1.

Introduction

The spinal cord is part of the central nervous system (CNS). It is formed by white matter located strictly aligned long nerve fibers, and by gray matter, which contains the neuronal cell bodies. The nerve fibers, named axons, are covered by sheaths of myelin, an insulating lipid-based substance responsible for the saltatory action potential propagation. Several types of glial cells perform essential functions both in white and gray matter and influence the neuronal function: astrocytes regulate homeostasis and interaction between synapses, oligodendrocytes form the myelin sheaths, microglia are the resident immune cells within neuronal tissue.¹

Spinal cord injury (SCI) is the result of an initial physical trauma followed by degenerative processes. Necrotic cell death and hemorrhage take place immediately after induction of the lesion. Secondary events including inflammation, edema, and ischemia lead to additional cell death and axonal degeneration, culminating in the formation of a scar.² The pathological changes in the white matter of spinal cord after injury involve gradual thinning and degradation of the myelin sheath.³ The axonal damage and the demyelination that follow the injury lead to permanent functional impairment.⁴

The clearance of myelin fragments after injury involves activated microglia and macrophages. Microglia rapidly responds to various kinds of CNS injury by transforming into large phagocytes, thereby removing debris. Additionally, invading macrophages migrate from the blood in the injured region after blood brain barrier damage and contribute to myelin clearance.⁵ Myelin has very high lipid content (70% to 75% of its dry weight).⁵ Demyelinated lesions are thus characterized by the presence of lipid-laden macrophages, which develop the distinctive morphology of foam cells by engulfment of large amounts of myelin-derived lipids.⁶ Foam cells originate both from activated microglia and infiltrating macrophages.⁷

It is well known that cytoplasmatic lipid droplets in foam cells mainly consist of lipid esters covered by a phospholipid monolayer. The lipid metabolism within the macrophages is complex and was investigated in several studies, partly leading to contradictory results.^8,9 Macrophages in SCI contain myelin, as demonstrated by analysis of histological staining.^10,11 This was also confirmed by in vitro studies: macrophages that were incubated with myelin transformed into foam cells mainly laden with intracellular droplets of myelin. However, the same study revealed the presence of neutral lipids, too.⁶ It has been widely demonstrated that the composition of lipid droplets in foam cells is very sensitive to the types of lipids that were engulfed by the cell. Macrophages that were cultured in vitro accumulated both saturated and unsaturated fatty acids,¹² as well as triglycerides¹³ and cholesteryl fatty acid esters,¹⁴ depending on the composition of the culture media. Microglia activation in vitro leads to the accumulation of lipid droplets and biosynthesis of triglyceride as well.¹⁵

Standard techniques for studying SCI-induced tissue alterations are based on the use of histological and immunohistochemical methods, which require extended tissue processing and cannot be applied in vivo. Chemical fixatives do not react with most lipids, so lipid bodies are not preserved in conventional morphologic methods and are vacant in histological sections.⁸ The use of noninvasive label-free techniques that can be applied on the fresh, unfixed tissue is thus beneficial for study of lipids in SCI and becomes mandatory for in vivo monitoring. Coherent anti-Stokes Raman scattering (CARS) microscopy is fast, intrinsically confocal, and effective in imaging of myelin and lipid bodies on native tissue.^16,17 Raman spectroscopy is an optical technique that can be applied both on fresh, unfixed tissue and in vivo to retrieve the entire biochemical information.

The overlap of the characteristic Raman signals from the different tissue biomolecules, such as proteins, lipids, nucleic acids, and carbohydrates, encodes a large amount of information potentially useful for diagnostic purposes¹⁸ and the combination of spectroscopy and microscopy enables the localized biochemical characterization of tissue.

