A comparative study on fluorescent cholesterol analogs as versatile cellular reporters

Cholesterol (Chol) is a crucial component of cellular membranes, but knowledge of its intracellular dynamics is scarce. Thus, it is of utmost interest to develop tools for visualization of Chol organization and dynamics in cells and tissues. For this purpose, many studies make use of fluorescently labeled Chol analogs. Unfortunately, the introduction of the label may influence the characteristics of the analog, such as its localization, interaction, and trafficking in cells; hence, it is important to get knowledge of such bias. In this report, we compared different fluorescent lipid analogs for their performance in cellular assays: 1) plasma membrane incorporation, specifically the preference for more ordered membrane environments in phase-separated giant unilamellar vesicles and giant plasma membrane vesicles; 2) cellular trafficking, specifically subcellular localization in Niemann-Pick type C disease cells; and 3) applicability in fluorescence correlation spectroscopy (FCS)-based and super-resolution stimulated emission depletion-FCS-based measurements of membrane diffusion dynamics. The analogs exhibited strong differences, with some indicating positive performance in the membrane-based experiments and others in the intracellular trafficking assay. However, none showed positive performance in all assays. Our results constitute a concise guide for the careful use of fluorescent Chol analogs in visualizing cellular Chol dynamics.


Preparation of GUVs
GUVs were prepared via the electroformation method as described previously ( 29 ). Briefl y, a lipid fi lm was formed on a platinum wire from a 1 mg/ml lipid mix of DOPC:SM:Chol (2:2:1) containing approximately 0.01 mol% of Chol analog. GUVs were formed in 300 mM sucrose solution at 68°C. Ten hertz, 2 V alternating electric current was used for electroformation.

GPMVs
GPMVs were prepared as described previously ( 31 ). Briefl y, RBL-2H3 cells seeded out on a 60 mm petri dish ( ‫ف‬ 70% confl uent) were washed with GPMV buffer [150 mM NaCl, 10 mM HEPES, 2 mM CaCl 2 (pH 7.4)] twice. Subsequently, 25 mM paraformaldehyde and 10 mM DTT (fi nal concentrations) were added to the GPMV buffer, and 2 ml of this fi nal buffer was added to the cells. The cells were incubated for 2 h at 37°C. Finally, GPMVs were collected by pipetting out the supernatant. Then, GPMVs were labeled by incubating 100 l of GPMV suspension with 1 l of 0.01 mg/ml [0.1 mg/ml for 6-dansyl-Chol (DChol) and fi lipin] fl uorescent lipid analog for 15 min.

Confocal microscopy of GUVs and GPMVs
GUVs and GPMVs were imaged with a Zeiss LSM 780 confocal microscope in BSA-coated 8-well Ibidi glass chambers (#1.5). Filipin III and D-Chol were excited with 405 nm and emission was collected between 420 and 480 nm. NBD-, TF-, and Star512-labeled analogs were excited with 488 nm and emission collected between 505 and 550 nm. KK114-and Atto647N-labeled analogs were excited with 633 nm and emission collected by a LP 650 Chol molecules by the Chol-binding peptides cannot be excluded. Filipin involves the least disturbance on the native Chol molecule, but its specifi city for Chol is questionable ( 13 ). In addition, proper labeling of cells with fi lipin requires their fi xation, such that pulse-chase or time-lapse studies with living cells are not practicable. Also, fi lipin labels all cellular Chol pools equally, such that endogenous Chol made by the cell cannot be discriminated from lipoprotein-derived Chol or from Chol circulating specifically between the plasma membrane and intracellular organelles. For such investigations, fl uorescent analogs of Chol are required. As has been shown before in various reports ( 10,11,(14)(15)(16)(17)(18)(19)(20)(21), natural Chol analogs such as DHE and CTL are, in principle, the most suitable molecules to mimic Chol. They can be inserted specifi cally into the plasma membrane or delivered to cells as part of lipoproteins for subsequent analysis of their transport itineraries and metabolism. Also, sterol-auxotroph cell lines and model organisms can use DHE or CTL as a sterol source and sterol transfer proteins can bind and transfer DHE and CTL between membranes similar to Chol ( 22 ). However, the unfavorable photophysical properties of DHE and CTL, which include UV absorption and UV fl uorescence, low quantum yield, and high photobleaching propensity, make their use for advanced microscopy involving prolonged and repeated imaging of the sample or single moleculebased fl uorescence fl uctuation analysis very challenging. Organic dye-labeled analogs are suitable for microscopic imaging of Chol ( 23,24 ), but despite their convenience, it is diffi cult to determine how well these analogs exhibit the properties of native Chol due to comparable sizes of the fl uorophores and Chol .
