Systematic Evaluation of Chemically Distinct Tissue Optical Clearing Techniques in Murine Lymph Nodes

Lymph nodes (LNs) are secondary lymphoid organs of high cellular density that play a key role in the development of adaptive immunity. Unlike many other organs in the body, LNs contain cells that perform their function through constant changes in their location. Ags and APCs enter the LNs via afferent lymphatics, whereas naive T and B cells enter via postcapillary high endothelial venules. Interactions between APCs and Ag-specific T cells are necessary prerequisites for development of adaptive immunity, and are facilitated by a stromal network of fibroblastic reticular cells and infiltrating lymphatics (1). During infection, LNs undergo a significant structural remodeling, cell proliferation and infiltration, creating an even more complex and tightly packed structure (2). Hence, visualization of LN at cellular level under both general and pathological states is as desirable as challenging (3).

So far, analysis of the entire LN structure was limited by either poor imaging resolution offered by gross imaging techniques, such as magnetic resonance imaging and computed tomography, or lack of spatial context in two-dimensional histological sections. The recently presented methodology of tissue optical clearing (TOC) serves as a compromise between high spatial and cellular resolution, offering millimeters to centimeters-wide volumetric imaging with subcellular resolution, within a reasonable time (4). Although first developed for brain imaging, TOC has already proved useful for every rodent organ of interest (5), even for clearing and imaging of the entire rodent bodies (6) in search of metastases (7) or sites of muscular damage accompanying dystrophy (8). Similarly to other peripheral organs, LNs turned out to be suitable to TOC, which facilitated a number of discoveries such as quantification of high endothelial venules remodeling (9–11) and changes in lymphatics after immunization (12) or inoculation of tumor cells (13); lymphotoxin-dependent cross-talk between B cells and fibroblastic reticular cells during de novo follicle formation in the course of helminth infection (14, 15); elucidation of a role for low-affinity effector T cells during early microbial containment (16) fate mapping of dendritic cells progenitors (17) or dissemination patterns of B cells during the germinal center reaction (18).

TOC methodology has rapidly expanded in the last decade and is now represented by nearly 30 original methods and dozens of their optimizations, all of which have their characteristic advantages and limitations. The vast majority of TOC methods were developed for optical clearing of brain tissue (19, 20), hence little is known about their suitability for clearing other organs, including LNs. Therefore, we chose 13 representative TOC protocols that were selected from all of the existing, chemically distinct TOC categories (i.e., solvent-based, water-based, tissue transformation and high refractive index [RI]–based clearing) and systematically screened for their suitability in optical clearing of murine LNs.

All experiments were performed on adult 8–12-wk-old wild-type C57B6L/6 mouse. Animals were housed in controlled environmental conditions of temperature (22 ± 2°C), humidity (55 ± 5%), and illumination (12/12 h of light–dark cycle), with water and food provided ad libitum. Animals were anesthetized with i.m. injection with 30 μl of 2:2:1 (v/v) ketamine/xylazine/0.9% NaCl solution. Next, mice were sacrificed by cervical dislocation followed by dissection of inguinal and mesenteric LNs. In the case of experiments employing adoptive transfer of CD8⁺ T cells (see below), mice were sacrificed 24 h after inoculation of T cells. Next, LNs were briefly rinsed in 0.01 M PBS (pH 7.4)/heparin (5 IU/ml final concentration) and subjected to 24 h fixation with 1% or 4% paraformaldehyde at 4°C. After fixation, LNs were rinsed three times for 1 h with PBS, and the residual connective and adipose tissues were gently removed under the stereomicroscope. The experiments were performed in accordance with the guidelines approved by the first Local Ethics Committee of the University of Warsaw (approval no. 391/2017) and in accordance with the requirements of European Union (Directive 2010/63/EU) and Polish (Dz. U. poz. 266/15.01.2015) legislation.

CD8⁺ T cells were isolated from LNs and spleens of Tg (CAG-DsRed*MST)1Nagy/J (hereafter referred to as DsRed) and Tg (UBC-GFP)30Scha/J (hereafter referred to as EGFP) transgenic mice (The Jackson Laboratory) that exhibit widespread fluorescent protein expression under the control of the β actin and ubiqutin C promoters, respectively. Isolation of CD8⁺ T cells was performed using EasySep Mouse CD8⁺ T Cell Isolation Kit (no. 19853; StemCell Technologies). Equal numbers of 0.8 × 10⁶ isolated CD8⁺ T cells were inoculated i.v. (via tail vein) to each C57BL/6 recipient mice. For the light-sheet fluorescence microscopy (LSFM) experiment seen at Fig. 10A, the isolated CD8⁺ T cells were stained with CellTrace Violet (CTV) cell proliferation kit according to the manufacturer’s instructions (no. C34557; Thermo Fisher Scientific) and inoculated i.v. at a number of 5.5 × 10⁶ per mouse.

