Organization: Contrary to the classical "layered" zonation model, these states were intermixed across the lobule. While some states (e.g., H5) were enriched near the central vein, others (H1–H3) coexisted in periportal regions.
• Positional Limitation: Spatial position (PV-CV axis) explained only ~60% of the variability in hepatocyte identity; the remaining ~40% was driven by local, non-positional factors, indicating a complex "sub-zonal" organization.
• Proteomic Correlation: Cross-modal alignment confirmed that these organelle-defined states correspond to distinct proteomic programs (e.g., nitrogen handling, oxidative stress response), suggesting neighboring hepatocytes perform complementary metabolic tasks.

C. Nutritional Perturbations

• Fasting (Acute Stress): Induced a coordinated remodeling of organelles (increased mitochondrial density/elongation, peroxisome expansion). This created new fasting-associated states (FH1–FH4) but preserved sub-zonal diversity. Intravital imaging showed fasting increased mitochondrial activity (TMRM intensity), particularly near the central vein, reducing the bioenergetic gradient.
• Western Diet (Chronic Stress): Led to a loss of heterogeneity. Hepatocytes collapsed into fewer states (WH1–WH2) with massive lipid droplet accumulation and rounded mitochondria. This "compression" of cell states suggests a loss of tissue plasticity and resilience.

D. Disease Progression and Prediction

• Early Detection: MLP models trained on organelle features could predict the stage of MASLD progression (0 to 50 days of Western Diet) with ~80% accuracy, detecting pathological remodeling before severe tissue architecture collapse.
• Nutritional Classification: Models predicted the specific nutritional regimen (Control, Fasted, WD) with ~99% accuracy, highlighting organelle architecture as a sensitive biomarker for metabolic state.

5. Significance

• Functional Readout: sOrganellomics provides a direct functional readout of cell state that is often missed by transcriptomics, as organelle morphology reflects the integrated output of transcriptional, translational, and post-translational regulation.
• New Paradigm for Tissue Organization: The findings suggest that tissue function is not just a result of spatial gradients but relies on local cellular diversity and the division of labor among intermixed cell states within a zone.
• Clinical Potential: The ability to detect early metabolic disease and nutritional status via organelle architecture offers a potential diagnostic tool. It suggests that the loss of organelle-defined diversity (as seen in chronic WD) may be a precursor to tissue failure and disease susceptibility (e.g., hepatocellular carcinoma).
• Scalability: The workflow is compatible with standard confocal microscopy and off-the-shelf machine learning, making it a cost-effective, scalable strategy for interrogating functional diversity in various biological systems.

In summary, this paper establishes organelle architecture as a fundamental, measurable dimension of cell biology, introducing sOrganellomics as a powerful tool to map tissue adaptability, metabolic transitions, and early disease progression in situ.