Here, we use several organic dye-labeled Chol analogs, those labeled with 7-nitrobenz-2-oxa-1,3-diazol-4-yl (NBD) ( 25 ) and TopFluor (TF) ( 26 ) at different positions or with the dyes AbberiorStar512 and AbberiorStarRed via PEG linkers ( 27 ), and demonstrate their performance in different experimental modalities: 1 ) their preference for more ordered membrane environments in phase-separated giant unilamellar vesicles (GUVs) and giant plasma membrane vesicles (GPMVs); 2 ) their localization following intracellular traffi cking in Niemann-Pick type C (NPC) disease cells; and 3 ) their applicability in fl uorescence correlation spectroscopy (FCS)-based measurements and superresolution stimulated emission depletion (STED)-FCSbased measurements of membrane diffusion dynamics. Our data revealed strong differences between the analogs, with some indicating high performance in the membranebased assay and others in the intracellular traffi cking NPCbased assay, but none showing positive performance in all these experimental modalities.

FCS and STED-FCS
FCS was carried out 1 ) for the green emitting analogs on a Leica SP8 system equipped with a pulsed white-light-laser (80 ps pulse width, 80 MHz repetition rate) for fl uorescence excitation at 488 nm, a 592 nm continuous-wave STED laser, an external avalanche photodiode (tau-SPAD) for fl uorescence detection, and the PicoHarp detection electronics (Picoquant, Berlin, Germany) for calculation of the correlation function with the option for a software-integrated gated detection at 1 ns ( 30 ); and 2 ) for the red-emitting analogs on an Abberior Resolft microscope (Abberior Instruments, Gottingen, Germany) equipped with an avalanche photodiode (SPCM-AQRH-13, Excelitas Techology), custom-extended with a pulsed 635 nm diode laser (Picoquant; 80 ps pulse width, repetition rate 80 MHz) for fl uorescence excitation, a pulsed Ti:Sa laser (MaiTai, Spectra-Physics/Newport; 200 ps pulse width at microscope, 80 MHz repetition rate synchronizing the excitation laser) ( 34 ). Measurements were done on SLBs ( 30 ). Each measurement lasted 10 s. STED-FCS was carried out with the same setups. Cells or model membranes were loaded with 100 nM fl uorescent molecules. Different STED powers were used for STED-FCS measurements.
The FCS data were fi tted by custom-written software that is freely available [https://github.com/dwaithe/FCS_point_correlator ( 35 )], using a one-component two-dimensional diffusion and one-triplet model ( 30 ), In this report, the data presented was fi tted in the time window of t c = 0.001 Ϫ 200 ms. Confocal measurements were used to determine T , and triplet correlation time T , which were then fi xed during further fi tting. The quality of each fi t was visually inspected.
In the case of the STED-FCS data, the transit time, D , was used to calculate the diameter, d (full-width-at-half-maximum), of the effective observation spot at different STED powers P STED .
fi lter. The multi-track mode of the microscope was used to eliminate the cross-talk between channels.

Percent liquid ordered partitioning calculation
ImageJ-Line profi le was used to calculate the preference of an analog to partition into the liquid ordered (Lo) environment of the GUVs and GPMVs, as described in ( 28,31,32 ). Briefl y, a line was selected which crossed opposite sides of the equatorial plane of the GUVs or GPMVs having environments of different order (Lo and Ld) on opposite sides. Opposite sides were chosen to eliminate polarization artifacts in the readout from the laser excitation. The Lo partitioning coeffi cient (%Lo) was calculated as: where F Lo and F Ld are the detected fl uorescence emission intensities in the Lo and Ld environment, respectively. If %Lo was >50%, a Chol analog was considered to prefer the Lo environment. We have shown negligible differences in the emission properties of the fl uorescent dyes in the different environments in a previous work ( 28 ).

