2.1. Mixed glial cultures for imaging

Neonatal mice (postnatal day 0–2) were sacrificed using Home Office-approved methods (proscribed by the Animals Act, 1986). Brains were dissected out and submerged in ice-cold dissection medium, consisting of DMEM:F12 (D6421, Sigma), 1% L-glutamine (A2916801, Gibco), and 1% Pen Strep (15140122, Gibco). The cortices of the brains were separated from the midbrain and cerebellum. The meninges were removed using fine forceps, and the brains were homogenized by trituration followed by digestion in trypsin (25200056, Gibco) and DNase (EN0521, Thermo Scientific) for 20 minutes at 37°C. Cells were diluted in DMEM (41965039, Gibco) with 10% FBS (F7524, Sigma), 1% Pen Strep (15070, Thermofisher) (serum medium), and centrifuged at 300 × g for 5 minutes at 10°C. Following resuspension of the pellet cells were seeded into poly-d-lysine coated imaging dishes (P35GC-0-14-C, Mattek) at a density of 5 × 10⁵ cells. The cultures were then incubated at 37°C/8% CO₂ for 16 hours to allow the cells to settle, followed by a complete media change. Cells were then cultured for 9–12 days, with a ⅔ media change (supplemented with 3 μg/mL insulin) performed every third day. Mixed glial cultures were used for imaging at 15 DIV.

2.2. OPC culture

Purified OPC cultures were isolated from mixed glial cells using the methods described above. Following centrifugation and resuspension, cell from two brains were plated into a poly-l-lysine (P9155, Sigma) coated T75 flask and cultured in serum medium at 37°C/8% CO_2. After 24 hours flasks received a full medium change with serum medium supplemented with insulin (2 μg/mL) (I5500, Sigma), with subsequent ⅔ media changes scheduled every third day (supplemented with 3 μg/mL insulin). Once an astrocyte monolayer had formed, and the proliferation of putative OPCs (appearing as small rounded profiles) confirmed, the flasks were secured on an orbital shaker mounted inside a humidified incubator (37°C/5% CO₂). The flasks were then shaken at 160 rpm for 30 minutes, after which the media was changed to remove microglia, and the flasks returned to the incubator and shaken for a further 16 hours at 220 rpm (37°C/5% CO₂) to dislodge OPC from the astrocyte monolayer. OPC were then isolated from contaminating microglia based on a differential adhesion protocol. Here, the resulting cell suspension containing OPC and microglia was transferred into a noncoated 10 cm dish, and left to incubate at 37°C for 30 min with gentle swirling every 10 min to prevent adhesion of OPCs. The cell suspension containing OPC was then collected and centrifuged at 1000 rpm for 5 minutes at 10.C. The pellet was resuspended in OPC media, containing DMEM, 1% N2 (17502001, Thermofisher), 1% B27 (17504, Thermofisher), 2 nM l-Glutamine, 1% Pen Strep, 20 ng/ml rhFGF2 (PHG0266, Thermofisher), 100 ng/ml IGF1 (PHG0078, Thermofisher), 20 ng/ml PDGF-AA (PHG0035, Thermofisher), 3.6 ng/ml Hydrocortisone (H0888, Sigma), and plated in imaging dishes at a density of 5 × 10⁵ cells.

2.3. OPC differentiation

In order to differentiate OPC, the cells were exposed to OPC media lacking fibroblast growth factor 2 (rhFGF2), PDGF-AA, and Insulin-like growth factor 1 (IGF1), and supplemented with Triiodothyronine (T3) (T6397, Sigma) (referred to as differentiation media). Cells were then maintained with media changes every other day for 1 week, and their differentiation status examined by immunofluorescent analysis of the OPC maker protein NG2. NG2 Immunostaining, performed according to methods described by Begum, Otsu (Reference Begum, Otsu, Ahmed, Ahmed, Stevens and Fulton31), revealed a robust NG2 signal in OPC cultures that was undetectable following differentiation treatment (data not shown) confirming the efficacy of this differentiation protocol.

