Extended Data Fig. 7: Reconstruction with estimated OPLD in beads, fixed cells, and live cells.

a, 4Pi-SIM image of 100-nm fluorescent beads. b, The x-z cross-sections along the dashed line in a, comparing the reconstruction results with an OPLD of 0 nm (upper) and an estimated OPLD of -160 nm (lower). c, Intensity profiles for a single bead reconstructed with OPLDs of 0 nm (blue, triangle) and -160 nm (red, circle) along the magenta dashed line in b, showing ringing artifacts when using an OTF with a mismatched OPLD. d, 4Pi-SIM image (maximum intensity projection) of immunolabeled ER in a COS-7 cell. e, The x-z cross-sections along the dashed line in d, comparing the reconstruction results with an OPLD of 0 nm (upper) and an estimated OPLD of -180 nm (lower). f, Intensity profiles of the single-layer ER indicated by the magenta dashed line in e, contrasting reconstruction with OPLDs of 0 nm (blue, triangle) and -180 nm (red, circle). g, 4Pi-SIM image (maximum intensity projection) of actin filaments in a live COS-7 cell. h, The estimated OPLDs varied over time, with green arrows marking the specific moments for i and j, alongside their estimated OPLDs. i, Comparative reconstructed images, with and without estimated OPLD, from the zoomed area within the yellow dashed box in g. The top panels show the x-y time-lapse images. The middle panels show the x-z cross-sections of reconstruction without estimated OPLD, corresponding to the dashed line marking the same actin filament. The bottom panels show x-z cross-sections of reconstruction with estimated OPLDs, demonstrating that our algorithm can effectively remove these axial artifacts. j, Comparative reconstructed images, with and without OPLD adjustment, from the magnified area within the magenta dashed box in g. The images from top to bottom represent time-lapse x-y cross-sections, and x-z cross-sections reconstructed without and with estimated OPLD, emphasizing the importance of OPLD-corrected reconstruction in distinguishing real signals from potential artifacts. All slices are 30 nm thick. Scale bars: 2 µm (a, d, e, g) and 1 µm (b, i, j). a.u., arbitrary units. Representative results are shown from three to six independent experiments.