Assembly of Bacillus subtilis Dynamin into Membrane-Protective Structures in Response to Environmental Stress Is Mediated by Moderate Changes in Dynamics at a Single Molecule Level

Dynamin-like proteins are membrane-associated GTPases, conserved in bacteria and in eukaryotes, that can mediate nucleotide-driven membrane deformation or membrane fusion reactions. Bacillus subtilis’ DynA has been shown to play an important role in protecting cells against chemicals that induce membrane leakage, and to form an increased number of membrane-associated structures after induction of membrane stress. We have studied the dynamics of DynA at a single molecule level in real time, to investigate how assembly of stress-induced structures is accompanied by changes in molecule dynamics. We show that DynA molecule displacements are best described by the existence of three distinct populations, a static mode, a low-mobility, and a fast-mobile state. Thus, DynA is most likely freely diffusive within the cytosol, moves along the cell membrane with a low mobility, and arrests at division sites or at stress-induced lesions at the membrane. In response to stress-inducing membrane leakage, but not to general stress, DynA molecules become slightly more static, but largely retain their mobility, suggesting that only few molecules are involved in the repair of membrane lesions, while most molecules remain in a dynamic mode scanning for lesions. Our data suggest that even moderate changes in single molecule dynamics can lead to visible changes in protein localization patterns.


Introduction
Dynamins are a class of GTPases whose major function lies in the remodeling of biological membranes, most prominently in the mechanical, GTP-driven pinching off of membrane vesicles. Other functions include organelle division and fusion of mitochondrial membranes [Ramachandran and Schmid, 2018]. Bacteria also contain dynamin-like proteins (DLP), whose functions lie in the involvement in bacterial cell division [Dempwolff et al., 2012;Schlimpert et al., 2017] and response to membrane stress [Sawant et al., 2016], likely in the resealing of holes in the membrane induced by chemical agents or phage evasion. Thus, the overarching scheme of dynamins are membrane fusion and fission [Antonny et al., 2016].
Mechanochemical membrane remodeling involves different domains of dynamins, mediating membrane binding and oligomerization, and GTP hydrolysis-driven This article is licensed under the Creative Commons Attribution 4.0 International License (CC BY) (http://www.karger.com/Services/ OpenAccessLicense). Usage, derivative works and distribution are permitted provided that proper credit is given to the author and the original publisher.
In this work, we wished to further characterize B. subtilis dynamin by studying its motion at a single molecule level and to analyze how the assembly of membrane-associated fluorescent foci of a functional DynA-fluorescent protein fusion is brought about by changes of molecule motion. We show that the dynamics of DynA can be best explained by assuming three populations of molecules with distinct diffusion coefficients. Induction of membrane stress and accompanying increase in membrane-associated assemblies are accompanied by only small changes in the size of DynA molecule populations, showing that visible rearrangements of proteins in a bacterial cell can be achieved by only small changes at a single molecule level.

A DynA-mVenus Fusion Forms an Increased Number of Foci in Response to Chemical Inducing Membrane Leakage
We have previously shown that DynA-GFP accumulates at sites of cell division, but can also form visible foci at the cell membrane away from the division site [Dempwolff et al., 2012]. The Bramkamp group has found that addition of nisin, which induces membrane leakage for ions, induces the formation of fluorescent foci at the cell membrane [Sawant et al., 2016]. Because DynA plays a protective role during membrane stress [Sawant et al., 2016], we wished to investigate if other environmental stresses might lead to an accumulation of DynA at the membrane, and treated cells with different chemical and physical stresses. Figure 1a shows that only a minority of exponentially growing cells (26%, Fig. 1b) showed fluorescent foci, predominantly close to mid-cell (septal area) and at the cell poles and that the number of cells with foci was similar during heat and ethanol stress. However, as quantified in Figure 1b, osmotic stress (sorbitol), osmotic and ionic stress (NaCl), and membrane stress (nisin) lead to a large increase in the number of cells with foci, especially under nisin and NaCl conditions. Compared with exponentially growing cells, where 25% contained a single focus and 1% several foci, 63% of nisin-treated cells contained a single focus and 26% several foci; a total of 50% of cells treated with sorbitol contained DynA-mV signals, and 86% of salt-treated cells (Fig. 1b). Following nisin treatment, the ratio of signals at the septum, the cell poles or the lateral membrane changed considerably, from roughly one-third for each position (many places at the Microb Physiol 2022;32:57-69 DOI: 10.1159/000521585 lateral cell membrane defining a "position") to 70% lateral membrane at the expense of septal and polar localization (Fig. 1b). These findings are in agreement with data from the Bramkamp laboratory showing a strong increase in membrane-associated foci after treatment with nisin [Sawant et al., 2016], and might imply that DynA assembles into visible structures after many kinds of stresses. Epifluorescence analysis of DynA-mVenus foci arising in response to different stress conditions. a YFP channels showing unstressed cells with foci mostly found at the septum, heat stressed cells (42°C), or cells after addition of 10% ethanol (EtOH), 1 M sorbitol, 1 M NaCl, 250 µg bacitracin, or 10 µg nisin. Cells were incubated for DOI: 10.1159/000521585