LNs were subjected to clearing with the following protocols according to original studies: SeeDB (21), SeeDB2 (22), 3DISCO (23), uDISCO (24), iDISCO (25), CUBIC (26), simplified clarity method (SCM) (27), 75% v/v glycerol (28), C_e3D (29), and FRUIT (30). The only modification made to the procedure was time of sample incubation as presented in Fig. 1A without any interference with chemical composition of original solutions. For each LN, 2 ml of a reagent was used at every step of clearing process. In case of CUBIC method, four protocols were tested: incubation in 1) CUBIC reagent-1, 2) CUBIC reagent-1 followed by CUBIC reagent-2, 3) CUBIC reagent-1A, or 4) CUBIC L/R reagents (7). In the case of SeeDB2, SeeDB2G protocol was implemented. Twenty milliliters of each of the SeeDB solutions (20, 40, 60, 80, 100%, and finally 101.25% w/v fructose in distilled water) were prepared in 50 ml conical tubes, and 80, 100, and 101.25% solutions were heated at 65°C for ∼ 2 h (with vigorous shaking every 30 min). After cooling down to room temperature (RT), 100 μl of α-thioglycerol was added to each solution to prevent Maillard reaction. Solutions are stable for ∼45 d and should be stored at RT. FRUIT solutions were prepared similarly (final FRUIT solution tends to crystalize within a week even at RT and, based on our experience, cannot be successfully melted by placing again at 65°C; thus this solution should be prepared prior to use). It is critical to add sodium azide (0.02% v/v) to SeeDB2 solutions to prevent rapid growth of microorganisms. In the case of uDISCO, BABB-D4 solution was used that consists of four parts of BABB (one part of benzyl alcohol to two parts of benzyl benzoate) and one part of diphenyl ether (and 0.4% v/v α-tocopherol). Even though LNs are very small clearing targets, incubation times in gradient of tert-butanol solutions should not be shortened, as this might lead to incomplete clearing during the RI matching step (LN will first become transparent and turn opaque within a few hours). Original CUBIC protocol consisted of two reagents: CUBIC reagent 1 for delipidation (and mild swelling) and CUBIC reagent 2 for RI matching and mild shrinkage of the initially swollen tissue. During preparation of CUBIC-R1 and -R2 solutions, excessive stirring (over 400 rpm) should be avoided, as this might generate foam and waste the solutions. CUBIC-R1 is ready to use after 2.5–3.0 h of stirring with a magnetic stirrer (no heating is required), and CUBIC-R2 after 3–4 h (heating at 43–45°C is recommended). CUBIC-R1A is an unpublished (but deposited; http://cubic.riken.jp) protocol made by authors of original CUBIC method for better stabilization of fluorescent signal that might be diminished with CUBIC-R1. Finally, CUBIC-L and CUBIC-R are second generation of CUBIC-R1 and -R2 for more efficient delipidation and RI matching, respectively. In contrast to original, these are easy miscible solutions of lower viscosity. However, based on our observations, CUBIC-R cannot be used for long-term storage of samples, as it gradually becomes yellowish, orange and finally reddish after ∼45 d from preparation at RT. In the case of SCM protocol, long-term storage of 4% acrylamide with 0.5% photoinitiator (VA-044; Wako Chemicals, which is double the concentration used in original PACT) should be avoided, as it polymerizes even at 4°C after ∼1.5–2.0 mo, and RIMS (RI matching solution for PACT-cleared samples that should be incubated at 37°C for ease and complete dissolution) must be supplemented with 0.02% v/v sodium azide. Finally, C_e3D, similarly to RIMS, contains Histodenz, which becomes highly viscous upon contact with liquids (here, N-methylacetamide); thus, we recommend a first addition of 1/2 final volume of N-methylacetamide, next the Histodenz, then the rest of the N-methylacetamide and dissolution at 37°C with shaking (which should take ∼4 h).

