2.1.

Experiment Design

In this study, we evaluate the feasibility of Brillouin spectroscopy to assess the mechanical properties of peripheral nerve ex vivo with the ultimate goal of using this emerging technology to assess viscoelastic properties of peripheral nerves in vivo. The workflow of the experiment is shown in Fig. 1.

Fig. 1

Workflow of Brillouin microscopy imaging for assessing peripheral nerves. (a) Sciatic nerve is harvested from a sacrificed mouse. (b) Nerve sample is chemically fixed and prepared for imaging. (c) Nerve sample is imaged with Brillouin microspectroscopy. Created with Ref. 19.

To determine the effect of nerve injury on the mechanical properties assessed by Brillouin imaging, we designed an experiment to assess both healthy and injured nerves. For the healthy nerve sample, the harvested sciatic nerves were suspended in free space without any applied stress. In the injured nerve model, the harvested sciatic nerve was compressed in a fixture mount, as described below, causing visible mechanical deformation of the tissue.

Orthogonal techniques were utilized to evaluate the responses from both intact and damaged nerves for Brillouin imaging, each designed to prevent sample deterioration and rapid evaporation of submersion media. For the healthy nerve, a partially open mount was designed [Fig. 2(a)], and for the injured nerve, a completely enclosed fixture was utilized [Fig 2(b)].

Fig. 2

(a) A photographic image of a custom-built open sample holder. (b) A photographic image of a lab-grade closed sample holder.

The open holder was designed and assembled in such a way that the sample, when placed inside, was free of any external forces that could alter its density or other properties. To avoid dehydration of the sample, the volume inside the holder was filled with phosphate-buffered saline (PBS) solution.

The closed fixture consisted of a quartz concave slide and a quartz microscope coverslip. This sample holder, albeit sealed from the atmosphere and filled with PBS, demonstrated visible compression of the nerve samples.

Validation of the interrogated region location was achieved through analysis of the signal from the Raman channel of the spectrometer, described in the text prior. The imaging plane field was scanned by means of sample table movement with fixed increments exceeding the spatial resolution of the spectrometer.

2.2.

Sample Preparation

All animal studies were conducted in compliance with the Washington University Institutional Animal Studies Committee and NIH guidelines. C57BL/6 male mice (Charles River) were used in these experiments. Mice were housed in a central animal care facility and provided with food and water ad libitum. Animals were kept in standard cages at room temperature and a 12:12 day/night light cycle. Each mouse was anesthetized with 2% (v/v) inhalational isoflurane and then sacrificed by cervical dislocation. The sciatic nerves on both sides were harvested through a mid-thigh incision.²⁰ A small vertical incision was made along the thigh, skin was retracted laterally, and the muscles of the posterior thigh were split with scissors to expose the sciatic nerve. The nerve was then gently lifted with forceps and removed by carefully cutting the connecting tissues to avoid deformation.

A murine model is commonly used to study neuropathic pain^21,22 and nerve regeneration.²³ This model is well studied and yields highly replicable results. The short sciatic nerve length is detrimental in some mechanical evaluation studies, whereas microscopic evaluation of small sections of sciatic nerve is not detrimental, but rather advantageous. In addition to this, the murine model is very common in chemotherapy response evaluation,²⁴ showing that such treatment can induce detrimental systemic response and peripheral neuropathy.²⁵

One of the nerve samples was subjected to mechanical physical damage ex vitro, which decreased the circumference of the sample by $\sim 20\%$ through the removal of the surrounding connective tissue. This sample is referred to as the “damaged” nerve sample.

Between measurements, nerve samples were stored in the fridge at 4°C in separate vials submersed in the PBS solution.

2.3.

Brillouin Imaging Spectrometer

This study was performed using a custom-built combined confocal Brillouin–Raman spectrometer (Fig. 3). The detailed description of the setup is provided elsewhere.^26–28 The confocal microspectrometer was optimized to acquire the Brillouin signal from highly scattering samples, which potentially makes it possible to image exposed peripheral nerves in the depth of the tissue (10 to $70\mu \mathrm{m}$).

