4Pi-RESOLFT nanoscopy

Abstract

By enlarging the aperture along the optic axis, the coherent utilization of opposing objective lenses (4Pi arrangement) has the potential to offer the sharpest and most light-efficient point-spread-functions in three-dimensional (3D) far-field fluorescence nanoscopy. However, to obtain unambiguous images, the signal has to be discriminated against contributions from lobes above and below the focal plane, which has tentatively limited 4Pi arrangements to imaging samples with controllable optical conditions. Here we apply the 4Pi scheme to RESOLFT nanoscopy using two-photon absorption for the on-switching of fluorescent proteins. We show that in this combination, the lobes are so low that low-light level, 3D nanoscale imaging of living cells becomes possible. Our method thus offers robust access to densely packed, axially extended cellular regions that have been notoriously difficult to super-resolve. Our approach also entails a fluorescence read-out scheme that translates molecular sensitivity to local off-switching rates into improved signal-to-noise ratio and resolution.

Figure 1. 4Pi-RESOLFT principle and sample optics.

(a) Coherent double-lens illumination cycles RSFP markers between dark (OFF) and bright (ON) states to generate spatial ON/OFF-contrast. For each pixel, an activation light pulse (focal pattern h_ac) induces two-photon activation of RSFP (state transition C->B) in a pattern S^B(r) with axial side-maxima (lobes) that are optionally suppressed by a subsequently applied deactivation pulse (h_zd, B<->A->C). Fluorescence generated by the ON-state A is detected during read-out (B<->A->C) by a pattern h_ro. Its mutual overlap with S^B(r) is constrained to the focal centre, resulting in an effective PSF h_ef∼S^ON that exhibits ≈100 nm axial FWHM and exceptionally low side-maxima. Profiles show on-axis values. (b) The upright 4Pi unit of the microscope. Cells are mounted on a ring-shaped sample holder (H), between two cover glasses fixed at 10 μm distance by spacer beads and epoxy resin (E). The set of refractive indices (in brackets) of the immersion and embedding medium, cover slip thickness and correction collar settings of the objective lenses (O₁, O₂) diminishes aberrations from the sample. The sample stage (S) is mounted on a vertically movable (Z) goniometer (G_S), accepts the sample holder (H) and provides five degrees-of-freedom for coarse xyz-positioning and z-scanning of the sample, as well as tip-/tilt-alignment (θ_S) of the cover slip normal (a_S) to the optic axis of O₁ (a₁). O₁ itself is mounted on a xyz-piezo stage (OS) that provides online fine control over the displacement of both foci. A triangular mount (M) allows for tip/tilt-(θ₂) and coarse xyz-alignment of O₂ (axis a₂) with respect to O₁, and can conveniently be detached to change the sample. Two polarizing beam splitter/quarter-wave retarder pairs (BR_1,2) clean up and tune the polarization of the incident beam pairs to opposing circular states. One beam splitter furthermore serves as a port for an alignment laser beam that provides optical feedback for online-stabilization of the axial sample position (Δz); the beam traverses the respective objective lens off-axis (solid red path), gets reflected at the embedding medium interface and is imaged onto a camera (dotted red path).

Figure 2. 4Pi-RESOLFT imaging exhibits only minor axial lobes.

4Pi-RESOLFT raw data (left) and volume renderings (right) of Dronpa-M159T targeted to (a) the lumen of mitochondria, (b) actin microfilaments and (c) intermediate filaments of the cytoskeleton. The sample in a was subject to PFA fixation to freeze the motion of mitochondria; the filament networks in b,c were recorded from living cells, and exhibit regions of reduced density adjacent to the cover slip (arrows). Estimates of the z-response (insets), measured as box profiles over extended structures, exhibit only minor axial lobes in the 15 % range. Fast-to-slow order of scan axes, xzy. Pulse parameters, E_ac, E_zd, E_ro=1.6 mW·50 μs, 18 μW·50 μs, 3.1 μW·50 μs. Scale bars, 1 μm.