Fluorescent labeling and imaging of NPC cells
In contrast to all other analogs, fi lipin III staining was done on fi xed cells. One microliter LysoTracker Red (Invitrogen; 100 M stock) was added into 1 ml of complete cell medium (fi nal concentration, 100 nM) and incubated for 2 h at 37°C. Cells were washed three times in medium without serum. One milliliter of 2% paraformaldehyde (in PBS) was added to the cells and the cells were incubated for 30 min at 4°C for fi xation. Afterwards, the cells were washed three times with PBS. Finally, 0.05 mg/ml fi lipin III was added to the cells on a shaker in the dark for 30 min, and the cells were washed three times before imaging.
For labeling of live NPC1-defi cient CHO cells with all other fl uorescent Chol analogs, 1 l LysoTracker Red (10 M stock) was added into 1 ml of complete cell medium (fi nal concentration, 10 nM) and incubated for 1 h at 37°C. Cells were washed three times in medium without serum, then incubated with approximately 5 g/ml fl uorescent Chol analogs in complete medium for 4 h at 37°C, and fi nally imaged on the confocal microscope, as described above. For simultaneous labeling of lysosomes, LysoTracker and fl uorescent analogs were added at the same time and incubated for 4 h at 37°C.
The labeling procedure was the same irrespective of genetic status, i.e., diseased (NPC1 Ϫ / Ϫ ) and healthy (NPC1 +/+ ) cells. Imaging was again performed with a Zeiss LSM 780 confocal microscope with the excitation and detection wavelengths for the fl uorescent Chol analogs chosen as for the GUV-and GPMVbased experiments. LysoTracker Red was excited with 543 nm and emission collected between 570 and 630 nm.

Colocalization analysis
Colocalization of the lysosome/late endosome marker, Lyso-Tracker Red, and the fl uorescent Chol analogs was analyzed from the confocal images of the doubly labeled NPC cells using the Coloc2 tool in FIJI (Fiji is just ImageJ, http://fi ji.sc), and quantifi ed by the Pearson correlation coeffi cient (PCC) ( 33 ). . Various proteins and lipids are thought to assemble within such small domains, thereby locally increasing the molecular order. An open question is how coordinated and active lipids reorganize into such structures. Unfortunately, due to their transient state and small size, lipid rafts have so far not directly been observed in the living cell. Consequently, membrane systems such as GUVs or GPMVs are often taken as a model to investigate lipid membrane ordering and, thus, potential lipid raft properties, especially because these model membrane systems tend to separate into phases of more ordered (Lo) and disordered (Ld) environments ( 44,45 ). Besides indicating coordinated lipid behavior, the Lo environment, specifi cally, due to its high molecular order, has been thought to reveal some physical properties of lipid rafts ( 45 ). While GUVs are fully artifi cial membrane vesicles, usually comprised of a ternary mixture of Chol and unsaturated and saturated lipids, GPMVs are derived from the plasma membrane of living cells, thus containing its lipid and protein diversity ( 32,46 ). Despite several recent contradictory reports ( 20,(47)(48)(49), Chol has been given an essential role in this lipid raft concept ( 2 ). This is mainly because Chol not only fl uidizes the ordered environment, but also orders the disordered environment, and Choldependent density fl uctuations are supposed to manifest at the nanoscale, i.e., beyond the resolution limit of classical imaging approaches ( 50,51 ). Thus, it is of great interest how Chol behaves in phase-separated membranes such as GUVs and GPMVs, specifi cally which environment (Lo or Ld) it prefers. The dominating opinion is that Chol enriches in the Lo domains because intrinsically fl uorescent Chol analogs (DHE and CTL) were found to partition with high preference into the Lo phase in model membranes ( 7,19,52 ). However, a recent hypothesis suggested that Chol is evenly distributed between domains ( 49 ). Studies using advanced fl uorescence microscopy are hampered by the necessity to use organic dye labels, which, due to their size, usually prevent the labeled lipids or Chol from effi ciently entering the Lo environment ( 28 ). We, therefore, investigated the partitioning characteristics of the different fl uorescent Chol analogs in both phase-separated GUVs and GPMVs.