DynA Moves as Three Distinct Fractions of Mobility within Cells
Because epifluorescence microscopy can only detect larger, static assemblies of molecules, but not diffusive molecules, we employed single molecule tracking to obtain more quantitative information on changes of DynA mobility. We investigated the dynamics of single DynA-mVenus molecules using "slim-field" illumination and bleaching of most fluorescent proteins until single molecules remain visible that can be efficiently tracked in real time [Reyes-Lamothe et al., 2010;Kleine Borgmann et al., 2013;Schibany et al., 2018]. We have previously shown that this approach leaves cells physiologically largely intact and harms them only to a small degree [Dersch et al., 2020;El Najjar et al., 2020]. We obtained an average track length of about 8 steps and used tracks of 5 steps and longer for the analyses of diffusion constants and possible existence of several distinct populations of molecules using squared displacement analysis (SQD). Figure 2 shows jump distance (JD) evaluation of SQD data, indicating that the observed step size distributions can be best explained assuming 3 distinct populations (Fig. 2a), leaving almost no residuals when comparing observed data and modeled data (Fig. 2b). When only 2 populations are used (Fig. 2c), observed JDs cannot be fully covered, and residuals remain in panel d, where the blue line deviates visibly from the dotted red line representing the modelled data. Although we do not want to overinterpret our data, we will continue assuming three diffusive populations, and will return to the idea of the existence of only two populations in the discussion section. For three population fits, we obtained the following diffusion constants (D) and population sizes for DynA molecules in unstressed cells, using SMTracker 2.0: 15% of the molecules had a D of 0.015 µm 2 s -1 , about 50% showed a D of 0.122 µm 2 s -1 , and the fastest population of about 35% showed a D of about 0.9 µm 2 s -1 ( Fig. 3a and b). The later D is characteristic of large cytosolic proteins [Schenk et al., 2017;Rosch et al., 2018], while that of 0.015 µm 2 s -1 reflects a rather static molecule, bound to a much larger structure, such as a ribosome [Sanamrad et al., 2014]. Thus, 15% of DynA molecules appear to be statically positioned within a larger subcellular structure, while 35% are freely diffusive. We will come back to this point later, when we consider the spatial distribution of static and diffusive molecules. Taking into account the findings that DynA is involved in the cell division process and is frequently observed at the septal area (see also Fig. 1a), a fraction of the static DynA molecules is likely bound to the Z ring containing a variety of cell division proteins. If DynA indeed diffuses as three distinct fractions, the mediummobile fraction might be molecules attached to the membrane, which would be expected to diffuse somewhat slower than cytosolic proteins, or molecules in a (pre) complex of an intermediate size. With the setup used, we cannot determine if molecules move along the membrane or within the cytosol, so the idea of an intermediate, membrane-associated DynA fractions remains a hypothesis.
To a first approximation, the JD distribution did not change markedly between exponentially growing cells and stressed cells (compare Fig. 2a with 2e-i), except for nisin-treated cells, in which JD visibly shifted left, towards slowed distances (Fig. 2j, compare with 2a). This implies that DynA molecules have lower diffusion constants after treatment with nisin. We quantified the percentage of molecules from the different diffusive populations and found slightly different values for D for each condition, as well as different population sizes (data not shown). In order to better compare changes, we kept D values constant (in fact, Ds were determined to best fit for all conditions), such that changes in molecule dynamics were only reflected in changes in population size. The bubble plot in Figure 3a shows that heat stress led only to a marginal increase in the fast-mobility fraction, while ethanol treatment led to a roughly 30% increase in this fraction, at the expense of the medium-mobile fraction. Sorbitol treatment reduced the static fraction from 15 to 11.3% (i.e., relatively by about 25%), and enlarged the fast-mobility fraction by 24%. The strongest decrease in static molecules was observed after addition of sodium chloride, accompanied by an increase in the fast-mobile fraction. Also following bacitracin treatment, DynA mol- Jump distance analysis (JD) of DynA-mVenus showing the distribution of particle displacements in a fixed time interval, plotted in a histogram (a + c + e-j). The quantile-quantile plots visualize the goodness of the fits, by comparing difference (blue line) between modelled (dashed line) and observed data (b + d). a Unstressed triple fit. b Quantile-quantile plot of the unstressed triple fit. c Unstressed double fit. d Quantile-quantile plot of the un-stressed double fit. Because the data imply a triple fit is the better option, the following conditions were analyzed with a triple fit: 42°C (e), 10% ethanol (f), 1 M sorbitol (g), 1 M NaCl (i), 250 µg bacitracin (j), 10 µg nisin (k). h, l Double fits of 1 M sorbitol treatment (h) of 10 µg nisin (l). m-n Examples of trajectories from cells expressing DynA-mV during exponential growth (m) or after addition of nisin (n). ecules became somewhat more mobile; however, as for the other stress conditions, this effect was not dramatic. Conversely, after the addition of nisin, the size of the static fraction, assumed to be DynA molecules engaged in cell division or in repair of membrane damage, almost doubled, while the number of highly mobile molecules strongly decreased (Fig. 3a). These data support the idea of a specific nisin-induced effect on DynA dynamics and suggest that changes observed by the other stress conditions are likely more unspecific. However, although DynA provides protection for the cell against nisin treat-ment, overall changes in the whole DynA molecule population are not profound, at least not as large as we would have expected. Only about 13% of all molecules move from a more mobile into a static mode, which we assume is the active form of DynA.

