a) and b) Schematic of nanopipette delivered triggering of TLR4 and Myddosome formation with LPS or Aβ aggregates. The LPS is delivered by nanopipette to the macrophage surface. When LPS binds to TLR4, it will trigger the dimerization of TLR4, which activates the recruitment of Mal, MyD88 and IRAK4&2 to form the Myddosome. Unstimulated TLR4 remains monomeric on the cell surface. c) Schematic of AFM cantilever light sheet 3D scan. The light sheet is reflected by an AFM cantilever to the target cell. Control by a piezo, the sample stage is moving up and down to achieve cell scanning.

Montage showing the time series of Myddosome assembly (small bright puncta) for different stimulations. a) PBS control shows no Myddosomes formed when PBS buffer is delivered. b) Aβ monomer control also showing no Myddosomes formed. c) Myddosome formation triggered by LPS, d) Myddosome formation triggered by sonicated Aβ fibrils. The scale bar is 5 μm.

a) 3D light sheet scanning of a live macrophage forming Myddosomes upon activation. The sheet scanning starts from the bottom surface of the cell and ends at the top surface. Z-stacks were acquired for each cell, consisting of 100 z-slices with 200 nm spacing. b) Images were rendered by 3D projection (brightest point method). Myddosomes could be visualised in the whole cell volume by rotating the 3D rendered imaged around the y-axis.

a) MyD88-YFP transduced iBMDMs were stimulated with 1 μg/mL LPS or 4 µM total monomer of sonicated Aβ fibrils delivered from a nanopipette for 30 mins. The timepoint when the first Myddosome formed was marked. (n = 77 across 5 independent experiments for each stimulation. The P-values are based on unpaired two-sided Student’s t-test.).

Histogram and cumulative distribution of MyD88 puncta lifetimes following nanopipette delivery of (a) LPS (1 μg/mL) or (b) sonicated Aβ fibrils (4 µM total monomer). c) Cumulative lifetime distributions after LPS and Aβ stimulation overlaid. The difference between the two lifetime distributions was significant (Kolmogorov-Smirnov test, p < 0.0001)

Normalised probability distributions plotted for a) LPS and b) Aβ stimulated Myddosomes. The histograms were fit assuming a 2-step degradation, with the rate constants k1 and k2, and mean lifetime < τ > calculated for each condition.

Total number of MyD88 puncta localized to the plasma membrane or cytoplasm, for Myddosomes which formed at the plasma membrane. Plots are normalised by cell number. a) LPS stimulation, b) Aβ stimulation. c) Overlay of total number of plasma membrane localised puncta, from membrane initialised trajectories after LPS and Aβ stimulation.

Example super-resolution images. Fixed macrophages stimulated by PBS (control), 100ng/ml LPS and 200nM sonicated Aβ fibrils. 488nm channel shows the MyD88-YFP signal. 641 nm channel shows the reconstructed super-resolved image of Alexa-647 conjugated anti-GFP antibody to represent the size of Myddosome. The merged channel shows signal overlap between the Myddosomes and Alexa-647 conjugated anti-GFP antibody. Only co-localized puncta (in yellow) correspond to real Myddosome signal. The scale bars are 5 μm (left figure) and 200 nm (right).

The size of Myddosomes triggered by LPS and sonicated Aβ fibrils at different times post triggering. MyD88-YFP transduced iBMDMs were stimulated with LPS (100 ng/mL) or sonicated Aβ fibrils (200 nM) followed by fixation and antibody labelling. FWHM of puncta were measured by 1D Gaussian fitting. (From 30mins to 24hrs: n = 212, 140, 99,77, 85 across 3 independent experiments for each stimulation. The P-values are based on unpaired two-sided Student’s t-test.)

a) Distribution of super-resolved MyD88 puncta shape factors after 30 mins LPS stimulation (, where A and p are area and perimeter respectively) n = 1 86 across independent experiments. Similar distribution is also observed for mins Aβ stimulation. b) Plot of super-resolved cluster area vs cluster shape factor for 30 mins LPS stimulations suggests that this corresponds to a populations of very small, spherical MyD88 clusters. Similar relationship was also observed for mins Aβ stimulation. n = 1 86 across 3 independent experiments. c) Magnified examples of 3 super-resolved MyD88 clusters with a range of different shape factors. Scale bar is 150 nm d) Plotting fractions of spherical clusters (shape factor = 1) compared to total puncta suggests a drop in the population of small, spherical MyD88 clusters after mins for both LPS and Aβ stimulation. There is a smaller fraction of these small spherical clusters following Aβ stimulation compared to LPS.

TEM images of amyloid-β 42 following sonication of incubated fibrils. Scale bar: 100 nm. Sonicated sample showed large population of sub-100 nm oligomers in background of image (as well as some larger fibrils).

(a) Nanopipette approaches the cell where the tip is 3 µm away from the membrane with an angle to delivery sonicated Aβ fibrils. Scale bar is 8.5 µm. (b) Projected view of tagged sonicated Aβ fibrils (red) after delivery to the cell membrane (green). Scale bar is 5 µm.

Cells were fixed following LPS triggering, and then imaging using the same parameters and microscope as previously described (Materials and Methods : Live cell scanning and 3d reconstruction). a) Time lapse of fixed cell following LPS stimulation. b) Representative Intensity trace of one puncta out of 14 in one cell with less than 40% of the puncta intensity lost over 30mins. c) Loss in spot detection due to photobleaching in fixed cells. Within the first 500s, there is a 6% loss of detected puncta; in comparison to the live cell stimulation (figure 5), there was an approximate 90% drop in the detected puncta for newly assembled puncta within the same time period, suggesting that signal loss during live cell imaging was mostly due to MyD88 puncta degradation rather than photobleaching.

Overview of 2D tracking analysis. From left to right: z-stacks for each time frame were projected into one frame. ImageJ’s find maxima plugin was used to identify MyD88 puncta which were subsequently tracked using a custom MATLAB code. Analysis was then carried out on identified trajectories.

Comparison of fitting different degradation kinetics to lifetime distributions after stimulation. Optimal fits at both long and short lifetimes (for both LPS and Aβ) stimulation were observed assuming that the degradation of Myddosomes was a two-step process, with different rate constants for each step (top row) – less optimal fits were obtained for a two-step process assuming the same rate constants for each step, and a one-step degradation.