Label-free characterization of single extracellular vesicles using two-photon fluorescence lifetime imaging microscopy of NAD(P)H

The heterogeneous nature of extracellular vesicles (EVs) creates the need for single EV characterization techniques. However, many common biochemical and functional EV analysis techniques lack single EV resolution. Two-photon fluorescence lifetime imaging microscopy (FLIM) is widely used to functionally characterize the reduced form of nicotinamide adenine dinucleotide and nicotinamide adenine dinucleotide phosphate (NAD(P)H) in cells and tissues. Here, we demonstrate that FLIM can also be used to image and characterize NAD(P)H in single isolated EVs. EVs were isolated using standard differential ultracentrifugation techniques from multiple cell lines and imaged using a custom two-photon FLIM system. The presented data show that the NAD(P)H fluorescence lifetimes in isolated cell-derived EVs follow a wide Gaussian distribution, indicating the presence of a range of different protein-bound and free NAD(P)H species. EV NAD(P)H fluorescence lifetime distribution has a larger standard deviation than that of cells and a significantly different fluorescence lifetime distribution than the nuclei, mitochondria, and cytosol of cells. Additionally, changes in the metabolic conditions of cells were reflected in changes in the mean fluorescence lifetime of NAD(P)H in the produced EVs. These data suggest that FLIM of NAD(P)H could be a valuable tool for EV research.


Supplementary Note 1: Phasor analysis of fluorescence lifetimes
The fluorescence lifetime, or average time between fluorophore excitation and fluorescent photon emission, is given by equation (1).
Here, kr is the rate of radiative (fluorescent) decay, and knr is the rate of non-radiative (quenching) decay. The number of fluorophore molecules present in the excited state (n) at a given time (t) can then be characterized with the following differential equations in equations (2a,b).
Thus, the time between excitation and fluorescence emission follows an exponential decay probability function for a given fluorophore at constant conditions. This means that the probability (p) of a photon being emitted after a given amount of time (t) follows equation (3).
The α variable is a constant for normalization. In a non-hypothetical environment, molecular interactions and nanoscale differences in variables such as temperature and pH cause differences in fluorescence lifetime. When a group of fluorophores in different conditions is probed, the corresponding probability function of time between excitation and emission will be the sum of many decaying exponentials, as shown in equation (4).
The α variable now represents the relative intensity of each species present. Another factor at play in fluorescence lifetime measurements is the detector time response function, d(t), since the detector does not provide an instantaneous response. The repetition rate of the laser, f, can also be represented in radians as ωo, or 2πf. Each pulse of the laser will cause the emission of photons that follow the shape of s(t) and are detected as the convolution of d(t) and s(t), as shown in equation (5) below, where the measured fluorescence is given by f(t).
From here, the challenge is to characterize the unknown variable s(t) with measured f(t) and knowledge of d(t) and as ωo. Equation (5) represented the time domain, but this equation can easily be converted into the frequency domain in equation (6) and rearranged in equation (7).
, ) * = = > ?@ > = A ?@ A = 3 @ + 0 @ (10a) These two basis components should fall within the unit square (but due to noise in experimental measurements, they occasionally do not) as represented in Supplementary Fig. S1. Any sum of exponential decays can be fully described by these two basis components, and the mean fluorescence lifetime of the sum of decays is calculated with equation (11).
Supplementary Figure S1 shows five different fluorescence lifetime profiles. In red, τ1 represents a pure fluorescence lifetime around 0.6 ns; τ2 in light blue is a pure fluorescence around 3 ns; τ3 in dark blue is a pure fluorescence lifetime around 4 ns; τ4 and τ5 both represent mixtures with mean fluorescence lifetimes of 1.6 ns. Mixtures of multiple fluorescence lifetimes are represented at a location between the fluorescence lifetimes present, weighted by relative intensity, α. For example, the τ4 location represents a mixture of τ1 and τ2, where the ratio of the intensity of fluorophores with fluorescence lifetime of τ1 to fluorescence lifetime τ2 is a1:a2.
The usefulness of phasor analysis of fluorescence lifetimes is highlighted by the ability to distinguish τ4 and τ5 as having different fluorescence lifetime components, despite having the same mean fluorescence lifetime. Figure S1. Example phasor plot with 5 fluorescence lifetime profiles represented. Measured fluorescence decays are decomposed into a two component basis (g and s) that can be visualized on this plot.

