High-resolution confocal and light-sheet imaging of collagen 3D network architecture in very large samples

Summary

Although notoriously difficult, imaging collagen network architecture, a key element affecting tissue mechanical properties, is of paramount importance in developmental and cancer biology. Here, we introduce a simple and robust method of whole-mount collagen staining with the ‘Fast Green’ dye that provides unmatched visualization of collagen 3D network architecture, via confocal or light-sheet microscopy, compatible with solvent-based tissue clearing and immunostaining.

Graphical abstract

Highlights

• •Fast Green staining yields unmatched visualization of collagen 3D network architecture
• •Fast Green stains collagen with high specificity in anhydrous conditions
• •Fast Green staining is compatible with tissue clearing and immunostaining
• •Fast Green staining will facilitate studies on morphogenesis and tumor progression

Molecular biology; Cell biology; Microstructure

Introduction

Collagen, the principal structural protein of the extracellular matrix, is the most abundant protein in animals, e.g., amounting to about one-third of the protein mass in vertebrates. The emergence of collagen I and collagen-binding integrins in the early ancestor of vertebrates is likely to have greatly facilitated the success of this lineage through the evolution of characteristic morphogenetic processes and anatomical systems requiring a stiffer extracellular matrix (ECM) and better mechanotransduction.¹ The collagen triple helix accounts for 96% of the composition of collagen I (the most abundant collagen type in vertebrates) but is also present in all other 27 members of the collagen superfamily, albeit in smaller proportions (e.g., <10% in collagen XII).² In collagen I, II, III, V, XI, XXIV and XXVII, series of overlapping triple-helix molecules assemble to form nanoscale molecular ropes, called fibrils, characterized by high tensile strength.^3,4

Collagen fibrils can further organize into so-called ‘fibers’ that are visible by optical microscopy. At the higher spatial scale of tissues, sheets of aligned fibrils or fibers can form a ‘twisted plywood’ organization, i.e., the direction of alignment is shifted among successive sheets by an angle that is variable within and among tissues.⁵ The avian cornea or coelacanth scales exhibit a yet more complex ‘double-twisted plywood’ organization in which pairs of orthogonal sheets jointly shift their orientations.^6,7 As another example, the most basal dermal layer of the skin in lizards and turtles consists of a collagenous ‘basement lamella’ forming an orthogonal plywood structure: fibrils or fibers of even and odd layers are oriented nearly perpendicularly to each other.^8,9 These various collagen network architectures endow different tissues with distinctive mechanical properties such as anisotropic stiffness and specific non-linear strain-stiffening responses. Note that the latter is because of a mechanical phase transition, from bending stress at low strain to stretching stress at larger strains, also observed in isotropic (random) collagenous networks; the exact strain-stiffness relation depending on the mean connectivity of the network.¹⁰ Through cell-ECM interactions, collagen architecture also modulates cell behaviors such as cell migration along aligned collagen fibers in developing tissues.¹¹ Note that the tangential disposition of straightened collagen fibers is considered a signature of desmoplastic tumor ECM, whereas perpendicular alignment of collagen fibers in contact with the tumor provides a route for invasion of cancer cells away from the primary cluster.¹² These signatures are often included among the parameters used to predict cancer progression, invasiveness and treatment response.^13,14

Given the importance of collagen architecture in developmental and cancer biology, transmission electron microscopy and histological techniques (including tissue contrast enhancement with hematoxylin and eosin, picrosirius red, van Gieson and Masson’s trichrome stainings) have been widely used for over half a century to visualize collagen in tissues. Image analysis approaches have also been developed to quantify bulk amounts of collagen, as well as the length, width, density, straightness and alignment of collagen fibers from 2D microscopy data.^{15,16,17,18,19,20,21,22,23,24} However, given that the global 3D architecture of the collagen network can typically vary extensively across a given tissue sample, it cannot be inferred effectively from these 2D approaches.²⁵

Design

Fast Green FCF (FG) is a synthetic additive used for coloring food products such as ice cream, fruit-based spreads and candies. For 2D histological applications, and similarly to Light Green SF and Aniline Blue, FG has been used in Masson’s trichrome staining (in aqueous glacial acetic acid) after treatment of sections with phosphotungstic or phosphomolybdic acids.²⁶ This acidic treatment has the advantage to improve the specificity of collagen staining with anionic dyes.²⁷ In addition, Sirius Red (SR) has been previously combined with FG in aqueous picric acid as FG counterstains collagen marked with SR in bright field microscopy of paraffin sections.²⁸ The combination of FG and SR also allows us to quantify (by spectrophotometry after NaOH-methanol elution of both dyes from stained paraffin sections) the relative proportions of green-stained non-collagen proteins and red-stained collagen.

Here, capitalizing on the fluorescence of FG, we report on a very simple and robust new method of highly-specific whole-mount staining of collagen fibers that provides unmatched confocal and light-sheet microscopy visualization of collagen 3D network architecture in tissue samples that vastly differ in their sizes as well as collection, fixation and storage conditions. Importantly, this new approach is compatible with solvent-based tissue clearing²⁹ as well as with immunostaining.

