Images acquired by TIRF microscopy were exported as OME-TIF files. Image processing and analysis were performed using Python (v3.9.7).
To correct for uneven irradiation by the laser, images were first processed using custom site package impy (
https://github.com/hanjinliu/impy; v2.1.1) as follows. For each channel, all acquired 4D image stacks (stage position, time frame, and XY) were projected along the axes of stage positions and frames by the median to create a 2D background image. This background image was fit to a 2D diagonal Gaussian function:
where
is the peak background intensity,
and
are the peak positions,
and
are the variations of the Gaussian function, and
is a constant. Subsequently, for each separate image slice, intensity values
were corrected by:
to extract the true fluorescence intensity. The corrected image
was averaged by the second to fifth frames and saved as 32-bit floating image stacks in TIF format. The first frame was not used for the analysis because some of them were not in focus because of the short buffering time after the stage movement.
To remove leakage of the fluorescence intensity between channels, blue-to-green, green-to-blue, red-to-green, and green-to-red leakage ratios were determined as follows. The green channel mean intensity along Alexa Fluor 488–labeled microtubules and Cy5-labeled microtubules were measured without D2-mCherry to calculate the blue-to-green and red-to-green leakage. Microtubules saturated with D2 (microtubules equilibrated with 4 μM D2 and 20 μM taxol) were used to measure the blue channel mean intensity along Cy5-labeled microtubules and red channel mean intensity along Alexa Fluor 488–labeled microtubules to calculate green-to-blue and green-to-red leakage. Leakage-corrected images were saved as 32-bit floating image stacks in TIF format.
Average fluorescence intensities along microtubules were quantified using Python image viewer napari (v0.4.16; Sofroniew et al, 2022 (
link)) and custom plugin napari-filaments (
https://github.com/hanjinliu/napari-filaments; v0.2.1). Briefly, fluorescence intensities (Alexa Fluor 488 channel or Cy5 channel) of the non-overlapping regions of microtubules were manually selected and fit to 2D spline curves. For the spline of length
,
sample points were placed at equal intervals including the edges (i.e., placed for every ∼1 pixel), and the intensities were interpolated by cubic interpolation. Average fluorescence intensities were calculated as the average of all interpolated intensities. All statistical tests were calculated using SciPy (v1.7.3), statsmodel (v0.13.2; Seabold & Perktold, 2010 (
link)), or scikit-posthocs (v0.7.0; Terpilowski, 2019 (
link)).
Liu H, & Shima T. (2023). Preference of CAMSAP3 for expanded microtubule lattice contributes to stabilization of the minus end. Life Science Alliance, 6(5), e202201714.