2.1.

Animals

Twenty male and female Sprague-Dawley rats (270 to 290 g) were obtained from the Jung Ang Laboratory Animal Company (Seoul, Korea). The animals were housed in a temperature-controlled environment at 23 °C with 60% relative humidity. All animals were exposed to a 12-h light–dark cycle and had ad-libitum access to food and water. Procedures involving the animals and their care conformed to the institutional guidelines of Seoul National University, which were in full compliance with current laws and policies.²¹

2.2.

Fluorescent Tracer Suspension

Cobalt-ferrite magnetic nanoparticles embedded in amorphous silica shells were synthesized. The shells contained a fluorescent dye, rhodamine B isothiocyanate (orange, γ_max = 555 nm), in the middle and a biocompatible polyethylene glycol on the outside. The average core size of the fluorescent magnetic nanoparticles was 9 nm, while the average size of the enclosing shell was 50 nm. The concentration of particles, which were suspended in a sterile saline solution at pH 7.4, was 10 mg/ml.²² The fluorescent tracer suspension was composed of fluorescent magnetic nanoparticles and ≥99.9% ethyl alcohol (Sigma, St. Louis, Missouri) in a ratio of 1:1.

2.3.

Stereotaxic Injection and Sample Preparation

A stereotaxic system was used to inject the fluorescent nanoparticles (FNPs) precisely into the lateral ventricle. The rats were anesthetized with an intramuscular injection of urethane (1.5 g/kg), and all surgical procedures were performed under general anesthesia. The rat's scalp was removed and a hole (diameter = 1 mm) was drilled in the skull to access the brain. The hole had the stereotaxic coordinates of 2.0 mm after the bregma, 2.0 mm to the right of the middle line, and 4 mm in depth.

FNPs (0.3 ml) were injected into the lateral ventricle by using a stereotaxic system with an injection needle (17-gauge syringe, 1.47 mm in diameter). Two hours later, the rat was decapitated, and the whole spine was isolated. The whole spine was immediately put into 0.4% neutral buffered formalin and kept in it for 24 h. After 24 h, the bone was removed from the spine, and the spinal cord was carefully exposed. After the entire spinal cord had been exposed, we measured the signal from the FNPs.

2.4.

Imaging Device and Image Processing

A fluorescent stereomicroscope (MVX 10; Olympus Co., Tokyo, Japan) equipped with a sensitive black/white charged-coupled device camera and a long-pass filter (600 nm; CVI, Seoul, Korea) was used to detect the fluorescence from the nanoparticles.²³ A sonicator (Digital Sonifier 450, Branson Co., Danbury, Connecticut) was used for breaking aggregates of nanoparticles in the suspension. Gray-scale images were analyzed using IMAGEJ software (Wayne Rasband, National Institute of Health, v1.43u). Gnuplot 4.4 was used to fit the functions.

2.5.

Calibration Material

Reference suspensions for calibration were made to quantify the line densities of the nanoparticles in primo vessels. In order to avoid the formation of aggregates of nanoparticles, we applied sonication for about a minute. The densities of the reference suspensions were 50, 20, 10, 1, 0.1, and 0 μg/ml. Three (3.0) or 2.7 ml of each reference suspension were poured into a small Petri dish whose diameter was 35 mm [Fig. 1a]. In order to reduce the background fluorescence, we placed the Petri dish on a piece of black leather. The same fluorescence filter was used to view the nanoparticles. Two fluorescent images of the reference suspension were taken for each density. The exposure times were from 250 ms to 12 s and were adjusted to obtain appropriate fluorescence intensities.

Fig. 1

(a) Inhomogeneous illumination of the nanoparticle suspension under a microscope. (b) Image intensity or brightness versus exposure time for the reference solution. The intensity of a sample is proportional to the exposure time. (c) Reference solution density versus brightness (intensity). The reference solution density is proportional to the brightness of an image. The c value was calculated by using a Lorentzian model fitting. (d) The Lorentzian model fitting. (e) The Gaussian model fitting. Both models fit the data within 2% accuracy with respect to the root-mean-square residual.