Immunostaining was performed prior to clearing in all but CUBIC-cleared samples. Dissected and fixed LNs were placed in 1 ml of blocking solution (2% w/v BSA [no. A3311; Sigma-Aldrich], 0.3% v/v Triton X-100 [no. X100; Sigma-Aldrich] in PBS) for 24 h at RT followed by 3 d of primary Ab staining in fresh blocking solution at 37°C with agitation. Next, samples were washed twice with 0.2% v/v Triton X-100 in PBS for 12 h each at RT with agitation. Secondary Ab staining was performed under the same conditions as for the primary Abs.

The following Abs were used in this study: primary anti-LYVE1 polyclonal (1:150 dilution; no. PA1-16635; Thermo Fisher Scientific), anti-mouse B220 conjugated with Alexa Fluor 594 (1:200 dilution; clone RA3-6B2, no. 103254; BioLegend), anti-mouse CD3 conjugated with Alexa Fluor 594 (1:200 dilution; clone 17A2, no. 100240; BioLegend), anti-mouse CD4 conjugated with Alexa Fluor 488 (1:200 dilution; clone GK1.5, no. 100423; BioLegend), anti-mouse CD8 conjugated with Alexa Fluor 647 (1:200 dilution; clone 53–6.7, no. 100724; BioLegend), anti-mouse CD31 conjugated with Alexa Fluor 647 (1:200 dilution; no. 102416; BioLegend), anti-mouse/human PNAd conjugated with Alexa Fluor 647 (1:200 dilution; no. 120808; BioLegend), and Alexa Fluor 546 Goat anti-Rabbit IgG (H+L) secondary Ab (1:500 dilution; no. A-11010; Thermo Fisher Scientific).

To represent values of transparency, the cleared LNs were placed in a petri dish filled with respective RI matching solution above grid paper and subjected to macrophotography using an Olympus E-M1 MarkII with M. Zuiko Digital ED 60 mm F2.8 Makro lens. Next, horizontal line was drawn through the middle portion of LN, and gray value was calculated using the Plot Profile function in ImageJ software (version 1.52n; National Institutes of Health). The coefficient of transparency was then calculated as schematically represented in Fig. 1B.

Size change quantification was based on the bright-field images of LN taken before and after TOC. The images were outlined by the polygon-selection tool in ImageJ software, and the area was measured (Fig. 1D).

The quantification of fluorescence was performed as previously described with some modifications (31). To compare EGFP fluorescence preservation after completion of TOC, LNs were first imaged uncleared in PBS solution, which enforces imaging depth for a maximum of ∼ 80–100 μm. As the first ∼40 μm of LNs imaged from the surface contains a significant autofluorescent/background signal coming from the capsular structures, LNs were sliced horizontally in the center into two fragments and then imaged, starting from the cut plane. The same LNs were imaged under the same conditions at three time points (i.e., before TOC, day after completion of TOC, and 14 d thereafter). After imaging, 60–80-μm z-stacks of T cells were semiautomatically segmented using surface function in Imaris (Bitplane) and the mean fluorescence intensity (MFI) was calculated for each T cell. As the fluorescence intensity of specific signal is a sum of fluorescence of this signal and the signal of the background pixels, the mean intensities of the background were calculated using ImageJ software by averaging the values of five regions of interest without a specific signal in the maximum intensity projection (MIP) image. Next, the MFI of background was subtracted from the MFI of the signal. Finally, the results were normalized by dividing resulting signal MFI (i.e., day 1 after TOC/before TOC and day 14 after TOC/day 1 after TOC).

Standard H&E protocol was used. In brief, cleared samples were subjected to autotechnicon tissue processor, sliced at 4 μm, and stained with hematoxylin for 6 min and eosin for 2 min.

For the assessment of immunostaining feasibility and imaging depth, images of LNs from CD31-stained or adoptively transferred with CD8⁺/DsRed or CD8⁺/EGFP T cells were taken using Zeiss (noncommercial setup) two-photon microscope (Upright Axio Examiner Z1) equipped with W Plan-APOCHROMAT 20×/1.0 DIC WD = 2.4 mm. For the excitation of anti-CD31 Ab-conjugated Alexa Fluor 647 and DsRed-positive T lymphocytes, a coherent Chameleon 690–1064-nm laser was used with 1050- and 820- nm excitation, respectively, zoom 3, and z-step = 10 μm, if not otherwise stated. Emitted wavelengths were detected with specific detectors and detection filters: PMT detector with HC 675/67 (640–708 nm) filter for Alexa Fluor 647 and GaAsP detector with HC 600/52 (574–626 nm) filter for DsRed.

