1. Introduction

Super-resolution microscopy has revolutionized biological research, offering molecular and cellular insights beyond the limitations traditionally imposed by the optical diffraction [1–3]. Over the past two decades, various super-resolution methodologies have emerged, each characterized by distinctive principles for resolution enhancement. In brief, these principles include illumination modulation (e.g., STED [4,5], SIM [6–9]), single-molecule localization (e.g., SMLM [10–13]), fluctuation analysis (e.g., SOFI [14], SRRF [15,16]), and physical sample manipulation (e.g., ExM [17,18]). These methodologies exhibit unique characteristics that have facilitated their extensive application spanning basic and translational disciplines, such as cell biology [19], immunology [20], neuroscience [21], and pathology [22].

Notably, amongst these techniques, structured illumination microscopy (SIM) stands out as a particularly versatile method for effective resolution doubling, longitudinal observation, and easy sample preparation [23]. Conventionally, SIM techniques employ patterned illumination to induce spatial-frequency mixing, thus realizing the retrieval of high-spatial frequencies containing structural details beyond the diffraction limit [9,24]. However, despite major advances, traditional interference-based SIM methods are contingent upon the accurate calibration of the illumination patterns, making them susceptible to aberrations and reconstruction artifacts [23]. Thus far, various efforts have been made to strengthen the realm of SIM technology [25–27], and in particular, increasing attention has been given to identifying reliable, cost-efficient, and adaptable SIM strategies. One such advance lies in image-scanning microscopy (ISM), which utilizes confocal scanning of diffraction-limited laser focus or multiple foci, achieving comparable resolution, enhanced optical sectioning, and signal-to-noise ratio (SNR) [28–35]. Nevertheless, current ISM configurations necessitate complicated alignment and implementations such as confocal spinning disks, galvanometric mirrors, or digital micromirror devices, despite recent developments to simplify the scanning mechanism by using translational sample stage while keeping the illumination static [36,37].

Alternatively, speckle-based SIM approaches have been exploited, which leverage the inherent statistical properties of speckles without a priori knowledge of illumination patterns to reconstruct super-resolution images [38–41]. This strategy, as a result, substantially simplifies alignment procedures and demonstrates robustness to optical aberrations and sample scattering [38,42]. In practice, optical elements such as spatial light modulators [38,43,44], digital micromirror devices [32,41,45], ground glass diffusers [39,46], and metamaterials [47,48] have been utilized to generate speckle illumination. Digital devices present precise manipulation at the pixel level, albeit at a higher expense; on the other hand, ground glass diffusers are cost-efficient but lack the capacity to adaptably modulate optical characteristics, while metamaterial-assisted illumination requires specialized fabrication facilities and protocols. Therefore, the development of new diffusive materials that provide an optimal balance between adaptability and cost efficiency is still an emerging area within the field of speckle-based SIM. In this context, hydrogels have emerged as a promising solution, as the inherent nanostructural heterogeneity of polymer networks contributes to the effective light scattering [49]. It has been reported that speckle patterns produced by hydrogels can be used to monitor the gelation process through dynamic light scattering [50].

In this study, we introduce hydroSIM, a super-resolution imaging technique incorporating hydrogel diffusers to generate random speckle patterns for illumination. Unlike conventional speckle-based systems, hydroSIM markedly simplifies imaging apparatus, realizing plug-in super-resolution microscopy via a commonly used epi-fluorescence platform. In particular, the diffusers are produced by embedding titanium dioxide (TiO₂) microparticles within an acrylamide-alginate hydrogel matrix, an optimized protocol to suppress ballistic transmission and ensure the generation of diffraction-sized speckles. Notably, this study demonstrates that our hydrogel diffusers can be readily inserted into the illumination pathway, mitigating the necessity for complex alignment procedures. The hydroSIM system is validated across a range of phantom and biological specimens, presenting effective 3D resolution doubling, optical sectioning, and image contrast. We foresee hydroSIM emerging as a user-friendly and versatile super-resolution approach, poised to significantly advance cell biological research.