FG produces a fluorescent signal in far-red wavelengths, facilitating its combination with antibody-coupled labels and/or nuclear stains emitting at shorter wavelengths, hence, making it possible to produce multicolor 3D images with at least four fluorescence channels (blue, green, red, and far-red). We found that FG staining is highly specific to collagen when applied and maintained in water-free conditions. This allows introducing FG staining in an anhydrous solvent as an additional incubation step after dehydration of a sample before clearing (e.g., in methanol when using iDISCO+ whole-mount immunostaining-clearing³⁰).

Results

Confocal 3D imaging of FG-labeled collagen exhibits multiple key advantages over second harmonic generation

So far, second harmonic generation (SHG) has been the gold standard technology for 3D collagen imaging. SHG occurs when an intense laser pulse undergoes a non-linear second-order polarization while passing through crystalline non-centrosymmetric material; this results in the generation of photons at twice the energy of the incident laser-light photons. Very few polymers in animal tissues, such as collagen, tubulin, or myosin in acto-myosin complexes, can produce SHG, explaining the very high specificity of this microscopy technique.³¹ Only crystalline collagens (i.e., fibril-forming, mostly type I and II) yield sufficient SHG signal because of its coherent amplification by the tightly-aligned harmonophores along the collagen triple helix and within fibrils.³² Although, in principle, SHG microscopy does not require staining and/or clearing, hence allowing for in-vivo visualization of fibrillar collagen,³³ refractive index matching through clearing of fixed tissues allows for improved scanning depth and produces much sharper images through reduction of light scattering.³⁴

One first key advantage of FG labeling over SHG is that FG allows for collagen network 3D imaging in highly heterogeneous samples (Figure 1A). This is explained by the fact that collagen fibers oriented parallel to the direction of laser light propagation (z axis) are efficiently visualized with FG labeling whereas they do not produce SHG signal because of their centrosymmetry.³⁵ Hence, if some segment of a fiber is oriented predominantly along the z axis, the continuity of this fiber on an SHG 3D image stack is lost in the corresponding region, whereas fiber continuity is correctly identified with FG staining (compare segments highlighted with arrowheads in middle and lower panels of Figure 1B). The issue becomes even more severe with smaller fibers as they become virtually undetectable by SHG (Figure 1C, bottom panel), whereas they are clearly identified with FG labeling (Figure 1C, middle panel). More generally, intensity of the SHG signal depends on the distribution of harmonophores within fibers, their respective orientations, and the direction of the laser light linear polarization. Hence, SHG signal is inhomogeneous along fibers even if they are approximately aligned with the scanning plane, causing again continuity loss (Figure 1D, lower panels), especially for thin fibers. Conversely, FG signal is homogeneous along fibers, irrespective of their orientation and thickness (Figure 1D, middle panels).

Figure 1.

Fast Green (FG) staining substantially improves collagen 3D network architecture visualization over Second Harmonic Generation (SHG)

(A) Confocal microscopy under FG staining (XY plane, maximum intensity projection, MIP, colour-coded by depth) reveals the fine collagen fibers 3D architecture in the upper dermal layer of seven embryonic skin scales (top left inset) in the African house snake (Boaedon fuliginosus). Panels (i), (ii) and (iii) show close-ups of three regions (highlighted with dotted frames) characterized by very different 3D network architectures.

(B–D) FG staining (upper and middle rows of images) provides substantial improvement in resolution over SHG microscopy (lower row). Overviews (top panels) show MIPs (B and C) or single optical section (D) with merged FG and nuclear staining; middle and bottom panels show magnified regions (dotted frames on top panels) with depth color-coded MIPs of FG or SHG signals. (B) Superficial dermal layer of a dorsal scale of a Madagascar giant day gecko (Phelsuma grandis): SHG signal is absent, even for thick fibers, in sections oriented out of plane of scanning (arrowheads). (C) Out-of-plane thin fibers in a P. grandis embryonic scale are visualized with FG staining but not with SHG. (D) Embryonic deep dermal layer of a ball python (Python regius): even when collagen fibers are oriented predominantly in the plane of scanning, heterogeneity of SHG signal (lower panels) allows following single fibers only when they are thick enough, whereas FG signal is much more homogeneous (middle panels). Nuclei were stained with antibodies against lamin B1 (Alexa Fluor 555) in (D), YoPRO1 in (B) and DAPI in (C) highlighting that FG can be used together with red, green and blue fluorescent stains. In (C) nuclei are visible together with SHG signal because of two-photon excitation of DAPI. Confocal and SHG microscopy were performed at the same resolution and with the same z-step.

Second, the required imaging time is typically much smaller (more than 10 times, although the exact ratio depends on the thickness of fibers) for FG-labeled collagen than for SHG (Figure 2). Therefore, contrary to SHG, FG-labeling allows producing very large tile scans, e.g., of whole embryonic organs. This point is illustrated by the top-left inset of Figure 1A, indicating that the full scan includes seven full embryoning scales (full sample size 1600 × 1100 μm). In the case of combining fluorescence with multiphoton imaging (such as SHG), it is often correctly emphasized that the latter greatly reduces out-of-plane bleaching. However, it must be noted that the high laser intensity required by SHG causes considerable in-plane bleaching of fluorescent stains.