Despite their potential, Raman spectroscopy was rarely applied to investigate SCI. Saxena et al.¹⁹ applied point Raman measurements to study demyelination and chondroitin sulfate proteoglycan upregulation after spinal cord hemisection. In a previous study, we applied infrared and Raman microspectroscopy in combination with multiphoton microscopy for the detection of lipid depletion and fibrotic scarring after SCI in a rat model of hemimyelonectomy,²⁰ as well as for assessing the effects of alginate implants on scarring and demyelination.²¹

Here, we focused on inflammation and retrieved the biochemical information about lipid alterations following SCI in rat models by application of Raman microspectroscopy. CARS microscopy was used to differentiate myelin sheaths, myelin debris, and lipid droplets. Endogenous two-photon fluorescence (TPEF) was used to localize inflammatory cells.²²

2.

Methods

3.

Results and Discussion

Five samples of SCI showing different degrees of inflammatory responses were selected according to the hematoxylin and eosin (H&E) staining and are shown in Figs. 1(a)–1(e). H&E is one of the primary stains used in histology and is the gold standard for pathology. Hematoxylin is a blue stain that binds to nucleic acids. Eosin is a pink stain that binds to acidophilic substances such as proteins. The samples in Figs. 1(a), 1(c), and 1(e) display extended inflammation inside and around the lesion core; the sample in Fig. 1(b) has areas of inflammation at the lesion site only, while the sample in Fig. 1(d) has less inflammatory cells that are mainly localized near the large cysts. Cysts are located mainly at the lesion site, which is indicated by the dotted line in the figure. Raman maps of several square millimeters in size covering the lesion site and the surrounding regions were acquired and used to visualize the distribution of lipid species inside the nervous tissue [Figs. 1(f)–1(k)] by plotting the intensity of the band at $1440{\mathrm{cm}}^{-1}$, which is attributed to deformation vibrations of ${\mathrm{CH}}_2$ functional groups.²⁴ The lesion core is generally poorer of lipids when compared to the surrounding tissue and therefore can be readily recognized. This indicates the lack of nervous tissue at the injury site and subsequent demyelination in adjacent regions, in agreement with previous results.²⁰ Areas with high lipid content are visible inside or immediately nearby the lesion in four samples [see arrows in Figs. 1(f)–1(h) and 1(k)]. However, the Raman maps did not enable single axons to be recognized (at least with the lateral resolution that characterizes our system) and therefore did not show whether the high lipid content is due to the presence of functional myelinated axons.

Fig. 1

Cryosections of SCI in rat models. (a–e) H&E staining; (f–k) Raman intensity maps of the band at $1440{\mathrm{cm}}^{-1}$ (${\mathrm{CH}}_2$ deformation vibration); (l–p) CARS image (tuned to excite the ${\mathrm{CH}}_2$ symmetric stretching vibration); (q–u) endogenous TPEF; (v–z) merged CARS (magenta) and TPEF (green) images. Scale bar: 0.5 mm. The dotted lines in (a–e) indicate hemisection positions. Arrows in (f–h) and (k) indicate lipid-rich areas in the injured region. The boxes in (l), (q), and (v) indicate the position of zoom-in images of Fig. 2.

CARS microscopy addressing ${\mathrm{CH}}_2$ symmetric stretching vibration was applied in order to combine the chemical information about lipid distribution with the cellular morphology of the tissue, and therefore to identify axons and cells. The amount of biochemical information is traded-off at the advantage of speed, and label-free imaging of whole sections is performed in minutes. The combination of the morphochemical information supplied by CARS microscopy with the biochemical fingerprint of lipid species provided by Raman spectroscopy was already assessed for investigation of lipid droplets.²⁵

