(C) PLOS One
This story was originally published by PLOS One and is unaltered.
 . . . . . . . . . .



Highly accurate image registration for 3D multiplexed cyclic imaging using dense labeling in expandable tissue gels [1]

['Hyunwoo Kim', 'Department Of Materials Science', 'Engineering', 'Korea Advanced Institute Of Science', 'Technology', 'Daejeon', 'Republic Of Korea', 'Joon-Goon Kim', 'Graduate School Of Medical Science', 'Jueun Sim']

Date: 2025-07 

Multiplexed cyclic imaging in expandable tissue gels has been extensively studied to visualize numerous biomolecules at a nanoscale resolution in situ. Previous studies have employed sparse labels, such as DAPI or lectin staining, as registration markers. However, these sparse labels do not adequately capture the full extent of deformation across the entire region of interest. To overcome this challenge, we propose the use of dense labels, specifically fluorophore N-hydroxysuccinimide (NHS)-ester staining, as registration markers to achieve highly accurate image registration. We first tested several NHS-functionalized fluorophores as fiducial markers and identified the proper candidates for three-dimensional (3D) multiplexed cyclic imaging. We analyzed the registration accuracy between DAPI and NHS-ester staining and illustrated that dense label-based registration provides a more accurate registration performance. In the multiplexed imaging of expanded specimens, we observed that repetitive expansion/shrinking processes and chemical treatments for signal elimination can induce 3D nonlinear distortion. This sample distortion can be mitigated by re-embedding the tissue gel or replacing the chemical de-staining process with photobleaching-based signal removal or computational signal unmixing. With such an optimized experimental setup, we demonstrated 3D multiplexed cyclic imaging with nanoscale precision image registration. Finally, we prove that dense biological structures, such as actin, can be used as registration markers to achieve high registration accuracy. We anticipate that the proposed dense labeling strategy will overcome the technical limitations of multiplexed cyclic imaging in expandable tissue gels, offering high-precision registration. We expect it to be widely adopted by the biological and medical communities.

Competing interests: The authors have read the journal’s policy and the following competing interests: J.-B.C., J.S., Y.-G.Y., and I.C. are co-inventors on patent applications for high accuracy image registration. H.K., H.N., S.B., Y.-G.Y., and J.-B.C. are co-inventors on patent applications for a multiplexed imaging technique. I.C., J.-B.C., and Y.-G.Y. are co-founders, shareholders, and employers of a company specializing in various imaging services; this research was conducted independently and is not affiliated with the company. J.-G.K., H.S., J.K., D.-H.S., and T.K. have declared that no competing interests exist.

Funding: This study was primarily supported by the National Research Foundation of Korea (NRF) with grants funded by the Korean government (MSIT) (NRF-2022R1A5A6000846 and RS-2024-00406240 to J.-B.C.; RS-2023-00209473 to Y.-G.Y.; RS-2023-00264980 and RS-2020-NR052130 to T. K.), the Bio & Medical Technology Development Program of the NRF with a grant funded by the Korean government (MSIT) (RS-2021-NR056586 to J.-B.C. and Y.-G.Y.), and the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (RS-2023-00302794 to J.-B.C.; RS-2024-00438788 to T.K.). Work involving expansion microscopy using DNA-conjugated antibodies was supported by the Samsung Research Funding & Incubation Center for Future Technology (grant SRFC-IT1702-09 to J.-B.C. and Y.-G.Y.). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Here, we introduce a dense label-based image registration technique for 3D super-resolution multiplexed cyclic imaging. This offers a substantially high level of registration accuracy, thus addressing the inherent drawbacks of conventional image registration methods based on sparse labels. The dense labeling of the target sample was realized by staining the sample with N-hydroxysuccinimide (NHS)-functionalized fluorophores, targeting the amine groups abundant in proteins. Several fluorophore NHS-esters were tested to identify suitable candidates for fiducial markers in multiplexed cyclic imaging. The optimal fluorophore NHS-esters should possess the following characteristics: (1) high fluorescent signals in all types of specimens, (2) heterogeneous structures at a scale of tens of nanometers, (3) resistance to photobleaching, and (4) compatibility with cyclic staining without any spectral shift, even after multiple rounds of staining, imaging, and signal removal. Based on these criteria, we screened various types of fluorophores and found the optimal ones. Subsequently, we illustrated that the registration accuracy obtained using these fluorophores was higher than that achieved using conventional fiducial markers, such as DAPI. Finally, we demonstrate the 3D multiplexed cyclic imaging of an expandable tissue gel with a dense label-based registration strategy, combined with sample treatment processes that effectively minimized sample distortion throughout imaging rounds. In addition, we proved that the accurate registration of the cyclic imaging of expandable tissue gel can be achieved not only with fluorophore NHS-ester staining but also with dense biological structures, such as actin. This finding expands the range of options for achieving precise registration in multiplexed cyclic imaging studies.

