Fig 1 Idea.
Published:31 July 2024,
Published Online:07 June 2024,
Received:28 January 2024,
Revised:06 May 2024,
Accepted:16 May 2024
Scan QR Code
Cite this article
MINFLUX has achieved extraordinary resolution in superresolution imaging and single fluorophore tracking. It is based on localizing single fluorophores by rapid probing with a patterned beam that features a local intensity minimum. Current implementations, however, are complex and expensive and are limited in speed and robustness. Here, we show that a combination of an electro-optical modulator with a segmented birefringent element such as a spatial light modulator produces a variable phase plate for which the phase can be scanned on the MHz timescale. Bisected or top-hat phase patterns generate high-contrast compact excitation point-spread functions for MINFLUX localization in the x, y, and z-direction, respectively, which can be scanned across a fluorophore within a microsecond, switched within 60 microseconds and alternated among different excitation wavelengths. We discuss how to compensate for non-optimal performance of the components and present a robust 3D and multi-color MINFLUX excitation module, which we envision as an integral component of a high-performance and cost-effective open-source MINFLUX.
MINFLUX
For optimal performance, a MINFLUX microscope requires fast and repeated scanning of the pattern to average over potential intensity fluctuations of the fluorophore. In most custom
Here, we overcome this challenge by developing a MINFLUX excitation module based on a novel variable phase plate, which enables 3D multi-color MINFLUX with high spatio-temporal resolution using a simple, robust, and economic setup.
This module generates high-contrast MINFLUX PSFs and scans them rapidly across a fluorophore. However, it is only one component of a fully functional MINFLUX microscope, which additionally requires real-time feedback of the position estimate on the scan pattern and a very stable microscope body combined with active sample stabilization with sub-nanometer accuracy. The construction of such a complete MINFLUX microscope is complex and out of scope for this article. Instead, we experimentally demonstrate the generation and fast scanning of optimized MINFLUX PSFs with a low-NA setup, show with simulations that this setup can be directly transferred to high-NA microscopes without loss in performance, and discuss how to overcome experimental imperfections that otherwise might limit the quality of the PSFs.
As shown by Wirth et al.
Fig 1 Idea.
a Phase patterns and point-spread functions (PSFs) for MINFLUX localization in the x-, y-and z-direction, respectively. The phase φ determines the position of the intensity minimum (see Supplementary Fig. 3). b Theoretical localization precision limit, the Cramér-Rao bound, for 2D MINFLUX for sequential localization with the bilobed x and y patterns (left) and 3D MINFLUX with the bilobed x and y patterns plus the 3D donut (right), in a region at the center of the PSF as denoted in a. Simulation parameters: scan range L = 50 nm in the lateral and Lz = 150 nm in the axial direction and an offset (imperfect zero and background) of 0.5% of the maximum intensity when φ = 0. The calculated localization precisions are normalized by
Phase differences of π give rise to symmetric patterns with the minimum on the optical axis (x, y pattern) or in the focus of the objective lens (z pattern), because here the π phase shift leads to destructive interference. Phase differences different from π displace the position of the minimum because the rays must acquire an additional path difference to reach a phase difference of π (Supplementary Fig. 3). Simulations show that for the bilobed PSFs a true zero at the minimum is preserved even for high-NA systems and large scanning, whereas for the 3D donut the contrast stays high over a z range of a few hundred nm (Supplementary Fig. 4).
The phase patterns can be created with a spatial light modulator (SLM), which, in principle, can scan the position of the intensity minimum by changing the phase. However, SLMs that produce continuous phase delays are too slow (< 1 kHz) for the rapid scanning required for MINFLUX, thus a fast scanner like an electro optical deflector would still be required.
Here we overcome this limitation by combining an EOM, an SLM and a polarizer to generate binary phase patterns with variable phase delay, where the phase delay can be changed on the sub-microsecond time scale by the EOM phase (
Using Jones matrices, we can calculate in the horizontal/vertical coordinate system that an input beam of
To experimentally validate our idea without the need of implementing a complete MINFLUX microscope, we designed a beam path in which we replaced the objective lens by an f = 400 mm achromat and the single fluorophore by a small pinhole and photo detector that reports the intensity at the putative single-fluorophore position (
Fig 2 Experimental demonstration with camera detection.
a Beam path of the test setup. The laser is polarized under 45°, passes the electro-optical modulator (EOM), a halfwave plate (HWP) to compensate for imperfections of the Spatial Light Modulator (SLM) and a beam expander comprising achromatic lenses and a pinhole. After the SLM, it passes another polarizer and is focused by an achromat either onto a CMOS camera or onto a small pinhole with a photo diode that detects the intensity a fluorophore would see. b PSFs recorded with the camera for different phases (as indicated) imposed by the EOM for the bilobed x pattern and the 3D-donut z pattern. The scale bar and values in nm are rescaled corresponding to an implementation in a microscope with an NA 1.35 objective lens. c Profiles through the x-PSF as indicated in b for different EOM phases with a difference of 5° between consecutive profiles. d Intensities at the center pixel of the z-PSF for different EOM voltages. Each profile was measured with the last lens positioned at a different distance from the camera (step size 2 mm) and corresponds to measurements with a fluorophore at different z positions (corresponding to a step size of 18 nm in z when integrated in a microscope). Inset: Cross-section through the PSF as indicated in B for different phases
The bisected phase patterns produced the expected bilobed PSFs (
To mimic the excitation signal seen by a single fluorophore, we placed a tiny pinhole (0.03 Airy Units) in the focus of the achromat, which transmits the intensity in the central part of the PSF (
Fig 3 Experimetal demonstration of fast scanning using a pinhole, see
a An experimental x-PSF acqiured by a camera with the position of the pinhole indicated by a circle. b Intensity (normalized to the intensity measured for a flat phase pattern) recorded through the pinhole for a linear ramp in the EOM voltage. The intensity reaches a minimum (contrast value indicated in the graph) when the minimum of the PSF is at the pinhole position. The measurement is repeated for the different phase patterns (insets) for y-and z-localization, respectively. Switching between phase patterns takes 60 µs. c Mimicking of a MINFLUX measurement. By imposing three different voltages, each lasting for 1 µs, the minimum of the phase pattern is positioned at three distinct positions around the fluorophore. x0 indicates the position of the minimum for the case when the setup is used in conjunction with a microsope objective, L is the diameter of the scan pattern. d By changing the amplitude of the voltage, the scan range L can be reduced to improve the localization precision. The dashed line indicates the zoom region for the subsequent panel. e By using two colinear laser beams of different colors which are alternated (in this case every 9 µs corresponding to 3 scan patterns), dual-color MINFLUX excitation can be realized without further modifications of the setup
Next, we tested the speed of our MINFLUX excitation module by repeatedly applying three different EOM phases, which positions the x-PSF at 3 positions around the pinhole that mimics the fluorophore. With excitation times as low as 1 µs per position we could reliably detect three different intensity values with high contrast (
Multi-color MINFLUX excitation can be realized with the same setup without modifications by using as an input co-linear laser beams. In case of small distances between the fluorophores and low spectral dependence of the phase delay
The simple schematic in
SLM
The binary Ferroelectric Liquid Crystal on Silicon (FLCoS) SLM used here has the birefringent axis of the 'on'-pixels oriented under 33.5° compared to the 'off'-pixels, instead of the ideal 45°. Additionally, the phase delay might deviate slightly from π. By calculations (Methods, Data and code availability) and experimentally we found that the addition of a HWP in the beam path before the SLM can perfectly compensate for both imperfections (
EOM
The EOM phase delay is wavelength dependent, leading to larger displacements of the minima for shorter wavelengths for a given EOM voltage. To compensate this, the static offset of the EOM phase, as well as the amplitude of the phase scan can be set separately for different excitation colors.
Any instability of the EOM phase directly causes a localization bias (1.4 nm/deg in x and y, 3.6 nm/deg in z, see Supplementary Fig. 7). We monitored the EOM phase drift using a second crossed polarizer and photo diode by scanning the EOM phase around the intensity minimum and found phase drifts below 0.2° (corresponding to 0.3 nm for an NA 1.35 system) within 10 s and below 1.1° (1.6 nm) over hours (Supplementary Fig. 7). This setup could be used in the future for passive phase monitoring or active phase stabilization. Alternatively, our low-NA setup using a pinhole and photo diode (
Silver mirrors
Most mirrors, including silver mirrors, lead to a phase delay between the s and p polarization components, which can spoil carefully engineered polarization states. By aligning the EOM in along the s/p coordinate system and placing the polarizer directly after the SLM, any phase imposed by the mirrors can be compensated by adjusting the static offset of the EOM phase.
Aberrated wave fronts
Laser sources, especially free space diode lasers, can have an imperfect beam profile. Additionally, optical components (wave plates, EOM, mirrors) can further deteriorate the wave front, leading to aberrated PSFs and a reduced contrast of the intensity minimum. Here, whenever possible, we place optical components (EOM, waveplate) before the mode cleaner, which then produces a close to ideal wavefront. To reduce astigmatism and coma introduced by a slightly curved SLM, we choose a small beam size on the SLM. In the future, we will insert a second SLM before the microscope to compensate for aberrations from the objective lens or sample.
Scan range
As MINFLUX is based on a confocal principle, and because EODs and varifocal lenses currently do not allow for de-scanning due to polarization dependency and auto-fluorescence, the scan range in all MINFLUX implementations is limited by the confocal pinhole and a secondary slow, but large range scanner (galvo
Polarization of MINFLUX PSFs
For high-NA microscopes, the x and y phase patterns require the polarization to be parallel to the phase boundary, which can be fulfilled only for one of the directions. Otherwise, the axial component of the electric field leads to incomplete destructive interference and an increased intensity in the minimum of the PSF (Supplementary Fig. 8). One solution is to reflect the beam for a second time on the SLM and rotate the polarization state only for the x pattern (Supplementary Fig. 8). The non-ideal direction of the birefringent axis can only rotate the polarization by 67° leading to a calculated deviation of at least 11.5° from the optimal polarization for both the x and y patterns. This results in a contrast of not better than 0.6% (Supplementary Fig. 8) for an NA 1.35 objective, which however is still acceptable (see Supplementary Fig. 2C). A more accurate alternative is to add a second EOM to the output beam path to generate an optimal polarization state for each phase pattern. This would also allow turning the direction of the phase pattern on the SLM and match it to the sample (e.g., direction of motion of a motor protein) for ultra-fast 1D MINFLUX
We developed an excitation module for 3D and multi-color MINFLUX that combines fast and precise positioning of the intensity minimum with a robust and affordable setup. Although we tested the module only in a low-NA setup, the integration into a full MINFLUX microscope is straightforward as long as the other challenges (fast FPGA-based position feedback, ultra-stable microscope) are met. We performed extensive vectorial wave simulations for a high-NA objective (Supplementary Fig. 4) that show that the high contrast of the PSFs will be retained.
MINFLUX relies on the brightness of the fluorophore to be constant during probing of intensities around the fluorophore and any intensity fluctuations on time scales longer than a fraction of the probing time at a single location can lead to a position bias. Most fluorophores show transient dark states on the microsecond to millisecond time scale
Whereas 1D localization is sufficient for a few applications such as tracking of linear motor proteins
The ability to switch quickly between different phase patterns allowed us to use an optimized PSF for each dimension. The bilobed PSFs that we use for lateral localization result in a high precision for a given number of detected photons (Supplementary Fig. 1). They have a smaller footprint compared to donut
Currently, multi-color MINFLUX is performed sequentially with different fluorophores, or using a single excitation laser in combination with fluorophores of slightly different emission wavelengths
As all beams are colinear and are not split up to generate different patterns, colors, or interference phases, they cannot misalign with respect to each other. Thus, our setup is intrinsically robust and stable, which is essential to reach sub-nanometer accuracies in MINFLUX and to use it for routine biological applications. Stability is further supported by the simplicity of the setup with few components and short beam paths.
Our MINFLUX excitation module is very cost-effective with the components in
We envision our excitation module to be a key element for the future development of an affordable open-source MINFLUX instrument with highest performance.
Calculation of polarization states
The polarization state of the beam was calculated in Mathematica (Wolfram) using Jones matrices (Data and code availability). Here, a linearly polarized beam is described by
in the horizontal/vertical coordinate system, and a beam with 45° linear polarization by
A waveplate (e.g., halfwave plate or SLM) with a phase delay of
Note that global phase factors are omitted as they do not affect the intensity distribution. The EOM with a phase delay of
A polarizer transmitting horizontal polarization is described by
The simple setup (
A realistic setup with experimental imperfections is modeled as:
Here,
Calculation of MINFLUX point-spread functions
For the numerical calculations of the electromagnetic field near the focus of an objective lens, we used MATLAB (R2022a, MathWorks) and a software package provided by Leutenegger et al.
The background offset is due to an imperfect PSF with a minimum larger than zero and autofluorescence generated by the sample (including out-of-focus fluorophores) and by the optics of the microscope, all of which scale with the total laser intensity. This is why we normalized all PSFs by the total integrated light. To have more interpretable values, we again normalize those so that a Gaussian beam has a maximum intensity of 1. This makes the different PSFs easily comparable in terms of contrast and background offset caused by the total intensity.
Calculation of Cramer-Rao-Bounds (CRBs)
For calculating the theoretically best possible localization precision for specific PSFs, scanning schemes, detected photons and signal to background ratios we followed Masullo et al.
the log-likelihood (after dropping of constant terms) can be written as
The Fisher information matrix can then be written as:
The CRB is then
To account for imperfect contrast and a fluorescent background, we modeled the background explicitly by adding an offset to the PSF. To make the background comparable for different PSFs, we calculated the maximum value
See
The EOM was driven by a voltage amplifier (HVA200, Thorlabs), which was controlled by a function generator (SDG1062X, Siglent). Intensity signals from the photo diode were acquired with an oscilloscope (SDS2104X-Plus, Siglent). The function generator and the oscilloscope were triggered by the SLM at the start of an image sequence. Additionally, the SLM produces a trigger signal when a valid pattern is established, which was used to switch off the lasers during the pattern switching via a TTL signal. Camera images were recorded asynchronously. For dual-color measurements, we used a custom TTL signal converter in combination with the function generator to alternate between the 561 nm and 638 nm laser line.
Note that our time resolution for 1D probing of 1 μs is currently limited by the electronics for EOM scanning and the bandwidth of the photo diode and can in principle be one order of magnitude faster. Such high speeds might not be necessary for MINFLUX but could be useful for other applications of the variable phase plate.
We used a polarization analyzer (PAX1000VIS/M, Thorlabs) to align the Glen-Thompson polarizer to 45°. Using a flat phase pattern and large iris diameter, the camera was positioned in the focus of the f = 400 mm achromat. The pinhole was then positioned at an equal distance from the achromat.
Using the top-hat phase pattern, the beam was aligned on the SLM to produce a symmetric PSF. Then the iris diameter and EOM phase were optimized to maximize the contrast of the 3D-donut in focus. The contrast was further maximized by aligning the angle of the half wave plate and the EOM phase.
The pinhole, mounted in an x-y translation stage (ST1XY-D/M, Thorlabs), was adjusted in the lateral directions to maximize the signal on the photo diode using a flat SLM phase pattern.
To monitor the phase drift of the EOM (Supplementary Fig. 7), we split off 10% of the laser light after the EOM with a non-polarizing beam splitter, passed that beam through a Glen-Thompson polarizer (GTH10M-A, Thorlabs) oriented orthogonal to the first one before the EOM and monitored the intensity with an amplified photo diode (PDA100A2, Thorlabs). We repeatedly scanned the EOM voltage around the intensity minimum and fitted the minimum with a quadratic function to extract the zero-point voltage V0.
We thank Giuseppe Vicidomini and Eli Slenders for their kind feedback on the manuscript, Luciano Masullo for help with the CRB calculations and the EMBL electronic and mechanic workshops for contributing to the setup. This work was supported by H2020 Marie Skłodowska-Curie Actions (RobMin grant no. 101031734 to T.D.); the European Research Council (grant no. ERC CoG-724489 to J.R.); and the European Molecular Biology Laboratory (T.D. and J.R.).