For this purpose, we labeled the membrane of phaseseparated GUVs [DOPC:SM:Chol (2:2:1)] and RBL-2H3 cell-derived GPMVs with the different fl uorescent Chol analogs, and imaged the equatorial plane of the vesicles using confocal microscopy, as representatively shown in Fig. 2A for fi lipin III and 3-C6-NBD-Chol. We used the fl uorescent phospholipid analog, Atto647N-DOPE, as a marker for the Ld environment ( 28 ), and for each Chol analog quantifi ed the preference of entering the Lo environment by calculating the partitioning coeffi cient %Lo from the fl uorescence intensities detected for the Chol analogs in the Ld and Lo environment using equation 1. Values of %Lo = 0 and 100 indicate the extreme cases of no partitioning and complete partitioning into the Lo environment, respectively. Figure 2B compares values of the partitioning coeffi cient %Lo for the different Chol analogs. Due to the compositional complexity, the molecular Here, d ( P STED ) and D ( P STED ) are the diameter of the observation spot and the average transit time at the STED power, P STED , and d ( P STED = 0) and D ( P STED = 0) the respective parameters at confocal recordings, where the diameter of the confocal spot [ d ( P STED = 0) = 200 nm for 488 nm and 240 nm for 635 nm excitation] has been determined from imaging fl uorescent beads or from theory ( 36 ). Shortening of the transit time D due to photobleaching from the STED laser is unlikely, as has been shown by several control measurements before, i.e., the decrease in d with STED power P STED is not a result of photobleaching by the STED laser [e.g., ( 37,38 )]. While such photobleaching would potentially prevent the detection of slowly moving fl uorescent molecules, because they would photobleach before entering the effective observation spot, a reduction of the transit time D due to photobleaching from laser irradiation has been shown to be more prominent for confocal recordings, i.e., without the STED laser ( 38 ).

RESULTS AND DISCUSSION
We used the following fl uorescent Chol analogs for their performance in three different assays testing their potential to measure membrane and intracellular traffi cking dynamics: the Chol binding fl uorescent dye, fi lipin III ( 12,39 ), and Chol analogs tagged with the organic dye, dansyl (DChol labeled at the steroid backbone) ( 40 ) Fig. 1 . Labeling at the alkyl chain is usually preferred over attachment of the dye to the 3 ␤ -OH group, because it has been indicated that the respective -OH group may play important roles in Chol functionality and interactions ( 41 ). Labeling lipids with linkers has previously been introduced as a way to decrease the infl uence of the dye label on the lipids' characteristics ( 27,42 ), as done with the C6-and C12-linkers in 3-C6-and 3-C12-NBD-Chol, or the even longer PEG2000-linkers, in the case of Star512-and KK114-PEG-Chol. We have not included measurements on DHE and CTL here due to the necessity to use UV irradiation, as well as due to their low brightness and low photostability, which make them unfavorable for advanced microscopy.

Partitioning in phase-separated model membranes
The organization and dynamics of molecules in the plasma membrane of the living cell play an important role in many cellular signaling events. Here, a crucial role is dedicated to the role of Chol, because many experiments indicated that signaling varies with varying Chol levels ( 43 ). Specifi c attention has been given to the lipid raft or nanodomain concept, which postulates transient, lipiddependent, and Chol-rich domains as an organizational principle of the plasma membrane, compartmentalizing Fluorescent cholesterol analogs 303 model membranes. Filipin III is an intrinsically fl uorescent compound that binds Chol ( 12,39 ), and it is obviously expected to occupy less space than the Chol analogs carrying an additional similarly sized organic dye ( 28,54 ). Consequently, it is not surprising that the %Lo values of all other dye-tagged analogs are smaller. Yet, TF-Chol and the PEG2000-linked Star512-and KK114-PEG-Chol revealed almost similar Lo preference with values of %Lo у 50% in both GUVs and GPMVs, as indicated before ( 21,26,28,42,55 ). All other analogs, i.e., those labeled with NBD or dansyl, mostly avoided the more ordered environments order in the Lo environment and, thus, the order difference between Lo and Ld environments is less in GPMVs than in GUVs, leading, in general, to an easier access of more ordered environments and, consequently, to larger %Lo values in GPMVs compared with GUVs ( 28,51,53 ). The Chol binding probe, fi lipin III, shows the highest preference for the Lo environment in both GUVs and GP-MVs with %Lo values of up to 80% ( Fig. 2A, B ). This observation is in accordance with earlier studies using CTL or DHE ( 19 ), and it provides strong evidence for ordered domain enrichment of native Chol in phase-separated  one for complete and zero for no colocalization. As expected, the PCC was close to one for NPC1 Ϫ / Ϫ and almost zero for NPC1 +/+ cells ( Fig. 3B ). Calculation of the difference parameter ⌬ PCC in PCC values (equation 2) further revealed this change in colocalization, because larger values of ⌬ PCC indicate the expected accumulation of Chol in lysosomes in NPC1 Ϫ / Ϫ cells, while values of ⌬ PCC around or even below zero indicate no change or an unexpected change.
