Targeted fluorescence lifetime probes reveal responsive organelle viscosity and membrane fluidity

Article abstract of DOI:10.1371/journal.pone.0211165

The only way to visually observe cellular viscosity, which can greatly influence biological reactions and has been linked to several human diseases, is through viscosity imaging. Imaging cellular viscosity has allowed the mapping of viscosity in cells, an

Full text of DOI:10.1371/journal.pone.0211165

RESEARCH ARTICLE
Targeted fluorescence lifetime probes reveal
responsive organelle viscosity and membrane
fluidity
Ida Emilie Steinmark 1*, Arjuna L. James1, Pei-Hua Chung1, Penny E. Morton2,
ID
2
3
1
Maddy Parsons , Cecile A. Dreiss , Christian D. Lorenz , Gokhan Yahioglu4¤*,
´
Klaus Suhling1*
1 Department of Physics, King’s College London, London, United Kingdom, 2 Randall Centre for Cell and
Molecular Biophysics, King’s College London, London, United Kingdom, 3 Institute of Pharmaceutical
Science, King’s College London, London, United Kingdom, 4 Department of Chemistry, Imperial College
London, London, United Kingdom
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
¤ Current address: Antikor Biopharma, Stevenage, United Kingdom
* ida.steinmark@kcl.ac.uk (IES); g.yahioglu@antikor.co.uk (GY); klaus.suhling@kcl.ac.uk (KS)
Abstract
OPEN ACCESS
The only way to visually observe cellular viscosity, which can greatly influence biological
reactions and has been linked to several human diseases, is through viscosity imaging.
Imaging cellular viscosity has allowed the mapping of viscosity in cells, and the next frontier
is targeted viscosity imaging of organelles and their microenvironments. Here we present a
fluorescent molecular rotor/FLIM framework to image both organellar viscosity and mem-
brane fluidity, using a combination of chemical targeting and organelle extraction. For dem-
onstration, we image matrix viscosity and membrane fluidity of mitochondria, which have
been linked to human diseases, including Alzheimer’s Disease and Leigh’s syndrome. We
find that both are highly dynamic and responsive to small environmental and physiological
changes, even under non-pathological conditions. This shows that neither viscosity nor flu-
idity can be assumed to be fixed and underlines the need for single-cell, and now even sin-
gle-organelle, imaging.
Citation: Steinmark IE, James AL, Chung P-H,
Morton PE, Parsons M, Dreiss CA, et al. (2019)
Targeted fluorescence lifetime probes reveal
responsive organelle viscosity and membrane
fluidity. PLoS ONE 14(2): e0211165. https://doi.
org/10.1371/journal.pone.0211165
Editor: Banafshe Larijani, University of Bath,
UNITED KINGDOM
Received: September 18, 2018
Accepted: January 8, 2019
Published: February 14, 2019
Copyright: © 2019 Steinmark et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Introduction
Changes in cellular viscosity have been found in a range of human diseases, including cancer
[1], Alzheimer’s disease [2], Huntington’s disease [3], Amyotrophic lateral sclerosis [3], and
Leigh’s syndrome [4], as well as during disease treatment [5]. On a macroscopic level, viscosity
is defined as the resistance to flow, but on the microscopic level—where the probe is compara-
ble in size to the solvent molecules—it can be described, after some simplification, as the extent
of free volume between solvent molecules and around the probe [6].
Data Availability Statement: All imaging and value
files are available from the King’s College London
Research Data Manager database (https://dx.doi.
org/10.18742/RDM01-403).
Funding: This research was funded by the King’s
College London through a LIDo PhD studentship
for author IES, and by EPSRC (EP/L000202 and
EP/P020194/1) to author CDL. The funders had no
role in study design, data collection and analysis,
decision to publish, or preparation of the
manuscript.
Only recently have the potentially extensive implications of altered viscosity and fluidity for
biological reactions on the molecular level been recognised [7]. Some biological reactions are
diffusion-limited, meaning that diffusion is rate-limiting, but practically all have a diffusion-
mediated step [8]. If the viscosity of the medium through which a chemical species must
PLOS ONE | https://doi.org/10.1371/journal.pone.0211165 February 14, 2019
1 / 20

Fluorescence lifetime probes reveal organelle viscosity
Competing interests: The authors have declared
diffuse is greatly altered, this will have a significant effect upon reaction rate [9]. Indeed it is
possible that, should the viscosity increase markedly, some reactions that were not previously
diffusion-limited will become so. It is even conceivable that some reactions, as a result of a
high viscosity and reduced diffusion, will be overtaken by other, competing reactions and
hence not even take place. This could have a cascading effect with consequences for the whole
cell, or potentially the whole organism. Additionally, some organisms such as bacteria are
known to adjust their membrane viscosity responsively [10]. Therefore, the ability to accu-
rately monitor and visually follow changes in viscosity and fluidity is greatly beneficial when
trying to understand the kinetics and implications of biological reactions.
that no competing interests exist.
Since mechanical viscometers cannot probe cellular viscosity, fluorescence-based methods
have been developed, most notably fluorescence anisotropy [11]. This method requires polari-
sation-resolved data and specialist knowledge, and anisotropy imaging lacks commercial anal-
ysis software. The combination of fluorescent molecular rotors (FMRs) and fluorescence
lifetime imaging (FLIM) is a recent alternative, providing easy and reliable imaging with vis-
cosity as the contrast [7]. Using fluorescence lifetime instead of intensity is beneficial because
intensity is concentration-dependent [12] but lifetime is not. This is particularly useful in bio-
logical samples where concentration homogeneity cannot be guaranteed. We believe that
FMRs are a versatile solution and that they offer the potential for extending lines of inquiry
within studies of disease and drug delivery.
FMRs are dyes that in the excited state can rotate around a single bond. This rotation gives
access to a non-radiative deactivation pathway, such that in low viscosity environments the
FMR will be very weakly fluorescent with a short lifetime (see Fig 1A). Conversely, in high vis-
cosity environments, rotation will be restricted, thereby increasing the fluorescence and pro-
longing the lifetime [13],[7].
Cells contain many microenvironments that are often determinants of disease [14], so tools
to investigate both organellar viscosity and membrane fluidity are sorely needed. This requires
specific targeting of organelles, ideally with sub-organellar specificity for lumena, matrices and
membranes.
One approach uses covalent bonding of an FMR to a protein [15]. This provides very versa-
tile targeting and means only one rotor (with a suitable ligand) must be synthesised. However,
the binding to such a comparably large structure lowers the viscosity sensitivity compared to a
standard FMR, and such a framework cannot be used for membrane fluidity probing, as the
FMR would not be able to embed into the membrane structure.
Instead we use a combination of chemical targeting and organelle extraction. Targeting is
achieved with a small chemical motif that will not interfere with sensitivity and ensures diffu-
sion through the targeted region. A range of targeting motifs already exist, such as triphenyl
phosphonium (TPP+) for mitochondria [16], morpholine for lysosomes [17] and benzyl boro-
nate [18] for the nucleus. Due to the high number of membranes in cells, specific probing of
organelle membranes is challenging but can be achieved with extraction for most organelles
(see Fig 1B) [19],[20].
Here, we focus on mitochondria. Mitochondrial matrix diffusion has been studied with
steady-state anisotropy [21], fluorescence correlation spectroscopy (FCS) [4] and fluorescence
recovery after photobleaching (FRAP) [22], and has been linked to complex I deficiency and
Leigh’s syndrome. Furthermore, a few matrix-targeting lifetime FMRs have been reported
[16],[23],[24]. Changes in mitochondrial membrane fluidity have been studied with steady-
state anisotropy and linked to a range of neurodegenerative diseases and ageing [2],[3]. How-
ever, the mitochondrial membrane fluidity has never to our knowledge been imaged on a sin-
gle organelle level, and potential dynamic and/or responsive viscosity and fluidity in non-
diseased organelles is an active area of research [10]. Studying the viscosity response in healthy
PLOS ONE | https://doi.org/10.1371/journal.pone.0211165 February 14, 2019
2 / 20

