Autofluorescence in Flow Cytometry

Autofluorescence in flow cytometry refers to the natural glow that cells emit when hit by a laser, caused by molecules already present inside them rather than by any added dye or label. For decades, this glow was treated almost exclusively as a nuisance, something that muddied the signal researchers actually cared about. That framing is shifting. Advances in spectral instruments and computational analysis have turned autofluorescence from background noise into a biological readout in its own right, one that can reveal metabolic states, identify cell types, and even sort aging cells from healthy ones without a single staining step.

Where the Glow Comes From

Every living cell contains molecules that absorb light at certain wavelengths and re-emit it at longer ones. These endogenous fluorescent compounds, sometimes called autofluorophores, span a wide range of biological roles. Some are metabolic workhorses: NAD(P)H and FAD, cofactors involved in energy production, fluoresce in the blue-green and green-yellow range, respectively. Collagen, the structural protein that holds connective tissue together, emits its own fluorescence. Then there are waste products like lipofuscin and advanced glycation end products, which build up in cells over time and are linked to aging and cellular dysfunction.¹ The combined output of all these molecules is what a flow cytometer picks up as autofluorescence whenever its laser hits an unstained cell.

Because different cell types carry different concentrations of these molecules, autofluorescence varies enormously from one cell to another. Alveolar macrophages, the immune cells that patrol the lungs, are famously bright: their high content of ingested debris and metabolic cofactors gives them a strong fluorescent signal. Lymphocytes, by contrast, are relatively dim. Even within a single cell type, autofluorescence intensity can shift depending on the cell’s metabolic activity, its age, and environmental conditions. This variability is both the source of the problem in conventional flow cytometry and the basis for newer label-free techniques.

Why Autofluorescence Causes Trouble in Standard Flow Cytometry

In a typical flow cytometry experiment, you stain cells with fluorescently labeled antibodies to identify surface markers or internal proteins. The instrument then measures how bright each cell is in specific detector channels. The challenge is that the cell’s own autofluorescence overlaps with many common fluorescent dyes, especially those that emit in the green-to-yellow part of the spectrum. When you are trying to detect a dim marker on a cell surface, the cell’s intrinsic glow can be loud enough to drown out or mimic the real signal. Background fluorescence from autofluorescence, spectral overlap between dyes, and unwanted antibody binding are the three main sources of background noise that can throw off accurate measurement of antigen-positive cells.²

The problem gets worse with certain sample handling steps. Fixation, a routine procedure used to preserve cells for later analysis, can dramatically increase autofluorescence. In one study tracking leukocyte surface markers after fixation, monocyte autofluorescence climbed roughly ninefold by 96 hours post-fixation.³ Other work has confirmed that fixed cells of various types show a noticeable increase in autofluorescence compared to fresh samples.⁴ This is something worth keeping in mind if you are working with stored or fixed samples. What seemed like a tolerable background in fresh tissue can balloon into a real obstacle once the fixation chemistry alters cellular proteins.

Strategies for Reducing Unwanted Autofluorescence

Researchers have developed several practical approaches to beat back autofluorescence when it gets in the way.

Chemical Quenching

Sudan Black B, a lipophilic dye historically used in histology, can physically and chemically block autofluorescence. It works by interacting directly with autofluorescent molecules and by forming a physical barrier that absorbs stray light.⁵ In tissue sections or isolated cells, a brief incubation with Sudan Black B (on the order of 8 to 15 minutes depending on sample type) can effectively mask intrinsic fluorescence without destroying the specific fluorescent labels you want to see.⁶ Other chemical quenchers exist, including trypan blue and copper sulfate solutions, each with trade-offs in terms of which wavelengths they suppress and how much they affect specific staining intensity.

Shifting to Longer Wavelengths

Most autofluorescence is strongest when excited by shorter-wavelength (blue and violet) lasers and emits primarily in the green-yellow range. One straightforward workaround is to design panels that rely on fluorophores excited by red or near-infrared lasers, where cellular autofluorescence is substantially lower.⁷ The signal-to-noise improvement can be considerable, especially for dim markers. Detector technology matters here too. Avalanche photodiodes offer better sensitivity than traditional photomultiplier tubes at wavelengths beyond about 650 nanometers, extending the useful working range of a flow cytometer past 1,000 nanometers.⁸ Pairing a red or near-infrared laser with an avalanche photodiode detector gives you both lower background and better sensitivity at those longer wavelengths.

Gating and Compensation

In conventional flow cytometry, the standard approach is to run an unstained control sample and use its fluorescence profile to set a baseline. Any signal above that baseline is attributed to the fluorescent label. Compensation algorithms then mathematically subtract spectral overlap between channels. These strategies work reasonably well for bright markers, but they start to break down for dim signals, because the autofluorescence of individual cells varies. A single threshold drawn from the population average will misclassify some naturally bright cells as positive and some genuinely positive but dim cells as negative.