Figure 2.

Effect of scanning time (pixel dwell time) on image quality for FG staining versus SHG

(A–E) A sample of P. grandis skin was stained with FG and YoPRO1 (nuclear staining), cleared and scanned in the same region with confocal (A–C) and with multiphoton (D and E) modes using the same resolution and z-step. Pixel dwell time is specified for each column of panels. (C) illustrates that FG fluorescence imaging with a 0.4 μs pixel dwell time (10 times reduced compared to optimal 4 μs in (A)) still identifies collagen fibers much better than SHG with an optimal pixel dwell time of 20 μs (D). Overviews (top panels) show maximum intensity projections (MIPs); successively magnified regions (middle and bottom panels) show depth color-coded MIPs.

FG staining overcomes limitations of collagen immunostaining

FG staining also proves to be substantially superior to whole-mount collagen immunostaining for reasons similar to DNA-intercalating-dyes nuclear staining being superior to nuclear-antigen immunostaining. Indeed, antibody-based staining is species-specific, sensitive to sample collection, to fixation and to storage conditions, and the penetration of antibodies can require weeks of incubation for embryonic tissues,²⁹ whereas it is often impossible for differentiated tissues³⁶ (Figure 3). Conversely, the low molecular weight of FG (808.9, i.e., about 200 times smaller than immunoglobulin G) greatly facilitates its penetration within tissues, despite that incubation is performed in an anhydrous solvent rather than in saline water with detergents. We found that FG deeply penetrates tissues when a potassium-hydroxide solution incubation step is introduced before classic permeabilization, hence, allowing for the imaging of very large samples. For example, Figure 4A shows the embryonic hand of a Madagascar giant day gecko (Phelsuma grandis), where FG staining reveals fine details of collagen deposition and alignment within developing cartilages/bones, ligaments, tendons, as well as collagenous sheets that compartmentalize these structures in the musculoskeletal system. Note that FG more intensively stains calcified bone tissue (cf. inset of Figure 4A with alizarin red staining, AR). As FG penetrates tissues beyond the centimeter scale, our method is easily applicable to very large samples undergoing light-sheet microscopy. For example, FG staining (1) marks, in an embryonic elephant trunk sample (Figure 4B), collagenous sheets of connective tissues surrounding muscle bundles as well as associated with nerves and blood vessels; (2) easily diffuses through the keratinized skin of newborn mice (whereas this tissue is impermeable to antibodies) and highlights tail vertebral bones and tendons (Figure 4C); and (3) counterstains embryonic structures and organs in a more distinguishable way that nuclear staining (Figure 4D). Note that the elephant sample (Figure 4B) had been fixed in formalin in the 1970s, and stored in the same fixative ever since, highlighting the robustness of the collagen FG staining.

Figure 3.

FG stains collagen much more efficiently than immunostaining

(A) Confocal microscopy imaging (single XY optical section in the top dermal layer) of P1 mouse skin stained with FG (top panel) versus anti-collagen antibodies (bottom panel). Antibodies penetration is impaired on one side by the keratinized epidermis and, on the other side, by densely packed dermal collagen fibers; this results in the absence of collagen immunostaining signal in the depth of the skin, even after 3 weeks of incubation. Conversely, FG stains collagen across the whole sample and highlights the pattern of hair follicles.

(B) Confocal microscopy imaging (MIP corresponding to 3 μm thickness in Z) of E16.5 mouse skin. The second and third rows of panels illustrate that the upper dermis thin and densely packed fibers are not distinguished by immunostaining while they are identified with FG staining. The two bottom rows of panels indicate that fewer fibers are distinguished with antibodies than with FG in the lower dermis. On the overview top row of panels (XZ), the dotted and continuous white outlines indicate the upper border of the epidermis and a developing hair follicle, respectively.

Figure 4.

FG staining is highly effective in multiple sample types

(A) Confocal microscopy imaging of the embryonic hand of a Madagascar giant day gecko (P. grandis) stained with FG, DAPI and antibodies against β-tubulin III (to mark nerves). Panels (i), (ii) (iii) and (iv) show close-ups of four regions (highlighted with dotted frames in the overview top left panel) illustrating the efficiency of FG to stain ligaments, tendons (arrows), dermal collagen fibers (asterisks) and developing bones/cartilages with brighter calcified regions (arrowheads). The same calcified regions are visible under a stereomicroscope after alizarin red staining of the same sample (AR inset). The full confocal image stack consists of 264 tiles, each with 256 optical sections; the ranges of images used for MIPs in different panels are specified between parentheses. The two sub-panels in (iv) show a single optical section through wrist bones/cartilages and highlight the efficiency of FG in defining borders of embryonic skeletal structures (top sub-panel) undetected with DAPI nuclear staining (bottom sub-panel).