While these results show that fi lipin is a robust reporter to observe Chol accumulation in NPC disease, it is not applicable to living cells because the cells need to be fi xed before labeling, precluding live-cell observations. In contrast, all other Chol analogs used in this study could be employed in live-cell investigations; their colocalization with lysosomes is also depicted in Fig. 3 . Unfortunately, the values of Pearson coeffi cients, PCC and ⌬ PCC, show less appropriate traffi cking properties for all of the livecell adaptable probes. Specifi cally, dansyl-Chol, TF-Chol, Star512-PEG-Chol, and KK114-PEG-Chol were the least sensitive analogs to NPC1 status. Among them, the inappropriate properties of TF-Chol are probably the most surprising ones because TF-Chol is considered as one of the most suitable Chol analogs at the present time ( 26,63 ). It accumulates in spherical structures inside the cell, which do not colocalize with LysoTracker (supplementary Fig.  1). It is likely that these structures resemble lipid droplets, in which TF-Chol has been shown to partition preferentially compared with other sterol probes, such as DHE, fi lipin, or Raman-active phenyl-diyne-Chol ( 21,65,66 ). All NBD-labeled analogs, specifi cally, 3-C6-NBD-Chol and 25-NBD-Chol, showed values of ⌬ PCC comparable to fi lipin ( Fig. 3A, B ), making these the most appropriate candidates for reporting live-cell traffi cking. Close comparison of the PCC values of these NBD-tagged analogs suggests that the labeling position is essential for the dynamics of the analog. For example, while 25-NBD-Chol was to a large extent localized to lysosomes in NPC1 Ϫ / Ϫ cells, 22-NBD was much less colocalized with lysosomes in NPC1 Ϫ / Ϫ cells (supplementary Fig. 3).
Despite the promising properties of Star512-PEG-Chol and KK114-PEG-Chol with regard to plasma membrane organization, only a small fraction of these probes gets internalized; rather, they were still enriched in the plasma membrane even after several hours of incubation (supplementary Fig. 2). Interestingly, the internalized pool of PEG-containing Chol probes colocalized with late endosomes/lysosomes to a similar extent in both NPC1 +/+ and NPC1 Ϫ / Ϫ cells. Related to this, in separate experiments (carried out in baby hamster kidney cells), we observed that KK114-PEG-Chol strongly colocalized with DHE in the plasma membrane and after some time in a few endocytic vesicles (supplementary Fig. 4A). In those baby hamster kidney cells, we also found a distinctive traffi cking of DHE and TF-Chol to the perinuclear endocytic recycling compartment (ERC), a major sterol storage organelle in healthy mammalian cells (supplementary Fig. 4B) ( 67 ). To a certain extent, the low fraction of internalized KK114-PEG-Chol also targeted the ERC (supplementary Fig. 4B).
