Autofluorescence lifetime flow cytometry rapidly flows from strength to strength

Abstract

[Color figure can be viewed at wileyonlinelibrary.com]

In flow cytometry, the fluorescence lifetime of the reduced form of co-enzymes nicotinamide dinucleotide (NADH) and nicotinamide dinucleotide phosphate (NADPH) can be used as a reporter of metabolic activity in single cells. Moreover, their metabolic state can be determined faster than imaging them with fluorescence lifetime imaging (FLIM), as reported by Samimi et al. [1].

Ever since Britton Chance highlighted the benefits of using autofluorescence of cells to study metabolism, respiration and associated redox reactions in cells and tissues in the 1960s, [2] this topic has attracted the attention the biophysics, biochemistry and life science research communities [3]. FLIM is now routinely used to study NADH and NADPH fluorescence lifetimes, but here the authors show that not only no fluorescence labelling is required but imaging can also be dispensed with.

Autofluorescence is the term given to the fluorescence from naturally occurring fluorophores in cells or tissue without adding any exogenous fluorophores or labels—the fluorescence originates from intrinsic fluorophores. There is a whole range of biological intrinsic fluorophores, such as some amino acids, flavins, lipofuscin, chlorophyll, porphyrins, carotenoids, collagen, elastin and others that fluoresce naturally when excited with light of an appropriate wavelength [4, 5]. While fluorescence of some amino acids can be excited around 280 nm, where glass is opaque, the autofluorescence of NADH and NADPH has an absorption peak around 350 nm and can conveniently be excited at 375 nm—a wavelength at which glass is transparent [6]. This is a big advantage facilitating uptake of this approach by the scientific community, especially considering the ubiquitous use of glass in the life sciences.

NADH and NADPH are pyridine nucleotides and the fluorescence originates from their nicotinamide ring, peaking at around 460 nm or so [7]. They are small molecules that play a big role in redox reactions, respiration and metabolism, and allow the optical investigation, via their fluorescence, of biochemical states and metabolic pathways in cells and tissues [3].

In general, fluorescence can be characterized by several features: intensity, wavelength, lifetime and polarization, and, in combination with microscopy, the position in the image where it originates from. The more of these parameters can be measured, the higher the biochemical resolving power of the measurement [8]. The fluorescence intensity can yield information about location, concentration and fluorescence quantum yield, the spectrum about the color of the emission (e.g., used to highlighting neurons in different colors in “brainbow” samples in neuroscience [9]), the lifetime about the time the fluorophores resides in the excited state (typically a function of the fluorophore's environment) and the polarization can yield information about the rotational mobility of the fluorophore, or proximity to its neighbors. The lifetime is the average time the fluorophore remains in the excited state after excitation, and this parameter is of interest in the context of NADH and NADPH, collectively expressed as NAD(P)H, due to their indistinguishable fluorescence features [6, 7].

Looking at the fluorescence lifetime of NAD(P)H in flow cytometry, as Samimi et al. do, [1] it is possible to distinguish between bound and unbound states of these co-enzymes. If it is unbound and in the free state, then its fluorescence lifetime is around 400 picoseconds, whereas upon binding to proteins, this value increases to several nanoseconds. A double-exponential fit to the fluorescence decay reveals the amplitudes and thus the contributions of bound and unbound NAD(P)H to the fluorescence decay [10]. When the metabolism becomes more oxidative, the amount of bound NADH increases. Indeed, autofluorescence FLIM is a useful and established method to look at metabolism in cells and tissues [3, 6, 7, 10-12].

The NAD(P)H fluorescence decays are acquired with a technique called time-correlated single photon counting (TCSPC). TCSPC is the best technique to measure fluorescence decays, with the main advantages stemming from its digital nature (either a photon is detected, or it is not, “yes” or “no”), high time resolution and its single photon sensitivity [13, 14]. TCSPC only needs a low illumination intensity, and its capability of resolving multi-exponential fluorescence decays is critical for the work presented by Samimi et al. [1] as it is the key feature to distinguish and quantify different metabolic states.

The great advantage of optical methods is that they allow non-invasive and non-destructive interrogation of the metabolic state in intact cells. This is often combined with imaging, for example, in fluorescence microscopy, a technique reviewed by Renz for readers of this journal [15]. Adding lifetime imaging capabilities, and thus performing FLIM, allows mapping of bound and unbound NAD(P)H in cells, as demonstrated by Lakowicz et al. some time ago [11]. Indeed, FLIM of NAD(P)H is a widely used analysis method in live cells and tissues, [16] ex vivo cells, [17] fixed cells, [18, 19] in clinical applications [12] and is being considered for fluorescence-guided surgery [20] and machine learning approaches for optical histopathology analysis in medical diagnosis and treatment [21].