(B) Light-sheet microscopy imaging of a piece of African elephant embryonic trunk (proboscis). FG efficiently stains collagenous sheets surrounding muscle bundles (dashed lines in panel (i) are separating muscle groups) as well as collagen layers associated with nerves (arrows) and blood vessels (arrowheads). Note the similarity of the patterns of nerves and blood vessels in the embryonic trunk with those described in the adult trunk using macroscopic serial sectioning.³⁷ (C) Light-sheet microscopy imaging of a postnatal day 1 (P1) mouse tail: the top-left overview panel (XZ plane) is a cross section at the level of an intervertebral disc and shows tendons arranged in four groups, whereas panels (i) and (ii) show near-frontal sections (XY planes) and illustrate efficient penetration of FG inside vertebral bones. The MIP of panel (ii) shows attachments of tendons on vertebral processes as well as the dense network of dermal collagen interrupted by the passage of hair follicles. (D) Light-sheet microscopy imaging of an embryonic day 14.5 (E14.5) mouse embryo. Whole-embryo MIP (left panels) show structures with highest FG accumulation, such as calcified collar bones (dashed circle) and skull bones (closeup in panel (i)) as well as the largest blood vessels (arrowhead shows the aorta). Single optical sections of FG-stained (middle panels) and nuclear-stained (YoPRO1, right panels) E14.5 mouse embryo indicates that FG is more efficient for counterstaining because it highlights borders of organs and collagen-rich structures such as fasciae, e.g., the diaphragm fascia is indicated with an arrow.

Note that the collagen binding protein CNA35³⁸ could also be an efficient alternative to immunostaining of collagen^16,39 as it exhibits better penetration than antibodies together with strong and specific binding.⁴⁰ However, CNA35 is not commercially available, such that researchers need to construct or obtain a plasmid encoding the protein, express it in E. coli, purify the protein and conjugate it with a fluorophore. Note that fluorophore-conjugated CNA35 has a much larger molecular weight than FG, making it unlikely to stain collagen as efficiently as FG in large samples.

FG staining in anhydrous conditions is highly specific to collagen

Apart from collagen, fibrous tissues that physiologically undergo repeated cycles of extension and recoil (e.g., in lungs, skin and blood vessels) can contain a substantial amount of so-called ‘elastic fibers’.⁴¹ Mature elastic fibers, which are big enough to be distinguished with diffraction-limited microscopy, are comprising an amorphous central core of the protein elastin, surrounded by an outer mantle of fibrils made of the glycoprotein fibrillin.⁴² Staining-free multiphoton microscopy allows us to simultaneously visualize collagen, through SHG, and elastic fibers, through their two-photon excited (2PE) autofluorescence.⁴³ Triggered by the same incident laser light, SHG and 2PE signals are reliably separable: as SHG energy is conserved, the resulting signal occurs at exactly half the wavelength of the incident light, whereas a partial loss of energy on 2PE fluorescence results in a longer wavelength of the emitted photons.

To examine the specificity of FG staining for collagen fibers, we first confirmed with multiphoton microscopy the presence of both collagen and elastic fibers (identified with SHG and 2PE, respectively; Figure 5A) in a label-free control sample of a bovine visceral pleura. Remarkably, when applied in aqueous conditions, FG stains elastic fibers much brighter than collagen fibers (Figure 5B). This result highlights the importance of proper sample dehydration for obtaining collagen-specific FG staining but also extends the utility of the method (for distinguishing collagen from elastic fibers). For example, the left panels of Figure 5C illustrate that both elastic and collagen fibers are detected with FG single-photon fluorescence when the sample is not fully dehydrated (e.g., because of shortened dehydration steps). In that case, elastic and collagen fibers can be distinguished by either (1) merging the collagen/elastin FG signal with collagen-specific SHG (Figure 5C, lower-left panel) or (2) using fluorescence lifetime imaging microscopy (FLIM;⁴⁴ Figure 5C, right panels). Different fluorophores exhibit different fluorescence lifetimes, but the lifetime of a given fluorophore also depends on its local microenvironment. Here, we demonstrate that collagen-bounded and elastic-fiber-bounded FG exhibit different fluorescence lifetimes (Figure 5C, right panels), allowing to distinguish these two types of fibers in partially hydrated samples.

Figure 5.

FG staining in anhydrous versus aqueous conditions discriminates elastic fibers

Confocal (intensity-based and FLIM) and multiphoton (SHG and 2PE) single optical sections of visceral pleura in whole-mount bovine samples. FG - Fast Green, SHG - second harmonic generation, 2PE - two-photon excitation, FLIM - fluorescence lifetime imaging microscopy.

(A) Control sample without staining: SHG and 2PE identify collagen and elastic fibers, respectively.

(B) FG staining in aqueous conditions shows much higher specificity for elastic than for collagen fibers (top panel); the latter can be imaged with SHG (lower panel).

(C) FG staining in a partially-dehydrated sample marks both collagen and elastic fibers which can be discriminated either by combining FG fluorescence with SHG (bottom left panel), or by the different fluorescence lifetimes of collagen-bound FG versus elastin-bound FG (right panels).