Stress Conditions Lead to Visible Changes in DynA Localization and Mobility
In order to further investigate changes in the localization and mobility of DynA, we generated heat maps, in which tracks from all cells are projected into a medium-sized cell Analysis of the dynamics of DynA-mVenus under different stress conditions. For the determination of the diffusion coefficients (D) and the fraction sizes, SQD analyses with simultaneous fits for all conditions were used. This leads to the determination of a single value for D that fits best to all conditions, in order to allow for a better comparison of changes in population sizes. a The bub-ble plot shows the size of the fraction where each bubble is proportional to the area of its corresponding diffusion coefficients. b In the table, population 1 ("pop 1") refers to D1, and other sizes accordingly. Note that standard errors refer to fitting errors, not to differences between the biological triplicates, whose data were pooled for the analyses. and are displayed for frequent or infrequent presence. Figure 4a shows a slight tendency for DynA to be present close to the cell membrane ( Fig. 4b is a twofold mirrored heat map), but even the frequent foci observed at the division site ( Fig. 1a) do not shift this preference to the cell center. These data suggest that DynA molecules are distributed throughout the cell, i.e. within the cytosol and at the membrane. Heat and sorbitol stress induced an accumulation of DynA at the subpolar regions ( Fig. 4c/d, and g/h), which might be due to some molecule aggregation due to un-or misfolding during heat or osmotic shock. Note that heat maps were not normalized to each other because of the strong subpolar signals observed after heat and osmotic stress induction. Heat maps of ethanol, sodium chloride, and bacitracin stress did not induce any clear changes ( Fig. 4e/f, i/j, and k/l). Conversely, DynA molecules showed a higher preference towards membrane-proximal accumulation following nisin treatment (Fig. 4m/n).
We also generated "confinement maps," in which molecules showing little displacement for an extended time (confined motion) and freely diffusing molecules are displayed in a standardized cell. This analysis must not be confused with SQD analyses, because molecules with a low diffusion constant do not necessarily represent those showing confined motion for some time, although these two parameters will strongly overlap, and the analysis does not distinguish between e.g., populations of medium or of high mobility. For confinement maps, we used a radius of 120 nm (about three times the localization error in the setup) and defined confinement as molecules remaining within this radius for at least 8 time intervals. Figure 5 shows that during exponential growth (Fig. 5a), confined motion (red tracks) occurred throughout the cell, and frequently close to the cell membrane. The distribution of tracks along the length of cells (long axis, or x axis) is shown in Figure 5h, left panel, distribution along the width (short/y axis) in Figure 5h, right panel. Sorbitol and heat stress induced subpolar areas of increased confined motion (best seen in Fig. 5i and k, long axis), possibly reflecting places of enhanced protein aggregation. For ethanol, sodium chloride and bacitracin treatment, no clear changes were observed. Nisin treatment induced the strongest change in spatial dynamics, in that confined motion was observed in larger areas at the lateral cell membrane (Fig. 5g) including stronger signals at the cell poles ( Fig. 5n, long axis), compared with unstressed cells (Fig. 5a and h). Accumulation of confined tracks at the poles changes the localization along the short cell axis seemingly away from the membrane (Fig. 5n, right panel, compare with 5h). Additionally, transitions between confined motion and free diffusion, indicated by green tracks, were visibly increased following nisin-induced stress. The appearance of larger areas of confined motion after treatment with nisin likely reflects the increased induction of foci seen in Figure 1.

