1. Lateral and axial resolution close to the coverslip

We first aimed at comparing seven commercially available mounting media with refractive indices ranging from 1.518 to 1.390, and determined the PSFs close to the coverslip. We imaged green fluorescent sub-resolution beads that were mounted directly on the coverslip, i. e. close to the objective lens (Fig. 2A, 2B). We determined a mean value of 255 nm for lateral and 523 nm for axial resolution with all tested mounting media observed. This set of data corroborated previous findings that close to the coverslip, a refractive index mismatch does not impair lateral and axial resolution in the media tested.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Lateral and axial resolution as a function of the mounting medium and focusing depth.

170 nm green beads have been mounted in CFM1, CFM3, 2’2-Thiodiethanol (TDE), Vectashield (VS), Fluoromount (FM), Mowiol (MW) or Prolong Gold (PG) mounting media as described in Fig. 1A, and measured directly at the coverslips (A, A’, B, B’) or in depth (B, B’, B”). Imaging was carried out with the 488 nm laser line collecting green emission. (A, A’) X, Z-projections of fluorescent beads are shown in A. Similar beads have been used to measure the lateral (dots) and axial (rhombus) resolution of the confocal microscope. Beads located at the coverslip show no significant difference in lateral (A’, mean value = 255 nm) and axial (A’, mean value = 523 nm) resolution through the different mounting media. Scale bar = 1 μm. (B, B’, B”) X, Z-projection of 170 nm beads mounted in CFM3 (B) or in Vectashield (B’) at the level of the coverslip, or at 58 μm or 116 μm depth. Axial resolution in high refractive index medium CFM3 (B”, triangles) remains constant with focussing depth, whereas axial resolution of beads increased with increasing focusing depth in Vectashield (B”, squares). Note that the lateral resolution remains unaffected in the two mounting media (B”, squares = CFM3 and circles = VS).

https://doi.org/10.1371/journal.pone.0121096.g002

2. Lateral and axial resolution at a distance from the coverslip

We next analysed how the refractive index mismatch impairs resolution when looking at beads at a distance from the coverslip. We restricted our work on the comparison of the high refractive index mounting medium CFM3 (ri = 1.518) and the widely used Vectashield (ri = 1.454). We built a sandwich by mounting green fluorescent sub-resolution beads on slides and coverslips that were separated by one or two spacers of 58 μm 3. 1A). We determined the lateral and axial resolutions of beads mounted in CFM3 and Vectashield at the level of the coverslip and at 58 μm and 116 μm depth. Using CFM3, the lateral and axial resolution remained constant throughout depth (mean values x,y = 224 nm and z = 538 nm). This means that spherical aberration is minimized when using CFM3 medium, and our observation corroborated previous findings of Hell and co-workers for a perfectly matched system [4, 5, 6]. Lateral resolution in depth is not impaired in a mismatched system [4] and we indeed observed a lateral resolution of 220–240 nm for beads mounted in Vectashield up to a depth of 112 μm. However, axial resolution was doubled to > 1.2 μm (depth = 58 μm) and tripled to 1.8 μm (depth = 112 μm) when using Vectashield. Furthermore, we observed a considerable loss of peak intensity at a depth of 60 and 112 μm and had to adjust the laser power accordingly in order to exploit the full dynamic range of the image. These results confirm previous theoretical and experimental data for glycerol (ri = 1.480), showing that spherical aberrations occur because of refractive index mismatches in the depth of the sample [4, 5].

3. Influence of the penetration depth on the axial resolution in mouse brain vibratome sections

We next aimed at performing a comparative study of different mounting media on a biological tissue. We carried out 100 μm thick vibratome sections of fixed mouse brains and mounted them in CFM3, Vectashield or the hardening mounting medium Mowiol. Images were taken with an Axiozoomer V16 equipped with a 1x objective in relief contrast mode (Fig. 3A-3C). This technique is based on Hoffman modulation contrast, which converts phase gradients into variations in light intensity [10]. We observed a reduction of phase contrast right after immerging brain sections in CFM3, which rendered the mouse brain almost completely transparent (Fig. 3A, 3A’). Instead, brain sections mounted in Vectashield or Mowiol remained quite contrasted (Fig. 3B, 3B’, 3C, 3C’). These differences appeared particularly striking in brain areas in which myelinated axon bundles are present, i.e. the corpus callosum, and the striatum (Fig. 3A-3I).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Axial resolution as a function of the mounting medium and focusing depth under brain sections.