2. System design and methods

2.1 Experimental setup

As depicted in Fig. 1(a) and S1, the hydroSIM system is constructed by incorporating a hydrogel diffuser into a standard epi-fluorescence microscope (Eclipse Ti2-U, Nikon), equipped with multiple laser lines (488 nm and 647 nm, MPB Communications) and objective lenses (CFI Plan Apochromat Lambda 100× Oil, 1.45 NA and CFI Plan Fluor 40× Oil, 1.3 NA, Nikon). The hydrogel diffuser was implemented on a motorized rotation mount (ELL14 K, Thorlabs) and placed conjugated close to the back focal plane of the objective lens (Fig. 1(b) and S1). Notably, we verified the optical transmissivity of the hydrogel diffuser is nearly an order of magnitude higher than the corresponding ground glass diffuser in our system without adding additional lenses, attributing to the inhomogeneous microstructures and scattering elements (Fig. S2). Speckle patterns are subsequently demagnified and delivered at the diffraction limit to the sample plane through the objective lens. Fluorescent signals generated by the speckle illumination are captured by an sCMOS camera (ORCA-Flash 4.0, pixel size = 6.5 µm, Hamamatsu). The axial position of the objective lens is controlled by a piezo nano-positioner (Nano-F100S, MCL) and can be synchronized with the camera for 3D acquisition.

 

Fig. 1. Speckle illumination microscopy with a hydrogel diffuser (hydroSIM). (a) Optical setup of hydroSIM and speckle pattern generated by a hydrogel diffuser. The orange dashed lines indicate the light rays without the hydrogel diffuser. (b) Photograph of the hydrogel diffuser. (c) Fabrication process of the hydrogel diffuser. PAAm-alginate hydrogel is prepared by mixing two components. 5-µm TiO₂ particles are added to increase the scattering effect. (d) Super-resolution image formation of hydroSIM. The reconstruction process includes raw image acquisition, averaging and deconvolving raw images, calculating covariance between speckles and raw images, and final deconvolution. The diffraction-limited image is obtained by averaging and deconvolving raw images. Scale bar: 10 µm (a).

Download Full Size | PDF

2.2 Fabrication of hydrogel diffusers

The hydrogel diffuser is fabricated using a polyacrylamide-alginate (PAAm-Alginate) hydrogel matrix, a material previously studied for stretchable optical fibers for optogenetics and sensing [51,52]. To enhance the scattering property and attenuate the ballistic transmission, 5-µm titanium dioxide (TiO₂) particles are incorporated into the hydrogel diffuser matrix (Fig. 1(c)). In brief, the alginate hydrogel component is synthesized by dissolving 0.16 wt% (weight percent) ammonium persulfate (APS; all obtained from Sigma Aldrich), 0.24 wt% calcium sulfate (CaSO₄), and 2 wt% sodium alginate into deionized water. Subsequently, 40 wt% acrylamide (AAm), 0.06 wt% N, N′-Methylenebisacrylamide (MBAA), 0.1 wt% N,N,N′, N′-Tetramethyl ethylenediamine (TEMED), and 0.01 wt% 5-µm TiO₂ particles are sequentially added to the alginate solution to formulate the hydrogel precursor. The precursor is then injected into a 1-inch lens tube (SM1L05, Thorlabs) to a thickness of 3 mm. The hydrogel diffuser is formed following a 30-minute exposure to near ultraviolet (UV) light at a wavelength of 405 nm.

2.3 Super-resolution image formation

The hydroSIM technique utilizes the fundamental principle of speckle illumination to facilitate the capture of spatial frequency components that are otherwise beyond detection and correspond to high-resolution details [39]. Theoretically, the achievable resolution of hydroSIM is determined by the range of spatial frequency contained within both the detection (${f_{det}}$) and illumination (${f_{ill}}$) paths as ${f_{max}} = {f_{det}} + {f_{ill}}$ [47]. Based on the convolution theorem, as the excitation and emission light utilize the full frequency bandwidth supported by the objective lens, the reconstructed image from the recorded images can reveal a two-fold improvement in spatial details over the wide-field counterparts. The detailed derivation of speckle characteristics can be found in Supplement 1.