Nisin-Induced Membrane Stress Leads Only to a Mild
Increase in Dwell Times of DynA SQD analyses for the dynamics, as well as heat and confinement maps for the localization, have shown that nisin addition leads to the clearest change in the behavior of DynA. The same is true when having a look at the residue time of DynA molecules. Figure 6 shows two possible fits for the molecules, which stay for a period of at least 8 time points within a radius of 120 nm. One-component fit with τ of 0.22 s fits well for DynA from non-stressed cells (Fig. 6a). The single decay curve already includes most of the data. Even if we assume two populations, the dwell times of τ 1 with 0.17 s and τ 2 with 0.23 s are close to each other. Note that due to molecule bleaching during acquisition, actual dwell times will be higher in cells, but for comparing dwell times between different conditions, estimates are sufficiently accurate. Around 88 ± 20% are included in τ 2 , while the other 12 ± 20% belong to τ 1 . This suggests that there is only one population of DynA molecules that arrests in motion for a considerable time. In comparison, the addition of nisin leads to two distinguish populations (Fig. 6b). The onecomponent fit does not include the whole data, while the two-component fit does. While τ 1 is with 0.18 s similar to the unstressed one, τ 2 increased strongly, to 0.35 s. This longer dwelling event occurred for roughly half of the molecules (48 ± 3.7%). Thus, we can assume that under nisin stress, there is a population which has a shorter average dwelling time (τ 1 ) similar to unstressed cells, while the long dwelling events (τ 2 ) refer to DynA molecules in the membrane that repair nisin-induced damage. Of note, the static population of molecules seen with SQD analyses of DynA, which almost doubled under nisin stress, will largely overlap with the population of molecules showing long dwelling events, but is not necessarily identical. Likewise, the popula- tion showing shorter dwell events likely conforms with the intermediate and fast populations in the SQD analysis, although even freely diffusive molecules can arrest for a short period of time. A single decay curve and thus a single average dwell time was also found for the other stress conditions, with no significant difference to the dwell time of exponentially growing cells (data not shown) compared to unstressed cells, except for bacitracin treatment (Fig. 6c), where two decay curves were determined, as well as a slightly higher dwell time for τ 2 of 0.28 s for half of the molecules (50 ± 9.1%). τ 1 on the other hand was similar to the nisinstressed condition with 0.22 s.
Thus, although a direct comparison of the SQD data with the dwell times cannot be made, it can be said that with both tools we are able to detect static DynA molecules. In both models, under nisin stress, DynA becomes more static, and a clear difference in its single molecule dynamics can be seen. While with SQD analysis, a transition to a more static behavior under bacitracin stress was not apparent, the change in the dwelling events suggests that also under this condition DynA is recruited to the membrane and induces membrane repair assemblies.