with values of down to %Lo < 10% in GUVs. On fi rst sight, this seems to be surprising, because both dyes are smaller in size than the other dyes, indicating, in this case, a sizeindependent property to be responsible for avoidance of highly packed membrane regions. It is known that NBD has the tendency to back-loop toward the membrane interface ( 56,57 ), hence, the side-chain tagged NBD-Chol analogs will, in turn, require more space in the membrane, which can squeeze the analogs out of the ordered phase ( 54 ). Our data shows that insertion of the long PEG2000 linkers between the Chol and a dye label improves the partitioning characteristics of the respective analog (as for Star512-and KK114-PEG-Chol). It is thought that due to the long PEG-linker, the dye label is excluded from the membrane ( 55 ). Yet, C6 and C12 carbon linkers, as introduced for the C6-and C12-NBD-Chol analogs, seem not to be suffi cient. However, it is worth noting that, similar to TF-and PEG-linker Chol analogs, 3-C6-NBD-Chol shows signifi cant %Lo (60%) in GPMVs compared with longer linker 3-C12-NBD-Chol ( Fig. 2A , lower panel). Thus, it is not trivial to draw a general rule to predict the partitioning of an organic dye-labeled analog.

Intracellular traffi cking: NPC1-null cells
Proper cellular function of Chol relies on its nonbiased insertion and localization into the plasma membrane (as outlined in the above example) and on its correct intracellular traffi cking, which are both related. Traffi cking of Chol between intracellular organelles and the cell membrane is tightly regulated ( 58 ). Shortcomings in these pathways may result in severe metabolic disorders. In NPC disease, caused by mutations in the genes coding either for the NPC1 or the NPC2 proteins, Chol traffi cking from late endosomes and lysosomes is impaired and Chol is "stored" in late endosomes ( 59,60 ). For NPC1-defi cient cells, not only transport of lipoprotein-derived Chol from endosomes is impaired, but also normal circulation of de novo synthesized-and plasma-membrane-derived Chol is affected ( 61,62 ). In fact, in a Chinese hamster ovarian cell line with mutated NPC1 protein, normal traffi cking of DHE and TF-Chol between the plasma membrane and recycling endosomes is disturbed, and the sterol probes were found to accumulate in lysosomal storage organelles over several hours ( 62,63 ). We used the same model cell system to determine the traffi cking itineraries of the Chol analogs in this study. NPC phenotyping has previously been performed by visualizing the localization of cellular Chol using fi lipin III ( 60,64 ). Therefore, we fi rst imaged healthy (NPC1 +/+ ) and diseased (NPC1 Ϫ / Ϫ ) NPC cells stained with fi lipin III and the lysosomal marker Lyso-Tracker using confocal microscopy, allowing us to examine the colocalization of fi lipin and LysoTracker. As expected, we observed a notable difference in the colocalization between NPC1 +/+ and NPC1 Ϫ / Ϫ cells ( Fig. 3A ).
While fi lipin mainly stained the plasma membrane of NPC1 +/+ cells, it was mainly in the late endosomes/lysosomes of NPC1 Ϫ / Ϫ cells. To quantify the extent of colocalization between Chol and LysoTracker, we calculated the PCC from their respective fl uorescence images, which is  ( 30 ).
To check the applicability of our fl uorescent Chol analogs in FCS, we recorded FCS data of the analogs diffusing in a SLB (100% DOPC on acid-cleaned glass). We selected only TF-Chol, Star512-PEG-Chol, KK114-PEG-Chol, and the 22-NBD-Chol for FCS analysis. Filipin III, as well as dansyl-Chol, did not give rise to any FCS curve, indicating their inappropriateness for these kinds of measurements due to low brightness and photostability. In the case of NBD, we only present the data of 22-NBD-Chol as a representative of all the NBD analogs, because we did not experience a remarkable difference between them. Also note that, while TF-Chol, Star512-PEG-Chol, and NBD-Chol were all excited at 488 nm, KK114-PEG-Chol was excited at 635 nm. From the FCS data, we could determine values of the average transit times, D (equation 3), and brightness or cpm (equation 4). Figure 4A, B compares the brightness values recorded at an excitation power of 5 W. Of all 488 nm excitable analogs, TF-Chol showed the highest brightness, followed by Star512-PEG-Chol and a very low brightness for the