(D) FG staining in anhydrous conditions is collagen-specific, as confirmed by the co-localization of the single-photon fluorescence signal of FG and the SHG signal (top panels). In the two bottom left panels, the 2PE fluorescence of both elastic fibers themselves (white arrowheads) and of collagen-bound FG is superposed to the single-photon FG fluorescence. The homogeneous fluorescence lifetime (bottom right panels), and absence of FLIM signal in the position of the elastic fibers confirm that, in anhydrous conditions, FG only binds collagen (and not elastic) fibers.

Conversely, in a properly dehydrated sample, FG staining is strictly specific to collagen fibers, as confirmed by SHG (Figure 5D, top panels). Although elastic fibers are present in the sample (as detected by 2PE; Figure 5D, bottom left panels), the homogeneous fluorescence lifetime across the sample (Figure 5D, bottom right panels) confirms that FG only binds to collagen (and not to elastic) fibers in anhydrous conditions.

To further test the specificity of FG staining for collagen in anhydrous conditions, we imaged a sample with thin collagen fibers (Figure 6A, left column of panels) and increased the image brightness until other structures became visible. After magnifying the brightness 10 times (Figure 6A, middle column of panels), hence, saturating the signal associated with collagen fibers, we could visualize red blood cells and the cytoplasm of fibroblasts. Note that very small collagen fibers (identified as such with FLIM; Figure 6A, right column of panels) also became visible at the border of small blood vessels. Conversely, when FG staining is applied in aqueous conditions, background fluorescence is substantial and the intensities of FG fluorescence are similar in collagen fibers and cells (Figure 6B).

Figure 6.

Low degree of unspecific FG staining in anhydrous conditions

Single optical section (confocal imaging, intensity-based and FLIM) in African house snake (Boaedon fuliginosus) embryonic skin.

(A) Optical section spanning the skin from the collagen-rich deep dermal layer to the hypoderm bearing blood vessels. FG staining applied in anhydrous conditions is very specific to collagen: full intensity range (0–255) shows dermal collagen fibers (left panels), whereas the visualization of red blood cells, the cytoplasm of fibroblasts, and very small collagen fibers bordering a blood vessel wall requires to magnify the image brightness 10× (middle panels). Different fluorescence lifetimes of bound FG, represented by different colors in the image (right panels) allow to distinguish red blood cells (cyan), the cytoplasm of fibroblasts (green) and small collagen fibers (orange).

(B) An optical section through the dermis stained with FG in aqueous conditions: FG binding is less specific to collagen (i.e., the FG fluorescence signal in cells is as intense as in collagen) than in anhydrous conditions.

Benefits of FG staining in aqueous conditions

Although specificity of FG binding to collagen is substantially reduced in aqueous conditions, the contrast is still adequate for distinguishing collagenous structures in the musculoskeletal system (Figure 7). Hence, when this level of specificity is acceptable, FG staining in aqueous conditions is very valuable because it can be (1) combined with other staining modalities that require non-dehydrated samples, such as phalloidin staining of actin (Figures 7C and 7D); (2) used to mark elastic fibers (Figure 5B), highly keratinized integumentary structures (Figure 7D) and/or calcified bone tissue (Figure 7E; instead of using alizarin red).

Figure 7.

FG staining in aqueous conditions

Confocal imaging of a P. grandis embryonic hand stained with phalloidin (marking actin filaments) and FG without dehydration.

(A) Single optical section.

(B) Close-ups of XY and XZ planes show that the contrast of FG staining in water is still sufficient to highlight collagen-rich structures, such as dermis (arrows), ligaments (arrowheads), tendons (orange arrows) and calcified regions of bones. Muscles are distinguished as slightly darker than the connective tissue around them. The collagenous extracellular matrix of bones/cartilages is not stained, while cell nuclei are marked.

(C) Embryonic ligaments can be distinguished from muscles with FG staining because FG signal is much more intense in the former (left panel). Conversely, actin staining can distinguish these two tissues only at magnification high enough to identify sarcomeres in the muscles (central panel); the right panel shows the merge between FG and actin staining.

(D) Highly keratinised tissues such as claws are strongly stained with FG in aqueous conditions, whereas they are not or weakly marked by actin staining.

(E) 3D representation of the full volume viewed from the ventral side of the embryonic hand, illustrating that calcified bones and tendons are well stained with FG even in aqueous conditions.

Discussion

Although further research is warranted to understand FG binding specificity in different collagen types, we have shown here that whole-mount FG staining is a simple and robust method that provides, in anhydrous conditions, unmatched visualization of collagen 3D network architecture because of a combination of characteristics not jointly observed in any other method: deep sample penetration, high specificity for collagen, insensitivity to fibers orientation, and bright fluorescence signal. Note that the method can also be combined with fluorescence staining by other dyes (including immunostaining) as well as with 2PE and SHG. This new method can be used for any type of fresh or historical sample and is easily applicable in any research institution with access to conventional confocal or light-sheet microscopy.