Discussion
Bacterial cells have traditionally been regarded as organisms whose protein dynamics are largely governed by free Brownian diffusion of enzymes. While this may be true for a majority of cellular proteins, many proteins show assembly at distinct subcellular sites at all times, or even at specific times during the cell cycle [Shapiro and Losick, 2000]. It has also been found that diffusion can be anomalous within the bacterial cytoplasm, and its magnitude is affected by metabolic activity of the cell [Parry et al., 2014]. Thus, mobility in non-compartmentalized bacterial cells is non-trivial, and it is still unclear for most proteins complexes in which fashion they are assembled. Diffusion capture has been described for some processes, but other modes appear to be possible.
In our work, we wished to address the question of how molecule dynamics change for a protein known to form assemblies, based on fluorescence microscopy, at a given subcellular site, at a single molecule level, in order to monitor and describe changes between diffusive and bound molecules. We chose DynA from B. subtilis, a soluble protein that can associate with the cell membrane via a transmembrane-like dimeric subunit, D1. DynA was shown to assemble at the septum in a subset of cells within a growing population, colocalizing with FtsZ [Dempwolff et al., 2012]. Based on genetic interactions, it was suggested that DynA plays a role in cell division, possible at the last step of fusion of the invaginating cell membrane. Increased membrane binding of DynA was inferred from wide field microscopy using a fluorescent protein fusion of DynA, when cells were treated with an agent inducing membrane leakage, lantibiotic nisin, a compound produced by lactic acid bacteria, with a broad antibacterial spectrum amongst Gram-positive bacteria. In order to obtain more insight into the dynamics of DynA assembly at the cell membrane, we used single molecule tracking, which not only localizes statically bound molecules, but also freely diffusing, mobile molecules. We found that DynA diffusion can be explained by assuming two distinct populations, and to an even better degree by assuming three populations. For a twopopulation model, one would assume a freely diffusing part and a statically, membrane-engaged part of molecules. We found minor changes in single molecule dynamics using two or three population models for general stress conditions, but considerable changes following nisin treatment. For the purpose of this study, we chose a model with three populations, which might be a better reflection of molecule dynamics within cells, keeping in mind that qualitatively, changes were similar regardless of whether two or three populations were assumed.
Nisin treatment induces or enhances the recruitment of DynA molecules to the cell membrane, and the formation of fluorescent foci as seen by observing a functional fluorescent protein fusion to DynA [Sawant et al., 2016] (Fig. 1). Rather than about 25% of cells showing a single DynA-mVenus focus at the cell membrane during exponential growth, about 80% of cells contain mostly several foci following nisin treatment. We show that this observation made by wide-field fluorescence microscopy involves only about 13% of the molecules, which change from being mobile (regardless of whether one or two mobile populations are assumed) to being static, using single molecule tracking. Accumulation of molecules moving in a confined manner occurred close to the cell membrane, as shown by confinement maps. Considerably longer dwell times that were observed between exponentially growing cells and nisin-treated cells also underline that DynA becomes more static, by assembling in membrane repair complexes.
Using general stress factors such as heat, ethanol, osmotic, and salt stress, we found a visual redistribution of molecules, in that especially after heat and osmotic stress, more molecules of DynA were found to cluster close to the cell poles. We interpret these findings to indicate that some molecules would denature and thus aggregate at the cell poles, possibly soon after translation, which occurs predominantly at the poles [Lewis et al., 2000;Mascarenhas et al., 2001]. However, we did not find any notable changes at a single molecule level after general and membrane stress, different from treatment with nisin. An argument can be made that also Bacitracin treatment leads to significant changes of molecule dynamics, which can be seen for the confined motion, and increased average dwell times that were observed. Both nisin and bacitracin are antibiotics that bind to and affect the cycle of lipid II during cell wall synthesis, and nisin induces membrane leakage [El Jastimi et al., 1999]. These findings suggest that DynA is recruited to membrane damage if there is a specific leakage, e.g. through interference with lipid II components.
The data suggest that for repair of membrane leakage or generally membrane fusion, a minority of DynA molecules becomes statically involved/assemble at the membrane. The remaining molecules continue to be mobile, likely scanning the cell membrane for lesions. Our work shows that assembly of a protein at the cell membrane (in order to mediate membrane integrity) involves a small DOI: 10.1159/000521585 exchange in molecule dynamics and thus an induced diffusion/capture mechanism. We speculate that a large majority of protein assembly mechanisms involve diffusion/ capture rather than specific membrane integration or proteolysis or other targeted mechanisms.

Materials and Methods
Bacterial Strains and Growth Conditions B. subtilis strain BG214 (also termed YB886) is a derivate of the strain 168. It is auxotrophic for tryptophan (trpC2) and methionine (metB5); therefore, minimal media were supplemented with 50 mg/mL tryptophan and methionine, respectively [Yasbin et al., 1980] [Jaacks et al., 1989] with the same temperature and speed. For the determination of the growth rate, it was measured with an optical density at 600 nm (OD 600 ). An OD 600 of 0.8 was used for microscopy. Selection of the strains was accomplished by using the antibiotics ampicillin (100 µg/mL) for E. coli and chloramphenicol (5 µg/mL) for B. subtilis. In BG214 cells the xylose promotor was induced by adding 0.1% (w/v) xylose to the cells.

