Robust Quantification of Live-Cell Single-Molecule Tracking Data for Fluorophores with Different Photophysical Properties

High-speed single-molecule tracking in live cells is becoming an increasingly popular method for quantifying the spatiotemporal behavior of proteins in vivo. The method provides a wealth of quantitative information, but users need to be aware of biases that can skew estimates of molecular mobilities. The range of suitable fluorophores for live-cell single-molecule imaging has grown substantially over the past few years, but it remains unclear to what extent differences in photophysical properties introduce biases. Here, we tested two fluorophores with entirely different photophysical properties, one that photoswitches frequently between bright and dark states (TMR) and one that shows exceptional photostability without photoswitching (JFX650). We used a fusion of the Escherichia coli DNA repair enzyme MutS to the HaloTag and optimized sample preparation and imaging conditions for both types of fluorophore. We then assessed the reliability of two common data analysis algorithms, mean-square displacement (MSD) analysis and Hidden Markov Modeling (HMM), to estimate the diffusion coefficients and fractions of MutS molecules in different states of motion. We introduce a simple approach that removes discrepancies in the data analyses and show that both algorithms yield consistent results, regardless of the fluorophore used. Nevertheless, each dye has its own strengths and weaknesses, with TMR being more suitable for sampling the diffusive behavior of many molecules, while JFX650 enables prolonged observation of only a few molecules per cell. These characterizations and recommendations should help to standardize measurements for increased reproducibility and comparability across studies.

The bacterial strain used for this study is derived from E. coli AB1157.We replaced the endogenous mutS gene with a HaloTag fusion, which enables covalent bonding with cell-permeable fluorophores such as the tetramethylrhodamine (TMR) and JFX650 used in these experiments.The MutS-Halo fusion was generated by Lambda Red recombination 1 .We inserted the HaloTag 2 sequence and a flexible 11-amino acid linker at the C terminus of the endogenous mutS gene, followed by a kanamycin resistance cassette.The allele was moved into the wild-type AB1157 strain by P1 phage transduction.The kanamycin resistance cassette was removed by expressing Flp recombinase from plasmid pCP20.Growing cells at 37 °C cured the temperature-sensitive pCP20 plasmid to produce the strain used for all experiments here.The presence of the gene fusion was confirmed by colony PCR and through fluorescence imaging for both the intermediate kanamycin-resistant strain and for the subsequent strain where kanamycin resistance has been removed.Full functionality of the MutS-Halo fusion protein was confirmed by measuring the frequency of spontaneous rifampicin resistance mutations.

S.1.2. Recipe for minimal M9 glucose growth and imaging medium
The following are added to a small volume (approximately 100 mL, to prevent precipitation) of MilliQ water in an autoclaved glass bottle: 500 µL 100 mM CaCl2, 1 mL 1M MgSO4, 100 mL 5x M9 minimal salts (Sigma-Aldrich), 10 mL 50x MEM amino acids ([-] L-Glutamine, Gibco), 5 mL 10 mg/ml L-Proline (dissolved in MilliQ water), 50 µL 0.5% Thiamine, 5 mL 20 % Glucose; this is then topped up with MilliQ water to make a total volume of 500 mL M9 medium.The medium is filtered using a Nalgene filter and a vacuum pump, and then stored at 4 °C.M9 media is warmed to room temperature prior to use.
The microscope has total internal reflection fluorescence (TIRF) illumination mode capabilities via a translatable stage (MB1530F/M, Thorlabs) onto which a series of lenses are mounted, such that movement of the stage translates the position that the laser is incident at the back focal plane of the objective; this stage is translated by a motorized actuator (z812, Thorlabs) and controller (Kinesis KDC101, Thorlabs).
For the single-molecule tracking of MutS-Halo in this study, we use an exposure time of 30 ms.Accounting for the readout time per frame, the total capture duration between subsequent frames is 0.030475 s, which is used for single-molecule tracking analysis.We use a frame size of 300 (H) x 512 pixels (W), where each pixel has a physical size of 16 µm for the Andor camera used; accounting for system magnification, our pixel resolution is 107 nm.EMCCD gain is also used for all fluorescence imaging and is set to a value of 300x.For camera control, we use Solis software (Andor).For controlling all other microscope components, we use micro-manager open source software 3 .

S.1.4. Data analysis: probability density function for histogram-fitting
To estimate the average diffusion coefficient D of diffusive states and their relative populations, we fit histograms of apparent diffusion coefficient (D*) measurements to a mixture of probability density functions 4,5 .
The distribution of * values takes the form of a Gamma distribution: Where  is the number of steps the mean-square displacement (MSD) is being calculated over (here we use 4-steps),  is the average diffusion coefficient, and  is the number of individual measurements made.
We believe there to be three diffusive states present in our single-molecule tracking data for MutS-Halo, so the distribution of * values is therefore fit to a summation of three probability density functions, where  1 ,  2 and  3 are the average diffusion coefficients of states 1, 2, and 3 respectively: and where:  1 +  2 +  3 = 1, corresponding to the relative occupancies of each diffusive state.