FG staining for studying tissue mechanical properties

It can be argued that morphogenesis reduces to the spatiotemporal modulations of three physical quantities: volumetric growth, tissue mechanical properties and active forces.⁴⁵ The quantification of 3D tissue growth (e.g., by identifying the spatial distribution of proliferating cells within tissues) has been greatly facilitated by advances in imaging and tissue clearing. Estimating mechanical properties in 3D remains difficult, although the recent development of so-called ‘optical elastography’ techniques provides potential new solutions: local variation of stiffness within a sample volume can be estimated by Brillouin microscopy⁴⁶ and optical coherence elastography.⁴⁷ Unfortunately, these new approaches poorly describe mechanical properties of fibrous tissues. For example, collagen fibers endow tissues not only with local stiffness properties, but also directional long-distance tensile strength. The latter is likely to play a major role in morphogenetic processes not only through anisotropic response to homogeneous stress but also by the transmission of active forces generated by cellular activities.^48,49 In tissues with simple spatial organization of fibers, e.g., in arterial walls, 2D histological data is sufficient for proper biomechanical modeling.⁵⁰ On the other hand, the precise quantification of fibers distribution and 3D orientation is a key prerequisite to proper biomechanical modeling in tissues with heterogeneous collagen network architectures (such as the embryonic skin of snakes, illustrated in Figure 1A). Hence, we anticipate that FG staining, by allowing high-resolution imaging of 3D collagen network architecture, will greatly facilitate the use of imaging data in the study of organs biomechanics, but also for the quantitative mathematical/physical description of complex processes in morphogenesis and tumor progression.

Limitations of the study

Although we have shown that specificity of FG staining in aqueous conditions is sufficient to image collagen-rich structures, such as tendons, ligaments and dermal layer of skin (Figure 7), signal intensity and specificity is substantially smaller than in dehydrated samples. Also, two characteristics of FG staining must be taken into account when planning an experiment:

• 1)As for virtually any fluorescent molecule, the emission spectrum of FG includes a long tail, beyond the wavelength of emission maximum intensity, that contaminates near-infrared fluorescent channels. Hence, when combining FG staining with other imaging modalities, it is advised to select florescence molecules (e.g., for immunostaining) emitting in the blue, green and/or red channels.
• 2)Accumulation of FG in calcified regions of bones can, in some instances, force the user to underexpose the signal in other regions of the sample. However, when problematic, this drawback can be mitigated by decalcification of bones before FG staining.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit polyclonal anti-Lamin B1	Abcam	Cat#ab16048, RRID:AB_443298
Rabbit polyclonal anti-Collagen I	Abcam	Cat# ab21286, RRID:AB_446161
Mouse monoclonal anti-Tubulin β 3	Biolegend	Cat#MMS-435P
Chemicals, peptides, and recombinant proteins
Fast Green FCF	Thermo Scientific	Cat#A16520.22
Alizarin Red S	Sigma-Aldrich	Cat#A5533
DAPI	Invitrogen	Cat#D1306
TO-PRO-3	Invitrogen	Cat#T3605
YO-PRO-1	Invitrogen	Cat#Y3603
Phalloidin-AlexaFluor488	Invitrogen	Cat#A12379
30% hydrogen peroxide	Sigma	Cat#216763
TritonX100	Sigma	Cat#X100
Saponin	Sigma	Cat#S4521
Gelatin	Sigma	Cat#G9391
Sodium azide	Sigma	Cat#32002
Dibenzyl ether	Sigma	Cat#108014
Dichloromethane	Sigma	Cat#270997
Bouin’s fixative	Roth	Cat#6482.1
Agarose	Roth	Cat#2267.4
RapiCLEAR 1.52	SunJin Lab	Cat#RC152001
Experimental models: Organisms/strains
Mouse: C57BL/6J	The Jackson Laboratory	RRID:IMSR_JAX:000664
Giant day gecko (Phelsuma grandis)	LANE (Univ. Geneva)	/
Ball python (Python regius)	LANE (Univ. Geneva)	/
African house snake (Boaedon fuliginosus)	LANE (Univ. Geneva)	/
Software and algorithms
Fiji/ImageJ	Schindelin et al.51	https://imagej.nih.gov/ij/
Imaris	Oxford Instruments	https://imaris.oxinst.com/
Xuv Stitcher	Emmenlauer et al.52	https://lmb.informatik.uni-freiburg.de/people/ronneber/
Other
Kwik-Sil glue	Microprobes for Life Science
iSpacer 0.05 mm	SunJin Lab	Cat#IS201

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Michel C. Milinkovitch (Michel.Milinkovitch@unige.ch).

Materials availability

This study did not generate new unique reagents.