How Spectral Flow Cytometry Changes the Game

Spectral flow cytometry captures the full emission spectrum of each cell across dozens of detectors simultaneously, rather than measuring a few discrete wavelength bands. This makes a fundamental difference in how autofluorescence is handled. Instead of treating autofluorescence as a flat background to subtract, spectral instruments extract the autofluorescence signature from unstained reference cells and then use that signature during computational unmixing, effectively separating the autofluorescence component from each fluorophore’s signal at the individual-cell level.⁹

The practical payoff is clearest when you are looking for markers expressed at low levels. In conventional instruments, those dim signals can vanish into the autofluorescent background. Full spectrum unmixing can unmask them, which is especially valuable for immunophenotyping projects that need to identify rare or faintly stained cell subsets.¹⁰ If you have ever struggled with a histogram where positive and negative populations overlap so much they look like one broad hill, spectral unmixing with proper autofluorescence extraction is likely where resolution improves most.

One subtlety that trips people up: the quality of your autofluorescence reference matters a lot. If the unstained reference does not accurately represent the autofluorescence of the cells in your stained sample, the unmixing algorithm can introduce artifacts rather than removing them. Cells that have been treated differently, stimulated, cultured, or fixed at different time points may have a genuinely different autofluorescence profile from the unstained control. Matching the reference to the conditions of the experiment is an underappreciated step that affects data quality more than many panel design choices.

Autofluorescence as the Signal, Not the Noise

The flip side of autofluorescence being a nuisance is that it carries real biological information. When the fluorescence comes from metabolic cofactors, changes in autofluorescence intensity directly reflect changes in a cell’s metabolic state. When it comes from waste products like lipofuscin, it tracks cellular aging. A growing body of work treats autofluorescence as the primary measurement rather than something to subtract away.

Metabolic Profiling Without Labels

NAD(P)H and FAD are the two metabolic cofactors most commonly targeted. Their ratio, often called the optical redox ratio, shifts depending on whether a cell is relying more on oxidative phosphorylation or glycolysis. Recent work has shown that these signals are measurable by flow cytometry without any staining step. Inhibiting mitochondrial respiration, uncoupling mitochondria, or changing glucose availability all produce detectable changes in the optical redox ratio, making it possible to profile the metabolic state of distinct cell populations in real time.¹¹ This approach has even been applied to stallion spermatozoa, where autofluorescence-based flow cytometry was used to assess how different energy substrates (glucose, lactate, pyruvate) affected sperm metabolic activity without any labels at all.¹²

The appeal is obvious: no labels means no staining artifacts, no worries about dye toxicity to live cells, and no cost per reagent. You can combine surface marker staining with autofluorescence-based metabolic readouts in the same tube, getting phenotype and metabolic state simultaneously. The limitation is sensitivity. The autofluorescence from NAD(P)H and FAD is not bright compared to synthetic fluorophores, so you need good optical design and careful instrument setup to resolve meaningful differences, especially in cells with low metabolic activity.

Tracking Cellular Senescence

Senescent cells, those that have stopped dividing and entered a state of permanent growth arrest, accumulate lipofuscin and other autofluorescent waste products. This makes autofluorescence a surprisingly reliable marker of senescence. In human mesenchymal stromal cells, researchers demonstrated that autofluorescence measured by label-free flow cytometry tracked tightly with standard senescence markers: high-autofluorescence cells showed increased expression of lipofuscin-associated proteins compared to their low-autofluorescence counterparts.¹³ Because the measurement is fast, nondestructive, and requires no reagents beyond a flow cytometer, it offers a practical way to screen cell populations for senescence, which is relevant for quality control in cell therapy manufacturing and for basic aging research.

Cell Type Identification by Autofluorescence Fingerprint

Different cell types do not just differ in how bright their autofluorescence is; they differ in its spectral shape. This opens the door to identifying or sorting cells based purely on their intrinsic glow. Alveolar macrophages, for example, have been separated from dendritic cells, lymphocytes, and granulocytes in bronchoalveolar lavage samples by sorting on autofluorescence intensity alone. The high-autofluorescence fraction was predominantly macrophages, while the low-autofluorescence fraction contained dendritic cells and other cell types.¹⁴ In mouse lung tissue, similar logic works: CD11c-bright cells split into a low-autofluorescence population with dendritic cell characteristics and a high-autofluorescence population with macrophage features, including weak T-cell stimulation and low expression of co-stimulatory molecules.¹⁵

More recently, machine learning methods have been combined with autofluorescence data to classify macrophage phenotypes. Each polarization state turns out to have a distinct autofluorescence fingerprint, suggesting this approach could supplement or even replace antibody-based phenotyping in some contexts.¹⁶ And a deep learning framework applied to imaging flow cytometry data achieved over 96% accuracy in classifying neural stem cell differentiation stages without any labels, relying instead on cellular morphology and autofluorescence captured in brightfield and darkfield images.¹⁷

Autofluorescence in Microbiology and Ecology

Autofluorescence-based flow cytometry is not limited to mammalian cells. Microalgae and bacteria both carry endogenous fluorophores that can be exploited for identification and monitoring.