On sections of spinal cord tissue, CARS microscopy provides the overall distribution of lipids, thereby visualizing demyelination at the lesion site in a way similar to Raman mapping based on intensity of a ${\mathrm{CH}}_2$ deformation band [compare Raman maps in Figs. 1(f)–1(k)] and CARS images in Figs. 1(l)–1(p). Due to the submicrometric lateral resolution and the inherent confocality, CARS microscopy allows the visualization of subcellular structures. Figure 2 shows in more detail the region with high lipid content immediately nearby the lesion of sample #1 and illustrates the tissue micromorphology. The H&E staining in Fig. 2(a) shows the microcystic morphology of damaged nervous tissue (on the left side of images), compared to the compact and highly ordered white matter in the contralateral tissue (on the right side of images). The CARS image in Fig. 2(b) provides additional morphochemical information: lipid droplets and fragmented myelin sheaths are localized in the damaged tissue, while the contralateral white matter is characterized by a strictly aligned axonal morphology. The simultaneous acquisition of TPEF [Fig. 2(c)] adds information about the extent of microglia/macrophages activation and therefore localizes the inflammation.^20,22 TPEF shows a clear overlap with the distribution of lipid droplets in the inflammatory region, while no fluorescent cells are present in the normal white matter [merged CARS and TPEF images in Fig. 2(d)]. Therefore, it constitutes an indicator for the presence of foam cells, i.e., activated microglia/macrophages that got laden with cytoplasmatic lipid droplets after engulfment of large quantities of myelin debris. The comparison with the H&E staining confirms the presence of foamy macrophages in the same regions.

Fig. 2

Region with high lipid content in more detail. (a) H&E staining, (b) CARS, (c) TPEF, and (d) merged CARS/TPEF images comparing morphology of normal white matter in preserved spinal cord and foam cells in the injured rat spinal cord (sample #1). Scale bar: $200\mu \mathrm{m}$.

The high lipid content regions indicated by arrows in the Raman maps of Figs. 1(f)–1(k) correspond to inflammatory regions with disrupted myelin sheaths, presence of fluorescent foam cells and devoid of integer myelinated axons, as identified by CARS-TPEF imaging. This clearly shows that the lipid distribution by spectroscopic mapping of the deformation band of ${\mathrm{CH}}_2$ groups is not exhaustive for evaluation of white matter injury. In the CNS, myelin clearance is slow:^26,27 as myelin debris and foam cells remain in the tissue for weeks after the injury, the sole evaluation of the lipid content does not allow distinguishing between preserved regions of white matter with properly myelinated axons and regions of ongoing fragmentation of myelin sheaths and inflammation.

After SCI, myelin debris is taken up by phagocytic cells. Myelin is composed by cholesterol, phospholipids, galactolipids, and plasmalogens in a molar ratio of approximately 2:2:1:1.⁵ Therefore, the metabolites of cholesterol and phospholipids are expected to determine the composition of lipid droplets in macrophages and their spectral markers were analyzed to retrieve the inflammation.

Intracellular lipid droplets are dynamic organelles,²⁸ where cholesterol is stored as cholesteryl esters. The sterol ring vibrations in free cholesterol and cholesteryl esters produce a sharp and characteristic Raman band at $697{\mathrm{cm}}^{-1}$.^29,24 However, a mapping of the sterol vibration band intensity does not enable to visualize the inflammation (data not shown). The esterification of cholesterol is opposed by neutral cholesteryl esters hydrolysis, which is critical for maintenance of cellular cholesterol homeostasis.³⁰ The free cholesterol is either released by macrophages in the intracellular space to be recycled on site, or is removed by migration of macrophages away from the injury, but no degradation of the sterol ring takes place in foam cells, since mammalian cells do not have this ability.³⁰ This explains why monitoring the sterol ring distribution by Raman spectroscopy is not useful for detection of degrading myelin and foam cells.

Lipid esterification is the key process for storage in lipid droplets. Also phospholipids are completely hydrolyzed in foam cells to produce a pool of free fatty acids³¹ that are subsequently esterified.³² The band of carbonyl group stretching vibration at $1743{\mathrm{cm}}^{-1}$ ^33,34 constitutes a marker of cholesteryl and fatty acid esters in Raman spectra. The carbonyl vibration intensity maps of all samples are shown in Fig. 3. The highest degree of lipid esterification is located in the injured regions and colocalizes with the foam cells as detected with CARS/TPEF microscopy (compare with Fig. 1). The average Raman spectrum calculated in an injured tissue region ($12\times 12\text{pixels}$) characterized by lipid esterification is shown in Fig. 3(f), compared with the average spectrum calculated in a region ($12\times 12\text{pixels}$) of preserved white matter. The carbonyl band is visible in the spectrum of the injured region and absent in the one of white matter.