Fluorescence lifetime probes reveal organelle viscosity
Fig 1. A) Schematic showing the concept behind FMRs, adapted from Levitt et al [13]. Normal absorption and
fluorescence occurs when rotation is restricted, but a non-radiative pathway can be accessed when rotation is free. This
influences the fluorescence lifetime. Note that while most FMRs are planar in the ground state, BODIPY rotors such as
FMR-1 and FMR-2 are not [7]. B) Diagram of confocal FLIM microscope set-up using the scanning to get position
and timing information. PMT: photomultiplier tube, TCSPC: time-correlated single photon counting. C) Diagram of
the step involved in mitochondrial extraction.
https://doi.org/10.1371/journal.pone.0211165.g001
cells is crucial as we cannot make sound inferences about pathology without understanding
the non-pathological baseline. Further, several vital antioxidant enzymatic systems in mito-
chondria are diffusion-limited [8],[25]. Therefore a change in viscosity could potentially have
a big impact on their function [10]. This adds to the importance of studying the viscosity/diffu-
sion aspects of mitochondria.
We introduce a new, chemically-targeted FMR, FMR-1 (see Fig 2), to image matrix viscos-
ity and to investigate the latent effect of varying, non-pathological Ca2+ exposure. Next, we
extract mitochondria from HeLa cells and image membrane fluidity using a well-known FMR,
BODIPY-C12 or FMR-2, this time exploring the effect of different nutrient conditions during
cell growth. The approach outlined in this paper provides a complete framework for imaging
both organellar viscosity and fluidity, with applications in both clinical and biophysical
studies.
Materials and methods
Chemical synthesis and characterisation
Synthesis and characterisation of FMR-2 were described previously [26].
PLOS ONE | https://doi.org/10.1371/journal.pone.0211165 February 14, 2019
3 / 20

Fluorescence lifetime probes reveal organelle viscosity
Fig 2. A) Chemical structure of the two FMRs used. FMR-1 is the mitochondria-targeting BODIPY-based FMR and a
TPP+ moiety, in the main text referred to as FMR-1. FMR-2 is a BODIPY-based FMR with a C12 carbon chain [26],
used to image the mitochondrial membrane fluidity, in the main text referred to as FMR-2. B) Normalised emission
and absorption spectra of FMR-1 in 80% glycerol and 20% methanol. Note that, while emission increases with
viscosity, the shape of the spectrum remains unaltered. Absorption remains the same. C) Synthetic scheme for the
synthesis of FMR-1. The synthesis of FMR-2 can be found in previous publications [26].
https://doi.org/10.1371/journal.pone.0211165.g002
For the synthesis of FMR-1, please refer to Fig 2. All chemicals including anhydrous sol-
vents were used as received from commercial suppliers. Analytical thin layer chromatography
(TLC) was carried out on glass backed silica gel 60 GF254 (Merck©) plates and flash column
chromatography was either performed on silica gel 60 (Merck©) or via automated chromatog-
raphy (Biotage Isolera) using SNAP KP-Sil columns with the indicated solvent system.
Nuclear magnetic resonance (NMR) spectra were recorded on 500 MHz spectrometers
at ambient probe temperature with predeuterated solvents as internal standards. 19F NMR
spectra are reported in parts per million referenced to hexafluorobenzene. Mass spectra
were carried out using ElectroSpray Ionisation (ESI) and only molecular ions are reported.
4-(4-bromobutyloxy)benzaldehyde 2 was synthesised according to a previous publication
[27] but using anhydrous DMF as solvent.
Synthesis of FMR-1. Synthesis of BODIPY 3 follows a previous description [28]. The
benzaldehyde 2 (1 g, 3.89 mmol) was dissolved in freshly distilled pyrrole (13.5 ml, 195 mmol)
and the solution degassed by bubbling N2 for approximately 20 minutes before the addition of
TFA. The mixture was stirred for 45 minutes at room temperature, diluted with DCM (100
ml) and washed consecutively with water (100 ml), NaHCO3 (100 ml, 0.5 M) and water (100
ml). The DCM layer was separated, dried over MgSO4, filtered and evaporated to remove the
DCM, excess pyrrole was removed by high vacuum to give the dipyrromethane as a green vis-
cous oil. TLC [Silica gel: DCM/Hexane 2:1] visualisation using Bromine vapour. The crude
was purified by Biotage automated chromatography [DCM/Hexane 2:1] giving the pure dipyr-
romethane as a viscous green-yellow oil 0.75 g (51.7%). The dipyrromethane (0.75 g, 2 mmol)
PLOS ONE | https://doi.org/10.1371/journal.pone.0211165 February 14, 2019
4 / 20

Fluorescence lifetime probes reveal organelle viscosity
was dissolved in DCM (60 ml) and DDQ (0.46 g, 2 mmol) was added and the reaction stirred
under N2, shielded from light for 45 minutes. Then triethylamine (0.87 ml, 5.88 mmol) was
added followed immediately by the addition of BF3. (OEt2)2 (0.62 ml, 5.02 mmol) and the reac-
tion was stirred at room temperature overnight. The reaction mixture was washed consecu-
tively with water (100 ml), NH4Cl (100 ml, 0.5 M), NaHCO3 (100 ml, 0.5M) and finally with
water (100 ml). The organic layer was dried over MgSO4, filtered and evaporated to give a dark
viscous oil which was purified by Biotage automated chromatography [DCM/EtOAc] to give
BODIPY 3 as an orange-green solid 0.13 g (15.5%) BODIPY 3 (0.1 g, 0.24 mmol) and triphe-
nyphosphine TPP (0.19 g, 0.72 mmol) were reacted as a solid melt at 90˚C, after 8 hrs TLC
[Silica gel: DCM/Hexane 8:2] still showed the presence of BODIPY 3 as well as an orange fluo-
rescent base-line spot which moved as a single spot on TLC [Silica gel: DCM/Acetone/MeOH
10:2:1]. A further portion of TPP (0.19 g, 0.72 mmol) was added and heating continued for a
further 8 hrs after which the reaction was complete by TLC. The reaction was cooled to room
temperature, resulting in a red-brown solid that was triturated with diethyl ether, filtered and
dried to give a red-brown powder. This was purified by flash chromatography [Silica gel:
DCM/Acetone/MeOH 10:2:1], the crude was loaded on top of the column as a concentrated
solution in DCM, eluted with 5 column volumes of DCM resulting in the separate elution of
both TPP and unreacted BODIPY 3. FMR-1 was then eluted with DCM/Acetone/MeOH,
coming off the column as a bright orange streaky band. Fractions were pooled and evaporated
to give pure FMR-1 as a bright orange solid 0.13 g (80%). 1H NMR (500 MHz, CDCl3) δH
7.89-7.84 (m, 8H), 7.8-7.76 (m, 3H), 7.7-7.66 (m, 6H), 7.5 (d, J = 8.8 Hz, 2H), 7.01 (d, J = 8.8
Hz, 2H), 6.95 (d, J = 4.3 Hz, 2H), 6.53 (dd, J = 4.3, 1.9 Hz, 2H), 4.25 (t, J = 5.8 Hz, 2H), 4.06-
3.95 (m, 2H), 2.28 (t, J = 6.4Hz, 2H), 1.9 (q, J = 7.8Hz, 2H); 13C NMR (500 MHz, CDCl3) δC
161.34, 147.39, 143.29, 134.99, 134.75, 133.74, 132.45, 131.34, 130.49, 126.24, 118.66, 118.24,
117.98, 114.64; 19F NMR (500 MHz, CDCl3) δF -145.11; MS (ESI) m/z found 601.2384
[M-Br]+ calculated for C37H3BF2N2OP, 601.2392.
Calibration
Methanol/glycerol fractions ranging from 0 to 90% glycerol were used for the calibration, con-
taining 13 μM of FMR-1. Absorption spectra were collected on a Hitachi U4100 Emission
Spectrometer, and emission and excitation spectra were collected using a Horiba FluoroMax-4
Spectrofluorometer. For all spectral measurements, Thorlabs 3500 μL quartz cuvettes with
four clear sides containing around 3mL of sample solution was used.
For quantum yield measurements, both PM546 (* 0.4 μM, quantum yield 0.95) and Alexa
Fluor488 (* 0.4 μM, quantum yield 0.92) were used as reference dyes. Methanol was used as
baseline for PM546 and FMR-1 (refractive index 1.3292), water for Alexa Fluor488 (refractive
index 1.333). The quantum yield was calculated using
2
Absref
I
n
QY ¼ QYref
2
nref Abs Iref
Where n is the refractive index of the solvent, I is the integrated fluorescence intensity, and
Abs is the absorbance at the excitation wavelength (kept between 0.02-0.05 to ensure linear
response) [29].
Fluorescence lifetime measurements were conducted on an inverted TCS-SP2 microscope
(Leica Microsystems, Germany) and SPC-150 TCSPC boards (Becker & Hickl, Germany). 200
μL of sample solution was pipetted into a well in an uncoated 8-well plate from Ibidi (Inte-
grated BioDiagnostics). The instrument response function (IRF) was made from NaI-
PLOS ONE | https://doi.org/10.1371/journal.pone.0211165 February 14, 2019
5 / 20