(A-I) 100 μm vibratome brain sections were mounted in different mounting media and bright field images were taken in relief-contrast mode at equal light levels with a 1.0 objective (Zeiss Axiozoomer). Overview image, taken with black and white pattern as background (A-C) and zoomed images (A’-C’, D-I) were carried out to visualise contrast changes in the striatum (S), corpus callosum (CC) and cortex (C). Sections were mounted in CFM3 (A, A’, H), ri = 1.518), Vectashield (B, B’, ri = 1.458), Mowiol (C, C’), glycerol (D, ri = 1.390), and CFM3–variants with tuned refractive indices, such as CFM3-1 (E, ri = 1.390), CFM3-2 (F, ri = 1.399), CFM3-3 (G, ri = 1.515) and CFM3-4 (I, ri = 1.522). Note that with fine-tuning of the refractive index to a value of 1.518, contrast of the brain section is very low. Scale bars = 500 μm. (J, J’) 170 nm red fluorescent beads were mounted in CFM3 (ri = 1.518) and Vectashield (VS), as described in Fig. 1B. Imaging was carried out with the 561 nm laser line collecting red emission. X, Z-projections (J) and axial resolution (J’) were assessed directly at the coverslip level (left), or at 60 μm depth, adjacent to brain sections (middle) and under the cortical region of the brain section (right). For beads mounted in CFM3 little variations in axial resolution were observed (J, upper panel, J’ circles, ri = 1.518). However, for beads mounted in Vectashield (J, lower panel, J’ triangles, ri = 1.454) a two-fold loss of resolution at 60 μm was observed. Scale bar = 1 μm. (K, K’) Red fluorescent beads were mounted in Vectashield (VS), as described in Fig. 1B. X, Z-projections (K) and axial resolution (K’) was measured directly at the coverslip level (left) and below the cortical region of brain sections, at 20 μm, 30 μm and 60 μm depth. Note that a considerable loss of axial resolution is already observed at 20 μm depth. Scale bar = 1 μm.

https://doi.org/10.1371/journal.pone.0121096.g003

We aimed to be sure that the transparency of mouse brain sections when mounted into CFM3 is due to refractive index matching of the fixed brain tissue with the medium and not to mounting medium components. We therefore used CFM3-variants (thereafter labelled CFM3-1 to CFM3-4) with refractive indices tuned to 1.390, 1.399, 1.515 and 1.522 by varying the amounts of glycerol and water to the CFM3 solution. Furthermore, we used a glycerol/water solution (ri = 1.390) to mount brain sections. We observed that glycerol (ri = 1.390), CFM3-1 (ri = 1.390) and CFM3-2 (ri = 1.399) resulted in comparable images of the brain sections (Fig. 3D-3F). Striatal fibres and corpus callosum were very contrasted. Fine-tuning of the refractive index of CFM3 showed that with the original CFM3 (ri = 1.518) the contrast is completely lost, whereas CFM3-variants with ri-values differing only slightly by 0,003 ± 0,001 from CFM3 increase notably the contrast (Fig. 3G, 3I). This set of data clearly demonstrates that the transparency observed with CFM3 is due to refractive index match of the fixed brain section with the medium.