The super-resolution image reconstruction process follows four main steps, as illustrated in Fig. 1(d), containing the acquisition of raw images, computation of the illumination pattern, creation of the covariance image, and execution of the final deconvolution [41]. In particular, raw images are recorded by the camera while the hydrogel diffuser rotates, and cumulative speckle patterns produce a uniformly illuminated field given sufficient frames (usually 100-200 frames). Therefore, a diffraction-limited wide-field image can be derived by averaging these raw images and deconvolving the resultant mean image using the system point-spread function (PSF). Given the high correlation between the intensity maps of the speckles and the corresponding raw images, calculating their pixel-wised covariance leads to enhanced fluorescent signals and suppressed noise, which thus improves the resolution and signal-to-noise ratio (SNR). Lastly, the Richardson-Lucy deconvolution is applied to the intermediate covariance image, typically over 5-10 iterations, to further increase the resolution and sectioning. For 3D hydroSIM, a volumetric intermediate image is first reconstructed, followed by 3D deconvolution using DeconvolutionLab2 software on the image stack, resulting in a 3D super-resolution volume [53]. Our lab-written algorithm for the automatic correction of the sCMOS noise (ACsN) [54] is used to enhance low-SNR raw images and mitigate noise artifacts.

3. Results

3.1 Characterization of hydroSIM

To characterize hydroSIM, we first synthesized a Siemens star with 60 spokes and conducted numerical illumination of the simulated object by 200 varied speckle patterns. Poisson noise was added accordingly to approximate actual experimental conditions. As shown in Fig. 2(a), compared with the wide-field ground truth and its deconvolved image, the intermediate and hydroSIM images exhibit improved resolution and SNRs of the synthetic pattern. The resolution and cutoff frequency (K_cd) of each image were quantitatively assessed using the decorrelation analysis [55] (Fig. 2(b)). The results indicated the values of DeARes = 138 nm, K_cd = 14.43 µm^-1 for hydroSIM, ∼2× enhancement in comparison with the wide-field image (DeARes = 265 nm, K_cd = 7.54 µm^-1). In addition, the frequency domain of the image data was analyzed (Fig. 2(c)). The cutoff frequency of the wide-field image (4.40 µm^-1) agrees with the theoretical value of 4.54 µm^-1, and the hydroSIM image displayed a consistent frequency doubling, expanding the bandwidth to 8.3 µm^-1.

 

Fig. 2. Characterization of hydroSIM. (a) Wide-field (WF), deconvolved wide-field (dec-WF), intermediate (INT), and hydroSIM images of a Siemens star with 60 spokes, simulated for the 515-nm emission wavelength and 100×,1.45 NA objective lens. (b) Decorrelation analysis of the four images in (a), resulting in the cut-off frequency (K_cd) and spatial resolution (DeARes). (c) Fourier analysis of the four images of (a), illustrating the cut-off frequency (black dashed lines). (d,e) Wide-field (d) and hydroSIM (e) images of 100-nm fluorescent beads (emission wavelength = 515 nm). (f-h) Zoomed-in images of the boxed regions as in the wide-field, intermediate, and hydroSIM images in (d,e). (i) Intensity profiles (black, red, blue) along the red dashed lines in (f-h), respectively. (j) Intensity profiles (black, red, blue) along the yellow dash lines in (f-h), respectively. (k) Wide-field (top right) and hydroSIM (bottom left) images of 6-µm dual-color surface-stained fluorescent microspheres (emission wavelengths = 515/680 nm). (l,m) Zoomed-in images of the boxed regions in (k). (n,o) Intensity profiles along the yellow dashed lines in (l,m) in the green (n) and dark red (o) channels. Scale bars: 10 µm (d, k), 500 nm (f), 5 µm (l).

Download Full Size | PDF

Next, we imaged 100-nm fluorescence beads (T7279, ThermoFisher) using hydroSIM. Experimentally, the microspheres were illuminated by 200 speckle patterns each at a 5-ms exposure time using the 488-nm laser (Fig. 2(d-j)). As seen, the resulting hydroSIM images exhibited the full-width at half-maximum (FWHM) values of individual sub-diffraction-limited microspheres at ∼120 nm, compared to the wide-field measurement of 235 nm (Fig. 2(i)). The improved resolution of hydroSIM permits the visualization of microspheres positioned as close as 195 nm (Fig. 2(j)), which was verified across multiple color spectra (Fig. S3). Consistent improvement was also demonstrated using 1-µm fluorescent thin-ring microspheres (F14791, ThermoFisher; Fig. S4). Furthermore, we demonstrated the multicolor imaging capability of hydroSIM utilizing the 488-nm and 647-nm excitations of 6-µm thin-ring fluorescent microspheres (F14806, ThermoFisher), captured with 200 frames per channel at a frame rate of 200 Hz. The reconstructed hydroSIM images showed a precise alignment of two co-stained structures, indicating the achromatic ability of the hydrogel diffuser (Fig. 2(k-m)). In contrast to wide-field microscopy, the super-resolution hydroSIM results displayed enhanced optical sectioning and resolution, revealing two adjacent microspheres separated 180-190 nm in both spectral channels (Fig. 2(n) and (o)).