NBD-Chol analogs. KK114-PEG-Chol ( Fig. 4B ) showed an even 5-fold larger brightness than TF-Chol at this excitation power, albeit at a different excitation wavelength, as already indicated. Figure 4C, D depicts the transit times, D , determined for increasing excitation powers. A decrease in D for large excitation powers indicates enhanced photobleaching, i.e., low photostability, because the fl uorescent molecules turn dark prior leaving the observation spot ( 30 ). Here, all analogs except the NBD-Chol showed a satisfactory performance for excitation powers of up to 40 W, while in the case of NBD-Chol, values of transit times had already dropped off at values of 5-10 W. In a nutshell, only TF-Chol, Star512-PEG-Chol, and especially KK114-PEG-Chol were suitable for FCS measurements, while fi lipin III, dansyl-Chol, and the NBD-labeled probes For DHE and TF-Chol, and also for Chol, it has been shown that part of the sterol transport between the plasma membrane and the ERC takes place by a nonvesicular mechanism ( 67,68 ). This transport mode requires fast trans-bilayer migration of the sterol, and indeed, the majority of DHE and CTL has been shown to reside after membrane labeling on the inner leafl et of the plasma membrane ( 69 ). We suggest that the large-sized PEG-linker in KK114-PEG-Chol and Star512-PEG-Chol prevents sterol fl ip-fl op and, therefore, only allows for vesicular uptake in these cells. Accordingly, their enrichment in Chol-containing organelles, such as the ERC in healthy cells or the lysosomal storage organelles in NPC cells, is lower than that of analogs with fast trans-bilayer movement ( 17,70 ).

FCS on Chol analogs
Fluorescence microscopy and spectroscopy are important techniques for observing molecular organization and dynamics in living cells, as indicated in the two previous examples for the localization of the Chol analogs. Besides the spatial localization, the diffusion and interaction dynamics are also parameters essential to the function of Chol. FCS is a useful tool for investigating diffusion and interaction dynamics of fl uorescently labeled molecules (71)(72)(73)(74). In FCS, temporal fl uctuations in the detected fl uorescence due to molecular transits through the microscope's observation spot are analyzed to, for example, determine average transit times and molecular mobility. Crucial parameters for accurate fl uorescence microscopy and spectroscopy, especially FCS measurements, are the brightness and photostability of the fl uorescent label ( 75 ). While a limited brightness results in noisy data, low photostability impedes the correct determination of molecular organization and dynamics; for example, in FCS, a limited photostability might result in the loss of fl uorescence before the molecules have fully traversed the observation STED-FCS also increases the demands on brightness and photostability of a fl uorescent label. On one hand, reduced observation spot sizes yield a smaller amount of total photons from a traversing single molecule. On the other hand, the additional STED laser irradiation might induce additive photobleaching ( 30,79 ). We checked the performance of those Chol analogs that we had already tested in the conventional FCS recordings for their usefulness for STED-FCS. We incorporated the Chol analogs into an SLB, as before, and measured their average transit times ( D ) as we increased the power ( P STED ) of the STED laser. From theory, the diameter ( d ) of the observation spot decreases with increasing P STED ( d ‫ف‬ C/ √ P STED , with C depending on, among other parameters, the photophysical properties of the fl uorescent label) ( 79 ). Figure 4E, F plots the dependencies [ d ( P STED )] obtained for the different Chol analogs. In all cases, we observed an expected decrease of d with P STED , differing between the different probes. Steep declines of d ( P STED ) entail reaching small observation spot sizes at low STED powers, thereby minimizing the laser irradiation on the sample. Shortening of the transit time ( D ) due to photobleaching from the STED laser is unlikely, as has been shown previously by several control measurements, i.e., the decrease in d with STED power were shown not to be useful. Not surprisingly, TF-Chol and, specifi cally, KK114-PEG-Chol have been used in previous FCS-based measurements ( 42,76 ).