When performing cyclic staining and imaging with expanded specimens, the most crucial step is registering images acquired in consecutive imaging rounds [ 6 ]. While nuclei or blood vessels are commonly employed as fiducial markers in this process [ 4 ], these sparse structures can only encode the degree of spatial deformation where they are present. Because the density of these structures varies even within a single organ type [ 7 ], such as the brain, the accuracy of measuring spatial deformation can vary between regions within the same specimen. In addition, because the scale of these structures is in microns, it is challenging to measure the spatial deformation of specimens with nanoscale precision [ 8 ]. This limitation is particularly problematic when imaging nanoscale subcellular structures, such as synapses. To address this challenge, nanoscale structures in close proximity to the target proteins, such as synaptic markers for imaging multiple synaptic proteins or fiber endpoints from different cell type markers for imaging fibrous structures, have been utilized as fiducial markers [ 5 , 8 ]. Nonetheless, this approach necessitates the use of numerous fiducial markers to visualize diverse protein structures within a single specimen, thus limiting multiplexing capabilities.

Spatially resolved proteomics has significant potential for a variety of biomedical applications, since it can provide deeper insights into biological phenomena and improve diagnostic and prognostic information [ 1 , 2 ]. For precise molecular profiling of biological samples, a high-resolution imaging modality is needed, with strong three-dimensional (3D) multiplexing capabilities. Tissue expansion techniques are attractive solutions for meeting such a need for high-resolution imaging because they visualize nanoscale features below the diffraction limit with the physical expansion of tissue gel (Expansion Microscopy [ExM] [ 3 ], Magnified Analysis of Proteome [MAP] [ 4 ], etc.). This approach does not necessitate specialized chemicals or equipment because it only modifies the scale of the target sample, making it simple to execute in a typical laboratory setting. Moreover, recent advances in tissue expansion techniques have enabled post-expansion antibody staining [ 5 ]. Post-expansion antibody staining enables highly multiplexed imaging by repeated staining, imaging, and fluorescence signal removal [ 4 , 5 ], simultaneously detecting multiple target molecules with enhanced spatial resolution.

Results and discussion

Various tissue expansion techniques have been developed for super-resolution imaging. In this study, we adopted the epitope-preserving magnified analysis of the proteome (eMAP), which allows repeated antibody staining after specimen expansion [4,5]. Multiplexed cyclic imaging in eMAP involves the following steps: (1) gelation and specimen homogenization, (2) multi-round staining, imaging, and antibody stripping, and (3) registration of the acquired images (Fig 1a). To conduct the image registration of the multiplexed images acquired from consecutive rounds, image channels containing fiducial markers were essential. Cellular-level structures, such as nuclei, blood vessels, and neurons, have been commonly used as fiducial markers [9]. However, these structures are not universally found in biological specimens at different scales. At the scale of conventional diffraction-limited microscopy, nuclei and blood vessels play the role of landmarks for the successful alignment of nearby structures [10]. However, at the scale of super-resolution microscopy, these cellular-level structures are too sparse to guide nearby sub-cellular structures. In particular, in regions of interest for imaging synaptic structures, cellular-level structures are rarely found, and image registration might be unsuccessful (Fig 1b). To overcome this limitation, in this work, we employed a pan-protein staining approach utilizing NHS-functionalized fluorophores that labeled all amines present in the tissue [11–13]. Because amine groups are distributed universally and densely in tissue at the sub-cellular level, we use these targets as fiducial markers for registering images acquired with super-resolution microscopy using tissue expansion (Fig 1c).

PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 1. General experimental procedure for super-resolution multiplexed imaging in an expandable hydrogel specimen. (a) Stepwise description of multiplexed imaging in an expandable hydrogel specimen. The entire experiment can be divided into three main steps: gelation and labeling, multi-round imaging, and image registration. The first step includes the gelation of the biological specimen, the labeling of biological molecules and fiducial markers, and the physical homogenization of the hydrogel–specimen complex. The sequence of such works can differ depending on the type of expansion method used. The second step is multi-round imaging by repeating the cycle of staining and de-staining. The de-staining step includes all signal elimination approaches, such as antibody stripping, DNA elution, or fluorophore bleaching. After acquiring images from each round, the third step, image registration, is performed to match the distorted pixel coordinates finely. (b) Conventional image registration using sparsely labeled markers. Representative sparsely labeled markers are nuclei or blood vessels. An image registered with sparsely labeled markers still includes severe pixel mismatches, especially in the region where fiducial markers are barely displayed. (c) Image registration using dense-labeled markers. Fluorophore NHS-ester staining can be utilized as a densely labeled fiducial marker. Since fiducial markers are densely localized in the entire field-of-view, this approach provides relatively high registration accuracy. https://doi.org/10.1371/journal.pbio.3003240.g001

As mentioned above, to evaluate the fiducial markers for 3D multiplexed cyclic imaging, we tested fluorophores based on four criteria: (1) High fluorescent intensity, (2) Uniform and high labeling density, (3) Resistance to photobleaching, and (4) High spectral stability. The staining patterns of each type of fluorophore NHS-ester varied depending on the hydrophobicity of the fluorophores [14,15]. Since the cyclic staining strategy inevitably required multiple rounds, the fluorescent signal of the NHS-ester should be stably visualized across repetitive imaging without significant photobleaching and the spectral shift of emission spectra such as red- or blue-shifts. We initially screened various NHS-esters of 405-nm excitable fluorophores. Although the NHS-esters of 405-nm excitable fluorophores, such as CF 405S, CF 405M, and ATTO 390, showed high labeling density, all of the tested fluorophores exhibited significant photobleaching over multiple imaging rounds. In addition, when excited multiple times by a 405-nm laser during the repeated staining and imaging process, the NHS-ester staining of these 405-nm excitable fluorophores was displayed not only in the 405-nm detection channel (422–468 nm) but also in the 488-nm detection channel (502–540 nm), which infers a spectral red-shift in the emission spectrum (S1 Fig). Such a red shift in the emission spectrum of these fluorophores could impede the use of 488-nm excitable fluorophores for staining proteins in specimens. To identify better candidates for fiducial markers, we screened additional 13 fluorophore NHS-esters (488-nm excitable: Alexa Fluor 488, CF 488A, and CF 514, 561-nm excitable: ATTO 565, ATTO Rho 101, ATTO 594, Cy3, and CF 568, 647-nm excitable: ATTO 633, ATTO 647N, ATTO 680, CF 660R, and CF 680R) that are excitable with other excitation laser wavelengths, such as 488-, 561- and 647-nm (S2 Fig). To investigate the resistance to photobleaching and spectral stability, we acquired images from the same field-of-view (FOV) twice, before and after 5-min illumination with corresponding excitation wavelengths (S3 Fig, see S1 Table for the specific intensity drop rate). The 488-nm excitable fluorophores, such as Alexa Fluor 488, CF 488A, and CF 514, showed a significant drop in fluorescent intensity after the 5-min illumination with a 488-nm laser. ATTO 594, a 561-nm excitable fluorophore, displayed much higher resistance to photobleaching, but initially had bleed-through across 561- and 647-nm detection channels. After the 5-min illumination with the 561-nm excitation laser, its signal in the 647-nm detection channel (660–737 nm) decreased, while its signal in the 561-nm detection channel (572–615 nm) showed a significant increase, implying a spectral blue-shift of the emission spectrum. ATTO 647N also showed higher resistance to photobleaching and initially exhibited a fluorescence signal only within the 647-nm detection channel. However, after 5-minute illumination with the 647-nm excitation laser, its signal was also gradually shown in the 561-nm detection channel, which is used for acquiring images of 561-nm excitable fluorophores, indicating a spectral blue-shift of the emission spectrum (S4 Fig). Such aforementioned red- and blue-shifts of the emission spectra of the fluorophores may result from a variety of interactions, such as changes in charge separation within the fluorophore and conformational changes in the fluorophore, including the fragmentation of dyes by photooxidation [16,17]. The characteristics observed in this experiment, in which brain slices stained with fluorophore NHS-esters were exposed to intense excitation lasers for an extended duration, may differ from those of fluorescent molecules under standard immunostaining and imaging conditions. After thorough consideration of the rest of the 7 fluorophore NHS-ester candidates, we narrowed down the selection to 3 fluorophore NHS-esters: Cy3, ATTO 565, and ATTO 680. To assess the impact of fluorophore NHS-ester staining on antibody staining, brain slices derived from the same mouse were prepared. One-half of a brain slice was stained with one of the selected fluorophore NHS-esters (ATTO 565 NHS-ester), while the other half-brain slice was left without NHS-ester staining. Subsequently, both slices were then stained with identical antibody to observe whether any significant difference was derived from NHS-ester staining. A total of 15 different antibodies were tested, and in all cases, no significant difference in antibody staining quality was observed between the samples with and without fluorophore NHS-ester staining. (see S5 Fig for the antibody compatibility test on the ATTO 565 NHS-ester stained mouse brain slices). This constitutes a promising option for the dense labeling strategy.