Mouse brain vibratome sections of 60 μm thickness were then introduced between slides and coverslips prepared with orange fluorescent beads on both sample directed sides (Fig. 1B). PSF-measurements were subsequently carried out in three positions, at the level of the coverslip, in a depth of 60 μm adjacent to the brain slice, and in a depth of 60 μm beneath the cortical region of the brain (Fig. 3J, 3J’). In the refractive index matched systems with CFM3 mounting medium, the axial resolution remained constant (mean value 635 nm) between the coverslip, and in depth adjacent and beneath the brain section. When using Vectashield as a mounting medium, refractive index mismatch between the oil-glass interface and between the mounting medium-sample interfaces occurs. This mismatch leads to a two-fold reduction in axial resolution at 60 μm sample thickness adjacently and beneath the brain section. These results further demonstrate that the refractive index of fixed brain sections and CFM3 matches perfectly, and that under these conditions spherical aberrations are absent.

When performing immuno-labelling of mouse brain sections, antibody labelling penetrates rarely deeper than 30 μm into the sample. We thus wanted to know if our findings were consistent in sample depths relevant to immuno-labelling. We therefore investigated the axial resolution at sample thicknesses of 20, 30 and 60 μm using Vectashield as a mounting medium. In this medium, we observed a resolution loss of 30% already in a depth of 20 μm (Fig. 3K, 3K’).

4. Improvement of axial resolution in immunofluorescence

We aimed at assessing the ability of our new protocol, involving CFM3 as a mounting medium, to provide higher axial resolution than conventional media like Vectashield in immunofluorescence labelling. We therefore decided to use the olfactory sensory axons of the mouse as a test system. These axons, originating from the olfactory epithelium lying in the nasal cavity, project to the brain at the level of the olfactory bulb. On their way towards their targets in the glomerular layer, these axons rearrange in the most peripheral layer of the bulb, called the olfactory nerve layer [11]. In this layer, olfactory axons are highly numerous and packed at a very high density, making them extremely difficult to resolve by light microscopy. To label olfactory axons, we used transgenic Gγ8-TTA x TetO-M72iresGFP mice [12, 13] expressing GFP in all immature olfactory sensory neurons, through the use of a Gγ8 promoter. To test whether our protocol would be suitable for co-localization studies, immunofluorescence against GFP and peripherin was performed. Peripherin is an intermediate filament known to be expressed in olfactory axons, and previously described to be expressed with a distinct spatiotemporal pattern in these axons [14, 15]. We document in Fig. 4 the difference, in terms of quality of imaging (in depth sensitivity and axial resolution) between Vectashield and CFM3 used as mounting media. All images were taken from the olfactory nerve layer of the olfactory bulb that contains essentially fascicles of olfactory axons en route towards their targets. As seen on Fig. 4A,B, CFM3-mounted sections allow the detection of the labels throughout the thickness of the section for both markers, whereas Vectashield-mounted sections allows the visualization of the labels only in superficial zones of the sections close to the coverslip. Immuno-labelling was carried out in a floating manner and labelling should thus be observed throughout the section even in Vectashield mounted preparations. Correct labelling of the opposite site was verified by turning the section (unpublished data). As shown in orthogonal views (Fig. 4C, 4D), while Vectashield-mounted sections display a poor resolution of the labelling, as observations are made deeper in the section, this is not the case with CFM3-mounted sections, which clearly display a very high resolution of labelling, with excellent and homogeneous sensitivity, throughout the thickness of the sections. Indeed, only CFM3 allows a satisfactory visualization of the thin (200 nm diameter) densely packed axons in both longitudinal and orthogonal views (Fig. 4A, 4C). A maximum intensity projection of 8 sections illustrates the high quality of images obtained from CFM3-mounted sections in which individual axons can be delineated (Fig. 4E). In contrast, Vectashield-mounted sections retrieved satisfactory resolution only in the few first μm of tissue depth (Fig. 4F). Since Peripherin localizes in the olfactory nerve layer with a distinct spatiotemporal pattern, with differential localization along these axons in the outer and inner olfactory nerve layer [15], and since in the transgenic mice GFP is expressed only in a subset of olfactory axons [12, 13], we expected to observe individual axon profiles expressing either marker, and some both. Using CFM3, regardless of the small diameter of olfactory axons, close to the limit of resolution of light microscopy, we can clearly resolve 3 categories of axons: GFP+/peripherin- (green) axons, GFP-/peripherin+ (red) axons, and GFP+/peripherin+ (yellow) axons (Fig. 4E).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Improvement of axial resolution in fixed mouse olfactory bulbs.