3.2 Imaging microtubules in mammalian cells

Next, we imaged immunostained microtubules in HeLa cells using hydroSIM with the 100× objective lens and 488-nm laser. The super-resolution cell images were formed with 100 speckle-illuminated frames at an exposure time of 100 ms (Visualization 1). The results of hydroSIM delineated microtubule structures, especially within densely packed regions, with increased clarity and sectioning (Fig. 3(a) and (b)). As seen, individual microtubule filaments, spaced as close as 134 nm, were discerned with hydroSIM, indicating a two-fold resolution enhancement over conventional diffraction-limited images (Fig. 3(c) and (d)). Accordingly, the decorrelation analysis identified the resolutions of the wide-field, deconvolved wide-field and hydroSIM images at 261 nm, 239 nm and 140 nm, respectively (Fig. 3(e-f), Fig. S5), consistent with the caliber measurements in Fig. 2. Furthermore, we imaged microtubules in HeLa using the lower-magnification 40× objective lens. The cells were immunostained in dark red and illuminated with the 647-nm laser using 100 speckle-illuminated frames at an exposure time of 100 ms. As observed, hydroSIM demonstrated higher resolution and optical sectioning as opposed to conventional wide-field microscopy (Fig. 3(g)). Specifically, hydroSIM images exhibited the FWHM values of individual microtubule filaments at 176 nm (over 385 nm in the wide-field images), as well as the clear visualization of otherwise undiscernible filaments in close proximity (Fig. 3(h-k)). These results validated the super-resolution ability of hydroSIM, as well as its usability across multiple objective imaging scales and color channels.

 

Fig. 3. Imaging microtubules in HeLa cells using hydroSIM. (a,b) Wide-field (a) and hydroSIM (b) images of immunostained microtubules in HeLa cells taken by the 100×, 1.45 NA objective lens. (c-f) Zoomed-in images of the boxed regions as marked in (a,b). The curves in (c,d) show the intensity profiles resolving filaments as close as 134 nm along the corresponding yellow dashed lines. (g) Wide-field (bottom left)and hydroSIM (top right) images of immunostained microtubules in HeLa cells taken by the 40×, 1.3 NA objective lens. (h, i) Corresponding zoomed-in wide-field (h) and hydroSIM (i) images of the boxed region in (g). (j) Intensity profiles along the red dashed line in (h, black) and (i, red). (k) Intensity profiles along the yellow dash lines in (h, black) and (i, red). Scale bars: 10 µm (a, g), 1 µm (c, e, h).

Download Full Size | PDF

3.3 Multicolor and live-cell imaging of actins and mitochondria

We subsequently imaged actins and mitochondria in living HeLa cells using hydroSIM. Given the critical role of actin filaments in mitochondrial dynamics—a relationship pivotal to understanding the cellular metabolism [56] —it is challenging to discern the delicate structures of these organelles using traditional widefield microscopy. Here, we captured actins labeled with CellMask Green, followed by mitochondria with Mitotracker Deep Red, at a frame rate of 200 Hz. Rolling reconstruction with an overlap of 100 frames was employed to better display the continuous movement [57]. hydroSIM recorded the subcellular organelles for thousands of frames without noticeable photobleaching (Visualization 2). As seen, the reconstructed super-resolution images revealed both subcellular organelles with significantly improved contrast and resolution (Fig. 4(a) and (b), Fig. S4). We employed Fourier ring correlation (FRC) [58,59] and decorrelation analysis (DeARes) [55] on the actin channel, obtaining a consistent characterization of the super-resolution image resolution of 140-160 nm. Notably, actin filaments separated 181 nm apart were well resolvable (Fig. 4(c) and (d)), and hydroSIM effectively captured mitochondrial movement without motion blur (Fig. 4(e)). These results demonstrate the capacity of hydroSIM for visualizing subcellular structures and dynamics with low photodamage, rapid recording, and exceptional clarity.

