1.

Introduction

The brain contains a high amount of lipids that are subdivided into a large variety of classes (such as phospholipids, sphingolipids, glycosphingolipids, and cholesterol), and play an important role in neuronal functions.¹ For example, cholesterol, phospholipids, and galactolipids are the main constituents of myelin, the insulating substance that covers the axons and enables saltatory conduction.² Several pathological states (such as tumors,^3–6 sclerosis,^2,7 Alzheimer,⁷ and injury⁸) are characterized by an alteration in quantity and/or quality of lipids.

Coherent anti-Stokes Raman scattering (CARS) microscopy is an ideal technique to visualize brain structures by addressing the distribution of lipids. CARS is a nonlinear variant of Raman spectroscopy based on the resonant excitation of a molecular vibration using ultrashort pulsed lasers.⁹ CARS imaging is performed by tuning the system on the Raman band at $2850{\mathrm{cm}}^{-1}$, which is produced by the symmetric stretching vibration of $\mathrm{C}─\mathrm{H}$ bonds in ${\mathrm{CH}}_2$ functional groups. Therefore, CARS is an emerging technique to probe the contributions of lipids,^10–13 enabling label-free imaging of brain,^14–16 and nervous structures^17,18 for the physiological or pathological studies. CARS microscopy was largely applied in vivo and to native nervous tissue,^17,19–21 but it is also emerging as a method for label-free histopathology, applied on histology preparations (e.g., tissue sections ^22–24 and smears²⁵).

Fixation is a standard procedure in histopathology preparation, necessary to preserve the tissue from autolysis and putrefaction.²⁶ During fixation, the tissue components are stabilized and made insoluble, the proteolytic enzymes are inactivated and bacteria are killed. The aim of fixation is to preserve the morphological and biochemical characteristics of the tissue, so that it can be further processed with the histological or immunohistochemical staining. The effects of fixation on histology and immunohistochemistry are well studied,²⁷ but are of limited use for the CARS imaging. As CARS directly probes the chemical composition of the tissue, the effects of all preparation procedures that might remove endogenous compounds or add exogenous compounds have to be verified.

The native tissue composition can be retrieved from unfixed cryosections, but they have drawbacks. They can be stored for only a limited period of time and the natural decay of tissue starts immediately after thawing. In other cases, the tissue from the pathology is already fixed in bulk as part of the standard preparation procedures. Therefore, it is important to know how the fixation procedures may alter the CARS signal intensity, modify the CARS image contrast of tissue structures with different compositions, and finally, choose the method that less affects the CARS image.

Although some research was performed to study the effects of tissue preparation on the vibrational spectroscopy of brain,^28,29 as well as of other organs^30–34 and cells,³⁵ no specific studies were performed to our knowledge about the consequences of fixation on the CARS images.

Here, we analyze the effects of two different methods of fixation for the brain tissue, namely with formalin and methanol–acetone, on the contrast and intensity of the CARS images, and interpret the results with Raman microspectroscopy. An incubation in aqueous solution of formaldehyde (known as formalin) is a common chemical method for fixation of bulk samples as well as of cryosections. It acts by creating covalent bonds between the proteins in the tissue³⁶ and fixation is achieved through cross-linking of amines, amides, aromatic rings, hydroxyls, guanidine, sulfhydryl groups, and reactive hydrogen atoms.³⁷ Another fixation method often used for thin brain cryosections is performed by incubation in a solution $1:1$ of methanol and acetone at $-18°\mathrm{C}$. It reduces the solubility of proteins, which precipitate and aggregate.²⁶ Monohydric alcohols, such as methanol, denature the proteins by breaking the protein tertiary structure and stabilizing the secondary structures.³⁸

2.

Methods

3.