Fifty-micron thick vibratome sections of olfactory bulb from adult transgenic mice (Gg8-TTA x Tet0M72iresGFP) were processed for double immunofluorescence detection of GFP (green) and Peripherin (red) and mounted subsequently with CFM3 (A, C, E) or Vectashield (B, D, F). Confocal datasets were deconvolved. (A, B) 3D-reconstruction side view of characteristic z-stacks of brain sections mounted in CFM3 (A) or Vectashield (B) with the coverslip position on the left (arrowhead). (C, D) Orthogonal slices of z-stacks taken from sections mounted in CFM3 (C) or Vectashield (D) to demonstrate axial resolution and signal intensity throughout a single section (x, y-view, x, z-view and y, z-view) and in a maximum projection of 20 images (maximum projection). (E, F) 3D maximum projection of 8 confocal sections through olfactory bulb slices mounted in CFM3 (E) or Vectashield (F), with the insert showing a zoomed view. Note that in brain sections mounted in CFM3, 3 categories of axons can be easily distinguished: GFP+/peripherin- axons (green arrowhead), GFP-/peripherin+ axons (red arrowhead), and GFP+/peripherin+ axons (yellow arrowhead). The images obtained using Vectashield do not allow such a clear distinction between these differently labelled subpopulations of axons. Scale bars = 10 μm.

https://doi.org/10.1371/journal.pone.0121096.g004

Our results clearly show that only CFM3-mounted sections allow excellent in depth imaging (Fig. 4A), and optimal and constant axial resolution throughout the entire thickness of the 50 μm sections (Fig. 4C, 4E). In addition, CFM3 demonstrated its full suitability for co-localization purposes, even at this 200 nm scale, throughout the entire thickness of the sections (Fig. 4E, inset).

5. Improvement of axial resolution when imaging a fluorescent protein

In order to estimate the loss of intensity at different depths of the tissue, sections with homogeneous distribution of fluorescence are required. We thus used cortical sections from VGLUT1^venus knock-in mice, in which the gene coding synaptic marker VGLUT1 is fused with gene coding fluorescent protein Venus, so all presynaptic boutons expressing VGLUT1 are fluorescent [9]. Since the density of VGLUT1-positive synapses within a layer of the neocortex is homogenous [16], we could estimate the signal intensity profile through section depth. We imaged z-stacks of 100 μm sections of fixed brain, mounted in CFM3, Vectashield and Mowiol (Fig. 5). VGLUT1 positive presynaptic boutons are blobs of 300 to 900 nm diameter and are evenly distributed through the depth of the section. Subsequent 3D-reconstructions show that depth penetration is best in CFM3 mounted sections, where well resolved synaptic boutons are observed throughout the section until a depth of 80 μm (Fig. 5A). In Vectashield mounted sections synaptic boutons are observed until a depth of 50 μm, and in Mowiol mounted sections depth penetration does not exceed 30 μm. We then monitored synaptic bouton shape throughout the first 30 μm of the sections (Fig. 5A’) where we observe that in CFM3-mounted sections the shape of the boutons is well preserved along the z-axis and that the boutons are well resolved. In Vectashield and Mowiol mounted sections boutons appear elongated already in a depth of 5–15 μm (Fig. 5A’, 5A”) and they appear to be less well resolved.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Depth penetration and resolution improvement in VGLUT1-Venus mouse brain slices.