The mature OLs observed were split into two categories, eOL and mOL. eOL were observed as multipolar cells with processes bearing small regions of flattened membrane similar to the membrane bubbles previously identified during the early stages of myelination.⁽Reference Ioannidou, Anderson, Strachan, Edgar and Barnett³²⁾ Importantly, these flattened membrane bubbles were quite distinct from the continuous areas of flattened myelin, often termed myelin sheets, often observed to surround mature myelin-forming OL in vitro. ⁽Reference Asakura, Miller, Pease and Rodriguez^33, Reference Barateiro and Fernandes³⁴⁾ In this article, we refer to the flattened membrane regions produced by eOL as membrane bubbles (Figure 1). For imaging purposes, the analysis of eOLs focussed exclusively on these regions. In our hands differentiation treatment failed to convert OPC beyond the eOL stage thus it was not possible to study myelin sheets using this approach. Therefore, to study the continuous membrane sheets associated with compact myelin we imaged mOL in mixed glial cultures, where mOL were frequently observed to exhibit the “fried-egg” morphology of membrane sheets associated with fully mature myelin in vitro (Figure 1).

Figure 1. MBP::Dendra2 expression and motion profiles in oligodendroglial cells at distinct stages of maturation. (a) MBP::expression in OPC processes (a-i), eOL membrane bubbles (a-ii), and mOL sheets (a-iii). Upper panels depict representative cell morphologies with imaging regions of interest highlighted by light blue circles. Green areas in (a-ii) and (a-iii) indicate putative regions of myelination. Lower panels display representative MBP::Dendra2 signals from ROIs depicted in upper panels. Scale bars 5 μm. (b) Motion Profiles for MBP particles imaged in OPC processes (b-i), eOL membrane bubbles (b-ii), and mOL membrane sheets (b-iii). Note, all particle tracks are presented for a given imaging field to provide an overview of directionality at each developmental stage. Scale bars 1 μm. (c) Vector diagrams derived from the motion profiles shown in (b-i) (OPC), (b-ii) (eOL), and (b-iii) (mOL). Thicker arrows indicate vectors with greatest particle density.

2.4. Viral induction

In order to visualize MBP in living cells a recombinant Semliki Forest Virus (SFV) vector of subtype A7(74) (SFVA7(74)) was used to transduce OL with a Dendra2::MBP construct. cDNA encoding Dendra2 was obtained from Dr Peter Dedecker (Ku Leven, Leuven, Belgium). Sequences for MBP were provided by Professor Mikael Simons (Institute of Neuronal Cell Biology, Technical University Munich, Munich, Germany), and cDNA for the SFVA7(74) vector were obtained from Markus Ehrengruber (Kantonsschule Hohe Promenade, Zurich). The Dendra2::MBP construct was assembled by standard PCR cloning methods and ligated into the SFVA7(74) vector. We note that our new Dendra2::MBP construct lacked the spacer region used by the Simons group in their studies of MBP diffusive mobility.⁽Reference Aggarwal, Snaidero, Pähler, Frey, Sánchez, Zweckstetter, Janshoff, Schneider, Weil, Schaap, Görlich and Simons⁶⁾ Infectious particles were generated from the resulting SFVA7(74)-Dendra2::MBP vector in BHK cells according to methods reported by Ehrengruber and Schlesinger (Reference Ehrengruber, Schlesinger and Lundstrom35). Unpurified viral preparations were then aliquoted and stored at −80°C. SFVA7(74)-Dendra2::MBP was applied to cells when more than 20% of putative oligodendroglia possessed processes approximately 1.5 times the length of the soma on visual inspection. The cells were submerged in viral solution containing 60,000 infectious units/ml and left to incubate for 1 hour at 37°C. Following the incubation, the viral solution was washed off and the cells were incubated in media lacking phenol red for 6 hours at 37°C to allow sufficient time for transgene expression, after which time the cells were used for imaging.