Results and Discussion

In order to verify the effect of tissue fixation on the CARS image contrast and quality, consecutive cryosections of mouse brain (i) without any fixation, (ii) fixed with methanol–acetone, and (iii) fixed in 4% PFA were imaged with the CARS system tuned on the Raman band at $2850{\mathrm{cm}}^{-1}$. Image acquisition was performed using the same parameters (laser power, beam focalization, and photomultiplier gain) so that the CARS images could be compared. The CARS images acquired in the brainstem of the unfixed sample are shown in Fig. 1 (left, as indicated) at two different levels of magnification, and matched the structure revealed by H&E staining, which was performed on the very same samples after CARS imaging. The overall structure of the tissue [Fig. 1(a), left] was recognizable: white matter, which is rich in myelin and has high lipid content, appeared brighter in the CARS image, while gray matter, poorer in lipids, appeared darker. After PFA fixation (Fig. 1, middle), the difference in the CARS signal intensity generated by white and gray matters was still clearly visible and the image brightness was not substantially modified. Nevertheless, the magnified image in panel B shows an alteration of tissue micromorphology. The effect of fixation with the methanol–acetone was dramatic (Fig. 1, right). The CARS intensity was strongly decreased within both the white and gray matters and the micromorphology appeared to be also strongly affected.

Fig. 1

Effects of fixation on the CARS images of mouse brain. (a) CARS images and H&E staining pictures of 20-μm thick consecutive tissue cryosections, unfixed, formalin fixed (4% PFA), and methanol–acetone fixed (Met–Ace). The H&E staining was performed on the very same sections and provides then a precise reference for the tissue morphology to be compared with the CARS images. (b) Zoom-in of the boxes in (a).

The analysis of the CARS signal intensity of white and gray matters confirmed that the difference existing between lipid-rich and lipid-poor structures was affected by fixation. In the unfixed samples, the ratio between CARS intensity in gray and white matters was on the average of 0.51, in the PFA fixed sample 0.49 and in the methanol–acetone fixed 0.79. The effects of fixation on contrast and intensity of the CARS image suggested a profound interaction of methanol–acetone fixation with the tissue lipids. This is consistent with the property of acetone to dissolve lipids.

In order to gain a better insight in the biochemical effects of fixations, Raman spectroscopy was used to directly probe the tissue composition. The spectra of maps composed by $31\times 21$ points covering an area of $(600\times 400){\mu \text{m}}^2$ were acquired on unfixed, formalin, and methanol–acetone fixed cryosections of mouse brain tissue. The Raman measurements were performed in comparable regions in the lateral lobe of the brain (Fig. 2, measurement position A indicated on the H&E staining shown on the right). An area was chosen, where the tissue appeared homogeneous, in order to avoid differences due to the changes in tissue structure and composition throughout the sections investigated. The averages and the standard deviations of the 651 spectra acquired were calculated for each sample and then normalized to the phenylalanine band at $1003{\mathrm{cm}}^{-1}$ (stretching of the $\mathrm{C}─\mathrm{C}$ bond⁴²); the spectra are shown in Fig. 2(a). The normalization on the intensity of the phenylalanine band was chosen to quantitatively compare the spectra of the different samples because fixation was not reported to alter this band, in contrast to the one of amide. Amide vibrations are influenced by the reaction of methylene glycol, e.g., hydrated formalin, that cross-links the nitrogen atoms of lysine with the nitrogen atom of peptide linkages.³⁰

Fig. 2

Raman spectroscopy on unfixed, formalin (4% PFA), and methanol–acetone (Met–Ace) fixed consecutive brain cryosections. (a) Average Raman spectra and standard deviations, displayed in stacked view; the normalization of average spectra was done on the intensity of phenylanine band (indicated by the star). The underlined bands are related to the lipid vibrations and numbered in accordance with Table 1. The inset shows in details the spectra region between 600 and $650{\mathrm{cm}}^{-1}$. (b) Difference spectra obtained by subtracting the average spectrum of the unfixed sample from the average spectra of the PFA and methanol–acetone fixed samples. The intensity scale was changed with respect (a) in order to better visualize small variations. (c) Raman maps of unfixed and fixed consecutive cryosections are already shown in Fig. 1. The maps were obtained by plotting the integral intensity of the ${\nu }_s({\mathrm{CH}}_2)$ band in the range 2835 to $2865{\mathrm{cm}}^{-1}$. All maps are displayed in the same gray scale.