Although seeing is believing, Greek philosopher Heraclius in the 5th century BC stated “panta rhei”—everything flows. Here, in the latest application of this philosophy, the authors dispense with imaging and use flow cytometry instead [22]. Rather than imaging a large number of cells with FLIM, they relinquish the spatial information that imaging provides—which, after all, is not really needed for their purposes: their aim is to analyze the metabolic state of as many cells as possible in a short amount of time. Flow cytometry is the technique allowing this: they just let the cells flow through a flow cytometer and interrogate each cell for its fluorescence decay.

Using a small compact 50 MHz pulsed UV diode laser emitting at 375 nm with 0.6 mW average power, a bandpass filter centred at 440 nm and a photomultiplier with a blue-sensitive photocathode, they show that the average fluorescence lifetime of Jurkat T cells flowing through the illumination spot clearly shortens upon treatment of the cells with sodium cyanide, an electron transport chain disruptor.

In addition, T cell activation increases glycolysis and shortens the average lifetime compared to quiescent T cells which could be observed in real time using a phasor plot and quantified by decay fitting in post-processing decay data analysis. The same was true for activated and quiescent primary neural stem cells. They acquired fluorescence decays from 1500 single cells in 10 minutes, and the loading of the cells into a syringe was the only user input required.

They use a real-time phasor plot representation to visualize the multiexponential NAD(P)H decays during the measurement. This does not require any decay fitting or assumption of a decay model—it is a simple Fourier transform of the fluorescence decay where the imaginary part is plotted against the real part [23, 24]. This can computationally be executed rapidly, much faster than decay fitting, and the phasor plot can be updated in real time. One cell yields one fluorescence decay and one data point in the phasor plot. Alternatively, fast analysis could also be achieved with the centre of mass method of fluorescence decay analysis [25].

They retain the full decays for post-processing by fitting a bi-exponential decay to determine the bound and unbound ratio of NAD(P)H. They compare this to two-photon excitation FLIM which provides spatial resolution and requires an operator for the microscope. They show a 300 × 300 μm² field of view FLIM image with around 100 cells, and state they needed 2 hours to acquire FILM images of 1500 cells. By comparison, their flow cytometry method was 10 times faster, and they estimate that they can go 100 times faster.

Although frequency domain fluorescence lifetime flow cytometry was reported as far back as 1993, [26-28] and has been used with NAD(P)H autofluorescence, [29, 30] TCSPC flow cytometry is a more recent approach [25] and this is the first report applying it to NAD(P)H autofluorescence. The great advantage of this label-free approach is that the samples do not have to be modified in any way, and with TCSPC-based lifetime acquisition, the illumination intensity can be lower than for frequency domain methods. Samimi et al.'s average laser power is 0.6 mW (pulsed, low flow rate with 2.5 cells/s), whereas frequency-domain approaches have been quoted as 60 mW (cw, fast flow rate with 1000 cells/s) [29, 30].

Imaging the metabolic state of cells with FLIM allows spatial resolution to be obtained, and to identify the position of bound and unbound NAD(P)H in the cell. Improving the spatial resolution and imaging properties of fluorescence microscopes has been a field of great activity lately, with the development of super-resolution fluorescence microscopy [31] and a mesolens [32] which allows imaging a large field of view with high spatial resolution. However, if the aim is to investigate the metabolic state of as many cells as possible in a short space of time, or to look for rare or unusual cells, irrespective of any morphological information, then flow cytometry with lifetime capabilities is the way to go [33]. (If really needed, FLIM and lifetime flow cytometry can even be merged, according to a recent report. [34])

The authors quote an analysis speed of 2.5 cells/s, but faster speeds are possible. Indeed, using a 32 × 32 pixel single photon avalanche diode (SPAD) array, TCSPC flow cytometry of 60,000 cell/s has been estimated for acquisition [35] similar to frequency domain flow cytometry [33]. While the current work only analyses the cells, the authors point out that in the future, it would be straight-forward to sort the cells as well, with a sorting mechanism based on the real-time phasor plot analysis of the fluorescence lifetimes.

Finally, the illumination intensity 530 mW/cm² at 375 nm is higher than sunshine (140 mW/cm², all wavelengths), but it is only applied for 5 milliseconds per cell, and with 2.7 J/cm² well below the viable light dose of 25 J/cm² at 375 nm [36].

Flow cytometry is used in many applications, from high-throughput analysis to sorting in cell biology and clinical research and diagnostics, drug discovery and other applications. The addition of lifetime capabilities, perhaps also time-resolved fluorescence anisotropy, [7, 37, 38] the development of novel detectors and perhaps deep-learning approaches to data analysis [21] will make it an even more powerful analytical tool to extract a maximum of biochemical and biophysical information from cells flowing through the illumination spot for only milliseconds.

Klaus Suhling: Conceptualization; writing – review and editing; writing – original draft.