Strain Construction
BG214 DynA_mVenus at the original locus was constructed by firstly creating the plasmid in E. coli DH5α. For this, the plasmid pSG1164 (ECE155) with a mVenus fluorophore was used. The last 500 bp of the 3′-end of DynA were cloned in front of the fluorophore. The cutting enzymes EcoRI and ApaI were used with the forward primer GCTAGAATTCGACAACAGCCTTA and the reverse primer GCATGGGCCCCATTTTTATTGTATTGTCTG, where the stop codon was removed. After transformation of this plasmid in E. coli, the plasmid was checked via sequencing and then used for transformation in B. subtilis BG 214. To get B. subtilis BG214 competent for DNA uptake, an overnight culture of this strain was grown at 30°C and 200 rpm and inoculated in 10 mL of modified competence media to an OD 600 of 0.08. This was followed by incubation at 37°C and 200 rpm for around 3 h to an OD 600 of 1.3-1.5. When this OD was reached, 2-3,000 ng of plasmid DNA was added to 1 mL of the competent B. subtilis 168 cells and incubated at 37°C with 200 rpm for 2 h. Last, the cells were streaked on a Luria-Bertani agar plates with the needed antibiotics (chloramphenicol) for selection and incubated overnight at 30°C. Colonies were screened for a positive clone via PCR (forward primer CATC-CATGGATGACAGATCAAAACAGAAAA, reverse primer CCTTACTCGAGTTTGGATCCTTACTTGTACAGCTCGTC-CATGCCG) and by antibiotic selection.
Preparation of BG214 DynA-mVenus for Microscopy B. subtilis BG214 DynA-mV cells were grown in S7 50 minimal medium at 30°C under shaking conditions to an OD 600 of 0.8. Afterwards, the cells were stressed with 1 M sorbitol, 1 M NaCl, 10% EtOH, 250 µg bacitracin or 10 µg nisin for 30 min at 42°C in a shaking bath for 30 min. Cells were spotted on coverslips (25 mm, Marienfeld) and covered with a 1% agarose pad (w/v), made of S7 50 medium and a smaller coverslip (12 mm, Marienfeld).

Epifluorescence Microscopy
A Zeiss Axio observer Z1 microscope with a Cascade II:512 Photometrics camera was used for imaging. The magnification of the objective was ×100, while the magnification of the optivar was ×1.6. The total magnification was ×160. The camera chip was 512 × 512 with a pixel size of 16 µm × 16 µm. Images were acquired using Metamorph, and acquisition was done with an exposure time of 3,000 ms using a YFP filter.

Single-Molecule Tracking, Data Acquisition, and Analysis
Imaging was performed with a Nikon Eclipse Ti microscope equipped with a high numerical aperture objective (CFI Apochromat TIRF 100XC Oil, NA 1.49), an EM-CCD camera (ImagEM X2, Hamamatsu) and a YFP filter set (BrightLine 500/24, Beamsplitter 520 and BrightLine 542/27). mVenus fluorophores were excited by the central part of a laser beam (TOPTICA Beam Smart, 515 nm, max. power 100 mW) with a laser intensity of 150 W cm -1 . Each movie consists of 3,000 frames and was recorded with an exposure time of 20 ms, using Nikon NIS-Elements BR.
First, the videos were cut with Fiji (ImageJ) [Schindelin et al., 2012] and the last 2,000 frames were used. Afterwards, the cell meshes were set with Oufti [Paintdakhi et al., 2016]. For particle detection, U-track [Jaqaman et al., 2008], MATLAB software, was used. Here, the minimal length of tracks was set to 5 and to link to points, no gaps for the particle detection was allowed. Data were analyzed using software SMTracker 2.0 [Oviedo-Bocanegra et al., 2021]. In this program, the Stationary Localization Analysis panel was used for dwell time analysis and heat maps, and the Square Displacement Analysis (SQD) panel for calculation of diffusion constants are population sizes and confinement maps. SQD uses the squared, 2-dimensional radial displacement of molecules, generating a cumulative distribution function (CDF) of squared. For a given number of diffusive states, the algorithm implemented in SQD performs a simultaneous nonlinear least square fit to the experimental CDFs. SMTracker estimates the most approximate number of diffusive groups found in the experimental data using the Bayesian information criterion, which also help to avoid overfitting of data. To obtain the uncertainty on the estimation of the fitting parameters, the main dataset is split into two subsets, train and test. Fitting estimation is run using only the training subset and evaluating onset of the test data. The training subset is methodologically split into 10 independent folds without any common data point, and the fitting is performed in each of these. Final estimation of the parameters is obtained after determining the mean, and the error is the sum of the standard error for the mean and the 95% confidence intervals obtained from the fitting procedure. Normalization is done according to standard procedures; sizes of bins in histograms were adjusted to yield the visually best representation of data.

Statement of Ethics
An ethics statement was not required for this study type, no human or animal subjects or materials were used.