The spectra of fixed samples differed from the one acquired on unfixed tissue. The band intensity changes are resumed in Table 1 and the affected bands are underlined in Figs. 2(a) and 2(b). The changes were found to be stronger in the case of methanol–acetone fixation, which produced a marked decrease of all lipid-related bands. For instance, when the ${\nu }_s({\mathrm{CH}}_2)$ band at $2850{\mathrm{cm}}^{-1}$ is considered, formalin fixation did not appear to alter its intensity, while a strong reduction was caused by the methanol–acetone fixation; this is in agreement with the results of the CARS experiments.

Table 1

Variations of the intensity for lipid-related Raman bands of fixed samples compared to the unfixed one. ↓↓: decrease to <50%; ↓: decrease from 90% to 50%; –: unchanged (variation in the range ±10%), ↑: increase up to 50%. Band assignment following Refs. 42, 43, and 44.

The Raman bands exclusively attributed to proteins and nucleic acids were generally found to be not affected by fixation, namely: the bands at 621 and $640{\mathrm{cm}}^{-1}$ [ring deformation of phenylalanine and tyrosine, respectively⁴²—see inset in Fig. 2(a)], the band at $754{\mathrm{cm}}^{-1}$ (symmetric breathing of tryptophan⁴²), the region around $1340{\mathrm{cm}}^{-1}$ where the vibrations of proteins and nucleic acids are reported,⁴² and the band at $1652{\mathrm{cm}}^{-1}$ attributed to amide II vibration.⁴² Contamination with formalin was not evident in the spectra. Formalin contamination was reported³¹ to generate bands at 907, 1041, 1254, and $1492{\mathrm{cm}}^{-1}$. In our study, only a weak band could be detected arising at $1254{\mathrm{cm}}^{-1}$.

To further verify the effects of fixation on lipid distribution inside the tissue, Raman mapping was performed on the same regions already imaged by CARS (Fig. 2, measurement position C). In Fig. 2(c), the Raman maps of unfixed and fixed samples obtained by plotting the ${\nu }_s({\mathrm{CH}}_2)$ band intensity in the same gray scale are reported. Although the maps on unfixed and formalin fixed samples were in a similar range of intensity, the one acquired on the methanol–acetone fixed sample was characterized by strongly reduced intensity. These measurements confirmed that fixation with methanol and acetone strongly decreased the content of tissue lipids. When compared to the CARS images in Fig. 1(a), the overall intensities agree quite well, even though they cannot be quantitatively compared because the intensity of the Raman bands has a linear dependence to the number of scattering centers, while CARS has a quadratic dependence and a contribution from the nonresonant background.⁹

The ability of CARS microscopy to assess the lipid content of tissue is relevant for tumor pathology and any fixation-induced alteration of tissue lipid content is expected to have a negative impact on tumor detection, analysis, and diagnosis. The effects of fixation on tumorous tissue are shown in Fig. 3 using a consecutive cryosection of mouse brain with an experimental melanoma. In the CARS image of an unfixed section [Fig. 3(b)], the tumor appeared as a compact mass surrounded by white matter (compare the H&amp;E staining in panel A); the high cellularity of the tumor could be discerned and the cell nuclei recognized as darker round shapes [arrowheads in Fig. 3(e)], in agreement with already reported results.⁴⁵ After formalin fixation [Fig. 3(c)], the tumor was still clearly distinguished from the surrounding tissue and cell nuclei remained visible, although less defined [arrowheads in Fig. 3(f)]. As expected, methanol–acetone fixation profoundly altered the intensity of the CARS signal in the overall sample [Fig. 3(d)] and was again consistent with a strong removal of lipids. The tumor and the surrounding white matter were characterized by a similar CARS signal intensity and only a few cell nuclei were barely visible [arrowhead in Fig. 3(g)]. The comparison of the resonant and nonresonant images of the unfixed [Figs. 3(h) versus 3(i)] and PFA fixed samples [Figs. 3(j) versus 3(k)] demonstrated that there is an appreciable amount of resonant CARS signal produced by the tissue constituents. The comparison of the resonant and nonresonant images of the methanol–acetone fixed sample [Figs. 3(l) versus 3(m)] shows that the resonant signal (now expected to be generated by $\mathrm{C}─\mathrm{H}$ bonds in proteins) was in fact strongly reduced.