100 μm vibratome sections of brain of adult transgenic mice (VGLUT1^Venus) were mounted in CFM3, Vectashield or Mowiol and imaged using the 514 nm laser line. Imaging was carried out using an optimal z-section. 3D-reconstructions of the entire z-stack (A) with the coverslip position at the top (arrowhead) and the first 20 μm (A’) of VGLUT1^Venus-labelled synaptic boutons were carried out to demonstrate depth penetration in the different media. (A”) is a magnified view of the insert in (B) showing synaptic boutons in a depth of 5–15 μm. Note that the boutons mounted in Vectashield appear elongated in comparison to CFM3 mounted sections. Scalebars are 10 and 3 μm.

https://doi.org/10.1371/journal.pone.0121096.g005

Mounting media

We compared seven mounting media with refractive indices below and above ri = 1.500, to test different refractive index mismatch conditions. Refractive indices for non-hardening mounting media were verified at 21°C using a refractometer (Pal-RI, Cat.-number: 3850, ATAGO, Japan) (Table 1).

Mounting solutions below ri = 1.500 were Vectashield (Vector laboratories) with a measured refractive index of ri = 1.454, Fluoromount-G (Southern Biotech Assoc, ri = 1.389), Prolong Gold (Life Technologies), Mowiol (Tris-MWL 4–88, Citifluor) and a glycerol/water-mixture (ri = 1.390). Following manufacturer’s recommendation, curing of the Prolong was carried out at room temperature for 48h so its refractive index increases and stabilizes around ri = 1.440. The refractive index of Tris-MWL 4–88, a mixture of Mowiol, Tris and glycerol/water seems to range between 1.410 and 1.490. The refractive index of many sulfoxides such as methyl phenyl sulfoxide is quite high (ri = 1.5775). The use of these and related compounds has been explored in an attempt to produce a widely applicable mounting medium (R S Davidson UK Patent 2006 GB2419427). Two of those mounting media above ri = 1.500 were glycerol-based CFM1 (Citifluor, ri = 1.515) and CFM3 (Citifluor, ri = 1.518) containing propyl-gallate as an antifadent. Furthermore, 2,2’-Thiodiethanol (Sigma) has been prepared as described in [8] to reach a refractive index of ri = 1.518. CFM3-variants with refractive indices tuned from 1.390–1.522 were produced by varying the amount of water and glycerol.

Preparation of Sub-resolution fluorescent beads and Experimental setup for evaluating the axial resolution in depth

Fluorescent beads (PS-Speck, Lifetechnologies), of a diameter of 170 nm, and loaded with yellow-green and orange fluorescent dye, were diluted in water (1/800 v/v). Drops of the water-diluted sample were put on the surface of the coverslip (Menzel Glaeser #1.5, Agar scientific) or slide and air-dried. Beads were then mounted with a drop of mounting medium. Two experimental setups for evaluating axial resolution in depth were designed. First, fluorescent beads were mounted on coverslips and slides that were separated by one or two layers of adhesive tape (Scotch, 3M) with a nominal thickness of 58 μm and the volume filled with a drop of the respective mounting medium (Fig. 1A). Secondly, orange fluorescent beads were mounted on coverslips and slides, and mouse brain slices of different thickness (15, 30, 60 μm) incubated in CFM3 or Vectashield were inserted in between (Fig. 1B).

Confocal laser scanning microscopy

8-bit Images were collected using a Leica 63x oil immersion objective (HCX Plan APO CS, NA 1.4, working distance 0.14 mm) with an inverted Leica laser-scanning confocal microscope TCS SP5 II (Leica Microsystems, Heidelberg, Germany) equipped with a GaAsP hybrid detection system at a sampling rate of 60 nm in x,y and 200 nm in z-direction. Fluorochromes were detected using laser lines 488 nm and 514 nm. Imaging was performed in a temperature-controlled room at 21°C.

Image acquisition of beads

Bead images were obtained as in [26] with the following modifications. GaAsP gain was set to 16% and the laser power adjusted so that the signal occupied the full dynamic range of the detector, but saturated voxels were carefully avoided. Beads were imaged starting and finishing the stack at least 5μm below and above the bead centre. Beads were visually checked and improper stacks were discarded before determining the microscope PSF. At least eight imaged beads were registered and averaged in order to increase the SNR for deconvolution. Measured PSFs presented in Figs. 2 and 3 show representative beads displayed on the same logarithmic scale so that the low intensity detail characteristics of diffraction patterns was enhanced.