Next, we attempted to validate NHS-ester staining-based registration and demonstrate multiplexed cyclic imaging using NHS-ester staining as fiducial markers. However, when performing the multi-cycle 3D imaging of expanded specimens, we encountered an issue with non-linear 3D distortion, as reported previously [18]. Specifically, when expanded specimens underwent repeated cycles of staining and antibody stripping in 1 PBS, followed by expansion in DI water, their images exhibited non-linear 3D distortions. The origin of these distortions remains unclear; however, they might be attributed to deformation of the hydrogels’ bottom surfaces on the glass substrate, with distortion propagating throughout the entire hydrogel. Such non-linear distortion necessitates computationally heavy 3D registration that sometimes required human intervention [18]. To address this issue, we performed subsequent multiplexed imaging in 1 × PBS to minimize z-plane distortion and to achieve highly accurate image registration performance. In this process, the specimens expanded only 2-fold and maintained a consistent size throughput, eliminating the need for computationally intensive 3D registration. To perform the entire process in 1 × PBS, we replaced the antibody stripping process used in the eMAP protocol, which involves repetitive high-temperature treatment in a denaturation buffer containing sodium dodecyl sulfate (SDS) and sodium sulfate, with two approaches. First, we photobleached the antibody signals after each round of imaging instead of stripping the antibodies, as reported previously [19]. Second, we replaced the antibody stripping process with computational signal unmixing, as reported previously [20]. These two alternative approaches significantly improved spatial distortion issues (S6 Fig).

