High-Throughput Single-Cell Spectroscopy Using Phasor Analysis of Spectral Flow Cytometry

This paper introduces phasor analysis for spectral flow cytometry (phSFC), a high-throughput method that successfully extends the fit-free interpretation of membrane order and dynamics from hyperspectral microscopy to single-cell populations, demonstrating its robustness through validation with lipid vesicles, live cells, and primary leukocytes.

The Gist

Imagine you are trying to understand the "personality" of a crowd of people. You have two ways to do it:

1. The Microscope Approach (HSI): You stand very close to a few people, looking at them one by one. You can see exactly where they are standing, what they are wearing, and how they interact with their immediate neighbors. It's like taking high-definition photos of a small group.
2. The Flow Cytometry Approach (SFC): You put everyone through a giant, high-speed turnstile. You can't see where they stand or who they are talking to, but you can scan thousands of people per second and measure their "vibe" (color, brightness, etc.) incredibly fast.

The Problem:
Scientists have a special tool called Phasor Analysis that helps them understand complex "vibes" (specifically, the color of light emitted by cells) without needing complicated math models. It's like a universal translator that turns messy light data into a simple map.

Until now, this translator only worked for the Microscope Approach (looking at a few people up close). Scientists wanted to use it for the Flow Cytometry Approach (scanning thousands of people fast), but the data looked different, so the translator didn't work.

The Solution (The Paper's Big Idea):
The authors of this paper built a new version of the translator that works for the high-speed turnstile. They call it phSFC (Phasor-based Spectral Flow Cytometry).

Here is how they did it and what they found, using simple analogies:

1. The "Magic Paint" (LAURDAN)

To test their new tool, they used a special dye called LAURDAN. Think of this dye as magic paint that changes color depending on how "stiff" or "fluid" the surface it's painted on is.

• If the surface is stiff (like a frozen pond), the paint glows one way.
• If the surface is fluid (like a swimming pool), the paint glows a different way.
• If the surface is a mix, the paint glows somewhere in between.

2. The "Traffic Jam" Test (Lipid Vesicles)

First, they tested the tool on Multilamellar Vesicles (MLVs). Imagine these are tiny, hollow bubbles made of fat (lipids).

• They made bubbles that were purely stiff, purely fluid, or a mix, and added different amounts of cholesterol (which acts like a stiffener, turning a fluid pool into a stiff ice rink).
• The Result: They scanned these bubbles with both the Microscope and the Turnstile. Even though the Turnstile saw them as a "cloud" of data and the Microscope saw them as distinct spots, both tools agreed perfectly. They could clearly see which bubbles were stiff and which were fluid. The Turnstile just did it 100 times faster and with more bubbles.

3. The "Cellular Dance Floor" (Live Cells)

Next, they looked at real living cells (Vero cells).

• They used a chemical (MβCD) to steal the cholesterol out of the cell membranes.
• The Analogy: Imagine a dance floor where the dancers are holding hands tightly (stiff membrane). When you steal the "glue" (cholesterol), the dancers start slipping and sliding (fluid membrane).
• The Result: Both the Microscope and the Turnstile saw the dancers start slipping. The Turnstile confirmed that the cells became more fluid, matching what the Microscope saw, but it analyzed thousands of cells at once instead of just a few.

4. The "Inflamed City" (Real-World Test)

Finally, they took this to the real world. They looked at immune cells (white blood cells) from mice with lung inflammation (caused by a bacterial trigger called LPS).

• These cells were a messy mix: they had the magic paint, antibodies (like ID tags), and their own natural glow (autofluorescence). It was like trying to hear a specific instrument in a loud orchestra.
• The Superpower: The new phSFC tool acted like a noise-canceling headphone. It could separate the "orchestra" into individual instruments. It successfully ignored the background noise and the ID tags to focus only on the membrane stiffness.
• The Discovery: They found that in inflamed lungs, the immune cells' membranes actually became stiffer (more ordered). This was a surprise, as some previous studies suggested they might get looser. The high-speed Turnstile gave them enough data to be sure about this new finding.

Why Does This Matter?

Think of this paper as building a bridge between two worlds:

• World A (Microscopy): High detail, low speed. Good for seeing where things happen.
• World B (Flow Cytometry): Low detail (spatially), high speed. Good for seeing how many things happen.

By creating phSFC, the scientists showed that you can now use the "magic paint" analysis on the high-speed Turnstile.

• Before: If you wanted to study membrane stiffness in a disease, you had to look at a few cells under a microscope for hours.
• Now: You can scan thousands of cells in minutes, get a statistically rock-solid answer, and still use the same simple "map" (Phasor) that scientists have trusted for years.

In short: They took a powerful, easy-to-use map for understanding cell membranes and upgraded it to work on a high-speed assembly line, allowing scientists to study diseases with much greater speed and confidence.

Technical Summary

1. Problem Statement

• Limitations of Current Spectral Flow Cytometry (SFC): Traditional SFC analysis relies on spectral unmixing based on reference controls and manual gating strategies defined by intensity thresholds. These methods are user-dependent, require accurate reference spectra, struggle with low-abundance components (<5%), and often fail to treat autofluorescence as a distinct spectral component. They also lack the precision to capture continuous or overlapping spectral shifts.
• Gap in Phasor Analysis: Phasor analysis is a well-established, fit-free tool for interpreting complex fluorescence in hyperspectral imaging (HSI) and fluorescence lifetime imaging (FLIM). However, it has historically been restricted to microscopy modalities. There was no framework to apply phasor analysis to high-throughput SFC data, despite SFC's superior statistical power for analyzing large cell populations.
• Need for Integration: Researchers needed a unified analytical framework that bridges the spatial resolution of microscopy with the statistical robustness of flow cytometry to study membrane dynamics (specifically using the solvatochromic probe LAURDAN) at a single-cell level without prior assumptions or spectral modeling.