Fluorescence lifetime probes reveal organelle viscosity
Fig 3. A) Representative single decays from FMR-1 in methanol/glycerol fractions for calibration. B) Intensity
correlation along the cross section of the merged image on the far right of Fig 3C. C) Confocal image of HeLa cell
stained with both FMR-1 and MitoTracker Deep Red, excited at 488 nm and 642 nm, respectively. On right, merged
image with position of cross-section in Fig 3C indicated. Intensity correlation between red and green channel across
the whole of the merged image on far right. Manders colocalisation coefficients M1 (green in red channel) and M2 (red
in green channel) are 0.92 and 0.96, respectively.
https://doi.org/10.1371/journal.pone.0211165.g003
quenched fluorescein/methanol solution. A pulsed 467 nm diode laser (Hamamatsu) with a 20
MHz repetition rate and 90 ps optical pulse width was used to excite the samples (see Fig 3).
Lifetime measurements controlling for effects from polarity (dielectic constant), ionic
strength and temperature were carried out in a similar fashion. For polarity measurements,
solutions of 70% glycerol and either 30% methanol, ethanol, butanol or isopropyl alcohol were
used. For ionic strength measurements, increasing concentrations of NaCl (from 0 to 0.26
mol) in 70% glycerol, 30% methanol were used. For temperature controls, 100% glycerol was
heated and measured at 22˚C, 30˚C and 37˚C using a microscope stage incubator.
Bulk viscosity measurements for the control experiments were carried out on a strain-con-
trolled rheometer ARES (TA Instruments), equipped with a 25 mm parallel plate geometry
and a temperature-controlling Peltier unit (C-PTD 200), with a gap between 0.6 and 0.9 mm.
As an exception, the viscosity values for the 100% glycerol sample at different temperatures
were calculated using an open-science GUI based on parameterisation from literature [30]
[31].
Cell work
HeLa cells were cultured in a 12.5 mL petri dish (Thermo Fisher Scientific) in DMEM with 10%
FBS, 1% 1X non-essential amino acid, 1 mM sodium-pyruvate and 0.1% penicillin/streptomy-
cin at 37˚C in a Hela Cell 150 incubator from Thermo Electron Corporation with 5% CO2. A
few days before imaging, a sterile and coated 8-well plate (Ibidi) was prepared. Cells were har-
vested using 3 mL 1X Trypsin/EDTA and 7 ml complete DMEM. On the day of imaging, cells
were washed three times with fluorescence imaging medium (FluoroBrite DMEM, Thermo
Fisher Scientific), before staining with FMR-1 (well concentration 1 μM). Cells were incubated
PLOS ONE | https://doi.org/10.1371/journal.pone.0211165 February 14, 2019
6 / 20

Fluorescence lifetime probes reveal organelle viscosity
for 30 min. A similar protocol was followed for HCE cells, generously gifted by Struan Bourke
from King’s College London and Min S Chang from Vanderbilt University [32].
Extraction of mitochondria
Mitochondrial extraction was done according to established protocol [19]. This mainly
includes stages of homogenization and centrifuging at ice-cold temperatures using an iso-
osmotic homogenization buffer and a hypotonic buffer. The protocol uses cell pellets. In this
case, cells were pelleted by washing in PBS-CMF twice, harvesting them using 2 mL 1X Tryp-
sin/EDTA and 8 mL of DMEM/F12 before transfer to a conical tube. Cells were then spun
at 1400 rpm at 4˚C for 5 minutes before the supernatant was decanted. A second round of
centrifuging and decanting followed re-suspension in PBS-CMF. Extracted mitochondria
were imaged by staining them with FMR-2 (including 30 min incubation) before suspending
them in an agarose-gel scaffold. This was achieved by adding 0.3% agarose to a hypotonic
buffer, bringing the solution to the boil to dissolve the agarose and while still hot to the touch,
adding the extracted mitochondria such that the gel set around them. This allowed them to be
kept in place during imaging. It should be noted that the mitochondrial fraction may contain
some other membrane domains, but only structures of the right shape and size were analysed.
Mitochondrial perturbation
Two methods were used to perturb the mitochondrial matrix viscosity. Firstly, HeLa cells were
treated with an ionophore with no uncoupling activity, nystatin. Nystatin was dissolved in
methanol at 28˚C before being diluted into FluoroBrite DMEM (ThermoFisher Scientific) for
cell treatment. Cells were incubated with a 10 μM well concentration for 30 min prior to stain-
ing with FMR-1. Secondly, HeLa cells were treated with histamine which induces Ca2+ release
from the endoplasmic reticulum to the mitochondria. Histamine was dissolved in de-ionised
water and diluted in FluoroBrite DMEM for cell treatment. Prior to staining, cells were incu-
bated for 30 min.
Co-localisation
HeLa cells were pre-stained with MitoTracker Deep Red (Thermo Fischer Scientific) with a
well concentration of 0.5 μM. The cells were incubated for 30 min. The colocalisation imaging
was done on a Nikon inverted confocal scanning microscope using a 488 nm and a 642 nm
diode laser (carried out at the Nikon Imaging Centre at King’s College London), see Fig 3C.
FLIM experiments
Cells were imaged in 8-well plates on a inverted TCS-SP2 microscope (Leica Microsystems,
Germany) and SPC-150 TCSPC boards (Becker & Hickl, Germany). FMR-1 and FMR-2 were
excited using a picosecond-pulsed (90 ps optical pulse width, 20 MHz repetition rate) 467 nm
diode laser (Hamamatsu, Japan) through a 63X 1.2 NA water objective, with a 485 nm dichroic
mirror and a 500 long pass filter. All imaging was done at room temperature. Each FLIM
image was acquired over 900 s to ensure as little noise as possible (much shorter acquisition
rates are possible). Lifetime histograms and FLIM images were analysed using SPCImage soft-
ware (Becker & Hickl, Germany).
Statistics
Welch-corrected two-tailed two-sample t-tests were used to estimate statistical significance.
For the control versus nystatin experiment, the t-value was -9.3 and had 46.8 degrees of
PLOS ONE | https://doi.org/10.1371/journal.pone.0211165 February 14, 2019
7 / 20

Fluorescence lifetime probes reveal organelle viscosity
freedom. For the low versus high glucose extracted mitochondria experiment, the t-value was
-3.2 and had 40.6 degrees of freedom.
Molecular dynamics simulations
As we have done previously for the environmentally sensitive fluorescent dye, PRODAN [33],
we have used all-atom classical molecular dynamics (MD) simulations to investigate the effect
that the FMR-2 dye has on the molecular scale properties of a model lipid bilayer. Therefore,
two systems were simulated: (a) a DOPC bilayer consisting of 288 lipid molecules (144 mole-
cules in each leaflet) and (b) the same bilayer interacting with a single FMR-2 molecule. The
lipid bilayer was built using the CHARMM-GUI membrane builder [34–36]. The membrane
was first minimised using a steepest descent algorithm in order to remove any steric clashes
that might exist in the structure obtained from the membrane builder. Then a series of equili-
bration simulations were carried out in the NVT (constant Number, Volume and Tempera-
ture) ensemble to equilibrate the temperature at 303.15 K and in the NPT (constant Number,
Pressure and Temperature) ensemble to equilibrate the volume of our simulation system at a
temperature of 303.15 K and a pressure of 1 atm. Finally, the DOPC bilayer system was simu-
lated for a further 100 ns in the NPT ensemble at a temperature of 303.15 K and a pressure of 1
atm.
Then we inserted a single FMR-2 molecule into the aqueous phase of the final configuration
of this 100 ns simulation of the DOPC bilayer, such that it was at least 1.0 nm away from the
interface of the bilayer. Then this system was run for 200 ns using the NPT ensemble at a tem-
perature of 303.15 K and a pressure of 1 atm. The FMR-2 molecule inserted into the bilayer
within the first 100 ns of the simulation. Then the final 100 ns of this simulation were used for
the analysis that was carried out. Likewise, the DOPC bilayer system (without the FMR-2 mol-
ecule) was simulated for a further 100 ns, which was used for all analysis of it.
All of the simulations were carried out using the GROMACS molecular dynamics package
[37, 38]. The interactions of the lipid molecules are described by the CHARMM36 lipid force-
field [39]. The interactions of the FMR-2 molecule were modeled using the parameters for
the BODIPY head group that had been previously used [40], and then the rest of the molecule
was parameterised using the CHARMM General Force Field [41]. The water molecules and
their interactions were modeled using the CHARMM version of the TIP3P model [42]. In all
cases, the nonbonded interactions were cut-off at 1.0 nm, and the PME algorithm was used to
account for long range electrostatic interactions. The LINCS algorithm was used to constrain
all hydrogen-containing bonds, and a 2 femtosecond timestep was used during the production
´
simulations. In all simulations, we used the Nose-Hoover thermostat to control the tempera-
ture and the Parrinello-Rahman barostat to control the pressure.
In order to investigate the dynamics of the lipid molecules within each bilayer, we calcu-
lated the mean square displacement (MSD) in the x-y plane of the phosphorous atoms in the
phosphatidylcholine headgroup of the DOPC lipid molecules. The mean square displacement
is calculated as follows:
1
2
2
MSD ¼ SNn¼1ððxnðtÞ À xnðtref ÞÞ þ ðynðtÞ À ynðtref ÞÞ Þ
ð1Þ
N
where N is the number of atoms in your system, xn(t) and yn(t) are the x and y coordinates of
atom n at time t, and xn(tref) and yn(tref) are the x and y coordinates of atom n at a reference
time tref.
The structure of the lipid membrane is characterised by the calculating the carbon-hydro-
gen lipid order parameter for both the sn-1 and the sn-2 chains of each lipid molecule. The
PLOS ONE | https://doi.org/10.1371/journal.pone.0211165 February 14, 2019
8 / 20