Fig. 3

Combined confocal Brillouin–Raman microspectrometer. HWP, half-wave plate; QWP, quarter-wave plate; BS, polarizing beamsplitter cube; BD, beam dump; O, microscope objective lens; S, sample; L, plano-convex spherical optics; PH, precision pinhole; D, dichroic mirror; VC, iodine vapor cell; CL, cylindrical lens; VIPA, virtually imaged phase array etalon.

Inelastic scattering was induced using a tunable single-mode 532 nm laser [1064 nm 20 kHz linewidth tunable single mode laser diode (NKT Photonics Koheras Adjustik), seeding an Ytterbium doped fiber amplifier (NKT Photonics Koheras Boostik HPA), and generating 100-mW of 532 nm in a MgO:PPLN crystal (Covesion Ltd.)] shown in a red dashed line box in Fig. 3. Both the excitation and back-scattered signal collection were achieved using CFI60 Nikon microscope objective lenses with a varying numerical aperture (NA) and magnification values. The laser power delivered to the sample was finely controlled by a combination of HWP and BS and did not exceed the 6-mW limit before the objective lens, which has about 70% transmission at the wavelength of 532 nm. The collected and collimated signal was diverted using polarization plane rotation via a double pass through a QWP to the spatial filter composed of a confocal pinhole assembly (blue dashed line box in Fig. 3). The pinhole size was selected to be less than a single Airy unit.

Residual elastically scattered photons were removed from spectra using intrinsic molecular absorption in a heated iodine cell and was then imaged into a custom-built high-resolution spectrometer, which is schematically shown in a green dashed line box (Brillouin Channel) in Fig. 3. A virtually imaged phase array (VIPA) etalon (FSR 29.98 GHz @ 532 nm, OP-6721-3371-2, Light Machinery, Inc.) was used as a dispersive element in this spectrometer. Brillouin spectra were acquired using a water cooled electron-multiplying charge coupled device (EMCCD) (Andor Newport 970, Oxford Instruments), and a longer wavelength signal was steered into a commercial Czerny–Turner spectrometer (Andor Shamrock 303i, Oxford Instruments) fitted with a charge coupled device (CCD) camera (Andor Newport 940P, Oxford Instruments).

The Brillouin spectrometer was calibrated with 10,000 data points. Brillouin standard spectra were acquired from neat acetone sample stored in a sealed quartz cuvette at 37°C heated up with a microscope slide warmer. The mean value for the Brillouin shift was 5.712 GHz with a standard deviation of 0.018 GHz.

The Raman channel was utilized to determine the region of the sample that we are evaluating. The system can acquire optical spectra in the range of 530 to 660 nm. Either of the three metrics are used to determine the sample location: residue backscattering signal, fluorescence signal, or Raman signal. The data acquired from the so-called Raman channel form surface plots, with the height of the surface encoding the spectral peak’s position. We validate nerve location in the field of view of the microscope using the residual backscattering signal.

3.1.

Non-Compressed Nerve Tissue Sample

Two nerve samples were scanned in the open sample holder to minimize the external disturbance. Data acquired with both $\mathrm{NA}=1.0$ (water immersion) and $\mathrm{NA}=0.75$ (air) objectives are shown in Fig. 4 and show that the Brillouin shift within the sample’s area is lower that than the one from the surrounding media. These data suggest that the nerve region is softer than the surrounding PBS solution. At the same time, the Brillouin linewidth [full width at half maximum (FWHM)] for nerves displayed lower values than the surrounding environment for the damaged sample and higher values for the intact sample, suggesting a lower viscosity of the interrogated region with respect to the immersion media.

Fig. 4

Non-compressed nerves imaged with $\mathrm{NA}=1.0$ objective (top row), pixel size is $220\mu \mathrm{m}$, image size is $11\mathrm{mm}\times 11\mathrm{mm}$; and $\mathrm{NA}=0.75$ objective (bottom row), pixel size is $150\mu \mathrm{m}$, image size is $5.25\mathrm{mm}\times 5.25\mathrm{mm}$. (a), (d) Signal from the Raman channel; (b), (e) Brillouin shift (elasticity) image. (c), (f) Brillouin line FWHM (viscosity) image. White bar represents 2 mm. * represents intact sample; ** represents damaged sample.