For the assessment of fluorescence intensity, images of LNs from adoptively transferred mice with CD8⁺/GFP or CD8⁺/DsRed T cells were taken using Leica TCS SP8 equipped with Dry HC PL APO CS2 10×/0.40 DRY objective WD = 2.2 mm. For EGFP excitation, a 488 nm argon laser was used. The emitted wavelengths were detected with a HyD detector at range 499–550 nm. For DsRed excitation, a 561-nm white light laser was used. The emitted wavelengths were detected with a HyD detector at range 571–680 nm. Images were obtained with zoom 3× and z-step set at 2 μm.

For the assessment of proper 1) timing of immunohistochemistry in CUBIC protocol (before versus after completion of clearing) and 2) concentration of PFA used, Leica TCS SP8 equipped with Dry HC PL APO CS2 10×/0.40 DRY objective WD = 2.2 mm was again used. Optical slices were acquired 150–200 μm from the LN surface during the same imaging session under the same imaging settings for particular TOC method.

To image whole LNs, a Zeiss Lightsheet Z.1. equipped with two 5× objectives (numerical aperture 0.1) form two independent lightsheets of 11.4 μm, from the left and right sides of the sample. Each lightsheet was aligned manually into the same imaging plane, according to the procedure provided by Carl Zeiss AG. The LNs were imaged using dual sided illumination with these lightsheets and the resulting images were automatically fused by ZEN software (Zeiss). Emitted fluorescent light was captured using 5× Plan-Neofluar objective, depth of focus: 32.42 μm, numerical aperture: 0.16. Fluorescence was detected using two 16-bit PCO edge 5.5 sCMOS cameras (PCO AG), a triband 405/488/561 laser blocking filter was also inserted into the light path. To detect CTV, Alexa Fluor 546 (anti-LYVE1 Ab), and Alexa Fluor 647 (anti-CD31 Ab) excitation, 405-, 561-, and 647-nm diode lasers were used, respectively. Emitted wavelengths were detected with the following filters: BP445/25 nm for CTV, BP595/20 nm for Alexa Fluor 546 (anti-LYVE1), and LP660 nm for Alexa Fluor 647 (anti-CD31). A zoom factor of 0.16× was used to fit the entire LN in a single field of view. The pixel size was 1.66 × 1.66 μm, field of view diagonal 4.51 mm. The optimal z-step size was automatically calculated by the ZEN software to be 5.31 μm. Because of imaging performed under the same conditions for several tested TOC methods, some of the confocal images were slightly adjusted in brightness and contrast to ensure that images are visible and the upper part of background in Fig. 10A was airbrushed to remove artifact coming from the Z.1 LSFM sample holder.

Based on the literature analysis, 13 potent TOC methods were chosen for the evaluation of their clearing capability to LNs. The selected approaches represent the entire spectrum of TOC: hyperhydrating solutions (CUBIC and its modifications described in materials and methods in detail), high RI aqueous solutions (SeeDB, SeeDB2, C_e3D, FRUIT, glycerol), organic solvents (3DISCO, uDISCO, iDISCO) and tissue transformation (SCM). The incubation times for each of the clearing steps were either implemented as described specifically for LNs in the original articles [e.g., 3DISCO (23), glycerol (28), C_e3D (29)] or adjusted to ensure at least a 12 h period of incubation in each solution and are summarized in Fig. 1A.

Although measurement of tissue transmittance with spectrophotometer is altogether feasible, a more reliable data acquisition requires tissue placement in the same region of cuvette as well as familiarity with the tissue thickness, both of which are hardly accessible in case of small, irregularly shaped organs, such as murine LNs. Thus, instead of measuring tissue transmittance by spectrophotometry, we have used an alternative approach (i.e, macrophotography of LNs on a grid paper under stable lightning conditions), and analyzed the obtained images with ImageJ, as presented in Fig. 1B. The analysis revealed that the LNs were most transparent following clearing with iDISCO (Fig. 1C). Less transparency was observed in the LNs cleared using uDISCO, CUBIC R1, and C_e3D. However, the performance of uDISCO, CUBIC R1, and C_e3D was superior to SeeDB, CUBIC-L/R, 75% glycerol, and CUBIC-R1A in terms of the transmittance index. Next, LNs were photographed before and after completion of TOC to assess how particular methods influence size of the studied tissues (Fig. 1D). All of the tested methods, except for SeeDB, induced substantial alterations in the LNs size (Fig. 1E). All of the organic solvents along with C_e3D caused tissue shrinkage, up to 58.9 ± 5.91% of the projected tissue area in case of 3DISCO. Treatment with the remaining methods (glycerol, CUBIC-L/R, -R1A, -R1, -R2, FRUIT, SCM, SeeDB2) caused volatile tissue expansion, the variability of which was the least pronounced in SCM group (expansion SD ± 9.46%) with ± 23.8, ± 32.4 and ± 11.3% for CUBIC-R1, CUBIC-R1A and CUBIC-R2, respectively.