2. Methodology

The authors developed phSFC (Phasor-based Spectral Flow Cytometry), a unified analytical framework adapting spectral phasor analysis for SFC data.

• Data Acquisition:
  • SFC: Used a Cytek Aurora spectral flow cytometer (405 nm, 488 nm, 633 nm lasers). Data was acquired for Multilamellar Vesicles (MLVs), Vero cells, and primary leukocytes from bronchoalveolar lavage (BAL).
  • HSI (Benchmark): Used a Zeiss LSM 880 confocal microscope with a 405 nm laser and 28 spectral channels (423–693 nm) to acquire hyperspectral images of the same samples.
• Sample Preparation:
  • MLVs: Prepared with defined lipid compositions (DOPC, DPPC, and mixtures) and varying cholesterol concentrations (0%, 33%, 50%) to create fluid, gel, and liquid-ordered/disordered phases.
  • Live Cells: Vero cells treated with Methyl-β-Cyclodextrin (MβCD) to deplete cholesterol.
  • In Vivo Model: CD1 mice injected with LPS to induce lung inflammation; BAL cells were isolated and stained with LAURDAN plus antibodies (anti-CD45, anti-CD11c).
• Analytical Framework (Python-based):
  • Preprocessing: SFC data (FCS files) was gated in FlowJo to isolate single events and specific populations (e.g., CD45+ cells). Spectral data (16 violet channels) was extracted. HSI data was masked to isolate regions of interest.
  • Phasor Transformation: A Fourier-based transformation was applied to spectral vectors ($I(\lambda)$) to calculate real ($G$) and imaginary ($S$) coordinates for the $n$-th harmonic:
    $$G = \frac{\sum I(\lambda) \cos(2\pi n \frac{\lambda - \lambda_{min}}{\lambda_{max} - \lambda_{min}})}{\sum I(\lambda)}$$
    $$S = \frac{\sum I(\lambda) \sin(2\pi n \frac{\lambda - \lambda_{min}}{\lambda_{max} - \lambda_{min}})}{\sum I(\lambda)}$$
  • Unmixing: Multi-component analysis (using 1st and 2nd harmonics) was employed to deconvolve mixed signals. For complex samples (BAL), a 5-component unmixing approach was used to separate LAURDAN (liquid order/disorder), antibody signals (CD45, CD11c), and autofluorescence.

3. Key Contributions

• First Implementation of phSFC: Established the first workflow to apply spectral phasor analysis to SFC, creating a fit-free, reference-free method for high-throughput single-cell spectral characterization.
• Unified Analytical Framework: Demonstrated that the same mathematical principles used in HSI can be directly translated to flow cytometry, preserving interpretative continuity between imaging and cytometric modalities.
• Advanced Unmixing Capability: Showed that phSFC can resolve complex spectral mixtures (probe + autofluorescence + antibody signals) using $n$-harmonic phasor analysis, overcoming the limitations of traditional spectral unmixing which often ignores autofluorescence.
• Statistical Robustness: Highlighted that SFC produces more compact phasor distributions than HSI due to the integration of signal across the entire cell (vs. pixel-by-pixel in HSI), offering superior statistical power for detecting subtle population shifts.

4. Key Results

• Validation with Model Membranes (MLVs):
  • phSFC successfully reproduced the distinct phasor signatures of fluid (DOPC), gel (DPPC), and liquid-ordered phases previously established by HSI.
  • Both modalities accurately resolved cholesterol-dependent spectral shifts. DOPC showed a linear trajectory with increasing cholesterol, while DPPC/mixed membranes showed non-linear, complex trajectories.
  • SFC distributions were tighter (more compact) than HSI, reflecting the averaging of signal over the whole event rather than spatial heterogeneity.
• Live Cell Membrane Dynamics (Vero Cells):
  • Both HSI and SFC detected the expected shift toward membrane disorder (green shift in phasor space) following MβCD-induced cholesterol depletion.
  • SFC confirmed the reduction in the liquid-ordered fraction (center of mass shift from 0.40 to 0.35) with high statistical confidence, despite the presence of autofluorescence.
• In Vivo Application (BAL Macrophages):
  • Applied to primary leukocytes from LPS-challenged mice, phSFC successfully separated LAURDAN signals from autofluorescence and antibody labels using 5-component unmixing.
  • Finding: LPS-treated macrophages exhibited a significant increase in the liquid-ordered membrane fraction (center of mass shift from 0.25 to 0.44) compared to controls, suggesting membrane rigidification during inflammation. This contrasts with some in vitro studies suggesting fluidification, potentially due to the "whole-cell" signal capturing internal organelle reorganization in primary cells.

5. Significance

• Bridging Modalities: The work establishes SFC as a robust, complementary extension of hyperspectral microscopy. It allows researchers to validate spatial findings from microscopy with high-throughput statistical power from flow cytometry.
• Overcoming Traditional Limits: By using a model-free, fit-free approach, phSFC eliminates the need for reference spectra and allows for the detection of continuous spectral transitions and rare subpopulations that are often missed by threshold-based gating.
• Complex Biological Insights: The ability to deconvolve multiple fluorescent sources (probes, antibodies, autofluorescence) in a single measurement makes phSFC ideal for complex in vivo studies and immunophenotyping where spectral overlap is common.
• Future Applications: The framework is applicable to any solvatochromic probe or spectral biosensor (e.g., FRET, calcium indicators) exhibiting activity-dependent spectral shifts, enabling new avenues for studying metabolic remodeling, membrane reorganization, and disease heterogeneity at the single-cell level.