2.5. Imaging

Regions of interest were centered on processes, membrane bubbles, and sheets since these are the key region in terms of MBP function. Imaging dishes containing the SFV-infected cells were imaged on an inverted wide-field microscope (Rapid Automated Modular Microscope, Applied Scientific Instrumentation) equipped with a 100× (1.4 NA) oil immersion TIRF objective (Nikon). The cells were first surveyed under bright-field to find cells with healthy-appearing processes, after which the microscope was converted to TIRF mode for OPC and eOL in purified cultures, and HILO mode for mOL in mixed glia cultures. Upon discovery of a process, the cells were imaged at 488 nm excitation to confirm the expression of Dendra2 in the selected process. The fluorescent cells were then primarily irradiated with 405 nm light for 5 seconds to induce Dendra2 photoswitching from a 488 nm excitation to 561 nm excitation, producing a spatially sparse red signal, which could be tracked (Figure 1). Time-lapse imaging was performed on the selected region of interest with excitation from a 561 nm laser line whose power was carefully adjusted to achieve effective excitation whilst minimizing photobleaching. Images were acquired using an Evolve Delta EM-CCD camera (Photometrics) set to 30 ms exposure times across 2500 frames. In the case of exceptionally sparse signals, or rapidly photobleaching samples, an image was taken at 488 nm to produce a reference image to compare the detected particles to, thus ensuring the resulting particle tracks were localized within a bona fide process, rather than existing as free-floating material within the dish. Following imaging the resulting time-lapse image-stacks were processed using ICY bioanalysis software.⁽Reference de Chaumont, Dallongeville, Chenouard, Hervé, Pop, Provoost, Meas-Yedid, Pankajakshan, Lecomte, Le, Lagache, Dufour and Olivo-Marin³⁶⁾

2.6. Data analysis

Image stacks were imported into ICY bioanalysis for particle detection and tracking.⁽Reference de Chaumont, Dallongeville, Chenouard, Hervé, Pop, Provoost, Meas-Yedid, Pankajakshan, Lecomte, Le, Lagache, Dufour and Olivo-Marin^36–Reference Chenouard, Smal, de Chaumont, Maska, Sbalzarini, Gon, Cardinale, Carthel, Coraluppi, Winter, Andrew, William, Rohr, Kalaidzidis, Liang, Duncan, Shen, Magnusson, Jalden, Paul-Gilloteaux, Roudot, Kervrann, Waharte, Tinevez, Willemse, Celler, Dan, Tsai, De Solorzano, Olivo-Marin and Meijering³⁸⁾ The images were opened and the contrast within the images adjusted to aid the detection of the processes/sheet by eye. Regions of interest were then constructed around the processes/sheet of the cell within the first image of the stack. The region of interest was verified with the other images within the stack to ensure it encompasses the process throughout the stack, and adjusted to ensure the region of interest was tightly confined to the process of the cell (Figure 1). In cases where the process moved during the imaging session, the longer segment where the process was stationary was taken forward for imaging, or the images were excluded from analysis. This region of interest was then connected to the Spot Detector function, which was set to expect detected particles with an approximate diameter of 3 pixels on a dark background, an assumption based on previous findings.⁽Reference Olivo-Marin³⁹⁾ Following particle detection, the detected particles were ported to the Particle Tracker and analyzed using the parameters shown in Table 1. Using the Particle Tracker, detected particles were compared to each other and the most likely particle track calculated for each detection.⁽Reference Chenouard, Bloch and Olivo-Marin³⁷⁾ Detected particle tracks were then filtered to remove exceptionally short tracks (< 10 frames) deemed unlikely to provide useful data, and verified by eye to remove any tracks which were not bound within the confines of the cell, resulting in between 650 and1500 tracks per process/sheet.⁽Reference Meijering, Dzyubachyk and Smal⁴⁰⁾ The precision of this tracking procedure was calculated from a series of 20 frames capturing stationary fluorescent particles. Tracking errors for X and Y coordinates were 23 (± 27) nm and 16 (± 17) nm, respectively for the TIRF imaging mode, and 21 (± 16) nm and 19 (± 16) nm, respectively, for images acquired in HILO mode. Thus, tracking errors were comparable between the TIRF and HILO configurations.