Figure 3. 4Pi-RESOLFT image formation with <100 nm isotropic resolution.

(a) A hollow switching pattern h_3d confines the central effective PSF to a spot with diameter d_ef by switching activated markers (B) back to their inactive state (C). Side-lobes due to inefficient switching at low off-centre h_3d amplitudes rise in relative strength as d_ef is reduced. μ, labelled structure. (b) Simulated z-response h_z(z) (laterally integrated h_ef) and axial MTF profile H(k_z) of the 4Pi-RESOLFT microscope (fourth on-order switching, solid lines) at different target resolutions d_ef. DL, diffraction limit. Graphs for an isoSTED microscope under similar conditions are included for reference (second on-order, dotted lines). (c) Normalized time-resolved, mean fluorescence signal collected from an xz-section through an actin fibre bundle (struct.) in a cell expressing Lifeact-Dronpa-M159T. Target resolution 50 nm, read-out pattern h_ro with a total power of P_ro=3.1 μW incident on the sample. An n-component multi-exponential fit to the data corresponds to n apparent switching speeds . A minimum of n=3 is required to adequately represent the data from the beginning of the read-out pulse t=0 up to 0.5 ms, (for up to 2.5 ms: n=4, ). Images Σ_0,1,3 integrated over time regimes that are dominated by fast (h_fast), slow (h_slow) and about constant PSF components (h_const) exhibit a declining resolution. (d) Rate-gated 4Pi-RESOLFT. Extrapolation of the initial contribution of h_slow (=S₀), based on integrated images Σ₁ (≈S₁) and Σ₂ (≈S₂), t₀=40 μs, provides an estimate of the partial image generated by h_fast (F₀≈ Σ₀−S₀, inset), improving resolution and image fidelity over Σ₀. Details are provided in Methods. (e) Rate-gated xz-sections through actin fibres, recorded with open pinhole to boost out-of-focus signal. The measured (y-integrated) side-lobe structure closely resembles the numerical prediction and can be further suppressed (right) by an additional z-donut h_zd (overlay, E_zd=1.8 μW·1.0 ms). Simulation parameters, numerical aperture 1.20, refractive index 1.362, pinhole diameter 0.5 airy units (e: open pinhole). Pulse parameters, E_ac, E_3d, E_ro=1.6 mW·0.2 ms, 1.3 μW·1.6 ms, 3.1 μW·2.5 ms (e: 0.5 ms). Scale bars, 250 nm.

Figure 4. 3D nanoscopy with strong optical sectioning.

xz-sections of live HeLa cells expressing Lifeact-Dronpa-M159T. (a) Overview (optical xz-section) of actin fibre bundles at an axial base resolution in the 100 nm range. Left inset, confocal reference. (b) Addition of a 3D deactivation donut (+h_3d, E_3d=2.6 μW·3.2 ms) to the RESOLFT pulse sequence reveals Dronpa patterns with apparent feature sizes well below 40 nm (inset, Gaussian reference spheres); (c) Lorentzian fits, plus a linear local background, to box profiles p1–3 over marked features in b along different directions. Numbers indicate full widths at half maximum (FWHM) over background. (d) Rendering of the volume surrounding a. (e–g) Time (T) evolution of an 8-μm-thick, densely labelled, vertical contact region between two adjacent cells (xz-section as marked in the xy-overview). Grayscale overlays of the preceding time step (f,g) aid in the tracking of individual features. A narrowed region of interest was generated online from initial overview scans (−h_3d) at each time frame and imaged at 50 nm target resolution (+h_3d, grey outline, E_3d, E_zd=1.3 μW·1.6 ms, 1.8 μW·0.5 ms). Despite the challenging imaging conditions, stacked actin structures are unambiguously resolved across the full axial extent of the cell layer. xz-panels depict rate-gated 4Pi-RESOLFT raw data, solely subjected to noise reduction. Fluorescence intensities I(r). Common pulse parameters (b,e–g), E_ac, E_ro=1.6 mW·0.2 ms, 3.1 μW·0.5 ms. Scale bars, 1 μm.