2.6.

Optical Contrast Measurement

The Weber contrast²⁴ is defined as (I _m − I _b)/I _b. Here, I _m means the intensity of a measured point, and I _b is the background intensity around the measured point. The larger the contrast, the more clearly we can see the image.

To measure the optical contrast of a fluorescent line, we measured the highest intensity over a specific part of the line and compared that value to the background. We analyzed the stereoscopic fluorescence images with bright fluorescent lines. We used IMAGEJ 1.44c software with the region of interest (ROI) manager function to analyze the images. First, we chose places to be analyzed. The brightness of a line varied over its length. Since several aggregates of FNPs made conspicuously bright spots, we avoided such spots. Second, we measured the background intensity. We chose five imaginary ovals near the fluorescent line as ROIs, and their mean intensities, I _b, were measured. Third, we measured the line intensity. Five imaginary rectangles covering the fluorescent line in the chosen place were drawn as ROIs. We chose the maximum value of the intensities in the ROIs as I _m. Fourth, we calculated the contrast for each ROI, the means, and the standard errors of the means.

2.7.

Multispectral Imaging for Removal of Autofluorescence

To ensure that the sample signal was from the nanoparticles and not from autofluorescence, we used multispectral imaging techniques to remove the autofluorescence. Usually, the frequency range of autofluorescence is broad enough to encompass the whole visible light range. In contrast, a nanoparticle emits more red fluorescence than green and blue fluorescence. Thus, if red, green, and blue filters are used to take images, the light signals from autofluorescence and from the nanoparticles can be differentiated. Filters were used with a black-and-white CCD and all images were in gray scale (black and white correspond to 0 and 255, respectively.). We followed a multistep process to remove the background noise due to autofluorescence. First, we evaluated the mean intensity of the background of an image for red, green, and blue filters. The intensity under the blue filter was negligible. Second, we calculated the ratio of the mean intensity of an image taken under the red filter to that taken under a green filter by using the ratio b = (mean intensity of background under the red filter)/(mean intensity of background under the green filter). Third, we multiplied the intensity value at each pixel under a green filter by b which was the effective red intensity due to autofluorescence. Fourth, we subtracted this effective red intensity of the background from the intensity of the image under a red filter at each pixel because the latter had contributions from both the nanoparticles and autofluorescence. In the resulting image [Fig. 3c], most of the background signal was removed, and we obtained a high contrast image of a primo vessel.

Fig. 3

Primo vessel in the spinal cord. (a) Illustration of the location of a vein just beside a primo vessel (the primo vessel is not shown in the illustration) in the subarachnoid space of the spinal cord. The illustration was modified from a reference book (Ref. 25). (b) Unprocessed picture of a primo vessel (arrow) accompanying a blood vessel (broken arrow) in the subarachnoid space of the dorsal surface of the spinal cord. The line density of nanoparticles in this sample was quantified. (c) The nanoparticle image of (b) after image processing to show only the nanoparticle signal by subtracting the background noise. The average contrast was 1.4.

4.1.

Calibration with Reference Suspensions of FNPs

Nanoparticles can be quantified by labeling with fluorescent, magnetic, or radioactive materials,²⁶ and multispectral imaging can be used to differentiate a target from the background.²⁷ Quantification of a photosensitizer, such as mesotetrahydroxyphenylchlorin, can be improved by considering the fluorescence of the background.²⁸ The concentration of fluorophores in vivo at the site of photodynamic therapy has been estimated through a comparison with a calibrated measurement or a standard solution.^{29, 30}