A rapid decay of fluorescent signal from proteinaceous fluorophores might strongly limit use of a particular TOC method. Thus, to evaluate preservation of one of the most commonly used fluorophores (EGFP and DsRed) upon clearing, we have adoptively transferred CD8⁺ T cells isolated from EGFP and DsRed mice to the wild-type syngeneic recipients. Sixty- to eighty-micrometer-thick z-stacks were acquired for each LN before TOC, the next day after tissue clearing completion, and 14 d thereafter (Fig. 2A, Supplemental Fig. 1). The results showing EGFP fluorescence intensity normalized for each time point (i.e., value for day 1 = fluorescence intensity on the next day after TOC divided by fluorescence intensity before TOC) are presented in Fig. 2B. All of the tested methods, besides organic solvents, retained substantial EGFP-related fluorescence after completion of the clearing protocol. Subsequent 2-wk observation of the tissues revealed major differences between methods: that is, the entire fluorescence signal was lost during a 2 wk period in case of FRUIT, glycerol, and SeeDB; stabilized in CUBIC-R1, CUBIC-R1A, and SeeDB2; and increased in SCM, CUBIC-R2, and C_e3D groups. The increase of normalized fluorescence intensity should be interpreted as a combination of both 1) high fluorescence preservation capability and 2) further increase of sample transparency during the studied period. Moreover, imaging depth for each TOC method was assessed using multiphoton microscopy (Fig. 3).

In contrast, DsRed signal was retained in every tested method, as presented by 100-μm-thick MIP images generated at the center of representative z-stacks acquired with multiphoton microscopy (Fig. 4A). However, the degree of signal retention was variable, with solvent-based methods quenching DsRed fluorescence within first 24 h and 75% glycerol and FRUIT within a 2-wk period (Fig. 5). Notably, the image quality was variable between methods, with 3DISCO, C_e3D-, SCM-, glycerol-, CUBIC-R1–, CUBIC-R1A–, and CUBIC-R2–based clearing resulting in low background (Fig. 4A). Further analysis showed that, similarly to EGFP, almost all of the tested methods allow imaging with multiphoton microscopy of the entire LN from top to the bottom (Fig. 4B). The imaging depth was limited only in case of SeeDB, SeeDB2, CUBIC-L/R, and glycerol (with the average imaging depths of 245.5, 592.5, 670.5, and 556.0 μm, respectively).

As TOC and confocal imaging of fixed LNs are complementary, not alternative, approaches to histological analysis, we have checked the compatibility of TOC methods for preservation of tissue structure in classical histology followed by H&E staining (Fig. 6, Supplemental Fig. 2). As expected, samples strengthened by hydrogel matrix after SCM retained the best gross structure of the LNs. However, similarly good results were obtained with C_e3D, CUBIC-R1, and glycerol treatments (the latter most probably resulted from the short incubation period). The remaining techniques were either prone to generate more frequent cutting artifacts (e.g., 3DISCO and SeeDB2) or induced deterioration of the tissue, resulting in some loss of histological detail in the interior of the LNs compared with that of the LNs immersed in PBS (Fig. 6) (e.g., CUBIC-L/R, uDISCO, Supplemental Fig. 2).