Fluorescence lifetime probes reveal organelle viscosity
lipid order parameter is calculated by:
1
2
2
ð2Þ
SCD
¼
ð3 cos b À 1Þ
where β is the angle that each C-H bond makes with the vector normal to the bilayer. The
larger in magnitude that SCD is, the more ordered the lipid molecules are. We have calculated
this quantity for each molecule. Then each lipid was placed into one of five different distance
(in x-y plane) ranges (d < 0.7 nm, 0.7 nm < d < 1.0 nm, 1.0 nm < d < 1.6 nm, 1.6 nm <
d < 2.0 nm, d > 2.0 nm), where the smallest range represents the distance of first neighbours
as determined from the radial distribution function of phosphorous atoms in the PC head-
group around the boron atom in the FMR-2 molecule (not shown). We then average this
quantity for each carbon in each tail over time and over each molecule within a given distance
range.
Finally, we measured the hydration of the phosphatidylcholine group as an indication of
the interfacial properties of the lipid membrane. In order to do so, we determined the radial
distribution functions of the oxygen atoms in water molecules around the nitrogen atom in
the onium group of the PC headgroup, the phosphorous atom in the phosphate group of the
PC headgroup and the double bonded oxygens in the ester groups of the PC headgroup (not
shown). From these radial distribution functions, we identified the distance at which first
neighbour water molecules are found. This distance was used to then count the number of
water molecules around each of these three groups of each lipid headgroup. The lipids were
once again placed into one of the previously stated five different distance ranges. Then we
averaged the number of water molecules found around each group of the lipid headgroup as a
function of time and the number of lipids within each distance range.
Results
Validation and co-localisation of FMR-1
FMR-1 is based on a simple BODIPY design, incorporating a mitochondria-targeting moiety,
triphenylphosphonium (TPP+). For FMR-1, the fluorescence lifetime increases from just over
200 ps in 100% methanol (viscosity 0.6 cP at 22˚C, quantum yield 2%) to 4.6 ns in 100% glyc-
erol (viscosity 1178 cP at 22˚C, quantum yield 55%). Low and high viscosity regions were fitted
with straight lines according to the Fo¨rster-Hoffman equation, yielding gradients of 0.25 and
0.63 respectively. This two-slope behaviour is common for FMRs [7]. For a more in-depth
explanation of the relationship between molecular rotors and viscosity, see the following refer-
ences [7][6]. For our purposes, the theoretical relationship is less important; this is simply an
empirical, observational relationship between fluorescence lifetime and viscosity. We addition-
ally tested for temperature, polarity and ionic strength dependence in the relevant region of
the calibration plot and found that, as expected, that FMR-1 is only sensitive to viscosity. This
is in line with previous publications on BODIPY FMRs [43][44].
It was co-localised with MitoTracker Deep Red in HeLa cells, yielding an average Pearson’s
r of 0.92, as well as average Manders Colocalisation Coefficients (MCC) M1 0.92 (green in red
channel) and M2 0.96 (red in green channel)—see Fig 3C. The average was taken over 16 co-
stained cells. As is evident from the MCC, FMR-1 stains slightly more regions than Mito-
Tracker Deep Red, and judging by the images, this weak additional staining is in the plasma
membrane which is not uncommon for this type of FMR [23].
FMR-FLIM of the HeLa cells showed that the viscosity of the mitochondrial matrix was
around 275 cP at room temperature as calculated from the average lifetime (see Fig 4A). It is
higher than that found using another BODIPY-based FMR (ca. 67.5 cP) [16]. This can be
PLOS ONE | https://doi.org/10.1371/journal.pone.0211165 February 14, 2019
9 / 20

Fluorescence lifetime probes reveal organelle viscosity
Fig 4. A) Representative fluorescence intensity and FLIM images of control and nystatin-treated HeLa cells, lifetime
range 2100-3500 ps. B) Representative average fluorescence lifetime histograms from images A-B, control in green and
nystatin-treated in red. C) Box plot showing the average lifetimes of cells in the control condition (n = 140, 1μM FMR-
1) and nystatin-treated condition (n = 33, 1μM FMR-1). The average control viscosity was found to be 274.7 54.4 cP,
and the average nystatin-treated viscosity was found to be 367 78.1 cP. The difference is statistically significant
(p < 0.01.). D) Representative fluorescence intensity decays from a control cell (in green) and a cell treated with
nystatin (in red), and a scaled instrument response function (IRF, in black). E) Calibration plot for FMR-1 based on
lifetimes in methanol-glycerol fractions. F) Control data for polarity, ionic strength and temperature plotted on top of
the calibration data, in the relevant region of the plot.
https://doi.org/10.1371/journal.pone.0211165.g004
explained by a difference in temperature and the fact that Yang et al. [16] fit their entire cali-
bration plot with one straight line, with a deviation from the data in the lower range—in which
their viscosity value falls.
To validate the FMR, we compared the results to cells which had been pre-treated with nys-
tatin (see Fig 4A). Nystatin is an ionophore with no uncoupling activity [45]. Ionophores are
known to push mitochondria from the orthodox to the condensed ultrastructural state, which
is dominated by high protein content and small volume, as evidenced by electron microscopy,
and high matrix viscosity, as found by fluorescence anisotropy [21]. As such, we hypothesised
that we would see an increase in matrix viscosity. After nystatin treatment, the matrix viscosity
increased from 274.7 54.4 cP to 367 78.1 cP, as calculated from average lifetimes, which
were 2473 119 ps and 2967 185 ps respectively. This confirms FMR-1’s ability to measure
changes in matrix viscosity.
PLOS ONE | https://doi.org/10.1371/journal.pone.0211165 February 14, 2019
10 / 20

Fluorescence lifetime probes reveal organelle viscosity
Fig 5. Representative fluorescence intensity and FLIM image of control HeLa cells (A) and of histamine-treated HeLa
cells (C), along with representative decays, showing fit and scaled IRF. Lifetime range for images is 2100-3500 ps. 1μM
FMR-1 in both conditions. B) Representative average fluorescence lifetime histograms from cells in each condition. D)
Box plot showing the average lifetimes of cells in the control (n = 140) and the histamine condition (n = 83). Each data
point is a cell.
https://doi.org/10.1371/journal.pone.0211165.g005
Matrix viscosity after Ca2+ exposure
Next we tested the effect of Ca2+ exposure on matrix viscosity. Ca2+ uptake into mitochondria
is a normal, physiological process, but is still not fully understood [46]. As such, we wanted to
explore any viscosity effects of non-pathological Ca2+ exposure. Histamine induces a short Ca2
+ release from the endoplasmic reticulum to the mitochondria in HeLa cells at low concentra-
tions (< 100μM), which lasts only a few minutes. We aimed to investigate subsequent viscosity
effects as a result of the Ca2+, hence avoiding the Ca2+ influx itself.
We pre-treated HeLa cells with histamine prior to FMR staining in order to observe genu-
ine viscosity changes, and not simply the influx of Ca2+. FMR-FLIM showed that after treat-
ment with 30μM histamine, the viscosity varied greatly and showed differences between
individual cells (see Fig 5). At lower concentrations, histamine appears to have little to no
effect on viscosity, but the variation is sustained at higher concentrations (see S2 Fig). This
could possibly be due to cells responding differently depending on pre-existing conditions,
such as stage in cell cycle. All in all, it shows the dynamic nature of the matrix viscosity, and
that it is both responsive to small physiological changes and highly variable between cells. We
conclude that matrix viscosity cannot be assumed to be fixed and underline the need for sin-
gle-cell viscosity imaging.
Effect of glucose on membrane fluidity
Previous studies of mitochondrial membrane fluidity, which have utilised steady-state anisot-
ropy, have found that the membrane is ridigified in neurodegenerative diseases. This has been
attributed to oxidative damage, leading to lipid peroxidation which in turn results in increased
membrane rigidification [2],[3].
To explore membrane fluidity during a non-pathological process, we assessed the effect of
nutrition, specifically glucose. Excess glucose leads to increased energy production which can
result in higher reactive oxygen species (ROS) production [47], but conversely, starvation can
PLOS ONE | https://doi.org/10.1371/journal.pone.0211165 February 14, 2019
11 / 20