In order to compare the density of FNPs in the primo vessel with that of the reference suspension, we took the effects of background, exposure time, and light sources into account. First, the background effect was considered by subtracting the intensity of a control suspension with 0 μg/ml of nanoparticles. The source of the background was mainly the fluorescence of the leather. Second, the effect of exposure time was considered. The exposure time was changed from 5 to 200 ms, and the mean intensity was measured. The intensity was shown to be proportional to the exposure time (t) [Fig. 1b]. If the reference suspension's exposure time was t ₁ and if the control's exposure time was t ₀, then the intensity of the control image was multiplied by t ₁/t ₀, and the adjusted control image was subtracted from the reference suspension's image. Third, the inhomogeneity of the illumination was taken into account because the center area of the reference suspension was more brightly illuminated than the peripheral area [Fig. 1a]. For this purpose, we used a shading correction without model functions^{31, 32} or a second order polynomial function³³ modeling the background image. Considering the intensity of the peak at the center which decreased toward the boundary, we tried Lorentzian and Gaussian fittings, instead of polynomial modeling, of the illumination intensity:

where x _m is the x(pixel) value of the brightest point, and y _m is the y(pixel) value of the brightest point, and Γ _x(pixel) and Γ _y(pixel) are the full widths at half maximum intensity. The Lorentzian intensity I _L and the Gaussian intensity I _G were adapted from those used by Lettington ³⁴ and Condon.³⁵ The parameter c _o was introduced to account for the deviation from a pure Lorentzian or Gaussian distribution. Thus, the parameters c _o, (x _m, y _m), and (Γ _x, Γ _y) describe the shape of illumination, and the overall factor c represents the peak intensity at (x _m, y _m). Both Lorentzian and Gaussian fittings turned out to be accurate enough. The root-mean-square residuals (RMSRs) were 3.33 [Fig. 1d] and 3.48 [Fig. 1e] for the Lorentzian and Gaussian, respectively. The degree of freedom was 3530, and the theoretical maximum RMSR value was 255. Therefore, our fittings were good to within a 2% accuracy. Fourth, a calibration of the nanoparticle density with fluorescence intensity was performed. The parameters c _o, (x _m, y _m), and (Γ _x, Γ _y) were determined by fitting the intensity curve I _G(r) or I _L(r) to the reference suspension of highest density. The overall factors, c, representing the brightnesses of reference suspensions of various FNP densities were determined by using the intensity at r = 0.

As demonstrated in Fig. 1c, the density of FNPs in the reference suspensions and their brightnesses c had a linear relation: ρV = a(c/t), where V is the volume of the reference suspension, t is the exposure time, and ais the slope of the linear relation. The coefficient a was determined by using a linear fitting of the line in Fig. 1c.

As an illustrative example, for a reference suspension, the parameters fitting the brightness profile are (x _m, y _m) = (687th pixel, 616th pixel), (Γ _x, Γ _y) = (1288 pixel, 1187 pixel), and c _o = −0.135 (dimensionless). The measured values of (t, V) in units of ms and ml were (930, 2.7), (980, 2.7), (250, 3.0), (320, 3.0). . ., for which the value c were determined by Lorentzian fitting to be, respectively, 131, 114, 180, 230, . . .. The corresponding densities of FNPs were measured to be 10, 10, 50, 50, . . . μg/ml. From these data we obtained the value of a by using a linear fitting as in Fig. 1c, and the value was 208 μg/ms.

4.2.

Quantification of the Density of Nanoparticles

The estimate of the line density of nanoparticles in the primo vessels was also a multistep process. The adjacent background intensity, which was due to tissue autoflurescence, was subtracted. The mass of the nanoparticles in the entire reference suspension, m, is given by m = ρV = ac/t, where ρ = density of nanoparticles, V = volume, a and c are previously defined. The nanoparticle density of the reference suspension is the volume density (mass over volume). Because we needed a relationship between the surface density of nanoparticles and the fluorescence intensity, we calculated the surface density, defined as the entire nanoparticle mass over the surface area of the reference suspension. The surface density is σ = m/A = ac/(tA), where A is the upper surface area of the reference suspension. Then, the nanoparticle mass in a primo vessel is given by m _p = ∫σdS = a∫cdS/(At), with the integral being a surface integrand and the line density is m _p/l, where l is the length of the primo vessel. For the integration, we first calculated c(x, y) from Eq. 1 for the Lorentzian fitting. The calculated nanoparticle line densities are given in Tables 1, 2.