To assess the epitope retention, anti-CD31 immunostaining was performed (Fig. 7A). As no consensus on the best fixative protocol for the LNs exists so far, we have chosen fixation with PFA, at either 1 or 4% concentration, as the most commonly used. We followed previous reports and for all the methods but CUBIC, we performed immunostaining before clearing. For CUBIC-related protocols, as it was published before (26), we did the immunostaining once the clearing was completed. We failed to achieve detectable anti-CD31 signal only after the iDISCO protocol (Supplemental Fig. 3), which is in accordance with the published data on the poor compatibility of iDISCO used to clear white adipose tissue (32). MIP images of the remaining protocols revealed the relatively good retention of the epitopes. Importantly, immunostaining completed after the TOC process resulted in severely impaired image quality, as compared with immunostaining prior TOC in case of CUBIC-treated samples (Fig. 7B). To further verify these CD31-based results, we have performed additional rounds of immunostaining against widely used immune cells epitopes (B220, CD3, CD4, and CD8, in particular). Using B220 and CD8 staining on LNs fixed with either 1% or 4% PFA, we observed the latter approach to result in better, or at least equally good, image quality (Fig. 8, Supplemental Fig. 4) (e.g., fixation step with 4% PFA was essential to retain anti-B220 immunostaining capability in case of SCM-treated samples). Importantly, only C_e3D-, CUBIC- and glycerol-treated samples were fully compatible with tested Abs. In the case of CUBIC-treated samples, image quality of LNs stained against B220, CD3, and CD4 before versus after completion of clearing was compared (Fig. 9). Similarly to CD31, in a group immunostained after completion of CUBIC, a wide range of artifacts were present that included uneven signal, disturbance of cell shape, and presence of very bright unspecific spots/blots.

The aforementioned observations are summarized in Table I. Based on the examined factors (time and ease to clear, imaging depth, preservation of the tissue structure, EGFP and DsRed fluorescent signals, compatibility with immunostaining), C_e3D and CUBIC-R1/R2 are optimal for LNs clearing and further rapid imaging of the entire structure using LSFM. The major differences between the mentioned methods include: 1) direction of sample deformation (Fig. 1E) (i.e., shrinkage in case of C_e3D and expansion after CUBIC-R1/R2) and 2) RIs of clearing solutions (RI = 1.49, 1.435 and 1.48–1.49, respectively). Considering that 1) the C_e3D promotes tissue shrinkage that might complicate further segmentation of closely packed T cells in low-magnification settings (33) (5–20×) and 2) our LSFM setup is optimized for RI ∼ 1.45, we decided to test the suitability of CUBIC-R1 as both a clearing and an imaging solution. This time we have adoptively transferred CTV-stained CD8⁺ T cells to also check whether 1) CUBIC is compatible with this widely used dye and 2) the proposed imaging will not be limited by tissue autofluorescence in such a short-wave excitation spectrum. As presented in Fig. 10, such approach allows to image the entire murine inguinal LN with great detail using both 20× (Fig. 10A) and 5× (Fig. 10C) magnification objectives. Quality of images, which are captured within 3 h and 60 s for the mentioned objectives, are sufficient for further discrimination of adoptively transferred CD8⁺ T cells (Fig. 10B) during the postprocessing (see also Supplemental Videos 1, 2).

Application of several TOC protocols to LNs clearing has already paved the way for an ample number of novel research possibilities by shifting LN immunohistochemistry to the third dimension (9, 10, 12–17, 28, 29). However, these reports lacked in-depth evaluation on how these methods affect the processed tissue. As the brisk development and evaluation of chemically distinct TOC approaches focus on brain clearing and imaging (34, 35), we aimed to perform such a systematic analysis for LNs. To this end, we have compared 13 different protocols, that span the entire range of chemically distinct TOC methods, for the resultant tissue transparency, size alteration, GFP and DsRed fluorophore retention, imaging depth, compatibility with immunofluorescence and postclearing histological sectioning, and H&E staining.

Similarly to the reports from other groups on various peripheral organs (36, 37), organic solvents applied to LNs clearing ensure the highest level of transparency, with iDISCO and uDISCO being superior to 3DISCO. This might be attributable to either 1) significantly shorter incubation time for clearing, taking only 140 min in case of 3DISCO (versus 4 d for iDISCO/uDISCO) or 2) higher degree of LN shrinkage observed with 3DISCO, which results in denser appearance of cells and stromal elements or combination of both. Nonetheless, it should be noted that the transparency obtained with the majority of protocols, but not with SeeDB, SeeDB2, CUBIC-L/R, and 75% glycerol, is sufficient to image the entire LN using multiphoton microscopy, as confirmed by imaging of adoptively transferred, EGFP- (Fig. 3) and DsRed-expressing CD8⁺ T cells (Fig. 4).