Damaged nerve data are presented in Figs. 4(a)–4(c), and the intact nerve is shown in Figs. 4(d)–4(f). According to the Brillouin shift values, both samples were softer than the surrounding environment with the average shift values being $6.48\pm 0.35\mathrm{GHz}$ for damaged and $6.68\pm 0.44\mathrm{GHz}$ for intact nerve samples. Brillouin line FWHM values, however, suggest that the viscosity values of the two samples are significantly different. FWHM value from the intact nerve is $0.983\pm 0.0816\mathrm{GHz}$ and from the damaged nerve is $3.14\pm 0.34\mathrm{GHz}$. This can be potentially explained by the lack of external tissues surrounding the neuronal fibers.

We should also note that the Brillouin shift value from the surrounding media is somewhat different for the two measurements. The immersion media, shown in Fig. 4(b), showed $7.48\pm 0.01\mathrm{GHz}$, and in Fig. 4(e), the shift value is $7.22\pm 0.02\mathrm{GHz}$. This is due to a slight variation of the water temperature.²⁶

As shown in Fig. 4, the non-compressed nerve displays a lower Brillouin shift value than the external environment (PBS), which can be explained by the nerve’s low density.

3.2.

Compressed Nerve Tissue Sample

To further elucidate the mechanical properties of nerves, the nerve tissues were imaged in the closed holder with a compressive force applied to simulate the external stress applied to a nerve. The intact nerve on the top (Fig. 5) had no myelin removed, whereas the one on the bottom underwent mechanical damage, as described in the previous section.

Fig. 5

Compressed nerves imaged with $\mathrm{NA}=1.0$ objective (top row), pixel size is $250\mu \mathrm{m}$, image size is $10\mathrm{mm}\times 10\mathrm{mm}$; and $\mathrm{NA}=0.75$ objective (bottom row) pixel size is $200\mu \mathrm{m}$, image size is $8\mathrm{mm}\times 8\mathrm{mm}$. (a), (d) Signal from Raman channel, (b), (e) Brillouin shift (elasticity) image, (c), (f) Brillouin line FWHM (viscosity) image. * represents intact sample, ** represents damaged sample. White bar represents 2 mm.

The images were acquired with $\mathrm{NA}=1.0$ and $\mathrm{NA}=0.75$ objectives and are presented in Fig. 5. In contrast with previous (non-stressed) measurements, the nerve structure displayed Brillouin shift values significantly higher than surrounding media, suggesting that stressed nerves are stiffer than immersed media. On the other hand, the Brillouin line FWHM images suggest that nerves are now more viscous than the PBS solution.

A statistical analysis of imaging data presented in Figs. 5(a)–5(c) shows that the intact nerve displays a higher Brillouin scattering shift value ($7.78\pm 0.08\mathrm{GHz}$ with 1.0 NA and $7.95\pm 0.05\mathrm{GHz}$ with 0.75 NA) than the immersion media and damaged nerve ($7.67\pm 0.05\mathrm{GHz}$ with 1.0 NA and $7.78\pm 0.05\mathrm{GHz}$ with 0.75 NA). Brillouin line FWHM values from the damaged sample are higher ($3.58\pm 0.19\mathrm{GHz}$ with 1.0 NA and $2.85\pm 0.31\mathrm{GHz}$ with 0.75 NA) than the intact sample ($2.89\pm 0.44\mathrm{GHz}$ with 1.0 NA and $2.38\pm 0.35\mathrm{GHz}$ with 0.75 NA). This can be explained, in part, by connective tissue increasing under external load, causing an increase in stiffness and a decrease in viscosity. One possible explanation of the observed changes of the Brillouin spectra from mechanically stressed tissues is related to a change in the sample’s density. It is also plausible that mechanical compression alters optical properties of the samples being imaged. Further investigation is required to interpret this observation.

Similar to the experimental results with the non-compressed sample, we observed a slight variation in the Brillouin shift from the media caused by the sample warming from 4°C to 20°C. We believe that this error is insignificant because the relation between the sample shift and FWHM values remains the same regardless of external temperature variations.

It is important to point out that tissue fixation is known to alter viscoelastic properties of biological samples²⁹ and neuronal tissues in particular.³⁰ However, comparing samples under the same environmental and imaging conditions yields insightful information about tissue’s viscoelastic properties.