 

Fig. 4. Two-color imaging of actins and mitochondria in living HeLa cells using hydroSIM. (a,b) Wide-field (a) and hydroSIM (b) images of actins (green) and mitochondria (red) in live HeLa cells. (c,d) Zoomed-in images of the boxed regions as marked in (a, b). The curves in (c, d) show the intensity profiles of actins along the yellow dashed lines, showing the resolution of filaments separated by 181 nm. (e) Time-lapse images of mitochondrial movement over a period of 15 sec. The yellow arrows facilitate a better visualization of representative moving components. Scale bars: 10 µm (a), 1 µm (c), 2 µm (e).

Download Full Size | PDF

3.4 3D Multicolor imaging of mitochondria and peroxisomes

Lastly, the hydroSIM system was expanded to encompass 3D super-resolution imaging of cellular details. Specifically, we targeted mitochondria and peroxisomes within HeLa cells, whose interactions have been known as a critical underlying aspect of cellular metabolism and signaling pathways [60,61]. The 3D image stacks were acquired sequentially by assembling speckle-illuminated frames at each depth and then stepping the objective lens axially across the specimen. Each stack was processed individually, following the same procedures to form 2D intermediate images. The final volume of the specimen was generated by the 3D deconvolution of the whole intermediate stacks using a simulated 3D PSF, resulting in the 3D super-resolution visualization of the cellular space. In practice, we captured 200 speckle frames with 5-ms exposure time in at each depth with an axial step size of 130 nm between layers. The reconstructed 3D hydroSIM image revealed the subcellular organelles with enhanced contrast, sectioning, and clarity (Fig. 5(a) and (b)). In particular, the hydroSIM images of individual, typically submicrometer-sized peroxisomes showed FWHM values of 120-150 nm and 300-400 nm in the lateral and axial dimensions, respectively, presenting a two-fold improvement over the diffraction limit in all three dimensions (Fig. 5(c-h) and Fig. S7). The multicolor images displayed the fine structural interactions between mitochondria and peroxisomes spaced as narrow as 450-550 nm, unveiling the subcellular features that are otherwise less distinguishable from conventional wide-field microscopy (Fig. 5(i-p)).

 

Fig. 5. Three-dimensional (3D) imaging of mitochondria (MT) and peroxisomes (PR) in HeLa cells using hydroSIM. (a,b) Maximum intensity projection of wide-field (a) and hydroSIM (b) images across a depth range of 4 µm. Depth information for the two organelles is coded as indicated in the color scale bar in (a). (c,f) Zoomed-in images of the boxed peroxisome in (a,b). (d,e,g,h) XZ (d,g) and YZ (e,h) views of the peroxisome in (c,f). (c-h) exhibit a nearly two-fold improvement of the FWHM values in the 3D hydroSIM images. (i,k) Zoomed-in images of the boxed regions containing mitochondria (red) and peroxisomes (green) as marked in (a,b). (j,l) XZ views of (i, k), respectively. (m,o) Zoomed-in images of the boxed regions containing mitochondria (red) and peroxisomes (green) as marked in (a, b). (n, p) YZ views of (m, o), respectively. Scale bars: 10 µm (a), 500 nm (c, f, i, m).

Download Full Size | PDF

4. Conclusion

In summary, we have developed hydroSIM, a 3D super-resolution system based on structured illumination using a hydrogel diffuser. The hydroSIM system distinguishes itself from traditional speckle-based methods by inserting cost-effective, easy-to-fabricate diffusive hydrogel material into commonly available epi-fluorescence microscopes without complex alignment or system modification. Incorporated with lab-written algorithms, hydroSIM presented effective 3D resolution doubling and enhanced optical sectioning and image contrast of various biological samples. The method can be further expanded by integrating a variety of single-cell techniques [62–64], super-resolution strategies [25,27,38,65], and computational frameworks [57,66–68]. We anticipate hydroSIM to become a widely adopted, user-friendly super-resolution method, offering substantial contributions to a wide range of cell biological discoveries.