Although the retention of fluorescent proteins upon TOC is not an utter necessity, as these can also be immunostained after quenching of signal (25), such retention inarguably reduces both time and cost of the procedure and does not rise the question of efficacy and signal specificity associated with the use of immunohistochemistry. Our analysis revealed that organic solvents (i.e., 3DISCO and uDISCO) cause a rapid decay of EGFP and DsRed fluorescence, as we could not detect any signal on the first day after completion of clearing procedure. Although this is strongly supported by other reports (38–40), such rapid decay could be also accelerated by the exposure of fluorochromes to light and mechanic stress during isolation and selection of T cells for the adoptive transfer. However, some reports (performed mainly on brain tissue) suggest that solvent-based clearing done at 4°C, rather than at RT, and/or with slightly alkaline dehydrating agents (e.g., THF or tert-butanol) and RI-matching solutions (DBE), might significantly prolong the lifetime of fluorescence signal (31, 41, 42). Very recently, EGFP signal was also greatly stabilized in murine brain tissue by Hahn et al. (43) who proposed first purification of peroxides and aldehydes in dehydrating (THF) and RI-matching solutions (DBE, BABB), and second addition of propyl gallate (efficient scavenger of peroxides that prohibits nascence of aldehydes that otherwise quench GFP) to DBE during RI-matching step. Preservation of EGFP signal in LNs treated with either glycerol or commercially available FocusClear was also evaluated by Song et al. (28), who observed that incubation in 75% v/v glycerol over 30 min further increases transparency but also reduces fluorescent signal. In our settings, however, 75% v/v glycerol was remarkably rapid, as 30-min-long incubation allowed us to achieve ∼550 μm imaging depth using multiphoton microscopy, but also did not cause any significant quenching of the fluorescent signal (from DsRed and EGFP) within the first 24 h. Unfortunately, glycerol turned out to be unsuitable for long-term storage, as it both highly deformed the samples and caused quenching of both EGFP and DsRed signal within 2 wk postclearing. In contrast, as presented in Fig. 2B, in case of some of the techniques, namely SCM, CUBIC-R2, and C_e3D, the normalized fluorescence intensity of EGFP raised over a 2-wk period. Hence, it is evidently the sum of two factors: éclat retention of EGFP and further increase in transparency caused by the prolonged incubation in imaging solutions, RIMS, CUBIC reagent 2 and C_e3D, respectively [as the process of TOC keeps going on with time (43)]. CUBIC-R1A, published online by Ueda and Susaki (available at http://cubic.riken.jp), was developed specifically to increase the stability of fluorescent proteins, that could be diminished by original CUBIC reagent 1. Indeed, we observed that CUBIC reagent 1A stabilizes fluorescent signal better than reagent 1 (Figs. 2B, 5B). Such treatment, however, also induces great and variable LN expansion (Fig. 1E) and should be applied with caution as the tissue size alteration is another important factor to be taken into account, especially during quantitative analysis of the data. Our investigation of projected tissue area of the same samples before and after completion of clearing (Fig. 1C) shows that there is no effective TOC protocol that, when applied to murine LNs, will fully preserve their size. Thus, a particular caution should be taken while evaluating and reporting of such size alterations by means of, for example, projected sample/cell area/volume, which are usually neglected, as in the case of quantification of high endothelial venules length and thickness after methanol (dehydrating agent) treatment (9). Whereas CUBIC-related protocols evidently expand LN volume, C_e3D causes minor tissue shrinkage. The latter observation contradicts the results obtained by authors of C_e3D protocol who, based on the morphology of myeloid dendritic cells in LNs and small intestines, concluded that this TOC technique does not influence LN/intestine tissue volume (29). However, supplementary data in this article show that C_e3D induces either statistically significant tissue shrinkage (brain, intestine, muscle) or tendency to induce shrinkage (liver and thymus) with only bone and lung being unaffected in size. Importantly, very recently Bossolani et al. (40) compared the suitability of nine TOC methods for intestine research and reported very similar shrinkage values for intestines, as we did for LNs. It should be noted that neither tissue shrinkage nor expansion are favorable under all circumstances (24, 33). For example, severe tissue shrinkage observed after DISCO techniques (Fig. 1E) can significantly lower both imaging time and size of related data (which is especially desirable while handling massive LSFM datasets). In contrast, physical compression of tissue also reduces the effective resolution and might impede data analysis. This issue is especially true for LNs characterized by high cellular density and has already limited multicolor fate mapping of dendritic cell progenitors by Cabeza-Cabrerizo et al. (17), who followed uDISCO protocol. Conversely, at the cost of data storage and acquisition time, tissue expansion facilitates elucidation of tissue three-dimensional microstructure and allows to perform super resolution imaging with diffraction-limited objective lens (44). Hence, we observed mild expansion of LNs after SCM and CUBIC-related clearing convenient during segmentation of T cells. However, easier/more accurate segmentation condition will only be true if the same level of transparency is achieved between particular swollen and shrunken tissue. Notably, during SCM, a rigid hydrogel mesh is being created that in our evaluation generated more predictable expansion. Certainly, novel approaches to perform such moderate and controllable expansion will be of great importance (45, 46).