Table 1. ICY software tracking parameters used within this study

Note: The parameters shown here were optimized on the oli-neu OL cell line and were utilized for all cells used within this study.⁽Reference Begum, Otsu, Ahmed, Ahmed, Stevens and Fulton³¹⁾

Data from tracks were extracted in ICY bioanalysis using the Motion Profiler, Mean Squared Displacement, and Export to XLS function plugins. These operations provided the mean squared displacements (MSD) for each track (the average squared movement of a particle between two points), total displacement (summed movement over the entire track), the net displacement (straight line movement between the start and end points of the track), the track lengths, and the raw track data.⁽Reference de Chaumont, Dallongeville, Chenouard, Hervé, Pop, Provoost, Meas-Yedid, Pankajakshan, Lecomte, Le, Lagache, Dufour and Olivo-Marin^36, Reference Chenouard, Smal, de Chaumont, Maska, Sbalzarini, Gon, Cardinale, Carthel, Coraluppi, Winter, Andrew, William, Rohr, Kalaidzidis, Liang, Duncan, Shen, Magnusson, Jalden, Paul-Gilloteaux, Roudot, Kervrann, Waharte, Tinevez, Willemse, Celler, Dan, Tsai, De Solorzano, Olivo-Marin and Meijering³⁸⁾ The diffusion coefficient was calculated from MSD values using (1):

(1) $$ D=\frac{<\mathrm{MSD}>}{qit}, $$

where D is the diffusion rate, and MSD is divided by the time (t) multiplied by the dimensionality constant (q_i). In this case, q_i is defined as “4” following the example of Ruthardt et al.⁽Reference Ruthardt, Lamb and Brauchle⁴¹⁾ for 2-dimensional imaging. We note that a single value of D was calculated for each track by averaging the values of D obtained from all available time intervals in the track. Although this approach provides a means to control bias caused by variability in the length of the captured tracks, we acknowledge it leaves open the possibility that D is over or underestimated for particles whose trajectories deviate from the linear MSD/q_it relationship expected by the equation, as would be the case for particles with restricted or enhanced diffusion. In this context, MSD versus time plots appear to show superdiffusive particle tracks within OPC processes and eOL bubbles, while the diffusion of particle tracks in mOL sheets appear relatively restricted (Supplementary Figure 1). These apparent deviations from normal diffusion may have introduced a degree of imprecision to our calculations of the diffusion coefficient. These limitations, and are our rational for selecting this equation, are considered in more detail in Section 4.

Equation (1) also assumes that particle tracking occurs in the absence of localization errors. As noted above, the localization methods used in this study are associated with a certain degree of error which will influence the diffusion coefficients. However, these localization errors are similar for both TIRF and HILO imaging modes, thus any imprecision in our calculations will have been equally distributed across the different cell types.

The confinement ratio was calculated using the track length and displacement values using (2):

(2) $$ r\mathrm{con}=\frac{d\mathrm{net}}{d\mathrm{tot}}\surd t\left(\mathrm{track}\right), $$

where net displacement d _net is divided by the total displacement d _tot, then multiplied by track duration t _(track) to correct for the influence of track length as discussed by Beltman et al.⁽Reference Beltman, Marée and de Boer⁴²⁾

The circularity ratio for each motion profile was calculated by tracing its outline and using the resulting polygon to determine values of area and perimeter. The circularity ratio was then calculated with (3):

(3) $$ r\mathrm{circ}=\frac{4\pi A}{p^2}, $$

where four multiplied by pi and the area (A) is divided by the perimeter (p) squared to yield a circularity ratio (r _circ) with a value between 0 and 1, with 0 being linear and 1 representing perfect circularity. A single motion profile was constructed for each imaged cell based on the motion of all of the particle tracks, hence circularity was assessed at the level of individual cells rather than the individual particle.