Fluorescence lifetime probes reveal organelle viscosity
Fig 6. A) Representative fluorescence intensity and FLIM images of extracted mitochondria, originally grown as HeLa
cells in high (4.5g/L) and low (1g/L) glucose DMEM respectively. Average lifetime range is 600-1000 ps. B) Histograms
of the third lifetime component for the images in Fig 5A, high glucose condition in blue and low glucose condition in
yellow. C) Representative fluorescence intensity decays from extracted mitochondria, high glucose in blue, low glucose
in yellow and scaled IRF in red. D) Calibration plot for FMR-2, re-produced from P.H. Chung’s thesis [50]. E) Box plot
showing the third lifetime components of mitochondria in the two conditions (n = 42 for high glucose and n = 48 for
low glucose). The difference is statistically significant (p < 0.01). On average, the viscosity for the high glucose
condition was found to be 536.8 303 cP (from an average third component lifetime of 2775 468 ps). For the low
glucose condition, it was found to be 869.5 659.8 cP (from average third component lifetime of 3517 877 ps) on
average.
https://doi.org/10.1371/journal.pone.0211165.g006
also lead to increased ROS production [48],[49]. The effect of small perturbations in glucose
on ROS, and hence fluidity, is therefore not obvious. Here, we explored the effect of high or
low glucose conditions.
HeLa cells were grown in either standard, high glucose DMEM (4.5g/L) or low glucose
DMEM (1g/L) prior to extraction. Once extracted, they were then stained with the membrane-
targeting FMR-2, suspended in a 0.3% agarose gel scaffold and imaged (see Fig 6). We found
that the low glucose condition lead to an increase in lifetime, and hence a decrease in mem-
brane fluidity, in accordance with trends observed for more extensive starvation [48],[49].
The change between the two conditions primarily took place in the third lifetime compo-
nent, yielding viscosities of the order expected for membranes [51]: ca. 536 cP (2.7 ns) for
the high glucose condition, and ca. 869 cP (3.5 ns) for the low glucose condition. As all other
variables between the two conditions (day of seeding, growth period, order and duration
of imaging) have been controlled for, this provides convincing evidence that the FMR is
indeed probing the mitochondrial membrane, and further that environmental cell conditions
prior to extraction can affect membrane fluidity. These experiments show that even slight star-
vation can induce membrane rigidification which is consistent with lipid peroxidation due to
increased ROS levels.
Effect of FMR-2 on model lipid bilayer
As FMR-2 is not a flat molecule, we wanted to investigate the effect of the FMR on a mem-
brane; specifically whether it could potentially bias our results by affecting the very property
we are trying to probe. To that end, we used all-atom MD simulations to investigate the effect
that the presence of FMR-2 has on the dynamic, structural and interfacial properties of a
DOPC bilayer. In order to assess the effect on the dynamic properties, we measured the mean
square displacement (msd) of the phosphorous atoms in the phosphatidylcholine headgroup
of the lipid molecules in a DOPC bilayer and a DOPC bilayer in which a FMR-2 has adsorbed.
Fig 7A shows the msd for the two different bilayers, and there is no significant difference in
PLOS ONE | https://doi.org/10.1371/journal.pone.0211165 February 14, 2019
12 / 20

Fluorescence lifetime probes reveal organelle viscosity
Fig 7. A) Mean square displacement of phosphorous atoms in the phosphatidylcholine headgroup of the lipid
molecules in a DOPC bilayer with and without FMR-2. B) Snapshot of MD simulation showing position of FMR-2
within the model lipid bilayer. C) Lipid order parameters for sn-1 in the same leaflet as FMR-2. D) Lipid order
parameters for sn-1 in the opposite leaflet as FMR-2. E) Lipid order parameters for sn-2 in the same leaflet as FMR-2.
F) Lipid order parameters for sn-2 in the opposite leaflet as FMR-2.
https://doi.org/10.1371/journal.pone.0211165.g007
the two curves. Therefore, it appears that there is no effect on the diffusion of the lipids within
the bilayer when the dye is present at this ratio of 1 dye molecule to 288 lipid molecules, as
compared to when it is not present.
In order to assess any effect the dye molecule has on the structure of the lipid bilayer, we
have measured the carbon-hydrogen lipid order parameter for lipid molecules whose phos-
phorous atom is within a given distance of the boron atom in the dye molecule. Fig 7C–7E
shows the lipid order parameters for sn-1 and sn-2 chains in the same and opposite leaflets as
the dye molecule. In doing so, we have shown these lipid order parameters for those lipids
that are first neigbours of the dye molecule and then at increasing distances in the x-y plane
from the dye molecule. In the same leaflet, we find that the sn-1 and sn-2 chains of the first
neighbour lipid molecules (d < 0.7 nm) are slightly more ordered than those chains at the
larger distances, but there is no effect on chains at any larger distances. While in the opposite
leaflet, there is no significant difference in the order parameter as a function of distance from
the dye.
PLOS ONE | https://doi.org/10.1371/journal.pone.0211165 February 14, 2019
13 / 20

Fluorescence lifetime probes reveal organelle viscosity
Table 1. Average number of first neighbour water molecules.
Onium group
Phosphate group
Same
Ester groups
Distance
Same
Opposite
17.8 (3.8)
17.5 (3.8)
17.5 (3.8)
17.4 (3.9)
17.5 (3.9)
Opposite
6.4 (1.4)
6.4 (1.4)
6.4 (1.4)
6.3 (1.4)
6.3 (1.4)
Same
Opposite
2.1 (1.0)
2.1 (1.0)
2.1 (1.0)
2.1 (1.0)
2.1 (1.0)
d <0.7 nm
16.8 (4.0)
17.5 (3.8)
17.6 (3.8)
17.4 (3.8)
17.4 (3.9)
6.1 (1.4)
1.9 (1.0)
2.0 (1.0)
2.0 (1.0)
2.1 (1.0)
2.1 (1.0)
0.7 <d <1.0 nm
1.0 nm <d <1.6 nm
1.6 nm <d <2.0
d <2.0 nm
6.3 (1.4)
6.3 (1.4)
6.3 (1.4)
6.3 (1.4)
Average number of first neighbour water molecules around the onium, phosphate and ester groups of the phosphatidylcholine headgroups of the DOPC lipid molecules
in the same and opposite leaflets as the FMR-2 molecule at different distances in the x-y plane from the dye molecule. The standard deviation of the values are shown in
parentheses.
https://doi.org/10.1371/journal.pone.0211165.t001
Finally, we have determined the average number of first neighbour water molecules around
the onium, phosphate and ester groups of the phosphatidylcholine head groups of lipids at dif-
ferent distances within the x-y plane from the dye molecule (as shown in Table 1). From these
values, we find that the DOPC lipid molecules that are first neighbours of the dye molecule
(d < 0.7 nm) in the same leaflet have their headgroups dehydrated by approximately 1 water
molecule as compared to the headgroups at further distances. Meanwhile, there is no effect of
the hydration of the headgroups of the lipid molecules in the opposite leaflet. Therefore, once
again, at this ratio of dye to lipid molecules, it appears that the dye will have very minor effects
on the interfacial properties of the lipid membranes.
Discussion
In this paper, we explored the effect of increased Ca2+ exposure on matrix viscosity and slight
nutrient deprivation on membrane fluidity. In both cases, viscosity/fluidity exhibited variation,
showing that these parameters are dynamic and not “fixed”, even in healthy cells—organelles
may be constantly responding to environmental and physiological fluctuations. We believe
this warrants further investigation because we cannot describe pathological states without
first understanding the range of viscosity variation in healthy controls. It is also important to
understand that, in cases where the viscosity/rigidity increases, diffusion-limited reactions will
be slowed down. Conversely, higher fluidity would likely cause a rate increase. This could have
potential knock-on effects for the whole organelle, cell and potentially organism. Changes in
viscosity could therefore play a significant role in the mechanism behind certain illnesses; this
again underlines the need to understand the physiological viscosity state. Further, it could be
relevant for drug delivery systems that inherently rely on diffusion. From a methodological
perspective, we have shown that a combination of chemical targeting and organelle extraction
enables easy imaging of organelle viscosity and membrane fluidity. To the best of our knowl-
edge, this is the first fully combined framework for imaging both matrix viscosity and mem-
brane fluidity using FMR-FLIM.
To thoroughly check that FMR-1 is only sensitive to viscosity, we have carried out a series
of control experiments, studying the effects of temperature, polarity and ionic strength in turn.
We find that, while some of these measures may alter the viscosity, they are not otherwise
related—the FMR reports on viscosity and nothing else. This is completely in line with the
findings of others on the topic of BODIPY FMRs [43],[44]. FMR-2 has already been used to
study membrane fluidity and membrane order in binary and ternary lipid mixtures, as well as
membrane cholesterol content, in what has been termed ‘molecular rheometry’ [52],[53]. One
such study found that, using FMR-2, the transition temperature of DPPC could be confirmed,
PLOS ONE | https://doi.org/10.1371/journal.pone.0211165 February 14, 2019
14 / 20