Fig. 3

Carbonyl vibration intensity maps of all samples. (a)–(e) Maps of intensity of the carbonyl stretching vibration at $1743{\mathrm{cm}}^{-1}$, illustrating the distribution of esterified lipids. Scale bar: 0.5 mm. (f) Raman spectra of regions with high lipid esterification (injury) and low lipid esterification (white matter); the spectra were calculated by average in regions $12\times 12\text{pixels}$ large, as indicated by boxes in (c).

For the interpretation of the maps in Fig. 3, it has to be considered that CNS lipids are particularly susceptible to oxidation by virtue of their high content of unsaturated fatty acids. Many studies have demonstrated increased oxidative damage and lipid peroxidation in CNS injuries.³⁵ Advanced lipid peroxidation end-products include a variety of reactive carbonyl compounds,³⁶ which could contribute to the intensity of band at $1743{\mathrm{cm}}^{-1}$. However, the time frame of lipid oxidation in rat SCI is limited to the acute phase and resolves at around 5 days after injury.³⁷ As no significant lipid peroxidation is expected one month postinjury, the maps of Fig. 3 represent the distribution of esterified lipids.

The intensity maps discussed above elucidate the distribution of lipids, sterols, and esterified lipids in the injury region versus normal nervous tissue. However, they do not indicate whether the injured and inflammatory regions are enriched of other types of lipids or fatty acids, which are eventually produced by myelin degradation inside the foam cells or as part of inflammatory pathways. This can be retrieved with multivariate methods, i.e., PCA.

When PCA is performed on the entire dataset (i.e., on the spectral data of all samples merged in one work space), the first to fourth components indicate tissue compounds that are not related to the distribution of inflammation, but represent other tissue constituents (as for example lipids and collagen, data not shown). The fifth component clearly appears to be associated to the presence of activated microglia/macrophages (Fig. 4). As it can be seen by comparing the maps of Figs. 4(a)–4(e) with the TPEF images of Figs. 1(q)–1(u), high pixel intensity areas—representing higher score—identify the inflammatory regions with accumulation of fluorescent macrophages. The colocalization with lipid droplets in the CARS image and with TPEF is shown with more detail in Figs. 4(g)–4(i). The fraction of variance associated to this component is as small as 0.13%, indicating a compound with very low concentration within the tissue. The corresponding loading vector [Fig. 4(f)] is consistent with saturated fatty acids. The positive narrow signals at 1060 and $1126{\mathrm{cm}}^{-1}$ are attributed to $\mathrm{C}─\mathrm{C}$ stretching vibration, and the ones at 1293 and $1434{\mathrm{cm}}^{-1}$ are attributed to ${\mathrm{CH}}_2$ and ${\mathrm{CH}}_3$ deformation vibration.^34,38 The negative spectral contribution at $1658{\mathrm{cm}}^{-1}$—attributed to $\mathrm{C}═\mathrm{C}$ stretching vibration³⁴—confirms that this component is related to a saturated type of lipid.

Fig. 4

PCA: (a–e) maps of score values for all samples and (f) loading vector for the fifth component. (g) Zoom-in of the box in the score map of sample #5 shown in (e); (h–j) CARS, TPEF, and merged images of sample #5 in the same region shown in (g). Scale bar: 0.5 mm. * in (e) indicates the alginate implant.

It was recently reported that the acute activation of microglial cells results in increased content of cellular total saturated fatty acids, especially of palmitic acid (C16:0) and stearic acid (C18:0).³⁹ However, a precise evaluation of the spectral pattern of the loading vector strongly suggests the presence of myristic acid (C14:0), matching with high precision the band positions as reported in Ref. 34. For instance, the four most intense bands of myristic acids are reported at 1433, 1294, 1125, and $1062{\mathrm{cm}}^{-1}$, while the same bands of both palmitic and stearic acids are all shifted 5 to $12{\mathrm{cm}}^{-1}$ toward the higher wavenumber. As no changes of myristic acid content in microglia are reported upon activation,³⁹ an increase in myristic acid might be explained by the increased density of inflammatory cells compared to other tissue regions.