As TOC is still a novel set of methodologies, it is not unexpected that relatively meager adjustments of, for example, temperature of incubation or pH can lead to substantially improved clearing protocols (31, 39). As different groups apply various concentrations of PFA (ranging from 0.4 to 4%) as a fixative (28, 29, 38), we decided to evaluate the influence of 1% versus 4% PFA fixation on LN clearing capability. Evidently, 4% PFA fixation led to improved epitope retention in case of anti-CD31 Ab, as represented by higher contrast of the obtained images with every tested protocol (Fig. 7, Supplemental Fig. 3). This effect was again true for the tested Abs against B220 and CD8 (Fig. 8, Supplemental Fig. 4), which also substantiated already reported observation that some TOC methods might impede recognition of particular epitopes (29) (such as loss of B220 with retention of CD8 signal in the case of DISCO-treated samples). Out of the tested TOC methods, only C_e3D, glycerol, and CUBIC allowed us to achieve satisfactory immunostaining. Moreover, we report that although processing of tissue with TOC before immunohistochemistry is desirable to increase diffusion rates of Abs, which is used in original (26) and following CUBIC-based articles (47), it might also highly limit the staining efficacy as shown in Fig. 7B and further in Fig. 9. Importantly, this might explain why Li et al. (29) obtained complete absence of B220 and almost complete absence of CD4 and CD8 with CUBIC clearing, whereas we constantly achieved excellent results using the same clones of Abs. Thus, as long as full penetration of Abs can be achieved without prior TOC, it is advised first to stain and then to apply clearing reagents. Our study also confirms previous observations that harsh iDISCO protocol, although developed specifically for the efficient immunolabeling compatible with further solvent-based clearing, may cause depletion of some epitopes, as CD31 in particular (32).

Recently, Nojima et al. (48) presented that CUBIC pipeline is applicable for lung and LN clearing of both mouse and human origin. The demonstrated approach was compatible with H&E staining and led to a discovery of metastases omitted during standard clinical evaluation. In our work, we report additional approaches compatible with subsequent histological sectioning and H&E staining that include C_e3D, SCM, and glycerol (if short periods of incubation with the latter are applied). Based on these findings, CUBIC, C_e3D, and SCM (and presumably other CLARITY-related methods) might even be used in TOC protocols applicable to the clinical material. This would, however, require imaging of the entire LNs, most probably with LSFM, for the balanced time-to-cost ratio. Recent reports revealed that such imaging of murine [C_e3D (49), CUBIC-R2 (16, 38)] and human LNs [CUBIC-R2 (48)] is possible. Unfortunately, both of these are viscous solutions of high RI, prone to formation of light-obstructing bubbles or precipitation in imaging chamber (with the latter being especially true for CUBIC reagent 2 in our experience). In this study, for the first time, to our knowledge, we show convincing data indicating that CUBIC reagent 1 by itself is an optimal clearing and imaging solution for LSFM. Although we have not successfully applied C_e3D to LSFM imaging because of the limitations mentioned above, this approach readily possesses genuine advantages for confocal-based histocytometry (50). Our analysis, similarly to a recent report by Bossolani et al. (40) demonstrated on intestines, confirmed C_e3D to greatly preserve both fluorescence of proteinaceous fluorophores and structure of tissue (with only minor LNs shrinkage and excellent retention of their morphology as presented by postclearing H&E staining). Moreover, clearing of inguinal LNs took only 24 h and was completed with one easy to prepare solution. A recent protocol for RNA detection in C_e3D-cleared samples extends its applicability even further (50).

Finally, it should be underscored that during selection of the ideal TOC approach to a particular experiment, the unique features of different methods should be also taken into account. For example, although SeeDB and FRUIT do not seem to be suitable for LN clearing at first glance, these are exclusively compatible with lipophilic dyes (21, 30) and could be applied to visualize distribution of liposomes. Otherwise, application of 1) C_e3D for confocal-based histocytometry, 2) CUBIC-R1 for LSFM-based inspections, and 3) DISCO for imaging of huge lymphoid tissue (e.g., murine mesenteric LNs) stained with stable fluorescent dyes, such as the Alexa Fluor family, is advised.

We thank Dr. Maximilian Gorelashvili from Zeiss for assistance with the Z.1 microscope.

The authors have no financial conflicts of interest.