Fluorescence lifetime probes reveal organelle viscosity
and that the reported viscosity values were in the expected range [52]. However, to test
whether any effect of FMR-2 on the membrane it was probing could possibly bias our results,
we have carried out MD simulations to explore a model lipid bilayer with and without FMR-2.
Our findings show that the presence of the dye has no significant impact upon lipid diffusion,
and that only very minor changes in lipid order and headgroup hydration are induced, and
only at very short distances from the dye. We find this convincing evidence that FMR-2 can be
used to study membrane fluidity at low concentrations without significant perturbation of the
system studied.
As mentioned, some other methods allow probing of viscosity and the related concept of
diffusion. One is FCS, which measures the intensity fluctuations of a small concentration of
dye molecules diffusing through a volume to gain diffusion information [54]. However it
requires significant knowledge of the studied system and extremely low dye concentrations
(pico- to nanomolar range) [54]. Another method, FRAP, includes bleaching a region of inter-
est and measuring the repopulation time [55]. In biological samples, the effect of organelle
shape and geometric boundaries can limit “back-diffusion” and hence recovery [56],[57], of
particular importance in mitochondria because of their cristae. FRAP cannot be used to create
“diffusion maps” analogous to the viscosity maps created by FMR-FLIM. Finally, Generalized
Polarization (GP) of suitable polarity-sensitive membrane dyes, which exploits a spectral shift
in emission between gel and liquid lipid phases, can be used to study membrane lipid order
[58][59]. Rather than measuring viscosity, it is based on sensing the penetration of highly
polar water molecules into the membrane. GP images are obtained by imaging in two spectral
regions, and is widely used to specifically study membranes.
The main alternative to FMR-FLIM is anisotropy—specifically Time-Resolved Anisotropy
Imaging Microscopy (TR-FAIM) [11]. TR-FAIM is usually conducted with a rigid dye mole-
cule and the method can indicate hindered rotation. However, anisotropy decays can be very
noisy, and they require high photon counts to fit properly—upwards of 1000 photons in each
pixel. In contrast, counts in the low hundreds can be used for FLIM [60],[61]. As a result,
much longer acquisition times are needed for scanning TCSPC TR-FAIM; this is problematic
for biological samples, dynamic events and/or photobleaching. It also means that FMR-
FLIM has bigger potential for viscosity movies than TR-FAIM. Finally, in contrast to FLIM,
TR-FAIM lacks commercial analysis software (manual analysis is computationally cumber-
some and non-trivial). Therefore, FMR-FLIM represents an easier, quicker alternative to
TR-FAIM, especially for non-specialists.
Clinical studies may in particular benefit from FMR-FLIM. We chose to focus on various
mitochondrial regions in part because the diffusion and fluidity of these have been the sub-
ject of intense clinical research. Altered protein diffusion in the mitochondrial matrix has
been linked to Complex I deficiency [4], and changes in mitochondrial membrane fluidity
have been found in the mitochondria of Alzheimer’s disease patients [2], as well as Hunting-
ton’s, Alzheimer’s and Amyotrophic lateral sclerosis mouse models [3]. FMR-FLIM could
uncover new insights in this area, underlining the wide range of potential applications for
this technique.
Conclusion
Cellular viscosity imaging has so far lacked a framework for imaging both matrix/lumenal vis-
cosity and membrane fluidity of organelles. In mitochondria, we have shown that the combi-
nation of chemical FMR targeting, organelle extraction and FMR-FLIM allow us to image the
viscosity of the major regions within the organelles, namely the matrix and the membranes.
We have shown that these properties are dynamic and responsive to small environmental and
PLOS ONE | https://doi.org/10.1371/journal.pone.0211165 February 14, 2019
15 / 20

Fluorescence lifetime probes reveal organelle viscosity
physiological changes, and further, we have tested the suitability of FMR-FLIM in membranes
through MD simulation. We believe that single-organelle FMR-FLIM has the potential to find
widespread use in biophysical, clinical and drug delivery studies.
Supporting information
S1 Fig. All spectra refer to FMR-1. For similar details on FMR-2, see previous publication
[26]. A) Fluorescence emission spectra with increasing viscosity (10% to 80% glycerol, 1.3 cP
to 380 cP). B) Peak emission intensity plotted against viscosity in a log-log plot. The gradients
of the low and high viscosity regions are 0.25 and 0.64, respectively. C) Quantum yield plotted
against viscosity in a log-log plot. The gradients of the low and high viscosity regions are 0.25
and 0.64, respectively, in excellent agreement with B. Note that the gradients of the fitted
regions for both log-log plots are in excellent agreement with the gradients achieved for the
log-log lifetime/viscosity plot (see main text), confirming that FMR-1 is perfectly in line with
theory D) Steady-state excitation anisotropy, measured at the main emission feature at 525
nm, against wavelength with viscosity increasing from 55% to 80% glycerol. It shows the char-
acteristic dip-and-rise anisotropy of FMRs [13]. The negative feature from roughly 380 to 490
nm is due to S0-S2 excitation, where the absorption transition is roughly orthogonal to the
emission dipole [13]. Past 460 nm the anisotropy increases with viscosity.
(TIF)
S2 Fig. A-D) Representative intensity and FLIM images of HeLa cells treated with an increas-
ing concentration of histamine (10μM, 20μM, 30μM, 40μM respectively). To the left of the
images are representative decays with fit and scaled IRF. E) Boxplot showing control (n = 140)
and the 10μM (n = 57), 20μM (n = 35) histamine conditions respectively. F) Boxplot showing
control (n = 140) and the 40μM (n = 120) condition.
(TIF)
S3 Fig. 1H NMR spectrum of FMR-1 in deuterated chloroform.
(TIF)
S4 Fig. 13C NMR spectrum of FMR-1 in deuterated chloroform.
(TIF)
S5 Fig. 19F NMR spectrum of FMR-1 in deuterated chloroform.
(TIF)
S6 Fig. Mass spectroscopy of FMR-1.
(TIF)
S7 Fig. Closer view of mass spectroscopy of FMR-1. S6 Fig and S7 Fig clearly show the rele-
vant m/z peak at 601.2384 [M-Br], matching the calculated mass for C37H33BF2N2OP,
601.2392.
(TIF)
S8 Fig. A) Intensity and FLIM images of HCE cells at 37˚C and 5% CO2. Staining protocol
was the same as the one described for HeLa cells. B) Representative decay from the FLIM
image with scaled IRF, fit and residuals. C) Normalised frequency histogram for all of the
image in A, clearly showing a Gaussian distribution. To show that FMR-1 performs equally
under more physiologically relevant conditions, we tested it in HCE cells, a human, non-
cancerous epithelial cell line, at physiologically relevant temperature and CO2 levels. S8 Fig
shows that here too, we can collect FLIM images and decays for easy analysis.
(TIF)
PLOS ONE | https://doi.org/10.1371/journal.pone.0211165 February 14, 2019
16 / 20

Fluorescence lifetime probes reveal organelle viscosity
S1 File. Brief description of the underlying theory of FMRs and their calibration.
(PDF)
Acknowledgments
I E Steinmark would like to thank the Nikon Imaging Centre for use of their inverted confocal
microscope; Struan Bourke for sharing his HCE cells, originally gifted to him by Min S Chang
[32]; and Alix Le Marois and Yurema Teijeiro-Gonzalez for their support with cell culture and
software. Via C D Lorenz’s membership to the UK’s HEC Materials Chemistry Consortium,
this work used the ARCHER UK National Supercomputing Service (http://www.archer.ac.uk)
and the UK Materials and Molecular Modelling Hub (MMM Hub) for computational
resources to carry out the MD simulations reported in this manuscript.
Author Contributions
Conceptualization: Ida Emilie Steinmark, Gokhan Yahioglu, Klaus Suhling.
Data curation: Ida Emilie Steinmark, Christian D. Lorenz.
Formal analysis: Ida Emilie Steinmark, Christian D. Lorenz.
Funding acquisition: Klaus Suhling.
´
Investigation: Ida Emilie Steinmark, Arjuna L. James, Pei-Hua Chung, Cecile A. Dreiss,
Christian D. Lorenz, Gokhan Yahioglu.
´
Methodology: Ida Emilie Steinmark, Cecile A. Dreiss, Christian D. Lorenz, Gokhan Yahioglu.
Project administration: Ida Emilie Steinmark.
´
Resources: Penny E. Morton, Maddy Parsons, Cecile A. Dreiss, Christian D. Lorenz, Gokhan
Yahioglu, Klaus Suhling.
´
Supervision: Maddy Parsons, Cecile A. Dreiss, Gokhan Yahioglu, Klaus Suhling.
Validation: Ida Emilie Steinmark.
Visualization: Ida Emilie Steinmark, Christian D. Lorenz.
Writing – original draft: Ida Emilie Steinmark, Christian D. Lorenz.
Writing – review & editing: Pei-Hua Chung, Penny E. Morton, Maddy Parsons, Christian D.
Lorenz, Gokhan Yahioglu, Klaus Suhling.
References
1. Guyer MF, Claus PE. Increased viscosity of cells of induced tumors. Cancer Research. 1942; 2(1):16–
18.
2. Mecocci P, Beal MF, Cecchetti R, Polidori MC, Cherubini A, Chionne F, et al. Mitochondrial membrane
fluidity and oxidative damage to mitochondrial DNA in aged and AD human brain. Molecular and Chemi-
cal Neuropathology. 1997; 31(1):53–64. https://doi.org/10.1007/BF02815160 PMID: 9271005
3. Eckmann J, Eckert SH, Leuner K, Mu¨ller WE, Eckert GP. Mitochondria: mitochondrial membranes in
brain ageing and neurodegeneration. The International Journal of Biochemistry & Cell Biology. 2013; 45
(1):76–80. https://doi.org/10.1016/j.biocel.2012.06.009
4. Koopman WJH, Distelmaier F, Hink MA, Verkaart S, Wijers M, Fransen J, et al. Inherited complex I defi-
ciency is associated with faster protein diffusion in the matrix of moving mitochondria. American Journal
of Physiology-Cell Physiology. 2008; 294(5):C1124–C1132. https://doi.org/10.1152/ajpcell.00079.2008
PMID: 18353897
PLOS ONE | https://doi.org/10.1371/journal.pone.0211165 February 14, 2019
17 / 20

