FILM: Mapping organellar metabolism by mid-infrared photothermal modulated fluorescence

1. Problem Statement

Metabolic processes in eukaryotic cells are compartmentalized within organelles, yet current technologies struggle to map these metabolic states in vivo with both high spatial resolution and chemical specificity.

• Limitations of Existing Methods:
  • Organelle-specific immunoprecipitation: Requires cell lysis, destroying spatial context and dynamic information.
  • Fluorescence microscopy: While excellent for structural dynamics, it relies on specific sensors that offer limited insight into complex metabolic mixtures.
  • Mid-Infrared Photothermal (MIP) Microscopy: Offers label-free chemical imaging based on vibrational signatures but traditionally suffers from low spatial resolution or requires long acquisition times.
  • Previous Fluorescence-detected MIP (F-MIP): While capable of organelle-level imaging, early implementations suffered from severe photobleaching and required exposure times three orders of magnitude longer than conventional fluorescence microscopy. This made comprehensive hyperspectral mapping of metabolic activity in living systems nearly impossible.

2. Methodology: The FILM System

The authors developed FILM (Fluorescence-detected mid-Infrared photothermaL Microscope), an upgraded system integrating three key innovations to overcome previous limitations:

A. Optical Boxcar Demodulation & Pulsed Excitation

• Principle: FILM detects IR absorption by measuring the photothermal heating of the sample, which induces dynamic quenching of a nearby fluorescent probe (reducing its quantum yield).
• Innovation: Instead of continuous-wave (CW) excitation, the system uses synchronized pulsed visible and IR lasers.
  • The IR pump (200 kHz) excites molecular vibrations.
  • The visible probe (400 kHz) is pulsed with a 30% duty cycle to gate the detection specifically during the "hot" (peak temperature) and "cold" states.
• Benefit: This "optical boxcar" strategy eliminates photons that do not contribute to the photothermal signal, reducing fluorescence exposure time by >100-fold compared to previous point-scan systems, thereby drastically mitigating photobleaching. It also shifts high-order harmonic signals to the detection frequency, enhancing signal amplitude.

B. AI-Assisted Data Denoising (SPEND)

• Challenge: Reducing exposure time to prevent bleaching resulted in a low Signal-to-Noise Ratio (SNR).
• Solution: The authors implemented SPEND (Self-permutation Noise2Noise Denoising), a self-supervised deep learning algorithm.
  • It splits a single noisy hyperspectral stack into two independent sequences (odd/even spectral frames) via permutation.
  • A 3D U-Net is trained on these pairs to learn noise statistics without requiring clean reference data.
• Benefit: SPEND improved image SNR by 26.9× and spectral SNR by 5.3×, enabling clear visualization of weak signals without increasing integration time.

C. Spectral Unmixing (MCR-LASSO)

• Challenge: Lysosomes contain hundreds of molecular species, making simple ratiometric analysis insufficient for quantitative decomposition.
• Solution: A hybrid unmixing approach:
  1. Augmented Multivariate Curve Resolution (MCR): Refines reference spectra using the actual lysosomal data to account for complex mixtures.
  2. LASSO (Least Absolute Shrinkage and Selection Operator): Performs quantitative decomposition using the refined references.
• Benefit: Allows for the quantitative mapping of specific biomolecular contents (e.g., proteins, lipids, amino acids) within individual organelles.

3. Key Results

A. Validation of FILM Performance

• The system successfully acquired hyperspectral data (1000–1800 cm⁻¹) of bacteria and organelles.
• Spectral fidelity was validated against scattering-based MIP and ATR-FTIR, confirming the ability to resolve characteristic peaks (e.g., Amide I/II, nucleic acids).

B. Revealing Lysosomal Metabolic Heterogeneity

• Chemical Specificity: FILM identified distinct spectral peaks in lysosomes corresponding to Amino Acids (AA) at ~1587 cm⁻¹ and Free Fatty Acids (FFA) at ~1711 cm⁻¹.
• Functional Proxies:
  • Proteolytic Activity: Defined by the ratio AA/Protein (1587/1649 cm⁻¹).
  • Lipolytic Activity: Defined by the ratio FFA/Lipid Esters (1711/1741 cm⁻¹).
• Heterogeneity: Even within a single cell, lysosomes exhibited significant metabolic diversity. Some were highly proteolytic, others lipolytic, and some both. This heterogeneity was not driven by organelle size or probe concentration.

C. Tracking Metabolic Changes During Aging

• Using C. elegans, the authors tracked lysosomal metabolism from Day 2 to Day 10 of adulthood.
• Early-Onset Dysfunction: Both proteolytic and lipolytic activities declined significantly as early as Day 4, preceding mortality.
• Spectral Shifts: Aging lysosomes showed a shift from AA/FFA dominance (Day 2) to increased accumulation of macromolecules like DNA, ceramides, triglycerides, and glycogen in later stages (Day 6–10).
• Increased Variability: Metabolic heterogeneity increased with age, suggesting a loss of homeostatic control.

D. Profiling Lysosomal Storage Diseases (LSDs)

• RNAi Models: Knockdown of genes associated with LSDs (e.g., nuc-1, asah-2, lipl-3) in C. elegans resulted in specific spectral signatures matching the expected accumulation of undegraded substrates (e.g., DNA, ceramides, glycogen).
• Mammalian Model: In Niemann-Pick Type C (NPC1 knockout) HEK293 cells, FILM detected reduced FFA and global accumulation of DNA, ceramides, triglycerides, and cholesterol esters, confirming impaired macromolecular degradation.

E. Generalizability

• The platform was successfully applied to image lipid droplets (dominated by ester C=O peaks) and mitochondria (protein-dominated with phosphate features), demonstrating its versatility beyond lysosomes.

4. Significance and Impact

• Technological Breakthrough: FILM solves the critical trade-off between spatial resolution, chemical specificity, and phototoxicity in live-cell imaging. It enables hyperspectral chemical imaging of organelles in living organisms with minimal photobleaching.
• Biological Insight:
  • It challenges the view of lysosomes as uniform organelles, revealing a metabolically heterogeneous population within single cells.
  • It identifies early metabolic dysfunction in aging (Day 4 in worms) that precedes phenotypic decline, offering new targets for anti-aging interventions.
  • It provides a label-free, quantitative method to diagnose and characterize Lysosomal Storage Diseases at the single-organelle level.
• Future Applications: This method establishes a "high-resolution chemical cellular atlas," paving the way for studying metabolic dynamics in neurodegeneration, cancer, and metabolic disorders in their native physiological context.

Conclusion

The FILM technique represents a paradigm shift in metabolic imaging. By combining optical boxcar demodulation, AI-driven denoising, and advanced spectral unmixing, it overcomes the historical barriers of photobleaching and low SNR in mid-infrared microscopy. This allows researchers to quantitatively map the metabolic fingerprints of individual organelles in living systems, providing unprecedented insights into the mechanisms of aging and disease.