As the lateral resolution of Raman spectroscopy in our experiments is not able to identify the lipid droplets from other cellular compartments, it is unclear whether the saturated fatty acids are a product of myelin degradation that are stored in the lipid droplets. Only the upregulation of saturated fatty acids in activated microglia was addressed in Ref. 39; their subcellular localization and a possible relationship to the presence of lipid droplets were not studied. Furthermore, no reference reporting an accumulation of saturated fatty acids inside the lipid droplet of immune cells was found, except in cases where they are cultured in media enriched with saturated lipids. The lipids of the CNS contain a very high amount of polyunsaturated fatty acids, particularly of arachidonic acid (C20:4), and docosahexaenoic acid (C22:6).⁴⁰ In fact, necrotic regions with presence of macrophages and foam cells in brain affected by glioblastoma were studied by Raman spectroscopy and the spectral signature of unsaturated fatty acids [i.e., oleic acid (C18:1)] was detected.⁴¹ Studies made by Raman microspectroscopy on macrophages in vitro indicated esterified unsaturated lipids as a main component of the lipid droplets.^42,43 Lipid droplets in leucocytes were found to be rich in unsaturated moieties by resonance Raman spectroscopy and provide a reservoir of arachidonic acid, which is involved in the activation pathway of immune cells.⁴⁴ Therefore, the available bibliography does not provide sufficient information to relate the overexpression of saturated fatty acids inside the injured spinal cord with the storage of myelin degradation products inside the lipid droplets of foam cells. However, the comparison with CARS and TPEF images provides further indication.

By inspecting in detail the spatial distribution of signals in the inflammatory regions, it is possible to clearly see that the score maps in fact mimic the pattern of TPEF more than the distribution of lipid droplets as seen by CARS [compare Figs. 4(g) with 4(h) and 4(i)]. This suggests a correlation with the activation state of the immune cells. It was hypothesized that palmitic and stearic acids may play a role in the activation of a pathway that modulates the expression of inflammatory cytokines.³⁹ It is also known that myristic acid has an important role in the cell lipid metabolism, by regulating the production of polyunsaturated fatty acid (e.g., of docosahexaenoic acid).⁴⁵ Moreover, recent findings suggest that myristic acid may have a specific regulatory role because of its ability to modify enzyme activities or protein functions through their N-terminal myristoylation. In particular, myristic acid regulates the activity of mammalian desaturases, which are enzymes involved in the metabolism of lipids.⁴⁶ Therefore, it is more likely that the small amounts of saturated fatty acids identified by PCA in foam cell-rich regions of spinal cord are not just stored in the lipid droplets, but play an active role in the metabolism of activated microglia/macrophages.

4.

Conclusions

SCI triggers degenerative processes characterized by scarring, demyelination, and inflammation. While demyelination and scarring were already addressed elsewhere by Raman spectroscopy,^20,19 inflammation was so far neglected in spectroscopic studies of SCI. Here we showed that: (i) an unspecific mapping of the lipid distribution based on ${\mathrm{CH}}_2$ (and on sterol ring) vibration band intensity does not discern between preserved white matter and inflammatory regions with myelin fragmentation and presence of lipid-laden microglia/macrophages; (ii) inflammatory regions with lipid-laden microglia/macrophages can be localized by addressing the distribution of esterified lipids (i.e., by mapping the intensity of carbonyl Raman band); (iii) these inflammatory regions are enriched with saturated fatty acids, which likely play a role in the lipid metabolism of activated microglia/macrophages. In conclusion, Raman spectroscopy specifically addresses the inflammation taking place in the nervous tissue after SCI, and the products of myelin degradation. Therefore, Raman spectroscopy is confirmed to be a very useful tool for characterization of degenerative events triggered by injury of CNS.