To aid the qualitative interpretation of these data, confinement ratios and diffusion coefficients were visualized as frequency distributions (% of particle population), with bins further categorized into color-coded groups. Confinement ratios naturally fell between 0 and 1, hence a set of 10 evenly sized frequency bins were selected. To determine the range of particle diffusion coefficients, all diffusion coefficient values from all conditions were combined to yield the range (9.34E-8 to 7.00E-2 μm²/s). These values were then divided between 12 equally spaced frequency bins. Confinement ratio frequency bins were categorized as being either light, medium, or highly confined. Similarly, diffusion coefficients were categorized into slow, intermediate, and fast diffusive particles, and the resulting groups are color coded in relevant histograms (Figures 3 and 5). These categorizations provide a simple aid to visualization and qualitative analysis of the data, but are not used on subsequent statistical analysis. The vast majority of MBP particles fell within the slowest diffusion bin (5.83E-04 μm²/s) (Supplementary Figure 2). Therefore, to aid visualization of faster diffusive particles, frequency distributions were graphed from the population falling within the remaining 11 bins. Nevertheless, statistical analysis for diffusion coefficients was computed on the entire population of particles.

2.7. Statistical analysis

Analysis focussed on comparing particle track circularity, particle confinement, and particle diffusion between the three stages of OL development: OPC processes, eOL membrane bubbles, and mOL sheets. Individual particle tracks were treated as biological replicates for analyses examining the behavior of individual particles (confinement and diffusion). For exploration of circularity, and for cell averaged analysis of particle confinement and diffusion (as noted in the text), individual cells were treated as biological replicates. Particle tracks were obtained from a total of 39 OPC processes, 23 eOL bubbles, and 27 mOL membrane sheets. Variance in the number of cells imaged for each cell type reflect the relative difficulty of generating and identifying cells at each maturational stage. A total of 24,721 OPC, 6,590 eOL, and 65,012 mOL MBP particle tracks were analyzed to assess particle confinement. For the diffusion analysis, a total of 32,320 OPC, 18,954 eOL, and 51,032 mOL particle tracks were analyzed. All statistical tests were computed with Prism 9.0 (Graphpad Software, El Camino Real, CA). Kolmogorov–Smirnov normality tests revealed that none of the data sets (particle track circularity, particle confinement, and particle diffusion, at all developmental stages) contained normal distributions. Therefore, nonparametric tests were used throughout the study and sample averages are reported as medians plus interquartile range (IQR, first to third quartiles). Multiple group comparisons (circularity data, confinement data) were investigated by the Kruskal–Wallis (KW) tests, and between-group differences checked by Dunn’s multiple comparisons test. For these tests, an alpha level of <0.05 was used for the detection of reliable differences. Frequency distributions for confinement and diffusion data were compared between pairs of groups by the Kolmogorov–Smirnov (KS) test for goodness of fit (e.g., OPC particle confinement data comparisons: OPC processes versus eOL bubbles; OPC processes versus mOL sheets). To correct for these multiple-group comparisons Bonferonni corrections were applied to alpha values as follows: original alpha value divided by the number of tests carried out on each sample (= 2), for example, original alpha p < 0.05 corrected to p < 0.025. Prism 9.0 software produced approximate p values for these KS tests, for example, p < 0.0001. Approximate p values were therefore referenced against the closest relevant Bonferroni corrected p alpha level, for example, observed p < 0.0001 referenced against p alpha 0.001 (corrected to p < 0.0005). Table 2 summarizes descriptive statistics (medians, IQR, 95% confidence limits for medians), and Table 3 summarizes statistical hypothesis tests for the data sets presented in the study.

Table 2. Summary of descriptive statistics aggregated from individual cells

Note: ID codes link table entries to in-text statistical results. Diffusion coefficient medians are expressed in μm²/s

Table 3. Summary of statistical tests

Note: ID codes link table entries to in-text statistical results. Diffusion coefficient medians are expressed in μm²/s.

Abbreviations: BF, Bonferroni correction for two comparisons; CL, confidence limits; D, Kolmogorov Smirnov goodness of fit test statistic; DF, degrees of freedom; KS, Kolmogorov Smirnov goodness of fit test; KW, Kruskal Wallace test statistic.