Super-resolution STED recordings
Recent developments in optical imaging have seen the rise of several super-resolution optical microscopy techniques. Using transitions between different states of a fl uorescent label (e.g., a dark and a bright state) and thereby modulating the fl uorescence (e.g., reversibly inhibiting fl uorescence emission), these techniques allow the observation of cellular structures at a spatial resolution below the diffraction limit of ‫ف‬ 200 nm of conventional microscopes ( 77 ). The fi rst of this kind was STED microscopy ( 77 ), where the size of the effective observation spot is reduced by the addition of a STED laser inducing reversible inhibition of fl uorescence emission through stimulated emission. A unique feature of STED microscopy is the possibility of tuning the size of the effective observation spot through the intensity of the STED laser, as well as the straightforward ability of combining it with FCS. Through recording FCS data at different observation spot sizes (with down to d < 60 nm diameters), STED-FCS realizes a detailed disclosure of molecular diffusion modes and interactions ( 37,78 ). Increasing the sensitivity of FCS, Chol. However, we kept them out of the scope of this study due to their fairly poor photo-physical properties (such as UV absorption and fl uorescence, low quantum yield, high photobleaching). Such properties make UV-optimized optics in the excitation path of a conventional wide fi eld microscope or multiphoton excitation and UV-sensitive detectors (which are all not easily available in cell biological laboratories) requirements for their visualization ( 80 ). In addition, their low brightness and high bleaching propensity make DHE and CTL unsuitable for single molecule studies or STED-based experiments, which were some important criteria we assessed in our analysis.
Overall, this study provides a useful guide for appropriate use of Chol analogs in advanced live subcellular imaging applications.
( P STED ) is not a result of photobleaching by the STED laser ( 37,38 ). Anticipated steep declines were revealed for KK114-PEG-Chol and Star512-PEG-Chol, making them the most appropriate for STED-FCS recordings. In contrast, TF-Chol did not show a signifi cant decrease of d ( P STED ). Note that in the case of the 488 nm excitable Chol analogs (e.g., Star512-PEG-Chol), where we employed a continuouswave (CW) STED laser (instead of the pulsed STED laser following 635 nm excitation as for KK114-PEG-Chol), the steepness of the decline was further increased using a gated detection scheme, as expected ( 30 ). While we could straightforwardly reach diameters of down to 50 nm for both KK114-PEG-Chol and Star512-PEG-Chol, as usually employed in STED-FCS recordings ( 42 ), with TF-Chol we could, even with gated detection, hardly reach 150 nm, at least with P STED < 60 mW of 592 nm CW STED laser light. Previous STED-FCS recordings of TF-Chol had to use gated detection in conjunction with >100 mW of 577 nm CW STED laser light to reach d < 100 nm ( 76 ). A current study has therefore employed KK114-PEG-Chol for STED-FCS membrane studies ( 42 ). Surprisingly, we realized rather good STED performance with NBD-Chol also. Unfortunately, its characteristics with regard to photobleaching in confocal FCS recordings also disfavor the NBD analogs for STED-FCS recordings.

CONCLUSION
Chol is of great importance in cellular structure and function. Therefore, it is crucial to visualize Chol in cells, which is often done using fl uorescent Chol analogs. Here, we analyzed various Chol analogs for their potential to image cellular Chol localization and dynamics. We tested their localization, i.e., their partitioning characteristics in phase-separated model membranes, their intracellular traffi cking in NPC1 disease cells, and their performance in confocal and advanced super-resolution FCS, for probing plasma-membrane dynamics. The Chol analogs performed very differently in these three assays. While Chol analogs with fl uorescent dye tags separated by a long PEG-linker proved superior for testing different membrane phases, as well as for use in FCS and STED-FCS measurements, they did not perform well in intracellular traffi cking experiments. Vice versa, NBD-labeled Chol analogs showed satisfactory performance in intracellular traffi cking, but exhibited poor performance in phase partitioning (except 3-C6-NBD-Chol, which partitioned into the ordered domains in GPMVs) and FCS experiments. TF-Chol, an analog that had so far been regarded to perform well in cellular assays, did not behave appropriately in the intracellular traffi cking assay and did not show good performance in STED-based experiments, yet indicated good properties for testing different membrane phases.
It is important to note that fl uorescently labeled Chol analogs carry a similar-sized fl uorophore as Chol, thus behavior of these analogs might be greatly ruled by the attached fl uorophore. Thus, in principle, natural Chol analogs, such as DHE and CTL, are better options to mimic at DEF-Univ Library of Southern Denmark, on May 24, 2017 www.jlr.org Downloaded from .html http://www.jlr.org/content/suppl/2015/12/29/jlr.M065326.DC1 Supplemental Material can be found at: