Three-dimensional multi-color optical nanoscopy at sub-10-nm resolution based on small-molecule organic probes

Summary

Achieving nanometer-scale resolution remains challenging in expansion microscopy due to photon loss. To address this concern, here we develop a multi-color expansion stimulated emission depletion technique based on small-molecule probes to realize high labeling density and intensity. Our method substantially lowers the barrier to visualizing diverse intracellular proteins and their interactions in three dimensions. It enables us to achieve sub-10-nm resolution in structures such as microfilaments, lysosomes, and mitochondria, providing new insights into cell biology.

Graphical abstract

Highlights

• •Small-molecule organic probes provide high labeling density and intensity
• •Small marker sizes decrease imaging artifacts caused by probe displacement
• •Adaptive optics configuration in STED compensates for sample-induced aberrations
• •The approach reveals subcellular structures and interactions at sub-10-nm resolution

Motivation

Combining expansion microscopy with other super-resolution techniques poses challenges due to the reduction in fluorescence density and intensity in expanded samples. Although employing multiple and post-expansion labeling of fluorescent proteins or antibodies can enhance the labeling density, attaining sub-10-nm resolution in super-resolution microscopy requires careful consideration of marker size. To address this issue, we have developed a multi-color 3D nanoscopy technique that employs small-molecule organic probes. The utilization of small-molecule organic probes serves to enhance both the density and intensity of labeling while reducing measurement error.

Achieving isotropic nanometer-scale resolution remains challenging when combining STED with expansion microscopy. Han et al. establish a 3D multi-color nanoscopy technique based on the high labeling density and intensity of small-molecule organic probes, revealing details of the structures and interactions of actin filaments, lysosomes, and mitochondria.

Introduction

The past decades have witnessed the explosive development of optical super-resolution microscopy, allowing visualizing ultrafine subcellular structures and interactions in cells. It has opened a pathway toward widespread applications in biological science. Different from other methods, expansion microscopy (ExM)¹ allows high-resolution imaging by the physical magnification of biological specimens in support of hydrogel. In principle, combining sample expansion with other super-resolution microscopy techniques would allow better resolution. Previous work in expansion stimulated emission depletion (ExSTED)² microscopy has reported ∼50-nm resolution in three dimensions (3D). However, there is a native contradiction between ExM and the other super-resolution methods in terms of resolution enhancement. For almost all super-resolution microscopy techniques,^3,4 the resolution heavily depends on the photon number (i.e., fluorescent brightness). In contrast, ExM induces loss of both labeling density and fluorescent intensity upon expansion of the specimen.^1,5 This contradiction compromises the endeavors to push the 3D resolution toward sub-10 nm. In the commonly employed techniques, the resolution dilemma is addressed through multiple utilization of antibodies (20–30 nm for primary and secondary antibodies combination) and fluorescent proteins (∼5 nm). However, a considerable proportion of inaccuracies arise in resolution owing to the substantial sizes of these labels.^2,6

Small-molecule organic probes (SMOPs),⁷ if applicable in ExM, can realize much higher labeling density due to their small sizes (<1 nm) and brighter fluorescent intensity, potentially improving the resolution. However, higher 3D resolution has yet to be implemented using this strategy. This is partly because traditional fluorescent probes usually cannot survive during the sample preparation without proper anchoring on the gel.⁶ It is also challenging for most of these probes to resist the high laser radiation required by super-resolution imaging. Here we develop a technique that achieves multi-color, sub-10-nm resolution by comprehensively using the ExSTED method and small-molecule organic probes (termed SMOP-ExSTED in this work).

Results

Design and characterization of probes

We designed a universal, modularized framework of the ExM probes based on SMOP. The structure of the probes contains at least four components: (1) a recognition module for the binding of target structures, (2) an anchoring module (methacrylic acid) to combine the probe to the expansion gel, (3) fluorescent dyes, and (4) linking modules to connect the other components (Figure 1A). The anchoring module, methacrylic acid,¹ is constructed into the probes to combine with the hydrogel during gelation, preventing the probes from falling off the labeled proteins during the expansion process. By targeting F-actin and lysosomal hydrolases,⁸ these probes are termed MAA-Actin-dye and MAA-Lyso-dye (Figure 1A). The lysine residues in the linking module were used to offer free amino groups that can be conjugated with the fluorescent dyes, and a variety of dyes can be incorporated into this construct, making the dye selection in multi-color imaging more convenient. Employing both strategies together allows more flexible choices of dye in multi-color imaging. Our test results suggest that Alexa Fluor (AF) 488, tetramethyl-rhodamine (TMR), AF 594, STAR ORANGE, and STAR RED perform well in SMOP construct. The results showed that MAA-Actin-TMR colocalized well with EGFP-Lifeact (Pearson’s coefficient: 0.94), and MAA-Lyso-Atto 565 colocalized well with LysoTracker Green (Pearson’s coefficient: 0.9), indicating the high specificity of the probes. Notably, not all the fluorescent dyes are suitable for constructing the actin probes, since strongly lipophilic and cationic dyes⁹ may cause nonspecific labeling (e.g., Atto 647N). Besides, cyanine derives (e.g., Alexa Fluor 647 and Cy5) should also be avoided because of their weak resistance to the following expansion process.⁵

Figure 1.

Design and characterization of our probes

(A) Structures of the probes. Left: actin probes; right: lysosomal probes.

(B–E) Confocal images of U2OS cells before (upper panels) and after (bottom panels) expansion. From left to right, the samples are labeled with MAA-Actin-AF488 (B), phalloidin-AF488 (C), anti-Actin AF488, (D) and EGFP-Lifeact (E). Scale bars, upper panels, 20 μm; bottom panels, 10 μm.

(F–H) Comparison of image quality when using different labeling strategy excited by 488-nm laser before and after sample expansion. Comparison of normalized intensity (F), signal-to-noise ratio (SNR) (G), and Signal-to-background ratio (SBR) (H) of the confocal and ExConfocal images of U2OS cells labeled with the above markers. Data are represented as mean ± SD. NA, not available. ns, p > 0.05; ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001; two-tailed t test, statistics were performed using Excel.

See also Figure S1.

To verify our SMOP method in ExM, we imaged the 3D distributions of actin filaments and lysosomes under confocal mode after ∼3.4× sample expansion with minimal distortions of 2.6% (Figure S1). Their maximum intensity projections (MIPs) are available in Figure S1G. These results confirm the labeling specification of our SMOPs and their compatibility with the dyes mentioned above.

We further compared the specimens labeled by our method with those by traditional fluorescent probes, fluorescent proteins, and immunofluorescence labeling (Figures 1B–1H). For quantitative analysis, we chose fluorophores excited by 488-nm excitation laser wavelength, i.e., AF 488 and enhanced green fluorescent protein (EGFP), for variable control. Other imaging parameters (e.g., laser intensity and camera gain) in parallel experiments were set the same, while all the labeling conditions were specifically optimized for each method. The results suggest that phalloidin-AF488 provides high fluorescence intensity, signal-to-background ratio (SBR),¹⁰ and signal-to-noise ratio (SNR) before expansion. However, it cannot survive the expansion treatment due to the lack of an anchoring module (Figures 1C and 1E–1G). Worse still, when labeled with actin antibodies, the images show discontinuous punctate markings both before and after expansion, irrespective of the antibodies and fixation buffer we used (Figure 1D). On the other hand, the actin filaments are continuous in structure within the cells transfected with Plasmid EGFP-Lifeact, whereas it is also accompanied by strong background. This phenomenon implies that it is critical but challenging to precisely control the protein expression level (Figures 1E and 1F–1H). In contrast, in cells stained with the probe MAA-Actin-AF488, fine details of filopodia, lamellipodia, and stress fibers are well reserved both before and after expansion (Figure 1B). Besides, using SMOP also allows the best fluorescence intensity and SBR, both before and after expansion (Figures 1E and 1G). To sum up, the higher labeling density of our probes leads to higher fluorescence intensity and density than antibody and fluorescent protein labeling both in confocal and in expansion confocal imaging.

Resolution measurement

To demonstrate our method on a more challenging target, we applied it in an STED nanoscope (Figure 2). The dyes featuring high brightness and photostability under STED imaging conditions, such as STAR RED, STAR ORANGE, and AF 594, were used. While our SMOP approach somewhat improved the image quality in confocal mode, the SMOP-related improvements in STED mode were striking: it enhanced the resolution to sub-10 nm (Figure 2), a level where nanoscale features of the actin fibers became clearly visible. A bundle of actin fibers in the expansion confocal image is replaced by several finer strands intertwined together in the SMOP-ExSTED image (Figures 2A–2D and Video S1). The minimum distinguishable distance between the adjacent actin molecules is ∼8 nm based on Gaussian fitting (Figures 2E and 2F). The resolution was quantified using decorrelation analysis¹¹ at roughly 6 nm (xy) and 9 nm (xz), which are over 17 times better than the resolutions in expansion confocal mode (106 nm for xy and 168 nm for xz, Figures 2C and 2D).

Figure 2.

Actin filaments in expanded cells

(A) Overview of actin filaments in a BSC-1 cell labeled with MAA-Actin-STAR RED. The left and the right parts in the figures represent the same field of view imaged under (left) ExConfocal and (right) ExSTED modes. Scale bar, 1 μm.

(B) Magnified image of the dashed box in (A). The red arrows indicate actin filaments resolved by (bottom) SMOP-ExSTED but indistinguishable in (upper) SMOP-ExConfocal image. Scale bar, 200 nm.

(C and D) xy and xz cross-sections of microfilaments imaged under (upper) SMOP-ExConfocal and (bottom) SMOP-ExSTED modes. Scale bars, 100 nm.

(E and F) Intensity profile at the positions denoted in the dashed boxes in (C) and (D). The raw data (black dots) are fitted with a Gaussian function (red curve). The uniform background is removed when plotting the graphs to enhance the contrast.

See also Figure S2.

Monochromatic 3D SMOP-ExSTED imaging in cells and tissues

The subcellular structures, e.g., cytoskeletons and organelles, spatially distribute not only in the perimembrane layers but everywhere across the whole cell. Therefore, it is desirable if high imaging penetration depth is possible. Compared with the area near the cover slip, imaging thicker samples poses a greater challenge for aberration-sensitive techniques such as STED nanoscopy. In particular, the physical expansion of the sample further increases the sample thickness,¹ amplifying any existing refractive index mismatch between the hydrogel (∼99% water in expanded specimens) and the objective immersion liquid. Using water objective may alleviate the mismatch but at the expense of a smaller numerical aperture (NA), that is, a poorer resolution.² To resolve this dilemma when imaging thicker, we employed a 1.35-NA silicon-oil objective and further integrated adaptive optics configuration into our instrument to dynamically compensate for the sample-induced aberrations.¹² This scheme ensures that sub-10-nm resolution remains almost constant throughout the whole imaging process.

Improved 3D resolution enables us to record the distributions of organelle proteins more precisely. For example, actin filaments that overlap in confocal mode are distinguishable in 3D SMOP-ExSTED mode (Figures 3A and 3B and Video S2). Besides, the 3D dataset of the lysosomes clearly reveals the sparse distributions of lysosomal enzymes (white arrows in Figure 3C) inside each enzyme cluster in SMOP-ExSTED (Figure 3C and Video S3), suggesting that uneven distribution of lysosomal enzymes routinely occurs. Mouse brain slices (30 μm) were also tested to verify the ability of our SMOP in labeling tissue samples. It is worth noting that the results showed paralleled actin fibers organized in groups in vascular smooth muscle cells¹³ and membrane-associated actins with a periodicity of ∼180 nm underneath the neuron membrane⁴ (Figure S2).

Figure 3.

3D rendering of mitochondria in expanded cells

(A) Color-coded SMOP-ExConfocal (left) and SMOP-ExSTED (right) xy MIPs of two actin bundles stained with MAA-Actin-STAR RED. Scale bar, 200 nm.

(B) 3D renderings of the fibers in (A) when the volume rotates 55° clockwise around y axis. Scale bar, 200 nm.

(C) Volume renderings of lysosomes stained with MAA-Lyso-STAR ORANGE in an expanded U2OS cell imaged via 3D confocal (left) and STED (right). Scale bar, 1 μm.

(D) Color-coded SMOP-ExConfocal (left) and SMOP-ExSTED (right) xy MIPs of a U2OS cell stained with Atto 647N NHS ester. Scale bar, 1 μm.

(E and F) xy (E) and xz (F) sections of SMOP-ExConfocal (left) and SMOP-ExSTED (right) images of mitochondria. Scale bars, 200 nm.

(G and H) Color-coded SMOP-ExConfocal (upper) and SMOP-ExSTED (bottom) xy (G) and xz (H) MIPs of mitochondria. Scale bars, 1 μm.

(I) Color-coded SMOP-ExSTED xy MIPs of mitochondria revealing cristae distribution. Scale bar, 1 μm.

(J) Intensity profile at the position denoted by the white box in (I). The uniform background is removed when plotting the graphs to enhance the contrast.

In another demonstration, we aim to image mitochondria. Atto 647N N-hydroxysuccinimidyl (NHS) ester has been demonstrated as a live-cell mitochondrial marker with its labeling capacity to both the inner and outer mitochondrial membrane through its NHS moiety even after fixation.⁸ Here we extended its application to expansion microscopy. The results showed that Atto 647N exhibits sufficient signal in the expanded hydrogel, indicating that this dye can survive the whole fixation and expansion procedure (Figure 3D and Video S4). The mitochondrial protein distributions, which blurred in expansion confocal mode, were visible in SMOP-ExSTED mode (Figures 3E–3H). Interspace between two adjacent mitochondria in the axial direction (white arrows in Figures 3G and 3H) and the boundaries of the voids between cristae groups (white dashed arrows in Figure 3G) were also revealed. From a particular perspective in 3D visualizations, it becomes possible to distinguish the distance between cristae as approximately 100 nm (Figures 3I and 3J), a value that conforms exceptionally well with prior investigations.¹⁴

Dual-color 3D SMOP-ExSTED imaging

For studying interactions of subcellular structures, multi-color imaging is of the essence. However, many more challenges are put forward for the labeling methods to realize multi-color ExSTED imaging. On the one hand, the complex label-fixation-expansion procedure increases the difficulty of performing multi-color staining; on the other hand, STED imaging requires enough photons for the fluorescent labels in each spectral channel in expanded cells. Before dual-color SMOP-ExSTED, co-labeling with MAA-Actin-AF594 and conventional antibodies in expansion microscopy was tested. Although double stained with both α and β tubulin primary antibodies, the microtubules showed obvious discontinuous signals in ExSTED, indicating the insufficient labeling density of antibodies (Figures S3A and S3B). Based on the performance of dyes we tested during our single-color ExSTED imaging, i.e., brightness, anti-bleaching property, and spectrum, we applied MAA-Actin-STAR RED, MAA-Lyso-AF594, and Atto 647N NHS ester in dual-color SMOP-ExSTED imaging (Figures 4 and S3C–S3G).

Figure 4.

Dual-color SMOP-ExSTED imaging

(A) Dual-color volume rendering of F-actin (MAA-Actin-STAR RED; cyan) and lysosomes (MAA-Lyso-AF594; yellow) in a BSC-1 cell. Scale bar, 1 μm.

(B) Magnified view of the dashed box region in (A). Scale bar, 200 nm.

(C) xy, xz, and yz sections of the bucket actin structure from the white box region in (A). Scale bar, 200 nm.

(D) Dual-color volume rendering of mitochondria (Atto 647N; magenta) and lysosomes (MAA-Lyso-AF594; green) in a U2OS cell after starvation. Scale bar, 200 nm.

(E) Frequency histograms of five types of interactions between mitochondrial and lysosomal proteins (M, mitochondria with no adjacent lysosomal proteins; <L>, lysosomal proteins wrapped with mitochondria; <M>, mitochondria wrapped with lysosomal proteins; L + n∗M, one lysosomal protein cluster attached with several smaller mitochondrial clusters; M + n∗L, one mitochondrial cluster attached with several smaller lysosomal clusters).

(F) Dual-color volume rendering (left and middle) and its 3D surface rendering (right) of mitochondria (Atto 647N; magenta) and lysosomal proteins (MAA-Lyso-AF594; green) in an autolysosome from different perspectives. During the 4-h starvation, a portion of the Atto 647N dyes undergoes gradual dissociation from the mitochondria without forming covalent bonds with the mitochondrial membrane. This process results in a significant decrease in fluorescence intensities. Conversely, the other mitochondrial signals exhibit enhancements attributed to the compression and folding of the mitochondrial membrane within autolysosomes. To determine the involvement of mitochondria in mitophagy, we consider those exhibiting high signal intensity and co-localization with lysosomes. Mitochondria that lack adjacent lysosomes or exhibit weak intensities are not considered due to insufficient evidence. Scale bar, 200 nm.

See also Figure S3.

We demonstrate our SMOP-ExSTED method on the interactions between microfilaments and lysosomes. Previous studies have reported that microfilaments play an essential role in regulating lysosomal fusion and traffic.^15,16 As the lysosomes are rear near the plasma membrane, the penetration depth of our imaging plane was ∼10–50 μm underneath the cover slip in the expanded sample. We were aware of nodal actin bundles forming around lysosomes, indicating lysosomal fissions occurring. This phenomenon generally appears in both U2OS cells (Figures S3C and S3D) and BSC-1 cells (red dashed arrows in Figures 4A, 4B, and S3E and Video S5). Interestingly, we discovered an actin structure shaped like a "zero" (Figure 4C) and several other actin bundles in similar shapes (red arrows in Figure 4A). There is no adjacent lysosome, which suggests these actin filaments were interacting with other organelles (red arrows in Figure 4).

Mitochondrial-lysosomal interactions are frequent and critical for apoptosis, autophagy, and many other physiological and pathological processes.¹⁷ It is not surprising that most lysosomes and mitochondria are separated if the cells are in a normal state before expansion (Figure S3F). However, although quite rare, their contacts do exist (Figure S3G). We further focused on mitophagy, a pathway for lysosomal degradation of mitochondria that occurs under cellular stress.¹⁸ For the study of this process using our nanoscope, living cells were stained with Atto 647N and MAA-Lyso-AF594 and then starved for 4 h before fixation and expansion. The results indicate that many mitochondria are crushed and enveloped in autolysosomes after starvation (Figures 4D–4F). Notably, the specific autophagic flux appears to be cell-line dependent. In our studies, the statistics from U2OS and BSC-1 cells are distinct. In U2OS cells, lysosomal protein clusters are much bigger than their adjacent mitochondria in autolysosomes. Yet, in BSC-1 cells, mitochondria occupy more space in autolysosomes than the labeled lysosomal proteins (Figures 4D–4F). Our results also reinforce the need to use high-resolution 3D data for analysis. For example, in Figure 4F, we tend to make incorrect judgments about spatial relationships between lysosomal proteins and mitochondria in a particular perspective or at a relatively low resolution (Figure 4F).

Discussion

In summary, SMOP-ExSTED nanoscopy reveals ultrafine structures with sub-10-nm isotropic resolution and demonstrates small-molecule organic probes for nanoscopy techniques. Compared with other approaches suitable for cell imaging, SMOP-ExSTED nanoscopy is especially advantageous with regard to 3D resolution. All STED instruments fundamentally stem from a confocal architecture. Thus, the SMOPs presented in this work are equally applicable to the ordinary confocal microscope (see Figure S1). Stimulated emission is only one of several mechanisms to realize reversible fluorescence inhibition. This underscores the potential of applying SMOPs to other nanoscopy techniques, such as single-molecule localization microscopy, to pursue comparable resolution. Notably, the MAA unit in our SMOPs binds to the hydrogel network through cross-linking of olefin. Therefore, our SMOPs are also applicable to other expansion microscopy techniques, including 10–20× expansion methods.^19,20,21

The results presented here were all obtained in fixed samples because the physical expansion process is not compatible with living cells. However, this does not hinder the potential to employ SMOPs for live-cell labeling. For those SMOPs which are not cell-permeable, intracellular targeting in living cells can be achieved in conjunction with cell-penetrating peptide.^8,22 Combined with cryo-fixation,²³ instead of chemical fixation, more precise and original morphology of the cells could be revealed using SMOPs. Given the modular design of our SMOPs demonstrated in this paper, the labeling of other structures can also be achieved expediently via recognition unit replacement. With the anticipation of developing more recognition units, the competitive nature of SMOPs in expansion and super-resolution microscopy will be further enhanced in the future. With the advantages of good biocompatibility, easily modifiable molecular structures, and adjustable fluorescence performance, SMOPs promise broad prospects for biomedical applications in 3D super-resolution imaging.

On the instrumental side, we employed adaptive optics to correct depth-dependent aberrations. The realization of the adaptive optics function in a nanoscope is reasonably flexible. For example, the deformable mirror in our instrument can readily be replaced by other devices, such as a liquid lens. Alternatively, while our scheme is sensorless but relies on an iterative process to obtain the aberrated wavefront, one may opt for a wavefront sensor if the overall dimensions of the nanoscope are not a major concern.

In conclusion, our development provides access to the 3D organization of subcellular structures at the nanoscale by fluorescence microscopy and represents an important step toward understanding open questions in cell biology.

Limitations of the study

In this work, we developed a multi-color nanoscopy based on small-molecule fluorescent probes in ExSTED, enabling sub-10-nm resolution. Achieving higher resolution should still be possible by further expanding the samples or by further improving the laser dose for stimulated emission. However, either strategy would result in a trade-off between the resolution and the SBR. This problem can partially be solved by binding more dyes onto the SMOPs, but the dye number is still limited by the possible binding sites and quenching effect within a limited probe size. From our perspective, our SMOPs are unlikely to yield sub-2-nm resolution without subsequent image processing,²⁴ a value that approaches the sizes of the labeled proteins themselves.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse monoclonal anti-Actin	Thermo Fisher Scientific	Cat#MA511869; RRID:AB_11004139
Mouse monoclonal anti-beta Actin	Abcam	Cat#ab8226; RRID:AB_306371
Goat anti-mouse STAR RED	Abberior	Cat#STRED-1001-500UG
Mouse anti-α tubulin	Beyotime Biotechnology	Cat#AF2827
Mouse anti-β tubulin	Beyotime Biotechnology	Cat#AF2835
Biological samples
Mouse coronal brain slices (30 μm; Male C57BL/6 mice, 8–12 weeks, 20–25 g)	Gong et al.25	N/A
Chemicals, peptides, and recombinant proteins
MAA-Actin-TMR	This paper	N/A
MAA-Actin peptide	This paper	N/A
MAA-Lyso peptide	This paper	N/A
Alexa Fluor 647 NHS ester	Thermo Fisher Scientific	Cat#A20006
Atto 647N NHS ester	Sigma-Aldrich	Cat#18373
Cy5 NHS ester	GE Healthcare	Cat#PA15101
STAR RED NHS ester	Abberior	Cat#STRED-0002-1MG
Atto 565 NHS ester	Sigma-Aldrich	Cat#72464
Cy3B NHS ester	GE Healthcare	Cat#PA63101
STAR ORANGE NHS ester	Abberior	Cat#STORANGE-0002-1MG
Alexa Fluor 594 NHS ester	Thermo Fisher Scientific	Cat#A20004
Alexa Fluor 488 NHS ester	Thermo Fisher Scientific	Cat#A20000
Atto 495 NHS ester	Sigma-Aldrich	Cat#00379
LysoTracker Green	Thermo Fisher Scientific	Cat#L7526
PV-1 peptide vehicle	Zhang et al.17	N/A
Alexa Fluor 488 Phalloidin	Thermo Fisher Scientific	Cat#A12379
Acryloyl-X, SE (AcX)	Thermo Fisher Scientific	Cat#A20770
Sodium Acrylate	Santa Cruz Biotechnology	Cat#sc-236893C
Acrylamide	Sigma-Aldrich	Cat#A9099-100G
N,N′-Methylenebisacrylamide	Sigma-Aldrich	Cat#M7279-100G
Ammonium Persulfate (APS)	Sigma-Aldrich	Cat#A3678-100G
N,N,N′,N′-Tetramethylethylenediamine (TEMED)	Sigma-Aldrich	Cat#T7024-100ML
proteinase K (proK)	Sigma-Aldrich	Cat#P4850-5ML
Paraformaldehyde (PFA)	Electron Microscopy Sciences	Cat#15710
Formaldehyde (FA)	Thermo Fisher Scientific	Cat#28908
Glutaraldehyde (GA)	Thermo Fisher Scientific	Cat#G7651-10ML
Experimental models: Cell lines
Human: U2OS cells	ATCC	HTB-96
Monkey: BSC-1 cells	ATCC	CCL-26
Recombinant DNA
EGFP-Lifeact	Michael Davidson Lab	Addgene Plasmid #54610
Software and algorithms
ImageJ/Fiji	Schneider et al.7	https://imagej.nih.gov/ij/
NIS Elements	Nikon	https://industry.nikon.com
Excel	Microsoft	https://www.microsoft.com/zh-cn/microsoft-365/excel
MATLAB	MathWorks	https://ww2.mathworks.cn/
FluoRender	University of Utah	https://www.sci.utah.edu/cibc-software/fluorender.html
Imaris	Oxford Instruments	https://imaris.oxinst.com/
Anaconda version 3	Anaconda	https://www.anaconda.com/

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Xiang Hao (haox@zju.edu.cn).

Materials availability

This study did not generate new unique reagents.

Experimental model and study participant details

Cell lines

U2OS (human osteosarcoma cell line; female) cells were purchased from American Type Culture Collection and cultured in McCoy's 5A medium (Thermo Fisher Scientific Inc.). BSC-1 (African green monkey kidney cell line; sex unspecified; BNCC102120; BeNa Culture Collection) cells were cultured in the MEM medium (Thermo Fisher Scientific Inc.). All media were supplemented with 10% (v/v) fetal bovine serum (FBS; Thermo Fisher Scientific Inc.), and the cultures were maintained at 37°C in a humidified 5% CO₂ environment.

Tissue slices

Mouse coronal brain slices (30 μm; Male C57BL/6 mice, 8-12 weeks, 20-25 g)²⁵ were a kind gift from Dr. Yicheng Xie (Zhejiang University, Hangzhou, China) and stored in an anti-freezing solution (PBS: glycol: glycerin = 4:3:3) at -20°C before use. The animal protocols were approved by the Animal Care and Use Committees at the Zhejiang University School of Medicine and were conducted following the policies of institutional guidelines on the care and use of laboratory animals.

Method details

Design and synthesis of the probes

The actin probe MAA-Actin-TMR was synthesized by solid-phase peptide synthesis and was purified by preparative high-performance liquid chromatography (HPLC) to a purity of >95%, and its mass was confirmed by electrospray ionization mass spectrometry (EI-MS).

The other actin and lysosomal probes were constructed from two parts:⁸ the peptide one and the dye one. The peptide parts were synthesized by solid-phase peptide synthesis and dissolved in bicarbonate buffer (0.1 M, pH 8.3) at a concentration of 1 mM and stored at 4°C before conjugation. The dye part, containing N-hydroxysuccinimidyl (NHS) group, conjugates with the peptide part by forming covalent binding to the free amino groups. Alexa Fluor 647 (Thermo Fisher Scientific Inc.), Atto 647N (Sigma-Aldrich Co., LLC), Cy5 (GE Healthcare Co. Ltd), STAR RED (Abberior GmbH), Atto 565 (Sigma-Aldrich Co., LLC), Cy3B (GE Healthcare Co. Ltd), STAR ORANGE (Abberior GmbH), Alexa Fluor 594 (Thermo Fisher Scientific Inc.), Alexa Fluor 488 (Thermo Fisher Scientific Inc.), and Atto 495 (Sigma-Aldrich Co., LLC) were used here. The dyes were divided into small aliquots before use.

For the conjugation, small aliquots of the peptide part solution were added to the dye aliquot, with a mole ratio of ∼1:2. For the dyes insoluble in water, anhydrous dimethyl sulphoxide (DMSO; Sigma-Aldrich Co., LLC) was added before conjugation. After incubation overnight in the dark, the mixture was purified using Pierce C18 Spin columns (Thermo Fisher Scientific Inc.), evaporated to dryness, and dissolved in Phosphate Buffered Saline (PBS, pH7.4; Thermo Fisher Scientific Inc.) to generate a stock solution.

Living cell labeling

Before labeling with the probes, the cells were seeded in glass-bottom dishes (Φ15 mm, NEST Scientific Inc.) at a density of 2-3 × 10⁴ per well in the growth medium. After overnight incubation, the cells were washed three times with PBS.

For LysoTracker Green (50 nM; Thermo Fisher Scientific Inc.) and Atto 647N NHS ester (1.5 μM), the cells were incubated with work solutions in a 5% CO₂ atmosphere at 37°C for 30 min before the supernatant was discarded. The cells were then washed three times with PBS and immersed in phenol red-free DMEM (Thermo Fisher Scientific Inc.) before imaging.

For our lysosomal probes, work solutions at different concentrations were prepared by diluting different volumes of the stock solution with PBS to a final volume of 100 μL. The cells were incubated with work solutions at 37°C for 30 min. For MAA-Lyso-STAR ORANGE, the work solution was supplemented with 1% PV-1 peptide vehicle.¹⁷ Then, the supernatant was discarded, and the cells were washed three times before fixation or confocal imaging.

For the plasmid EGFP-Lifeact (Addgene # 54610), Cells were grown overnight in 24-well plates at 37°C in a 5% CO₂ atmosphere before use. After reaching ∼80% confluence, the plasmid was transfected into the cells using Lipofectamine 3000 (Thermo Fisher Scientific Inc.) according to the manufacturer's instructions. The cells were then seeded in glass-bottom dishes at a density of 2-3 × 10⁴ per well before fixation or confocal imaging.

Cell fixation and subsequent labeling

Three kinds of fixation buffers were used here: 4% paraformaldehyde (PFA; Electron Microscopy Sciences), 3% PFA + 0.1% glutaraldehyde (GA; Thermo Fisher Scientific Inc.), and 4% formaldehyde (FA; Electron Microscopy Sciences). The cells were incubated with the pre-warmed fixation buffer for 13 min at 37°C and washed three times with PBS. Solutions of 1% sodium borohydride (Sinopharm Chemical Reagent Co., Ltd.) was used for 7 min to quench the background when using GA.

For our actin probes and Alexa Fluor 488 Phalloidin (Thermo Fisher Scientific Inc.), staining was performed after fixation with 4% PFA. Work solutions of the probes at different concentrations were prepared by diluting different volumes (i.e., 20-100 μL) of the stock solution with PBS to a final volume of 100 μL. The cells were then incubated with the work solution of the probes in a 5% CO₂ atmosphere at 37°C for 30 min, and the supernatant was discarded. The cells were washed three times with PBS before confocal imaging or expansion.

For immunofluorescence staining, the cells were fixed with 3% PFA + 0.1% GA, quenched with 1% sodium borohydride, and incubated with 0.2% Triton X-100 + 5% goat serum for 1 h at room temperature before staining. Mouse anti-actin antibody (Cat. No. MA511869; Thermo Fisher Scientific Inc. or Cat. No. ab7291; Abcam Plc.) 1:100 in PBS overnight at 4°C and goat anti-mouse STAR RED (Abberior GmbH.) 1:200 in PBS for 1 h at room temperature. Samples were washed three times with PBS for 3 min before confocal imaging or expansion.

Dual-color staining

For dual-color imaging of F-actin and lysosomes, living cells were stained with MAA-Lyso-AF594, fixed with 4% PFA, and then labeled with MAA-Actin-STAR RED. The gelation should be immediately performed.

For dual-staining of lysosomes and mitochondria in normal cells, living cells were incubated with work solutions of MAA-Lyso-AF594 and Atto 647N NHS ester at the same time and then fixed. For mitophagy imaging, living cells were stained with Atto 647N and MAA-Lyso-AF594, washed, and incubated with growth medium without serum for 4 h fixation and expansion.

For dual-staining of tubulin antibodies and our actin probe, the cells were fixed with 3% PFA + 0.1% GA and then quenched, permeabilized, and blocked as described above. The cells were then incubated with mouse anti-α and β tubulin primary antibodies (Cat. No. AF2827 and AF2835; Beyotime Biotechnology) 1:1:50 in PBS overnight at 4°C and goat anti-mouse STAR RED (Abberior GmbH.) 1:100 in PBS for 1 h at room temperature. Samples were then washed three times with PBS for 3 min. Labeling with MAA-Actin-AF 594 afterward was performed as described above.

Cell expansion

For the actin probes and antibody staining, the cells were fixed, incubated with the succinimidyl ester of 6-((acryloyl)amino)hexanoic acid (AcX; Thermo Fisher Scientific Inc.) solution (1:100 in PBS), and washed before labeling (Figure S1C). For the other labeling methods, the cells were transfected or stained and then fixed (Figure S1C). Notably, for our actin probes, the gelation should be performed immediately after staining, due to the noncovalent binding between Lifeact and F-actin.

The gelation solution was prepared by mixing 180 μL of the U-ExM Monomer solution²⁶ (19% (wt/wt) sodium acrylate (SA; Santa Cruz Biotechnology), 10% (wt/wt) acrylamide(AA; Sigma-Aldrich Co., LLC), 0.1% (wt/wt) N,N′-methylenebisacrylamide (BIS; Sigma-Aldrich Co., LLC) in 1× PBS), 10 μL of 0.5% ammonium persulfate (APS; Sigma-Aldrich Co., LLC), and 10 μL of 0.5% TEMED tetramethylethylenediamine (TEMED; Sigma-Aldrich Co., LLC) in a 1.5-mL Eppendorf tube at 4°C. APS was added last. Then the solution was quickly transferred into the pre-cooled glass-bottom dishes where cells were cultured. The dishes were then shifted to 37°C in the dark for 1 h for gelation. The sample dishes were then filled with the pre-heated denaturation buffer (200 mM Sodium dodecyl sulfate (SDS; Sigma-Aldrich Co., LLC), 200 mM sodium chloride (NaCl; Sigma-Aldrich Co., LLC), and 50 mM Tris (Thermo Fisher Scientific Inc.) in ultrapure water, pH 9) and incubated at 70°C for 1 h. After denaturation, gels were transferred into a new Petri dish filled with ultrapure water and incubated for 20 min. This step was repeated 3-5 times in fresh water until the size of the gel plateaued. The expanded gels were then cut into appropriate sizes with custom-made knives, sopped with absorbing paper to remove the exterior moisture, and placed in Nunc glass-bottom dishes (Φ 27 mm, Thermo Fisher Scientific Inc.) before imaging.

For the comparison in Figure S1A, denaturation buffer with or without proteinase K (proK; diluted 1:100 to 8 units/mL; Sigma-Aldrich Co., LLC) were incubated at room temperature overnight, or at 70°C or 95°C for 1 h.

Tissue labeling and expansion

The mouse brain slices were washed three times with PBS, incubated with AcX solution overnight at room temperature, and washed before use. For SMOP labeling, the slices were then incubated with the work solution of the probes in a 5% CO₂ atmosphere at 37°C for 30 min. The supernatant was discarded. The slices were washed thrice with PBS for 15 min before imaging or expansion.

The operation time for slices was extended in contrast to that for cells. Slices were incubated with the gelation solution at 4°C for 15-30 min and then moved to a 37°C incubator for 2 h. The excess gel was trimmed off, and the slices were transferred into the pre-warmed denaturation buffer and incubated at 70°C for 2 h. The subsequent expansion is identical to that for the cell samples.

Confocal laser scanning microscopy

The images were obtained using a confocal laser scanning microscope (C2, Nikon, Inc.) equipped with a 60×/1.49 numerical aperture oil-immersion objective lens.

STED microscopy

The STED images were obtained using a custom 3D-STED nanoscope¹² equipped with an Apochromat 100×/1. 35 numerical aperture, silicone-oil objective lens (Olympus).

For the depletion beam, we applied two orthogonal polarization components with a pulse delay from the same 775-nm laser (MPB) to generate the lateral (STED_xy) and the axial (STED_z) depletion PSFs at the sample plane without interfering with each other. To allow nearly isotropic resolution, the STED_xy beam power is 1/4 that of the STED_z beam.¹² A double-pass spatial light modulator (SLM, X13139-02, Hamamatsu) configuration²⁷ is used to encode the vortex or top-hat phase on each polarization component.

For the excitation beams, the light from a supercontinuum laser (SuperK Extreme EXR-20, NKT) is merged with the depletion beam via a dichroic mirror (Chroma). The depletion and excitation beams then scan across the sample using a 16 kHz resonant mirror (EOPC) combined with two synchronized galvanometer mirrors (Cambridge Technology Inc.).²⁸ For the detection path, fluorescence is collected by the same objective lens, descanned, and directed to a single-photon counting avalanche photodiode (APD, Excelitas). The measured fluorescence signals from the APDs are relayed to circuit boards for gated detection.²⁹ Meanwhile, a deformable mirror (DM, Boston Micromachines) is placed into the detection path to correct the sample-induced aberrations.

For multi-color imaging, two pulsed excitation beams and one depletion beam are used at 590, 647, and 775 nm wavelengths, respectively, with a 78-MHz repetition rate.

After correcting aberrations in the optical system, the shape of DM is further adjusted to compensate for the sample-induced aberrations. During the imaging, the DM is further tuned to automatically compensate for the depth-dependent aberrations,³⁰ an essence for imaging thick samples.

Image analysis

The confocal images were analyzed with NIS Elements (Nikon, Inc.) and ImageJ/Fiji software (National Institutes of Health). The STED images were analyzed with ImageJ/Fiji software. Pearson's coefficient was quantified using the Coloc 2 plugin, and the image resolutions were measured with the Image Decorrelation Analysis plugin¹¹ for ImageJ/Fiji. The SBR analysis was accomplished with the NoiSee plugin,¹⁰ and the SNR was analyzed by dividing the max intensity by divided by the standard deviation (noise) in the images. All the Box plots and histograms were drawn with Excel (Microsoft, Corp.) and the Gaussian fitting was performed with MATLAB (MathWorks, Inc.). The rendering of the 3D images was finished with FluoRender (University of Utah) and Imaris (Oxford Instruments).

Expansion factors and distortions were measured by comparing the pre-expansion and post-expansion confocal images of U2OS cells stained with MAA-Actin-TMR. The images were processed with the Python script given in the previous report,²¹ and then the matched images came out with the correlation coefficient, expansion factor, and rotation angle. The similarity-registered images were then processed with the bUnwarpJ plugin.³¹ The post-expansion images were non-rigid-registered to the pre-expansion one, and the distortions for every pixel were output.

Quantification and statistical analysis

For Figures 1F–1H, data are represented as mean ± S.D. (n ≥ 5 field of views for each test). NA, not available; ns, P > 0.05; ∗, P < 0.05; ∗∗, P < 0.01; ∗∗∗, P < 0.001; ∗∗∗∗, P < 0.0001; two-tailed t-test, statistics were performed using Excel.

For Figure S1A, the intensity data are represented as mean ± SD (n ≥ 5 field of views for each test).

For Figure S1B, the intensity data are represented as mean ± SD (n = 4 gels for each test). ns, P > 0.05; ∗, P < 0.05; ∗∗, P < 0.01; two-tailed t-test, statistics were performed using Excel.

For Figure S1E, the root-mean-square (RMS) error was calculated across different measurement lengths (n = 39 field of views; N = 6 cells).

For Figure 4E, the frequency was calculated by dividing the number of events that different types of interactions occurred by the total number of interaction events (total number = 55 for U2OS cell and 27 for BSC-1 cell).

Published: August 14, 2023

Data and code availability

• •All data reported in this paper will be shared by the lead contact upon request.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.