Fluorescence lifetime probes reveal organelle viscosity
5. Kuimova MK, Botchway SW, Parker AW, Balaz M, Collins HA, Anderson HL, et al. Imaging intracellular
viscosity of a single cell during photoinduced cell death. Nature Chemistry. 2009; 1(1):69–73. https://
doi.org/10.1038/nchem.120 PMID: 21378803
6. Valeur B. Molecular Fluorescence: Principles and Applications. Wiley VHC, Weinheim; 2002.
7. Kuimova MK. Mapping viscosity in cells using molecular rotors. Physical Chemistry Chemical Physics.
2012; 14(37):12671–12686. https://doi.org/10.1039/c2cp41674c PMID: 22806312
8. Northrup SH, Boles JO, Reynolds JC. Brownian dynamics of cytochrome c and cytochrome c peroxi-
dase association. Science. 1988; 241(4861):67–70. https://doi.org/10.1126/science.2838904 PMID:
2838904
9. Zheng K, Jensen TP, Savtchenko LP, Levitt JA, Suhling K, Rusakov DA. Nanoscale diffusion in the syn-
aptic cleft and beyond measured with time-resolved fluorescence anisotropy imaging. Scientific
Reports. 2017; 7:42022. https://doi.org/10.1038/srep42022 PMID: 28181535
10. Budin I, de Rond T, Chen Y, Chan LJG, Petzold CJ, Keasling JD. Viscous control of cellular respiration
by membrane lipid composition. Science. 2018; p. eaat7925. https://doi.org/10.1126/science.aat7925
PMID: 30361388
11. Vinegoni C, Feruglio PF, Gryczynski I, Mazitschek R, Weissleder R. Fluorescence anisotropy imaging
in drug discovery. Advanced Drug Delivery Reviews. 2018;(In Press). https://doi.org/10.1016/j.addr.
2018.01.019 PMID: 29410158
12. Domaille DW, Que EL, Chang CJ. Synthetic fluorescent sensors for studying the cell biology of metals.
Nature Chemical Biology. 2008; 4(3):168. https://doi.org/10.1038/nchembio.69 PMID: 18277978
13. Levitt JA, Chung PH, Kuimova MK, Yahioglu G, Wang Y, Qu J, et al. Fluorescence anisotropy of molec-
ular rotors. ChemPhysChem. 2011; 12(3):662–672. https://doi.org/10.1002/cphc.201000782 PMID:
21328515
14. Dean KM, Palmer AE. Advances in fluorescence labeling strategies for dynamic cellular imaging.
Nature Chemical Biology. 2014; 10(7):512. https://doi.org/10.1038/nchembio.1556 PMID: 24937069
´
15. Chambers JE, Kubankova M, Huber RG, Lopez-Duarte I, Avezov E, Bond PJ, et al. An Optical Tech-
nique for Mapping Microviscosity Dynamics in Cellular Organelles. ACS Nano. 2018; 12(5):4398–4407.
https://doi.org/10.1021/acsnano.8b00177 PMID: 29648785
16. Yang Z, et al. A self-calibrating bipartite viscosity sensor for mitochondria. Journal of the American
Chemical Society. 2013; 135(24):9181–9185. https://doi.org/10.1021/ja403851p PMID: 23713894
17. Wang L, Xiao Y, Tian W, Deng L. Activatable rotor for quantifying lysosomal viscosity in living cells.
Journal of the American Chemical Society. 2013; 135(8):2903–2906. https://doi.org/10.1021/ja311688g
PMID: 23409947
18. Tang R, Wang M, Ray M, Jiang Y, Jiang Z, Xu Q, et al. Active Targeting of the Nucleus Using Nonpepti-
dic Boronate Tags. Journal of the American Chemical Society. 2017; 139(25):8547–8551. https://doi.
org/10.1021/jacs.7b02801 PMID: 28598151
19. Clayton DA, Shadel GS. Isolation of mitochondria from cells and tissues. Cold Spring Harbor Protocols.
2014; 2014(10). https://doi.org/10.1101/pdb.prot080010
20. Storrie B, Amadden E. Isolation of subcellular organelles. In: Methods in Enzymology. vol. 182. Cam-
bridge, Massachusetts, United Stated of America: Elsevier; 1990. p. 203–225.
21. Scalettar BA, Abney JR, Hackenbrock CR. Dynamics, structure, and function are coupled in the mito-
chondrial matrix. Proceedings of the National Academy of Sciences. 1991; 88(18):8057–8061. https://
doi.org/10.1073/pnas.88.18.8057
22. Dieteren CEJ, Gielen SCAM, Nijtmans LGJ, Smeitink JAM, Swarts HG, Brock R, et al. Solute diffusion
is hindered in the mitochondrial matrix. Proceedings of the National Academy of Sciences. 2011; 108
(21):8657–8662. https://doi.org/10.1073/pnas.1017581108
23. Su D, Teoh CL, Gao N, Xu QH, Chang YT. A Simple BODIPY-Based Viscosity Probe for Imaging of Cel-
lular Viscosity in Live Cells. Sensors. 2016; 16(9):1397. https://doi.org/10.3390/s16091397
24. Song X, Li N, Wang C, Xiao Y. Targetable and fixable rotor for quantifying mitochondrial viscosity of liv-
ing cells by fluorescence lifetime imaging. Journal of Materials Chemistry B. 2017; 5(2):360–368.
https://doi.org/10.1039/C6TB02524B
´
25. Demicheli V, Moreno DM, Jara GE, Lima A, Carballal S, Rıos N, et al. Mechanism of the reaction
of human manganese superoxide dismutase with peroxynitrite: nitration of critical tyrosine 34.
Biochemistry. 2016; 55(24):3403–3417. https://doi.org/10.1021/acs.biochem.6b00045 PMID:
27227512
26. Kuimova MK, Yahioglu G, Levitt JA, Suhling K. Molecular rotor measures viscosity of live cells via fluo-
rescence lifetime imaging. Journal of the American Chemical Society. 2008; 130(21):6672–6673.
https://doi.org/10.1021/ja800570d PMID: 18457396
PLOS ONE | https://doi.org/10.1371/journal.pone.0211165 February 14, 2019
18 / 20