Simulated PSFs, scripts to calculate CRBs and a script to calculate the polarization state are available at: https://github.com/ries-lab/MINFLUXexcitation.
The EMBL has deposited the European patent application 23193790.5 on the 28 July 2023 to protect this work. T.D. and J.R. are co-inventors.
Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41377-024-01487-1.
Balzarotti, F. et al. Nanometer resolution imaging and tracking of fluorescent molecules with minimal photon fluxes.Science355, 606–612 (2017).. [Baidu Scholar]
Gwosch, K. C. et al. MINFLUX nanoscopy delivers 3D multicolor nanometer resolution in cells.Nat. Methods17, 217–224 (2020).. [Baidu Scholar]
Pape, J. K. et al. Multicolor 3D MINFLUX nanoscopy of mitochondrial MICOS proteins.Proc. Natl Acad. Sci. USA117, 20607–20614 (2020).. [Baidu Scholar]
Ostersehlt, L. M. et al. DNA-PAINT MINFLUX nanoscopy.Nat. Methods19, 1072–1075 (2022).. [Baidu Scholar]
Eilers, Y., Ta, H., Gwosch, K. C., Balzarotti, F.&Hell, S. W. MINFLUX monitors rapid molecular jumps with superior spatiotemporal resolution.Proc. Natl Acad. Sci.115, 6117–6122 (2018).. [Baidu Scholar]
Wirth, J. O. et al. MINFLUX dissects the unimpeded walking of kinesin-1.Science379, 1004–1010 (2023).. [Baidu Scholar]
Deguchi, T. et al. Direct observation of motor protein stepping in living cells using MINFLUX.Science379, 1010–1015 (2023).. [Baidu Scholar]
Schmidt, R. et al. MINFLUX nanometer-scale 3D imaging and microsecond-range tracking on a common fluorescence microscope.Nat. Commun.12, 1478 (2021).. [Baidu Scholar]
Masullo, L. A.&Stefani, F. D. Multiphoton single-molecule localization by sequential excitation with light minima.Light Sci. Appl11, 70 (2022).. [Baidu Scholar]
Masullo, L. A. et al. Pulsed Interleaved MINFLUX.Nano Lett.21, 840–846 (2021).. [Baidu Scholar]
Zhan, Z. et al. Simultaneous super-resolution estimation of single-molecule position and orientation with minimal photon fluxes.Opt. Express30, 22051 (2022).. [Baidu Scholar]
Huang, Z. et al. Spectral Identification of Specific Photophysics of Cy5 by Means of Ensemble and Single Molecule Measurements.J. Phys. Chem. A110, 45–50 (2006).. [Baidu Scholar]
Widengren, J.&Schwille, P. Characterization of Photoinduced Isomerization and Back-Isomerization of the Cyanine Dye Cy5 by Fluorescence Correlation Spectroscopy.J. Phys. Chem. A104, 6416–6428 (2000).. [Baidu Scholar]
Haupts, U., Maiti, S., Schwille, P.&Webb, W. W. Dynamics of fluorescence fluctuations in green fluorescent protein observed by fluorescence correlation spectroscopy.Proc. Natl Acad. Sci. USA95, 13573–13578 (1998).. [Baidu Scholar]
Leutenegger, M., Rao, R., Leitgeb, R. A.&Lasser, T. Fast focus field calculations.Opt. Express14, 11277–11291 (2006).. [Baidu Scholar]
Engelhardt, T. MINFLUX Nanoscopy with Interferometric Approach. PhD Thesis, University of Heidelberg. 2023.https://doi.org/10.11588/heidok.00031212. [Baidu Scholar]
Masullo, L. A., Lopez, L. F.&Stefani, F. D. A common framework for single-molecule localization using sequential structured illumination.Biophysical Rep.2, 100036 (2022).. [Baidu Scholar]
14
Views
12
Downloads
0
CSCD
Related Articles
Related Author
Related Institution