Data analysis

Determination of the lateral and axial resolution was carried out with the ImageJ-based MetroloJ plugin for ImageJ [27, 28]. The plugin generates maximum intensity projections of the stack along the x, y and z-axis resulting in 1D intensity profiles. The (x, y) coordinates of the maximum intensity pixel (mip) are then collected. A x,z cross-section is generated along a line passing through the previously determined 2D mip. From this image, the z coordinate of the mip is defined. The z slice is set to the z mip coordinate. The x profile and y profile are collected along the line passing through the mip. The z profile is collected on the x,z view, along the line passing through the mip. All three profiles are fitted to a Gaussian curve, using ImageJ’s built-in curve fitting function. The full width at half maximum (FWHM) of the gauss curve is calculated for each profile, based on the parameters retrieved from the fitting.

Deconvolution and image treatment

Confocal images of biological data of Figs. 4 and 5 were deconvolved with the Huygens 3.7 software (Scientific Volume Imaging, Hilversum, Netherlands) using a measured PSF (see above) and the Classical Maximum Likelihood Estimation algorithm with 100 iterations. Signal-to-noise ratios (SNR) lower than 20 or up to 50 are recommended by the manufacturer for noisy confocal images or low noise wide-field images. Since we used a confocal microscope with a low noise detector, we tested signal to noise ratios (SNR) of 15, 17, 19, 21 and 25 and visually inspected the results. A signal to noise ratio of 19 gave best results after visual inspection of the raw versus the deconvolved images. In images of Fig. 4 brightness and contrast were adjusted equally for all images after deconvolution and before 3D reconstruction, orthogonal slicing and volume rendering using ImageJ [27].

Preparation of mouse brain sections

Adult Gg8-TTA x TetO-M72iresGFP mice [10, 11] and VGLUT1^venus mice [9] were anesthetized with sodium pentobarbital (50 mg/kg i.p.) and fixed by intracardiac perfusion of 4% paraformaldehyde in 0.12 M phosphate-buffer saline (PBS), pH 7.4. The brain was removed, then postfixed for 3 h in the same fixative and kept in PBS at 4°C until vibratome sectioning. The olfactory bulbs of Gg8-TTA x TetO-M72iresGFP mice were cut in the frontal plane, and processed for immunolabeling. Frontal sections of VGLUT1^venus mice were directly observed since these mice express synaptic marker VGLUT1 fused to the fluorescent protein Venus. Sections of various thicknesses were collected serially in PBS. For their mounting, sections were incubated for 5 minutes in the respective mounting medium and mounted directly on the coverslip. Coverslips were then mounted on slides and exceeding mounting medium was squeezed out by tapping onto the coverslip. Confocal x,z-scans were performed to make sure that the brain sections were positioned directly adjacent to the coverslip.

Animal care was conducted in accordance with standard ethical guidelines (NIH publication no. 85–23, revised 1985 and European Committee Guidelines on the Care and Use of Laboratory Animals 86/609/EEC), and the experiments were approved by the local ethic committee “Comité d’Ethique en Expérimentation Animale Charles Darwin C2EA–05” (Ministère de l’Enseignement Supérieur et de la Recherche, Paris, France).

Immunocytochemistry

Sections were blocked for 1 h in PBS containing 0.25% Triton-X, 0.2% gelatine, 0.1% sodium azide and lysine (0.1 M) before applying primary antibody in PBS containing 0.25% Triton-X100, 0.2% gelatin, 0.1% sodium azide overnight on floating sections. Primary antibodies were chicken anti GFP (1/500, Aveslab) and rabbit anti-Peripherin (1/100, Charles Greer’s gift).

Sections were washed in PBS containing 0.25% Triton-X, and then incubated for 2 h with species-specific secondary antibodies, donkey anti-chicken linked to Alexa-488 (1/400; Invitrogen) and donkey anti-rabbit linked to Cy3 (1/200, Jackson) in PBS containing 0.25% Triton-X100, 0.2% gelatine, 0.1% sodium-azide. The sections were thereafter washed several times in PBS and mounted in CFM3 or Vectashield.