With minimized sample distortion conditions, we compared the registration accuracy of dense labels (fluorophore NHS-esters) with that of sparse labels (DAPI) when used as fiducial markers. The initial assumption was as follows: Since the staining pattern becomes much sparser after expanding the tissue gel, conventional sparse labels, such as DAPI, would provide limited information on the deformation of specimens (S7 Fig). This registration issue might worsen in regions with few cells. In contrast, dense labels, such as fluorophore NHS-ester staining, consistently provided substantial spatial information regardless of the target region of interest, enabling highly accurate image registration performance. We first compared the registration accuracy of these two labels in a small FOV over two consecutive imaging rounds. Mouse brain slices were processed with the eMAP protocol, stained with DAPI and ATTO 680 NHS-ester, and expanded 2-fold in 1 × PBS. Then, they were stained with a primary antibody against pre-synaptic marker (vGluT1) and Alexa Fluor 488-conjugated secondary antibody targeting the primary antibody. After imaging, the fluorescence signals of the specimens were photobleached with prolonged exposure to the excitation laser. The specimens were then stained with a primary antibody against Alexa Fluor 488 and an Alexa Fluor 488-conjugated secondary antibody targeting this primary antibody. To minimize the effect of chromatic aberrations, we stained the sample with an identical fluorophore (Alexa Fluor 488) and acquired images of the 488-nm channel. The two images, one acquired after anti-vGluT1 staining and the other acquired after anti-Alexa Fluor 488 staining, were registered using either DAPI or ATTO 680 (Fig 2a and 2b). For each imaging round, 10-µm thick z-stack images with 1-µm stepsize were acquired from six different samples. Notably, in the regions where the DAPI signal was absent, noticeable pixel mismatches were observed in the post-registered vGluT1 images when the two images were registered using DAPI as a fiducial marker. Conversely, when NHS-ester staining was used as a fiducial marker, it allowed precise registration between the two consecutive round images (Fig 2c; see S8 Fig for more comparison results between DAPI-based and NHS-ester-based registration). Pearson correlation coefficients (PCCs) between the first and second round images were estimated to quantitatively compare the registration accuracy of DAPI staining and NHS-ester staining (Fig 2d; see S9 Fig and S1 Data for a detailed dataplot of PCC among DAPI-based, NHS-ester-based and vGluT1-based registration). The results clearly showed that using dense labels through fluorophore NHS-ester staining as fiducial markers markedly enhanced registration accuracy compared to DAPI staining-based sparse labeling. Furthermore, no significant difference was observed when the NHS-ester-based registration results were compared to those obtained by registering with the target protein image, vGluT1, which was expected to yield the most accurate registration results (see S9 Fig for dataplot of the PCC).

PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 2. Validation of a dense label-based registration. (a) Experimental procedure for validation. The image registration accuracy was estimated from sparsely labeled (DAPI) and densely labeled (ATTO 680 NHS-ester) markers. The first round was stained and imaged with a rabbit anti-vGluT1 antibody and an Alexa Fluor 488 (AF 488)-conjugated goat anti-rabbit secondary antibody. Signal of the first round was photobleached and the second round was stained and imaged with anti AF 488 antibody. Images acquired from the first and second round were finally registered through the DAPI staining channel and the NHS-ester staining channel, respectively. (b) Initial region of interest (ROI) with DAPI (blue) and NHS-ester staining (light blue) channels. Target ROI where DAPI signal is barely visible (magenta-highlighted region). (c) Magnified view of post-registered vGluT1 fluorescent signal registered by DAPI and NHS-ester, respectively (first round: red, 2nd round green). (d) Box plot of Pearson correlation coefficient between DAPI registered images (1st–2nd round) and NHS-ester registered images (1st–2nd round) within initial ROI and magnified target ROI. The brain region where validation has been conducted was between CA3 and dentate gyrus regions. Please see S1 Data for individual numerical values of the Pearson correlation coefficient. Scale bars: (b) 20 µm; (c) 1 µm. All length scales are presented in pre-expansion dimensions. Number of sample N = 6, Number of datapoints M = 11 for each sample. https://doi.org/10.1371/journal.pbio.3003240.g002

Following the thorough validation of the highly accurate dense label-based registration, we conducted 3D multiplexed cyclic imaging on mouse brain slices through the first approach, which used photobleaching to remove signals. To avoid host species cross-reactivity and to use multiple same host antibodies, we adopted the preformed antibody complex (preassembly) strategy, utilized in our previous study [21,22]. For photobleaching-based multiplexing, we first validated the robustness of the photobleaching strategy and confirmed the absence of antibody crosstalk in the preformed antibody complexes (S10a Fig). Various antibodies were properly stained across multiple rounds of photobleaching, and the preformed antibody complexes visualized the corresponding target protein expression without antibody crosstalk. After validation, we demonstrated multiplexed imaging as follows: eMAP-processed specimens were stained with DAPI and ATTO 680 NHS-ester. Then, antibody staining, 3D imaging, and photobleaching were repeated for seven rounds, with two antibodies stained in each round, resulting in 16-color 3D multiplexed imaging (Fig 3; see S10b and S10c Fig for the prior validation result of photobleaching in an intact mouse brain slice). Notably, glial fibrillary acidic protein (GFAP), the astrocyte marker, showed high spatial colocalization with SRY-box 2 (SOX2) while ionized calcium-binding adapter molecule 1 (Iba1) did not colocalize with either GFAP or SOX2 (S11 Fig). The resulting multiplexed image matched the previously reported protein colocalization patterns [23–27].

PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 3. Registration of multiplexed images via ATTO 680 NHS-ester staining and photobleaching in an eMAP-processed mouse brain slice. (a) Experimental schematic of the registration of multiplexed images. Target proteins for each imaging round were stained, imaged and removed with photobleaching treatment. A pair of images from adjacent rounds are registered by the ATTO 680 NHS-ester staining image. (b–r) Seven-round cyclic staining images of an eMAP-processed mouse brain slice. (b) Merged 3D 16-plex image with 30 µm z-stacks. Blue, DAPI; gray, ATTO 680; white, Laminin; cyan, NF-H; yellow, Iba1; green, Lamin B1; orange, vGluT1; orange-red, Homer1; magenta, GFAP; crimson, Synaptophysin; dark purple, SV2A; purple, GM130; pink, SOX2; gold, ABAT; light green, Alpha-tubulin; green-yellow, MBP. (c–r) Single-channel images of the target proteins. Scale bars: (b–r) 20 µm. All length scales are presented in pre-expansion dimensions. Number of sample N = 2 acquired from two independent mouse brain slices. https://doi.org/10.1371/journal.pbio.3003240.g003

We next demonstrated 3D multiplexed cyclic imaging using the second approach, which replaced the antibody stripping process with computational signal unmixing employing the PICASSO technique [21]. In this approach, we consecutively stained the next round of target antibodies without any additional signal removal steps [20]. Then, the N + 1th round image was unmixed with the Nth image using the blind unmixing algorithm reported in the previous works [20,21]. To quantitatively assess the accuracy of signal unmixing, we calculated the PCC between the unmixed output images and the corresponding ground-truth images (S12 Fig; see S2 Data and the Experimental Section for details of the experimental design). Brain slices were consecutively stained for two or three rounds without expansion, followed by image acquisition and signal unmixing. The PCC between the unmixed images and their respective ground-truth images was then computed, yielding a value of approximately 0.98, indicating a high degree of correspondence with the ground-truth images. Using this approach, we demonstrated 10-color 3D multiplexed imaging (Fig 4).

PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 4. Registration of multiplexed images via ATTO 565 NHS-ester staining and signal unmixing in an eMAP-processed mouse brain slice. (a) Experimental schematic of the registration of multiplexed images. Target proteins for each imaging round were consecutively stained without signal removal. A pair of images from adjacent rounds are registered by the ATTO 565 NHS-ester staining image. The registered images were unmixed to extract target protein expression. (b–m) Five-round cyclic staining images of an eMAP-processed mouse brain slice. (b) Merged 3D 10-plex image with 20 µm z-stacks. Blue, DAPI; gray, ATTO 565; brown, calnexin; green, lamin B1; yellow, SOX2; cyan, NF-H; red, CALB2; white, laminin; gold, vGluT2; magenta, GFAP. (c–m) Single-channel images of the target proteins. Scale bars: (b–m) 20 µm. All length scales are presented in pre-expansion dimensions. Number of sample N = 6 acquired from two different mouse brain slices. https://doi.org/10.1371/journal.pbio.3003240.g004