Fluorescence lifetime probes reveal organelle viscosity
27. Sreenath K, Clark RJ, Zhu L. Tricolor emission of a fluorescent heteroditopic ligand over a concentration
gradient of zinc (II) ions. The Journal of organic chemistry. 2012; 77(18):8268–8279. https://doi.org/10.
1021/jo3016659 PMID: 22924325
´
28. Lopez-Duarte I, Vu TT, Izquierdo MA, Bull JA, Kuimova MK. A molecular rotor for measuring viscosity in
plasma membranes of live cells. Chemical Communications. 2014; 50(40):5282–5284. https://doi.org/
10.1039/c3cc47530a PMID: 24266030
29. Søren Preus. Calculate fluorescence quantum yield—FluorTools.com; 2015. Available from: http://
www.fluortools.com/software/ae/documentation/qy.
30. Cheng NS. Formula for the viscosity of a glycerol- water mixture. Industrial & engineering chemistry
research. 2008; 47(9):3285–3288. https://doi.org/10.1021/ie071349z
¨
31. Volk A, Kahler CJ. Density model for aqueous glycerol solutions. Experiments in Fluids. 2018; 59(5):75.
https://doi.org/10.1007/s00348-018-2527-y
32. Terry SJ, Zihni C, Elbediwy A, Vitiello E, San IVLC, Balda MS, et al. Spatially restricted activation of
RhoA signalling at epithelial junctions by p114RhoGEF drives junction formation and morphogenesis.
Nature cell biology. 2011; 13(2):159. https://doi.org/10.1038/ncb2156 PMID: 21258369
33. Suhaj A, Le Marois A, Williamson DJ, Suhling K, Lorenz CD, Owen DM. PRODAN differentially influ-
ences its local environment. Physical Chemistry Chemical Physics. 2018. https://doi.org/10.1039/
c8cp00543e PMID: 29850681
34. Jo S, Kim T, Iyer VG, Im W. CHARMM-GUI: a web-based graphical user interface for CHARMM. Jour-
nal of computational chemistry. 2008; 29(11):1859–1865. https://doi.org/10.1002/jcc.20945 PMID:
18351591
´
35. Wu EL, Cheng X, Jo S, Rui H, Song KC, Davila-Contreras EM, et al. CHARMM-GUI membrane builder
toward realistic biological membrane simulations. Journal of computational chemistry. 2014; 35
(27):1997–2004. https://doi.org/10.1002/jcc.23702 PMID: 25130509
36. Lee J, Cheng X, Swails JM, Yeom MS, Eastman PK, Lemkul JA, et al. CHARMM-GUI input generator
for NAMD, GROMACS, AMBER, OpenMM, and CHARMM/OpenMM simulations using the
CHARMM36 additive force field. Journal of chemical theory and computation. 2015; 12(1):405–413.
https://doi.org/10.1021/acs.jctc.5b00935 PMID: 26631602
37. Van Der Spoel D, Lindahl E, Hess B, Groenhof G, Mark AE, Berendsen HJ. GROMACS: fast, flexible,
and free. Journal of computational chemistry. 2005; 26(16):1701–1718. https://doi.org/10.1002/jcc.
20291 PMID: 16211538
´
38. Abraham MJ, Murtola T, Schulz R, Pall S, Smith JC, Hess B, et al. GROMACS: High performance
molecular simulations through multi-level parallelism from laptops to supercomputers. SoftwareX.
2015; 1:19–25. https://doi.org/10.1016/j.softx.2015.06.001
39. Klauda JB, Venable RM, Freites JA, O’Connor JW, Tobias DJ, Mondragon-Ramirez C, et al. Update
of the CHARMM All-Atom Additive Force Field for Lipids: Validation on Six Lipid Types. The Journal
of Physical Chemistry B. 2010; 114(23):7830–7843. https://doi.org/10.1021/jp101759q PMID:
20496934
40. Song KC, Livanec PW, Klauda JB, Kuczera K, Dunn RC, Im W. Orientation of Fluorescent Lipid Ana-
logue BODIPY-PC to Probe Lipid Membrane Properties: Insights from Molecular Dynamics Simula-
tions. The Journal of Physical Chemistry B. 2011; 115(19):6157–6165. https://doi.org/10.1021/
jp109629v PMID: 21513278
41. Vanommeslaeghe K, Hatcher E, Acharya C, Kundu S, Zhong S, Shim J, et al. CHARMM general force
field: A force field for drug-like molecules compatible with the CHARMM all-atom additive biological
force fields. Journal of Computational Chemistry. 2009; p. NA–NA. https://doi.org/10.1002/jcc.21367
42. MacKerell AD, Bashford D, Bellott M, Dunbrack RL, Evanseck JD, Field MJ, et al. All-Atom Empirical
Potential for Molecular Modeling and Dynamics Studies of Proteins. The Journal of Physical Chemistry
B. 1998; 102(18):3586–3616. https://doi.org/10.1021/jp973084f PMID: 24889800
ˇ
43. Vysniauskas A, Qurashi M, Gallop N, Balaz M, Anderson HL, Kuimova MK. Unravelling the effect of
temperature on viscosity-sensitive fluorescent molecular rotors. Chemical science. 2015; 6(10):5773–
5778. https://doi.org/10.1039/c5sc02248g PMID: 28791085
ˇ
´
44. Vysniauskas A, Lopez-Duarte I, Duchemin N, Vu TT, Wu Y, Budynina EM, et al. Exploring viscosity,
polarity and temperature sensitivity of BODIPY-based molecular rotors. Physical Chemistry Chemical
Physics. 2017. https://doi.org/10.1039/c7cp03571c PMID: 28718466
45. Soltoff SP, Mandel LJ. Potassium transport in the rabbit renal proximal tubule: effects of barium, oua-
bain, valinomycin, and other ionophores. The Journal of Membrane Biology. 1986; 94(2):153–161.
https://doi.org/10.1007/BF01871195 PMID: 3031306
46. Duchen MR. Mitochondria and calcium: from cell signalling to cell death. The Journal of Physiology.
2000; 529(1):57–68. https://doi.org/10.1111/j.1469-7793.2000.00057.x PMID: 11080251
PLOS ONE | https://doi.org/10.1371/journal.pone.0211165 February 14, 2019
19 / 20

Fluorescence lifetime probes reveal organelle viscosity
47. Yu T, Robotham JL, Yoon Y. Increased production of reactive oxygen species in hyperglycemic condi-
tions requires dynamic change of mitochondrial morphology. Proceedings of the National Academy of
Sciences of the United States of America. 2006; 103(8):2653–2658. https://doi.org/10.1073/pnas.
0511154103 PMID: 16477035
48. Visagie MH, Mqoco TV, Liebenberg L, Mathews EH, Mathews GE, Joubert AM. Influence of partial and
complete glutamine-and glucose deprivation of breast-and cervical tumorigenic cell lines. Cell & Biosci-
ence. 2015; 5(1):37. https://doi.org/10.1186/s13578-015-0030-1
49. Chen Y, Azad MB, Gibson SB. Superoxide is the major reactive oxygen species regulating autophagy.
Cell Death and Differentiation. 2009; 16(7):1040. https://doi.org/10.1038/cdd.2009.49 PMID: 19407826
50. Chung PH. Advanced Fluorescence Lifetime Imaging and Spectroscopy Techniques for Biological
Samples. King’s College London; 2012.
51. Feinstein MB, Fernandez SM, Sha’Afi RI. Fluidity of natural membranes and phosphatidylserine and
ganglioside dispersions: Effects of local anesthetics, cholesterol and protein. Biochimica et Biophysica
Acta (BBA)-Biomembranes. 1975; 413(3):354–370. https://doi.org/10.1016/0005-2736(75)90121-2
ˇ
52. Wu Y, Stefl M, Olzyńska A, Hof M, Yahioglu G, Yip P, et al. Molecular rheometry: direct determination of
viscosity in L o and L d lipid phases via fluorescence lifetime imaging. Physical Chemistry Chemical
Physics. 2013; 15(36):14986–14993. https://doi.org/10.1039/c3cp51953h PMID: 23912893
53. Pal S, Chakraborty H, Bandari S, Yahioglu G, Suhling K, Chattopadhyay A. Molecular rheology of neu-
ronal membranes explored using a molecular rotor: implications for receptor function. Chemistry and
physics of lipids. 2016; 196:69–75. https://doi.org/10.1016/j.chemphyslip.2016.02.004 PMID:
26879380
54. Ries J, Schwille P. Fluorescence correlation spectroscopy. BioEssays. 2012; 34(5):361–368. https://
doi.org/10.1002/bies.201100111 PMID: 22415816
55. Carisey A, Stroud M, Tsang R, Ballestrem C. Fluorescence recovery after photobleaching. In: Cell
Migration. Springer; 2011. p. 387–402.
56. Sbalzarini IF, Mezzacasa A, Helenius A, Koumoutsakos P. Effects of organelle shape on fluorescence
recovery after photobleaching. Biophysical Journal. 2005; 89(3):1482–1492. https://doi.org/10.1529/
biophysj.104.057885 PMID: 15951382
57. Sniekers YH, Van Donkelaar CC. Determining diffusion coefficients in inhomogeneous tissues using
fluorescence recovery after photobleaching. Biophysical Journal. 2005; 89(2):1302–1307. https://doi.
org/10.1529/biophysj.104.053652 PMID: 15894637
58. Lakowicz JR. Principles of Fluorescence Spectroscopy. Springer, New York; 2006.
59. Yu W, So P, French T, Gratton E. Fluorescence generalized polarization of cell membranes: a two-pho-
ton scanning microscopy approach. Biophysical journal. 1996; 70(2):626–636. PMID: 8789081
60. Le Marois A, Labouesse S, Suhling K, Heintzmann R. Noise-Corrected Principal Component Analysis
of fluorescence lifetime imaging data. Journal of Biophotonics. 2017; 10(9):1124–1133. https://doi.org/
10.1002/jbio.201600160 PMID: 27943625
61. Rowley MI, Coolen ACC, Vojnovic B, Barber PR. Robust Bayesian fluorescence lifetime estimation,
decay model selection and instrument response determination for low-intensity FLIM imaging. PLOS
ONE. 2016; 11(6). https://doi.org/10.1371/journal.pone.0158404
PLOS ONE | https://doi.org/10.1371/journal.pone.0211165 February 14, 2019
20 / 20