Spatial EGFR Dynamics and Metastatic Phenotypes Modulated by Upregulated EphB2 and Src Pathways in Advanced Prostate Cancer

Advanced prostate cancer is a very heterogeneous disease reflecting in diverse regulations of oncogenic signaling pathways. Aberrant spatial dynamics of epidermal growth factor receptor (EGFR) promote their dimerization and clustering, leading to constitutive activation in oncogenesis. The EphB2 and Src signaling pathways are associated with the reorganization of the cytoskeleton leading to malignancy, but their roles in regulating EGFR dynamics and activation are scarcely reported. Using single-particle tracking techniques, we found that highly phosphorylated EGFR in the advanced prostate cancer cell line, PC3, was associated with higher EGFR diffusivity, as compared with LNCaP and less aggressive DU145. The increased EGFR activation and biophysical dynamics were consistent with high proliferation, migration, and invasion. After performing single-cell RNA-seq on prostate cancer cell lines and circulating tumor cells from patients, we identified that upregulated gene expression in the EphB2 and Src pathways are associated with advanced malignancy. After dasatinib treatment or siRNA knockdowns of EphB2 or Src, the PC3 cells exhibited significantly lower EGFR dynamics, cell motility, and invasion. Partial inhibitory effects were also found in DU145 cells. The upregulation of parts of the EphB2 and Src pathways also predicts poor prognosis in the prostate cancer patient cohort of The Cancer Genome Atlas. Our results provide evidence that overexpression of the EphB2 and Src signaling pathways regulate EGFR dynamics and cellular aggressiveness in some advanced prostate cancer cells.


Contents
Method S1 | 2D single-particle tracking ..  Method S1 | 2D single-particle tracking Single-particle trajectories were determined from the raw data sets using a three-step process: (i) Identifying contiguous regions of pixels; (ii) Gaussian fitting; (iii) building trajectories from coordinates. This approach is similar to those described previously[1-3].

Identifying contiguous regions of pixels.
A series of 2D images in time trace was processed independently to find FN-IgG-EGFR coordinates. The contiguous regions of pixels, which represent the images of fluorescent particles, were identified on the basis of two criteria: (i) pixels had intensities greater than 3-fold the standard deviation of pixel intensities from areas defined as background (background offset algorithm [4]) and (ii) pixels were above a threshold [4]. Then, a high pass filtering was applied to the image with a 2D Gaussian filter (σ = 5). The binary image of pixels passing both criteria was later processed by Gaussian fitting.

Gaussian fitting.
To find out the center of the fluorescent particles, the center of mass of each contiguous region in the binary image was set as the starting point in a Gaussian fitting routine. The highest intensity pixel in a small region around the starting point (5 pixels square) was used as an updated starting. Fits were performed in a square region, of size ~ 2*σpsf, around the updated starting point. The σpsf defines the size of 2D Gaussian approximation to the point spread function. After convergence of the fitting routine (a change in location of fewer than 10 -5 pixels), a normalized cross-correlation was calculated to verify the 2D Gaussian-fitted coordinates. The found coordinates were only considered as positions of FN-IgG-EGFR and used in the further analysis if they exceeded a cross-correlation value of 0.7.

3.
Building trajectories from coordinates. The probability of finding a diffusing particle with diffusivity D in two dimensions at a distance greater than d from its starting point after a time Δt is given by [5] ( , ∆ ) = ∆ (1) Trajectories were built from the set of 3D coordinates (x,y, and t) in two steps. First, coordinates identified at time t were compared with coordinates at time t+Δt using Eq. 1 where Δt is the inverse frame rate of data acquisition. If P(r, Δt) was found to be greater than .05, the coordinate at t+Δt is associated with the coordinate at t in a trajectory. This process builds short, un-interrupted trajectories. Second, to connect these short trajectories originated from the same targets, the end coordinate of all trajectories is compared with all later starting coordinates of other trajectories using Eq. 1, where Δt is now the time interval between the end of the first trajectory and the beginning of the second. The later trajectory with the smallest Δt that has a P(r, Δt )> 0.01 is connected with the first trajectory. This process is continued until there are no remaining pairs of trajectories that satisfy the criteria. The reconstructed trajectories are further processed into mean-squared displacement to estimate diffusion coefficient.

Method S2 | Extracting dynamic parameters from MSD
The typical approach to analyze a single-particle trajectory starts with the calculation of mean-squared displacement (MSD) [6,7], which describes the average squared distance (d 2 , r is the position vector) that the particle has explored in space at a given time lag (Δt): For L, the MSD curves were fitted with an equation for confined diffusion [8,9], and L was defined as the MSD fitting result of the first 10 MSD points: Confined diffusion is featured by an abrupt change of slope in the MSD curve [10] after a characteristic equilibration time τ.
Method S3 | 3D single-particle tracking TSUNAMI (Tracking of Single particles Using Nonlinear And Multiplexed Illumination) is a feedbackcontrol tracking system which employs a spatiotemporally multiplexed two-photon excitation and temporally demultiplexed detection scheme. Sub-millisecond temporal resolution (under high signal-tonoise conditions) and sub-diffraction tracking precision in all three dimensions have been previously demonstrated [11][12][13]. Tracking can be performed in a live cell to monitor the movements of fluorescent nanoparticle-tagged EGFRs [13,14] or ballistically injected fluorescent nanoparticles [15]. The TSUNAMI microscope has been described in detail in our previous study 10. In brief, excitation of 800 nm from a

Method S5 | Data processing for 3D single-particle tracking and 2P scanning imaging
All data processing was performed in MATLAB (Mathworks). Saved in a binary format, the trajectory raw data contained photon counts and voltage outputs from the actuators (i.e. the xy scanning galvo mirrors (6125H, Cambridge Technology) and the objective z-piezo stage (P-726 PIFOC, PI)) at each 5 ms time point. Trajectories were plotted by simply connecting particle positions of consecutive time points.
2P-LSM raw images were read into MATLAB from binary files and denoised with a median filter before a 1D interpolation along the z dimension. To segment the cellular compartments, we used a simple intensity threshold technique that converts the image to a binary. Thresholds were selected at each z plane to account for variation in noise and brightness through the z-stack. The binary images were used as a mask to plot the cell isocontour. Because trajectories were measured with the same analog output device as the 2P-LSM images they can be directly overlaid with the cell compartment isocontour with no conversion or scaling required.

Method S6 | 3D Inward movement analysis
The 3D inward movement analysis is adapted from Picco's 2D inward movement analysis [16], and it has been used to analyze the anti-PDL1 antibody-induced PD-1 internalization [17]. The 3D inward movement analysis is a two-step procedure based on the 2P scanning imaging and 3D single-particle tracking. Firstly, the 3D-stacked live-cell image was acquired by 2P-LSM. Secondly, the fluorescently labeled EGFR was tracked by the TSUNAMI microscope. Then the EGFR trajectory was coregistered with the 3D cell image. The distance and velocity of the inward movement were then derived from the collected trajectories.
The 3D inward movement analysis is a two-step procedure based on the 2P scanning imaging and   The error bar represents the standard error of the mean from 9-20 measurements. All statistical analysis was performed using the unpaired t-test. The asterisk represents the level of statistical significance for ttest: *** p < 0.001.  Figure S4 and Figure S10.