Experimental model and subject details

Giant day geckos (Phelsuma grandis), ball pythons (Python regius), and African house snakes (Boaedon fuliginosus) were housed and bred in Milinkovitch’s laboratory, Department of Genetics and Evolution, University of Geneva, Switzerland. Maintenance of, and experiments on animals were approved by the Geneva Veterinary Cantonal authorities (authorisation 33145/GE24) and performed according to the Swiss law. These guidelines meet international standards. Fertilised eggs were incubated on a moistened vermiculite substrate at 29.5°C and embryos were collected at specific developmental stages. C57BL/6J mice samples were donated by the Laboratory of Neurogenetics (Prof. Ivan Rodriguez), Department of Genetics and Evolution, University of Geneva. The sample of African elephant embryonic proboscis was provided by Prof. Nigel Bennett (University of Pretoria, South Africa). Visceral pleura samples were dissected from a bovine lung bought in a butcher shop.

Method details

Whole-mount staining and clearing

We integrate our FG staining method with existing protocols of immunostaining,⁵³ tissue clearing³⁰ and depigmentation,⁵⁴ supplemented, when needed, with a potassium hydroxide (KOH) treatment for additional tissue permeabilisation and/or dissolution of light-reflecting guanine nano-crystals in snake and lizard iridophore cells. All samples were fixed overnight in 4% paraformaldehyde, except the elephant embryonic proboscis, that was fixed and stored in 10% formalin for about 45 years. All samples were dehydrated in a series of methanol-in-PBS solutions (with 30%, 50%, 70%, and 100% methanol; 1 hour each). The mouse embryo at embryonic day 14.5 (E14.5) was fixed overnight with Bouin’s fixative and dehydrated using a larger number of steps in the methanol series (10% increments) to avoid deformation of skin surface geometry.⁵⁵ All samples were bleached in a mix of 30% hydrogen peroxide water solution (Sigma) and 100% methanol (1:2 volume to volume) on a rocking table illuminated with a white LED lamp, and washed in 70% methanol for 30 min to remove hydrogen peroxide. Bleaching time (cf. Table S1) was adapted to each sample. In case of prolonged bleaching, the hydrogen peroxide solution was changed every 12 h. Samples were then rehydrated with 70%, 50%, and 30% methanol-in-PBS solutions (1h each) before a final step of 100% PBS. Rehydrated samples were then rinsed in water, incubated in a solution of 0.1M KOH with 0.5% Triton X-100 (SIGMA) in water (cf. Table S1), and rinsed again in water. Samples were next transferred either to PBSGST (1x PBS with 0.2% gelatine, 10 mg/L saponin, 0.02% sodium azide and 0.5% Triton X-100) for permeabilisation and antigen blocking, or in PBST (1x PBS containing 0.5% Triton X-100) if immunostaining was not needed. PBSGST was then replaced with fresh PBSGST containing the primary antibodies (anti-Lamin B1, ab16048, Abcam, 1:500; and/or anti-Tubulin β 3, MMS-435P, Biolegend, 1:200; and/or anti-Collagen I, ab21286, Abcam, 1:200) for an incubation time indicated in Table S1, washed twice 3 h in PBST and incubated (Table S1) in a filtered solution (Millipore 0.22 μm) of secondary antibodies (AlexaFluor488- and/or AlexaFluor555-conjugated, Invitrogen, 1:250) in PBST. Nuclear staining (Table S1) was performed with DAPI, TO-PRO-3 or YO-PRO-1 (Invitrogen, 1:1000 for confocal or 3:1000 for light-sheet) in PBST. All steps were performed on a tube rotator at 30 rpm and protected from light with aluminium foil (except for bleaching, see above). Sodium azide (0.02% final concentration) was added to solutions during PBS(GS)T, antibody and nuclear staining steps longer than 12 h. Samples undergoing confocal microscopy were rinsed in water, placed on a coverslip, gently flattened to maximise contact with the latter, covered with a 0.5 mm-thick piece of 1.2% agarose (Roth) whose edges were secured on the coverslip with two narrow stripes of paper tape. Note that agarose was prepared in water instead of PBS in order to avoid appearance of crystals that cause imaging artefacts. Samples were then dehydrated directly on this montage (sample positioned on the coverslip under the agarose layer) using a series of methanol-in-water (not PBS) solutions (30%, 50%, 70%, and 100% methanol; 30 min each). A final step of overnight 100% methanol incubation was performed to ensure full dehydration. Samples were then incubated (cf. Table S1) in a 2 μg/mL solution of Fast Green FCF (FG; Thermo Scientific) in methanol on a rocking table for confocal samples and on a tube rotator for light-sheet samples. As shown in Figure S1, each sample was transferred after FG staining to a glass container in which the standard iDISCO+ clearing protocol was used 30: incubations (Table S1) in 66% dichloromethane (DCM, Sigma) in methanol then 100% DCM; followed by incubation in dibenzyl ether (DBE, Sigma) until transparency. Fresh DBE was used for storage. For samples undergoing confocal microscopy, we removed the agarose layer, submerged the sample in a drop of DBE, applied a thin layer of Kwik-Sil glue (Microprobes for Life Science) all along the perimeter of the coverslip, and covered the sample with a glass slide (making sure that the glue doesn’t come in contact with the sample); see Figure S1. The coverslip was cleaned from excess of DBE before imaging with confocal microscopy.