Microalgae are particularly well suited to this approach because their photosynthetic pigments are intensely fluorescent. A study examining 32 representatives from nine major algal groups found that each displayed distinct autofluorescence patterns on a full-spectrum cytometer. The most striking division was between taxa containing phycobiliproteins (like cyanobacteria, cryptophytes, and red algae, which emit in the 570 to 650 nanometer range) and those without (like green algae and diatoms, which fluoresce primarily in the 650 to 720 nanometer chlorophyll band).¹⁸ For environmental monitoring and aquaculture, this means you can distinguish algal groups in mixed water samples without staining, at the speed of flow cytometry.

Bacteria are dimmer and trickier, but still detectable with the right equipment. A high-sensitivity flow cytometer was used to detect and quantify autofluorescence in eight bacterial strains at the single-cell level. The green autofluorescence was attributed mainly to endogenous flavins and ranged from about 80 to 1,400 FITC-equivalent molecules per cell, varying by species, with substantial cell-to-cell variation even within the same strain.¹⁹ That kind of variation is itself informative: it may reflect differences in metabolic state across a population, which is relevant for understanding antibiotic susceptibility and dormancy.

Practical Considerations for Your Own Experiments

If you are designing a flow cytometry experiment and wondering how to deal with autofluorescence, the strategy depends on whether you view it as signal or noise.

When autofluorescence is the enemy, your best tools are wavelength selection and careful controls. Choose fluorophores that emit in the red or near-infrared range where autofluorescence is weakest. Run fixation-matched unstained controls if your protocol includes fixation, because a fresh unstained control will underestimate the background in your fixed stained samples. Consider chemical quenching with Sudan Black B or similar agents when working with highly autofluorescent tissues. And if you have access to a spectral cytometer, take advantage of autofluorescence extraction during unmixing rather than relying on simple threshold-based gating.

When autofluorescence is the point, pay attention to excitation wavelength and detector sensitivity. NAD(P)H is typically excited in the ultraviolet (around 340 to 380 nanometers) and emits in the blue, while FAD absorbs blue light (around 450 nanometers) and emits in the green. You need UV or violet laser lines to excite NAD(P)H effectively, and your instrument’s optical path needs to be clean enough to resolve the modest signals these cofactors produce. Spectral cytometers are increasingly popular for this work because they can capture the full emission profile and separate autofluorescence from any overlapping dye signals computationally.

One thing that catches people off guard: cell culture conditions alter autofluorescence. Cells grown in media with phenol red, riboflavin, or folic acid will have different baseline fluorescence than cells grown in minimal media. Passage number matters too, as late-passage cells accumulate more lipofuscin and glow brighter. If you are comparing autofluorescence across conditions, you need to control for these variables as tightly as you control antibody concentrations or staining times.

Machine Learning and the Future of Label-Free Cytometry

The combination of high-parameter spectral data and modern computational tools is what makes autofluorescence-based cytometry newly practical. A human looking at a two-parameter dot plot might struggle to see meaningful structure in autofluorescence data, but algorithms trained on full spectral profiles or imaging flow cytometry data can pick up patterns that are invisible to the eye. The neural stem cell classification study mentioned earlier is one example, but the same principle is being applied across cell biology: train a model on cells whose identity you know from conventional staining, then use the model to classify unknown cells by their intrinsic optical properties alone.

The appeal for clinical applications is significant. Label-free methods sidestep reagent costs, staining variability, and the risk that the staining process itself alters the cells you are trying to study. For cell therapy products, where you need to characterize large numbers of cells and ideally leave them unaltered for transplantation, autofluorescence-based quality control is an attractive prospect. The same goes for any setting where speed matters more than deep phenotyping: rapid screening of blood or tissue samples for metabolic abnormalities, monitoring bacterial contamination in bioreactors, or tracking algal community composition in real time.

The field is still young enough that standardization is a real gap. There is no widely agreed-upon protocol for reporting autofluorescence measurements, and instrument-to-instrument variability can be large. A cell that looks “high autofluorescence” on one spectral cytometer may look different on another with a different laser configuration or detector array. Efforts to develop reference standards and reporting frameworks are underway, but for now, cross-lab reproducibility remains a challenge that limits how quickly label-free methods can move from proof-of-concept to routine use.