Fig. 3

Effects of fixation on the CARS images of mouse brain tumor (melanoma metastasis). (a) H&E staining of a 10-μm thick cryosection, the box indicates the area imaged by CARS. (b–d) CARS images of consecutive cryosections unfixed, fixed with 4% PFA, and with methanol–acetone, respectively. The H&E staining was performed on the very same section shown in (b). (e–g) Details of the tumor showing cell nuclei (arrowheads). (h, j, l) Resonant images of the tumor border. (i, k, m) Nonresonant images of the regions shown in (h), (j), and (l).

Moreover, CARS images of tumor samples clearly showed an alteration of tissue morphology after fixation [e.g., small tissue cracks in Figs. 3(e) versus 3(f) and 3(g)].

The presented results provided evidence that fixation with methanol and acetone decreases the tissue lipid content and therefore, strongly modifies the CARS signals acquired on cryosections of normal and tumorous mouse tissues, while formalin fixation might constitute an alternative to unfixed samples. In the study of nervous tissue, CARS imaging is widely used for imaging of lipid-rich axonal structures and may help to localize lipid-loaded cells by the detection of intracellular lipid droplets (LD). In order to further assess the applicability of formalin fixation, the potential fixation-induced changes on those microstructures were analyzed.

CARS imaging of LDs was performed on the cryosections of human glioblastoma. In the normal brain, LDs cannot be likely observed, but in high grade astrocytic tumors extracellular LDs are a consequence of necrosis, while cytosolic LDs are expressed by the tumor cells.⁴⁶ As shown in Fig. 4, only minor differences could be detected between the images acquired before (panel A) and after fixation in PFA (panel B), which are likely due to the small differences in the focal plane position. To enable a better visualization of LDs, the section was imaged with the rehydration in PBS.

Fig. 4

Effects of formalin fixation on the CARS images of human glioblastoma cryosection. (a) Unfixed, rehydrated cryosection. (b) Formalin fixed, rehydrated cryosection.

Finally, the effects of formalin fixation were investigated using the fresh bulk samples of mouse brain cerebellum. The axonal structures of white matter tracts were imaged on the fresh, native tissue [Fig. 5(a)], and after formalin fixation of the same tissue specimen [Fig. 5(b)]. CARS imaging of comparable regions did not underline alterations in the myelin sheath appearance, demonstrating that formalin fixation preserves axonal microstructure and is a viable solution also when applied to bulk tissue samples.

Fig. 5

Effects of formalin fixation on the CARs images of white matter structures in bulk mouse cerebellum. (a) Fresh, unfixed bulk sample. (b) Bulk sample after formalin fixation.

4.

Conclusions

We analyzed the effects of two common methods of fixation (formalin and methanol–acetone) on brain tissue, showing how the procedure modifies the biochemistry of the tissue and therefore the CARS image intensity and contrast. Fixation with methanol and acetone strongly decreased the tissue lipid content and therefore it was considered to be incompatible with the CARS imaging. On the other hand, the results show that fixation with formalin constitutes a viable alternative that does not alter substantially the CARS image contrast generated by the different tissue structures.