For FG staining in aqueous conditions (P. grandis embryonic hand, Figure 7), the PFA-fixed sample was washed in water, incubated in a water solution of 0.1M KOH with 0.5% Triton X-100, washed again in water, incubated in PBST for 1 day, incubated overnight in a PBS solution of 100 μg/mL FG with 1:100 Phalloidin-AlexaFluor488 (Invitrogen), washed in PBS for 1 h and placed on a coverslip. Excess of liquid was removed with paper and 30 μL of RapiClear 1.52 (SunJin Lab) was added on the sample for a 10 min clearing step. The coverslip was then attached to a glass slide using 250 μm-thick ‘double-sticky iSpaser’ (SunJin Lab). All incubation steps were performed at room temperature, on a tube rotator (30 rpm) and protected from light with aluminium foil.

Alizarin-red (AR) staining of calcified bone regions was performed on a sample (P. grandis embryonic hand, inset of Figure 4A) that was already stained with FG, DAPI and antibodies, cleared and imaged via confocal microscopy. The AR staining was performed as follows: the sample was removed from the glass slide, incubated in clearing solutions in a reverse order: DCM, methanol:DCM (1:2), pure methanol (2 times); each step for 10 min. Bleaching was then performed for 1 day in a solution of hydrogen peroxide in methanol (1:2) on a rocking table with illumination to remove previous stains. The sample was then rehydrated (70%, 50%, and 30% methanol-in-PBS solution before a final step of 100% PBS; 10 min each) and stained for 2 days (room temperature, on a rocking table, protected from light) in a water solution of 0.1M KOH with 0.5% Triton X-100 and 0.003% Alizarin Red S (Sigma-Aldrich).

Microscopy

A SP8 DIVE (Leica Microsystems) with a 60x oil-immersion objective (numerical aperture 1.4) was used for both confocal (intensity-based and FLIM) and SHG microscopy. FG was excited at 627 nm (laser pulse repetition rate = 80 MHz) and the signal was detected in the 630–730 nm range with the ‘HyD’ detector in ‘photon counting’ mode for both intensity acquisition and FLIM. In multiphoton regime, the tissue was excited at 860 nm, backward SHG was detected at 430 nm (detection window = 10 nm), and 2PE autofluorescence of elastic fibres was detected in the 500–600 nm range using an external ‘HyD’ detectors (i.e., the signal is not going through the pinhole). Note that, to cover the available 8 bit intensity range (that, in photon counting mode, corresponds to counting 0 to 255 single photons) while, at the same time, insuring a linear regime (limited to 60 counts per μs), pixel dwell time should be > 255/60 = 4.25 μs. However, FG staining gives such a bright fluorescence that the speed of confocal scanning was never limited by the FG emission intensity. Hence, for very large tile-scans (e.g., Figures 1A, 4A, and 7), we reduced the full time of scanning to a range of 1–2 days, by decreasing pixel dwell time to 0.8μs at the cost of compromising the detection linear regime and covering only two thirds of the intensity range. For FLIM, the scanning speed was reduced to ensure proper fluorescent lifetime measurement according to the threshold indicated in the LAS X version 3.5 (Leica Microsystems) software.

The speed of scanning under SHG microscopy is limited because of the much weaker signal intensity (in comparison to that of FG). To maximise SGH signal, we set the multiphoton laser intensity to two thirds of the level at which thermal damage of samples appeared during the acquisition of a single z-stack. Then, all SHG scans were performed with that laser intensity, but different pixel dwell time (8–20 μs,i.e., 10 to 25 times longer than for FG-stained samples) to optimise results, i.e., to cover the available 8 bit intensity range. For single optical sections in multiphoton regime, higher laser intensity was used.

An UltraMicroscope Blaze (Miltenyi Biotec) was used for light-sheet microscopy, with a 630/30 (30 nm band-pass window around 630 nm) excitation filter and a 680/30 emission filter for FG. A 4x objective lens was combined with an internal lens magnification of 0.6x to 2.5x for different samples. Stacks of 16 bit 2048 × 2048 images were acquired with a CMOS camera, with a 3 μm z-step.

Image processing

We used Imaris version 9.9.0 (Oxford Instruments) and Fiji⁵¹ to visualise volumetric data and produce maximum-intensity projections (MIPs). We compared the SHG and FG signals after a voxel-precision alignment also performed in Imaris. The ‘Temporal-Color Code’ function of Fiji was used to generate depth colour-coded MIPs. For tile scans, stitching was performed with Xuv Stitcher, version 1.8.⁵² As this software operates with Imaris format (.ims) files, the original imaging data was converted to .ims with the Imaris File Converter version 9.6.0 (Oxford Instruments). FLIM images were generated in the LAS X version 3.5 (Leica Microsystems) software.

Published: March 20, 2023

Data and code availability

• •Full resolution images used for Figures 1, 2, 3, 4, 5, 6, and 7 are available at https://www.lanevol.org/projects/FGstaining. Raw volumetric data used in this study are available from the corresponding author upon reasonable request.
• •This paper does not report original code.
• •Any additional information required to reanalyse the data reported in this paper is available from the lead contact upon request.