Inducible fluorescent speckle microscopy

Abstract

The understanding of cytoskeleton dynamics has benefited from the capacity to generate fluorescent fiducial marks on cytoskeleton components. Here we show that light-induced imprinting of three-dimensional (3D) fluorescent speckles significantly improves speckle signal and contrast relative to classic (random) fluorescent speckle microscopy. We predict theoretically that speckle imprinting using photobleaching is optimal when the laser energy and fluorophore responsivity are related by the golden ratio. This relation, which we confirm experimentally, translates into a 40% remaining signal after speckle imprinting and provides a rule of thumb in selecting the laser power required to optimally prepare the sample for imaging. This inducible speckle imaging (ISI) technique allows 3D speckle microscopy to be performed in readily available libraries of cell lines or primary tissues expressing fluorescent proteins and does not preclude conventional imaging before speckle imaging. As a proof of concept, we use ISI to measure metaphase spindle microtubule poleward flux in primary cells and explore a scaling relation connecting microtubule flux to metaphase duration.

Figure 1.

Principle and proof of inducible speckle imaging. (a) A single transverse mode laser beam is scattered by a diffuser before the microscope objective, creating a pattern composed of speckles with a characteristic size (correlation length) at the diffraction limit. This high-contrast pattern is generated at all planes: at, before, and after the objective’s focal plane. The beam induces a heterogeneous photoswitch pattern across the sample. (b) A PSF volume comprises a large number of potentially tagged slots. In FSM, slots are randomly occupied, leading to faint density fluctuations at the PSF scale. Increased fluctuations can be induced by narrowing the PSF or, conversely, by clustering fluorescent molecules, as does ISI, to emulate PSF-sized slots. (c) Comparison between simulated and observed fluorescence pattern structure after an ISI pulse on a uniform fluorescent layer (concanavalin A–conjugated Alexa Fluor 488) at different bleaching strengths (γ). Bars, 2 µm. (d) ISI on GFP–α-tubulin in a fixed Drosophila S2 mitotic cell. To produce a roughly uniform fluorophore pool, microtubules were depolymerized using colchicine. Bars: 5 µm; (inset) 1 µm.

Figure 2.

ISI imprinting and FSM on the mitotic spindle. (a) Photoconversion ISI mode in a human transformed cell line (U2OS) and photobleaching ISI mode in a Drosophila S2 cell. Bar, 5 µm. (b) Orthogonal views of a metaphase Drosophila S2 cell before and after ISI show the 3D character of the imprinted speckles. Bar, 5 µm. (c) An FSM and an ISI-prepared metaphase cell. ISI cells (n_ISI = 20) displayed an approximately sixfold increase in speckle amplitude compared with FSM cells (n_FSM = 20, chosen and imaged by an independent experimenter) generated by expression of leaky GFP-tubulin levels in the absence of the promotor. Error bars represent SD. ADU, analog-digital units. Bar, 5 µm. (d) Post- to pre-ISI fluorescence signal ratio image in pseudocolor in a Drosophila S2 cell. ISI-induced fluorescence fluctuations, but not mean levels, are dependent on the dynamics of the structure to which the fluorophore is bound. Cytoplasmic pool dynamics induces a fast wash out of the speckle pattern already at the first post-ISI frame (red box). Bar, 2 µm.

Figure 3.

Optimal speckle generation. (a) In a uniform fluorescent layer, increasing γ leads initially to a spreading of the histogram, resulting in increasingly visible speckles. For γ ⪢ 1, the tendency is reversed, ultimately driving all fluorophores toward the bleached state. Labels in the bottom row correspond to laser exposure time in seconds. (b) Experimental data for the dependence of speckle amplitude on γ, obtained by performing the experiment outlined in a. Dashed curve corresponds to fitting of the model curve (Eq. 3) using the theoretically calculated baseline for the loss of contrast factor caused by imaging, M_img = 2.12, and the instrumental factor, M_instr.\, which arises as the single-fitting parameter in the model. Both the shape and the peak position are intrinsic to the model and not dependent on fitting. The solid curve applies a γ-dependent correction to M_img, as obtained by computer simulation to correct for subdiffraction leakage, which becomes relevant for γ ⪢ 1 (Fig. S4 d and Materials and methods).