S.2. Quantification of localization error from TMR and JFX650
The localization error can be inferred from measurements of the apparent diffusion coefficient of molecules which are in an immobile state.
The apparent diffusion coefficient is calculated as follows: where MSD is the mean-square displacement, ΔT is the time between successive frames, and σloc is the localization error.We describe the use of vbSPT to determine the average diffusion coefficients of the three diffusive states of TMR-and JFX-labelled MutS-Halo (main text, Figure 4).Knowing that the MSD of immobile molecules is approximately zero, we can calculate the localization error from D* measurements of the immobile diffusive state (state 1 in our study).We see no significant differences between the estimated mean localization errors for TMR-and JFX650-labelled MutS (Figure S1).

S.3. Confirmation of dead E. coli cells in microscopy images and details of methods used for excluding them from single-molecule tracking analysis
During single-molecule tracking experiments, we noticed a subpopulation of cells in the sample which took longer than average to photobleach (i.e.longer than 500 frames, see main text, Figure 2 for general photobleaching characteristics of TMR).We hypothesised that these cells were potentially dead, and should be excluded from analysis.
To confirm this, we performed a dual-labelling experiment to add SYTOX™ Green nucleic acid stain (Invitrogen) to E. coli cells alongside the TMR-labelling of MutS-Halo; we followed the same sample preparation protocol as described in the main text section 2.1, except that SYTOX was added to the cell culture at the same time as TMR to a final concentration of 5 nM.To image, we used the same procedure as for the single-molecule tracking protocol (see main text, methods), with an additional imaging step.Briefly, we used an exposure time of 100 ms, whilst keeping the 488 nm laser at very low power density (< 1 Wcm -2 ).We acquired a short movie (1000 frames) from which a mean intensity projection could be calculated through the 1000 frames (example shown in Figure S2a), indicating the locations of SYTOX-positive (dead) cells.To test for correlation with TMR photobleaching rate, we acquired a movie of 1000 frames from laser turn-on (in the 561 nm channel), and calculated a median intensity projection through the 1000 frames (example shown in Figure 2b).Comparing figures S2a and S2b, we see that SYTOX-positive cells correlate overwhelmingly to the cells in which TMR fluorescence bleaches slowly, and we therefore use this median intensity projection method to exclude these cells from our single-molecule tracking analysis, without the need for SYTOX labelling for every experiment.
Similarly to the TMR experiments, we also find a subpopulation of dead cells in the JFX650 experiments (Fig S2c).We correlated SYTOX-positive cells in this case to cells which were still fluorescing in the 640 nm channel at the end of our usual 5000-frame JFX650 single-molecule tracking acquisition.To exclude these cells from analysis, we therefore generate a mean intensity projection of the final 500 frames of our single-molecule tracking movie (see example in Fig. S2d), highlighting the locations of these cells without the need for SYTOX staining in every experiment.

S.5. vbSPT analysis of partitioned trajectories
In our study, we tested how partitioning the single-molecule trajectories affects the diffusion analysis when using the MSD calculation method.For completeness, we tested the HMM analysis method to see whether the result would differ depending on whether non-partitioned or partitioned trajectories were used (Figure S4).Performing vbSPT analysis on tracks partitioned into 5-frame sections produced similar outputs to the non-partitioned tracks (shown in Figure 4, main text), indicating that HMM analysis is less sensitive to biases caused by trajectory length or number.

Figure S1 :
Figure S1:The mean localization error for MutS-Halo-TMR and MutS-Halo-JFX650, inferred from vbSPT measurements of the immobile state diffusion coefficient.Bar heights show the mean values and errorbars show the standard error of the mean from 3 and 6 independent experimental repeats for TMR and JFX650, respectively.

Figure S2 :
Figure S2: Correlation of SYTOX-positive (dead) cells to TMR-and JFX650-labelled "slow-to-bleach" cells.a) Example image of E. coli cells which have been labelled with SYTOX Green.Numbered cell outlines are shown in yellow and SYTOX-positive cells are highlighted in red.b) The same cells as in a) have been labelled with TMR.Image shows a median intensity projection of the first 1000 frames from laser turn-on in the TMR channel."Slow-to-bleach" cells are highlighted in red, with normal cells in yellow.c) Example image of E. coli cells which have been labelled with SYTOX Green.SYTOX-positive cells shown in red.d) The same cells as in c) have been labelled with JFX650.Image shows a mean intensity projection of the final 500 frames of the single-molecule acquisition in the JFX650 channel, in which "slow-to-bleach" cells are highlighted in red.

Figure S4 :
FigureS4: A comparison of TMR and JFX650-labelled MutS-Halo diffusive state analysis using HMM, where the input trajectories have been subjected to imposed partitioning.Shown are the diffusion coefficient values and the state occupancies for the 3-state model.Error bars show the standard error of the mean across 3x (TMR) and 6x (JFX650) independent experimental repeats and bar heights are weighted means according to the total number of trajectories per repeat.

Figure S6 :
Figure S6: Standard deviation across experimental repeats for measurements of state 1 occupancy (immobile state) for TMR and JFX650-labelled MutS-Halo.Values are shown separately for the three analysis methods used on our datasets: HMM, MSD and MSD on partitioned trajectories (MSD-part).