To compare the registration accuracy between dense label-based and sparse label-based registration within an approximately 4× fully expanded state while minimizing the variation in the expansion factors throughout the rounds, we re-embedded an expanded tissue gel in a neutral gel and performed cyclic imaging. In this experiment, we used a proteinase-based ExM technique demonstrated in 2015 [3]. After re-embedding, we conducted a series of steps: hybridizing fluorophore-conjugated oligonucleotides (imager DNA) to the gel-anchored oligonucleotides (tertiary DNAs), imaging, and de-hybridizing the imager DNA from the hydrogel. This cycle was repeated three times. To precisely measure the registration error not affected by chromatic aberration, we used the same fluorophore to label the same DNA in different imaging rounds. Images were then registered using either the DAPI or fluorophore NHS-ester channel as fiducial markers, and registration errors were measured by analyzing line profiles. We found that registration using DAPI as a fiducial marker yielded a registration error larger than 50 nm, especially when the nucleus density was low, even though the gel was re-embedded in a neutral gel and did not undergo repeated shrinking and expansion (S13a and S13b Fig). This expanded brain slice underwent non-linear deformation, resulting in a registration error where the nucleus was absent (S13c and S13d Fig). However, when Cy3 NHS-ester staining was used as a fiducial marker for the repetitive imaging of an expanded mouse brain slice, a registration error of less than 10 nm was achieved, as shown in S13e–S13h Fig (see S3 Data for detailed analysis of the line profiles). For the further quantitative evaluation of registration accuracy, we cropped several sub-regions of interest (ROIs) of Bassoon and GFAP structures acquired from the first and second rounds and the second and third rounds, respectively (S13i–S13l Fig). PCCs were calculated to measure two variables’ linear correlation. The PCCs acquired from the Bassoon and GFAP images both showed that NHS-ester-based registration outperformed DAPI-based registration by approximately 20% (S13m Fig, see S4 Data for detailed analysis of the accuracy of computational signal unmixing).

After validating the high registration accuracy of dense label-based registration, we turned our attention to exploring alternative dense labels as registration markers. While fluorophore NHS-esters visualize diverse structures, their lack of molecular specificity prompted us to consider dense biological structures for this purpose. We specifically chose actin as our dense label for registration. Actin fulfills the four essential requirements for effective registration markers. First, it exhibits high heterogeneity on a scale of tens of nanometers, which helps achieve precise registration [28]. Second, actin is present in all types of specimens, making it a versatile choice for various biological samples [29]. Third, actin demonstrated uniform expression for all the specimens, ensuring consistent and reliable labeling for registration [29]. Last, actin is compatible with cyclic staining in expanded specimens, allowing repeated imaging cycles without compromising the registration process [30]. To validate this idea and showcase its potential applications, we performed multiplexed imaging in cultured cells using simple DNA-based barcoding. To achieve this, we employed the actin-ExM technique, which we recently developed for simple actin staining and expansion [30].

Briefly, cultured cells were labeled with fluorescein isothiocyanate (FITC)-conjugated phalloidin, primary antibodies against FITC, oligonucleotide-conjugated secondary antibodies against primary antibodies, and four proteins: vimentin, lamin A/C, clathrin heavy chain (CCP), and cytokeratin 8/18. Following tertiary DNA hybridization, the cells were expanded 4.2-fold using previously established pro-ExM protocols [3,31]. The expanded specimen was then embedded in an uncharged polyacrylamide gel to maintain its expanded state during the two-round imaging process. We then repeated the hybridization and dehybridization of the fluorophore-conjugated oligonucleotides to generate a distinct color barcode for each protein. To analyze the barcoded image completely, we registered two images taken from sequential rounds using the actin channel as a fiducial marker and found that all proteins were clearly designated by their own color codes (Figs 5a, S14a, and S14b). For example, vimentin was labeled with Alexa Fluor 488 in the first round (displayed in blue) and the second round (displayed in green), showing cyan-labeled fibers in the resultant image. CCP was also labeled with Alexa Fluor 488 in the second round (displayed in green) but with ATTO 565 in the first round (displayed in red), resulting in yellow-labeled ring structures in the cell (Figs 5a, S14c, and S14d). As visualized in Fig 5b and 5c, we observed single fibers of different types of cytoskeleton—vimentin and cytokeratin—in assorted colors after a 4.2-fold expansion. While the actin channel was used as a fiducial marker, it also visualized relative spatial distribution information with other structures as an important cytoskeleton.

[END]
---
[1] Url: https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3003240

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/