Figure 4.

Speckle contrast fadeout is a measure of spindle microtubule turnover. (a) The ISI pulse induces clustering of the fluorescent molecules, increasing contrast. Any subsequent declustering process, such as diffusion and structural turnover, randomizes fluorophores’ positions, progressively driving contrast down to the FSM-like baseline. In the context of microtubule turnover, it is depolymerization (catastrophe) and polymerization of new microtubules that drives the system toward the low-contrast baseline. (b) A GFP–α-tubulin–tagged Drosophila S2 cell in metaphase before and ∼2 and ∼40 s after the ISI pulse (γ ≅ 2). Progressive contrast fadeout is observed because of microtubule turnover. The bottom row shows the same experiment after treatment with a low dose of Taxol (10 nM), which increases microtubule stability and, consequently, speckle persistence. Bar, 5 µm. (c) ISI-induced speckle fadeout is driven by the complementary effects of decreasing-intensity bright speckles and increasing-intensity dark speckles, a time window during which poleward speckle motion (flux) can be observed. (d) Contrast–time curves (lines) and their time point mean (dots) measured in rectangular areas enclosing the spindle in Drosophila S2 metaphase cells (n_control = 44, n_taxol = 31, $\langle \gamma \rangle \cong 1.5-2.0$). Microtubule half-lives for a double-exponential fitting are t_fast = 2.9 ± 0.4 s and t_slow = 21 × 2 s (R² = .999) for control and t_fast = 3.4 ± 0.8 s and t_slow = 17 ± 3 s (R² = .975) for Taxol-treated cells, which displays a baseline term y₀ = 0.38 ± 0.01 (see Materials and methods), indicating that a significant microtubule population (40%) persists for a time much longer than the duration of the experiment. This baseline compares to y₀ = 0.05 ± 0.01 in control cells.

Figure 5.

Mechanical spindle relaxation time correlates to metaphase duration. (a) Drosophila melanogaster brain tissue before and after ISI. A rectangular ROI is used to generate guided kymographs from which flux velocity is calculated. Bars, 5 µm. (b) MT poleward flux velocity (F) versus spindle length (2L). Center and semi-axes of the shaded ellipses correspond to the mean and SD of the data point cloud for each cell type. (2L ± SD, F ± SD, n), Ganglion mother cells (6.4 ± 1.0 µm, 0.8 ± 0.4 µm/min, n = 27 cells); ovarian follicles (6.8 ± 1.4 µm, 1.7 ± 0.3 µm/min, n = 10 cells); neuroblasts (12 ± 2.6 µm, 1.4 ± 0.5 µm/min, n = 8 cells); and S2 cells (9.8 ± 1.2 µm, 1.0 ± 0.4 µm/min, n = 15 cells). (c) Metaphase duration (τ_meta) matches relaxation time (τ_relax = L/F) across cell types, meaning that time is allowed for ≈1 flux-driven translocation cycle along the half-spindle (green cone) before anaphase onset. Relaxation time is obtained from the data points in panel b, through τ_relax = L/F. Metaphase duration is, for Ganglion mother cells: 3.3 ± 1.3 min, n = 9; ovarian follicles: 1.9 ± 0.7 min, n = 7; neuroblasts: 5.9 ± 2.2 min, n = 9; and S2 cells 5.5 ± 3.2 min, n = 22. (*) A Drosophila embryo (cycles 11–13) data point is added by collecting data from the literature for the three parameters involved: τ_meta = 2.5 min (Minden et al., 1989); 2L = 12 µm (Brust-Mascher et al., 2009); F = 2.6 µm/min (mean value of available data; Brust-Mascher and Scholey, 2002; Brust-Mascher et al., 2004, 2009; Rogers et al., 2004; Wang et al., 2010).