Supplementary Note 2: Negative control sample
To verify that the particles being imaged with FLIM were cell-derived EVs, the differential centrifugation, sample preparation, and image processing procedures were performed on naïve serum-free media that was not incubated with any cells. As can be seen in Supplementary Fig. S2, only four particles were found in the automated blob detection segmentation, and of those only one contained a fluorescence lifetime profile falling on the phasor plot. Likely, these particles are not EVs containing NAD(P)H, but are some other type of particulate or contaminant that is scattering light, and not fluorescing. Similar results were obtained using EV diluent (sterile PBS that was previously passed through 50 nm PES syringe filter twice). These data provide evidence that the particles imaged throughout the main text are cell-derived.   Figure     NADPH, NADH plasma membrane 15+ Gene names, NADH or NADPH affiliation, primary cellular location, and references for various proteins found in EVs. All cited studies used mass spectrometry for protein identification unless indicated with * for Western Blot or + for flow cytometry.

Supplementary Note 3: Cell concentration and viability
Prior to experiments, cell confluency was determined visually to be about 70% on the day cells were switched to serum-free media. On the day of EV isolation, cell viability was measured with Vi-CELL Cell Viability Analyzer (Beckman Coulter, Horsham, PA). Immediately after serum-free media was removed for EV isolation, cells were removed from culture flasks and rediluted in complete media. Consistent with ATCC protocols, 0.25% trypsin in HBSS was used to lift MDA-MB-231 and U87 MG cells from their flasks, followed by adding complete media to the trypsin and cell solution, centrifugation of that mixture at 125×g for 5 minutes to pellet cells, and finally resuspending the cells in complete media. A cell scraper was used to remove J774A.1 cells from their flasks. Once cells were resuspended in complete media, they were diluted in a 1:2 ratio with phenol red-free 1x TrypLE Select Enzyme (Gibco, Waltham, MA) and set in an incubator for 5 minutes to prevent clumps of cells. After 5 minutes, the cells were run through the Vi-CELL analyzer using 50 images to calculate cell concentration and viability using a software algorithm for Trypan Blue exclusion assay. Cell concentration, experimental conditions, viability, and concentration are given in Supplementary Table S4.

Supplementary Note 4: FLIM of large vs. small EVs
Serum-free cell culture fluid was sequentially centrifuged at 800×g for 10 minutes to pellet cells, then at 2,000×g for 30 minutes to pellet particles larger than 1000 nm, at 12,000×g for 60 minutes to pellet large EVs (LEVs), and at 100,000×g for 60 minutes to pellet small EVs (SEVs) to compare the capability of FLIM to measure LEVs and SEVs. Supplementary Table S5 provides information on the size, concentration, and number of EVs detected in FLIM for large and small EVs. Concentration is calculated to correspond to mL of initial cell culture fluid used. Mean diameter and concentration were determined with NTA as described in the main text. Equivalent samples of SEVs and LEVs showed more LEVs than SEVs in FLIM, but more SEVs than LEVs in NTA, which means that a higher proportion of LEVs are detected with NAD(P)H FLIM.
Volume is cubically related to diameter, so as EVs get smaller, they contain much less material. For example, this could cause an EV of diameter 50 nm to appear 8 time less bright than an EV of diameter 100 nm if both EVs had the same concentration of NAD(P)H. Furthermore, it is possible that small EVs contain less NAD(P)H or less bound NAD(P)H, which could further account for fewer of them being visible. These two factors are suspected to be the cause for fewer SEVs being detected. Future work will need to better characterize the NAD(P)H content of SEVs. 147 Size and concentration were determined with NTA and concentration is calculated to correspond to EVs/mL of cell culture fluid in T175 cell culture flasks with 30 mL of media per flask.

Supplementary Note 5: EV characterization and validation
For validation, EV samples were imaged with TEM (Phillips CM200, FEI Company, Thermo Fisher Scientific, Hillsboro, OR) to examine their morphology. The initial EV sample after isolation was diluted further in PBS in a 1:99 ratio for TEM. Carbon-coated copper grids were placed in the EV and PBS solution for 3 minutes, then gently dabbed to dry on sterile filter paper. Next, the grids were negatively stained with 2% uranyl acetate for 30 seconds, after which the excess uranyl acetate solution was removed. Grids were then left to dry for 10-15 minutes prior to imaging. Transmission electron microscopy images (Supplementary Fig. S5a) showed the expected cup-like morphology and lipid-bilayer membrane enclosed spheres 16,17 . As described in the main text, NTA was also used to characterize samples. NTA size distribution consistently showed particles ranging from 50 to 600 nm with most of the particles concentrated in the 100 to 300 nm range; an example histogram of estimated particle sizes is shown in Supplementary Fig. S5b.